class v5: the first class v5 pack, Devnet 3's epoch 0 (genesis 4020cb43... as epoch and era seed, day 20,733, the genesis state's 11 leaves under root 7e37a9fb...), generator 5, program id e5a4ac5978462156, exported on igneum-build-1 at 18:06:41Z; its Metal fingerprint 82b19cbde8557ea5 on the M5 Max at 18:07:09Z (vectors 3 of 3, cache and dataset PASS); the page's time line
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
igneum-labs
parent
99c8782c64
commit
3efd339786
15 changed files with 3824 additions and 0 deletions
|
|
@ -27,6 +27,7 @@
|
|||
| 11:47 to 11:59 | The flip run on the quiet box: the chain did everything the design says (the flip, four v5 epochs, a root per epoch, the stale miner off on 52 of 52, the stateless node stopped at 480); six `FAILED CHECK`s were the harness's own scope (it compared the stateless node's sink with the others, counted the stale miner's rejections as a fault, and read the flip epoch's stream after the executor had dropped it: captures are kept two epochs); the checks fixed |
|
||||
| 12:00 to 12:11 | PASS on every check (section 10) |
|
||||
| 12:12 to 12:23 | The no-flip case PASS (section 10); both branches pushed; master merged for the build-2 test route (d1285cef) |
|
||||
| 18:5x to 19:07 | Resumed on the project lead's first-priority order (build class v5 now, on the 0.3.23 line, byte 6): the frozen sub-version 3 (ca3-v4-amend 017e7037) merged into class-v5 and release-0.3.23-node into class-v5-node; the draw's and the acceptance rules' class keys set the state flag aside (the acceptance key had judged a v5 candidate under the pre-amendment rule and the draw diverged at instruction 0; fixed, 70 unit tests green on build-2 at 18:04Z); class v5 pinned to object byte 6 counted exactly (byte 7 is v4 sub-version 3); the gap list as section 13; the FIRST CLASS V5 PACK `proto-cuda/packs-ca3-v5/v5-dn3-epoch0` (Devnet 3's genesis 4020cb43... as epoch and era seed, day 20,733, the genesis state's 11 leaves under root 7e37a9fb...), exported on build-1 at 18:06:39 to 18:06:41Z, program id e5a4ac5978462156, its Metal fingerprint on the M5 Max under the measure lock at 18:07:04 to 18:07:09Z: 82b19cbde8557ea5, the three vector warps bit-exact, the build 29.8 ms GPU; the Metal pack bench takes `leaves.bin` |
|
||||
| 13:2x | The coordinator's correction: the hot-set bound is NOT satisfied by sub-version 1 (F8's re-gate: nine of the first 30 seeds over 1.2x, p31 at 29.3x, 0.19 percent hot-set programs; saturated values pass saturation-preserving writers); v5 inherits the gap by merge and takes sub-version 2 by merge when it lands; the AP-F4-1 and AP-F1-1 rules come in the next round; the lane stops here |
|
||||
| 12:5x to 13:21 | The amended class v4 taken in: ca3-v4-amend 8c728ca3 merged into class-v5 (4e737543) and release-0.3.20-node into class-v5-node (699db5a2); the source rule's class key sets the state flag aside so v5 and its rungs draw under it (a unit test pins the v5 chain draw equal to the amended v4's with no lossy-sourced load, the unamended v3 stream the known-failed case); generator 5 keeps `program_id(5, seed, attempt)` with no sub-version (the hash lane's reading: the suffix separates two streams inside generator 4); class v5 is object byte 6 above the amended v4's 5 (the fork, the harness, the default byte); igneum-pow 67 unit and 39 integration tests on igneum-build-2, the packs test with the re-exported control pack, the flip case PASS again with bytes 6,6,6 (the flip at epoch 8 on 3 of 3, 183 v5 blocks, the stale miner off on 59 of 59, the stateless node stopped at 480, equal roots) |
|
||||
| 10:4x | The Counter ASIC coordinator relays the project lead's ruling of option A on AP-F8-1: the injecting-source load draw lands in class v4 itself on the hash lane's branch ca3-v4-amend (0.3.19); class v5 takes it by merging that branch once its commit is up, not by a second implementation; the page's section 11 row stands as the record of the bound and its gate |
|
||||
|
|
|
|||
545
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel.cl
Normal file
545
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel.cl
Normal file
|
|
@ -0,0 +1,545 @@
|
|||
// Generated by igneum-pow export (generator v2) for seed "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000". Do not edit by hand.
|
||||
// OpenCL C twin of the Metal kernel for the same seed (see proto-opencl/README.md, WAVEFRONT.md and program.metal).
|
||||
// Built from source at runtime by proto-opencl/host.c, which passes these defines:
|
||||
// IGNEUM_GROUP work-group size of igneum_hash, a multiple of 32 (default 32: one work-group = one 32-lane unit)
|
||||
// IGNEUM_EXCHANGE 0 = local-memory exchange with a barrier (any device, any wave width; the default)
|
||||
// 1 = sub_group_shuffle_xor (cl_khr_subgroup_shuffle), only with IGNEUM_GROUP 32 and a sub-group size of exactly 32
|
||||
// 2 = intel_sub_group_shuffle_xor (cl_intel_subgroups), same condition
|
||||
// The verification unit is always 32 lanes. A 64-wide hardware wave (AMD GCN/CDNA, RDNA in wave64) runs two units;
|
||||
// the exchange masks are 1, 2, 4, 8, 16, so every partner lane lies inside the lane's own aligned run of 32.
|
||||
#ifndef IGNEUM_GROUP
|
||||
#define IGNEUM_GROUP 32
|
||||
#endif
|
||||
#ifndef IGNEUM_EXCHANGE
|
||||
#define IGNEUM_EXCHANGE 0
|
||||
#endif
|
||||
#ifdef __OPENCL_VERSION__
|
||||
#define IGNEUM_KERNEL_HASH __kernel __attribute__((reqd_work_group_size(IGNEUM_GROUP, 1, 1)))
|
||||
#define IGNEUM_LOCAL_WORDS(name, n) __local uint name[n]
|
||||
#if IGNEUM_EXCHANGE == 1
|
||||
#ifdef cl_khr_subgroups
|
||||
#pragma OPENCL EXTENSION cl_khr_subgroups : enable
|
||||
#endif
|
||||
#ifdef cl_khr_subgroup_shuffle
|
||||
#pragma OPENCL EXTENSION cl_khr_subgroup_shuffle : enable
|
||||
#endif
|
||||
#elif IGNEUM_EXCHANGE == 2
|
||||
#pragma OPENCL EXTENSION cl_intel_subgroups : enable
|
||||
#endif
|
||||
#else
|
||||
// Not an OpenCL compiler: proto-opencl/emu compiles this file as C++ and supplies the built-ins and these two macros.
|
||||
#include "emu_opencl.h"
|
||||
#endif
|
||||
|
||||
#if IGNEUM_EXCHANGE == 1
|
||||
#define IGNEUM_SHFL_XOR(dst, a, m) dst = sub_group_shuffle_xor((a), (uint)(m))
|
||||
#define IGNEUM_BCAST0(dst, a) dst = sub_group_broadcast((a), 0u)
|
||||
#elif IGNEUM_EXCHANGE == 2
|
||||
#define IGNEUM_SHFL_XOR(dst, a, m) dst = intel_sub_group_shuffle_xor((a), (uint)(m))
|
||||
#define IGNEUM_BCAST0(dst, a) dst = sub_group_broadcast((a), 0u)
|
||||
#else
|
||||
// Local-memory exchange. Two buffers of IGNEUM_GROUP words alternate (xk counts exchanges), so one barrier per
|
||||
// exchange is enough: a lane can only overwrite buffer b at exchange k+2 after passing barrier k+1, and every lane
|
||||
// reaches barrier k+1 only after its read of buffer b at exchange k. The partner lid ^ m stays inside the lane's
|
||||
// aligned run of 32 because m < 32. Control flow is uniform, so every work-item reaches every barrier.
|
||||
#define IGNEUM_SHFL_XOR(dst, a, m) { xch[(xk & 1u) * IGNEUM_GROUP + lid] = (a); barrier(CLK_LOCAL_MEM_FENCE); dst = xch[(xk & 1u) * IGNEUM_GROUP + (lid ^ (uint)(m))]; xk += 1u; }
|
||||
#define IGNEUM_BCAST0(dst, a) { xch[(xk & 1u) * IGNEUM_GROUP + lid] = (a); barrier(CLK_LOCAL_MEM_FENCE); dst = xch[(xk & 1u) * IGNEUM_GROUP + (lid & ~31u)]; xk += 1u; }
|
||||
#endif
|
||||
|
||||
static inline uint splitmix32(uint x) {
|
||||
x ^= x >> 16; x *= 0x7feb352du;
|
||||
x ^= x >> 15; x *= 0x846ca68bu;
|
||||
x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
// n is a literal in 1..31 at every call site. OpenCL rotate() rotates left by n modulo 32.
|
||||
static inline uint rotl_imm(uint x, uint n) { return rotate(x, n); }
|
||||
// Right rotation by n modulo 32 as a left rotation by (32 - n) modulo 32; n == 0 gives x.
|
||||
static inline uint rotr_var(uint x, uint n) { return rotate(x, (0u - n) & 31u); }
|
||||
static inline uint ds_elem(uint i, uint d0, uint d1) {
|
||||
uint x = i ^ d0;
|
||||
x *= 0x9E3779B1u; x ^= x >> 15;
|
||||
x += d1;
|
||||
x *= 0x85EBCA77u; x ^= x >> 13;
|
||||
x *= 0xC2B2AE3Du; x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
|
||||
// Memory-hard dataset core (MEMHARD.md). Cache: 2^26 words in 2^16 segments of 64 chained ChaCha12 lines.
|
||||
// Item: 8 rounds of 8 x seed-parameterised mixer + one 64-byte cache read, then 8 x final mixer (class v3, mixer multiplier 8,
|
||||
// docs/plans/mixer-x4.md: the round key of application j of round r is 0x9E3779B9 * (r * m + j + 1)). All parameters are literals.
|
||||
#define MH_CACHE_LINE_MASK 0x003fffffu
|
||||
#define MH_SEGMENT_LINES 64u
|
||||
#define MH_QR(a, b, c, d, r1, r2, r3, r4) { a += b; d ^= a; d = mh_rotl(d, r1); c += d; b ^= c; b = mh_rotl(b, r2); a += b; d ^= a; d = mh_rotl(d, r3); c += d; b ^= c; b = mh_rotl(b, r4); }
|
||||
static inline uint mh_rotl(uint x, uint n) { return (x << n) | (x >> (32u - n)); } // n in 1..31 at every call site
|
||||
|
||||
// y = ChaCha12 core(x) + x
|
||||
static inline void mh_chacha_block(const uint* x, uint* y) {
|
||||
for (uint i = 0u; i < 16u; ++i) y[i] = x[i];
|
||||
for (uint r = 0u; r < 6u; ++r) {
|
||||
MH_QR(y[0], y[4], y[8], y[12], 16u, 12u, 8u, 7u) MH_QR(y[1], y[5], y[9], y[13], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[6], y[10], y[14], 16u, 12u, 8u, 7u) MH_QR(y[3], y[7], y[11], y[15], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[0], y[5], y[10], y[15], 16u, 12u, 8u, 7u) MH_QR(y[1], y[6], y[11], y[12], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[7], y[8], y[13], 16u, 12u, 8u, 7u) MH_QR(y[3], y[4], y[9], y[14], 16u, 12u, 8u, 7u)
|
||||
}
|
||||
for (uint i = 0u; i < 16u; ++i) y[i] += x[i];
|
||||
}
|
||||
|
||||
// One cache segment: 64 chained lines written at cache[seg * 1024]. in_j = prev ^ (sigma || K || seg || j || tag), prev_0 = 0.
|
||||
static inline void mh_cache_segment(__global uint* cache, uint seg) {
|
||||
uint prev[16]; uint x[16]; uint y[16];
|
||||
for (uint i = 0u; i < 16u; ++i) prev[i] = 0u;
|
||||
for (uint j = 0u; j < MH_SEGMENT_LINES; ++j) {
|
||||
x[0] = 0x61707865u ^ prev[0]; x[1] = 0x3320646eu ^ prev[1]; x[2] = 0x79622d32u ^ prev[2]; x[3] = 0x6b206574u ^ prev[3];
|
||||
x[4] = 0xe23f9008u ^ prev[4];
|
||||
x[5] = 0xde33e763u ^ prev[5];
|
||||
x[6] = 0xc5ba415du ^ prev[6];
|
||||
x[7] = 0x8ddf6786u ^ prev[7];
|
||||
x[8] = 0x59f4a4beu ^ prev[8];
|
||||
x[9] = 0x8a3bc680u ^ prev[9];
|
||||
x[10] = 0x701b8e40u ^ prev[10];
|
||||
x[11] = 0x3b59025au ^ prev[11];
|
||||
x[12] = seg ^ prev[12]; x[13] = j ^ prev[13]; x[14] = 0x49676e65u ^ prev[14]; x[15] = 0x756d4d48u ^ prev[15];
|
||||
mh_chacha_block(x, y);
|
||||
__global uint* line = cache + ((seg * MH_SEGMENT_LINES + j) * 16u);
|
||||
for (uint i = 0u; i < 16u; ++i) { line[i] = y[i]; prev[i] = y[i]; }
|
||||
}
|
||||
}
|
||||
|
||||
// M_r: per word (s ^ (RC + rk)) * MUL, then a column round and a diagonal round with the seed-drawn rotations.
|
||||
static inline void mh_mixer(uint* s, uint rk) {
|
||||
s[0] = (s[0] ^ (0xe5bef5a3u + rk)) * 0xd1d79e7fu;
|
||||
s[1] = (s[1] ^ (0xdb2a7d90u + rk)) * 0xbcb61f8bu;
|
||||
s[2] = (s[2] ^ (0xcd1fa7e1u + rk)) * 0x970e91abu;
|
||||
s[3] = (s[3] ^ (0x30289419u + rk)) * 0xd749d96du;
|
||||
s[4] = (s[4] ^ (0x0a730d58u + rk)) * 0x143d7339u;
|
||||
s[5] = (s[5] ^ (0x432f8579u + rk)) * 0x2cde0d69u;
|
||||
s[6] = (s[6] ^ (0x6ab978a5u + rk)) * 0x12d0a8c1u;
|
||||
s[7] = (s[7] ^ (0x8c984f49u + rk)) * 0x2d335be9u;
|
||||
s[8] = (s[8] ^ (0x788c3c9eu + rk)) * 0x80a8aae9u;
|
||||
s[9] = (s[9] ^ (0x051eef02u + rk)) * 0x7ac896c7u;
|
||||
s[10] = (s[10] ^ (0x05e77db9u + rk)) * 0x9de23db7u;
|
||||
s[11] = (s[11] ^ (0x8b3cd4f3u + rk)) * 0xc362827du;
|
||||
s[12] = (s[12] ^ (0x6948d6cfu + rk)) * 0x5f4cdb5bu;
|
||||
s[13] = (s[13] ^ (0x0d8677f6u + rk)) * 0xfc6c5097u;
|
||||
s[14] = (s[14] ^ (0xf9505c8cu + rk)) * 0x6f547f83u;
|
||||
s[15] = (s[15] ^ (0x4513c163u + rk)) * 0x1ac31b47u;
|
||||
MH_QR(s[0], s[4], s[8], s[12], 28u, 11u, 18u, 18u) MH_QR(s[1], s[5], s[9], s[13], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[2], s[6], s[10], s[14], 28u, 11u, 18u, 18u) MH_QR(s[3], s[7], s[11], s[15], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[0], s[5], s[10], s[15], 13u, 28u, 5u, 13u) MH_QR(s[1], s[6], s[11], s[12], 13u, 28u, 5u, 13u)
|
||||
MH_QR(s[2], s[7], s[8], s[13], 13u, 28u, 5u, 13u) MH_QR(s[3], s[4], s[9], s[14], 13u, 28u, 5u, 13u)
|
||||
}
|
||||
|
||||
// Item t: 16 words. s = (K, t * MUL[i] + RC[i]); 8 rounds of 8 x mixer + cache line s[0] & mask; 8 x final mixer.
|
||||
static inline void mh_item(__global const uint* cache, __global const uint* leaf, uint t, uint* s) {
|
||||
s[0] = 0xe23f9008u;
|
||||
s[1] = 0xde33e763u;
|
||||
s[2] = 0xc5ba415du;
|
||||
s[3] = 0x8ddf6786u;
|
||||
s[4] = 0x59f4a4beu;
|
||||
s[5] = 0x8a3bc680u;
|
||||
s[6] = 0x701b8e40u;
|
||||
s[7] = 0x3b59025au;
|
||||
s[8] = t * 0xd1d79e7fu + 0xe5bef5a3u;
|
||||
s[9] = t * 0xbcb61f8bu + 0xdb2a7d90u;
|
||||
s[10] = t * 0x970e91abu + 0xcd1fa7e1u;
|
||||
s[11] = t * 0xd749d96du + 0x30289419u;
|
||||
s[12] = t * 0x143d7339u + 0x0a730d58u;
|
||||
s[13] = t * 0x2cde0d69u + 0x432f8579u;
|
||||
s[14] = t * 0x12d0a8c1u + 0x6ab978a5u;
|
||||
s[15] = t * 0x2d335be9u + 0x8c984f49u;
|
||||
for (uint i = 0u; i < 16u; ++i) s[i] ^= leaf[i];
|
||||
for (uint r = 0u; r < 8u; ++r) {
|
||||
for (uint j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (r * 8u + j + 1u));
|
||||
__global const uint* line = cache + ((s[0] & MH_CACHE_LINE_MASK) * 16u);
|
||||
for (uint i = 0u; i < 16u; ++i) s[i] ^= line[i];
|
||||
}
|
||||
for (uint j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (64u + j + 1u));
|
||||
}
|
||||
// Class v5 (docs/design/class-v5-stored-state.md): leaf(t) = leaves[t mod nLeaves], 16 words per leaf (leaves.bin).
|
||||
static inline __global const uint* mh_leaf(__global const uint* leaves, uint nLeaves, uint t) { return leaves + ((t % nLeaves) * 16u); }
|
||||
// Era layout (docs/plans/era-layout.md 1.2): dataset word w holds word mh_j(w) of item mh_t(w); j's bits sit at positions 1 9 13 14 of w.
|
||||
static inline uint mh_j(uint w) { return ((w >> 1u) & 1u) | (((w >> 9u) & 1u) << 1) | (((w >> 13u) & 1u) << 2) | (((w >> 14u) & 1u) << 3); }
|
||||
static inline uint mh_t(uint w) { w = (w & 0x00003fffu) | ((w >> 15u) << 14u); w = (w & 0x00001fffu) | ((w >> 14u) << 13u); w = (w & 0x000001ffu) | ((w >> 10u) << 9u); w = (w & 0x00000001u) | ((w >> 2u) << 1u); return w; }
|
||||
static inline uint mh_addr(uint t, uint j) { uint w = t; w = ((w >> 1u) << 2u) | (w & 0x00000001u) | (((j >> 0u) & 1u) << 1u); w = ((w >> 9u) << 10u) | (w & 0x000001ffu) | (((j >> 1u) & 1u) << 9u); w = ((w >> 13u) << 14u) | (w & 0x00001fffu) | (((j >> 2u) & 1u) << 13u); w = ((w >> 14u) << 15u) | (w & 0x00003fffu) | (((j >> 3u) & 1u) << 14u); return w; }
|
||||
// dataset[w] without the dataset: derive item mh_t(w) and take word mh_j(w).
|
||||
static inline uint mh_word(__global const uint* cache, __global const uint* leaves, uint nLeaves, uint w) { uint s[16]; mh_item(cache, mh_leaf(leaves, nLeaves, mh_t(w)), mh_t(w), s); return s[mh_j(w)]; }
|
||||
|
||||
// Memory-hard dataset (MEMHARD.md). One work-item per cache segment; one work-item per 64-byte dataset item.
|
||||
// The same constants as memhard.h in this pack (one emitter, three dialects).
|
||||
__kernel void igneum_cache_fill(__global uint* cache, uint nSegments) {
|
||||
uint seg = (uint)get_global_id(0);
|
||||
if (seg < nSegments) mh_cache_segment(cache, seg);
|
||||
}
|
||||
// Class v5: the window's leaves (leaves.bin, IGNEUM_STATE_LEAVES x 16 words) and their count.
|
||||
__kernel void igneum_build(__global uint* ds, __global const uint* cache, __global const uint* leaves, uint nLeaves, uint nItems) {
|
||||
uint t = (uint)get_global_id(0);
|
||||
if (t < nItems) {
|
||||
uint s[16];
|
||||
mh_item(cache, mh_leaf(leaves, nLeaves, t), t, s);
|
||||
for (uint i = 0u; i < 16u; ++i) ds[(ulong)mh_addr(t, i)] = s[i];
|
||||
}
|
||||
}
|
||||
|
||||
// One hash per work-item. IGNEUM_GROUP is a multiple of 32; lane = lid & 31 and every exchange stays inside the
|
||||
// lane's own aligned run of 32 work-items, exactly like simd_shuffle_xor inside a 32-wide Metal SIMD group and
|
||||
// __shfl_xor_sync inside a CUDA warp. Control flow is uniform (no branches at all).
|
||||
IGNEUM_KERNEL_HASH void igneum_hash(__global const uint* ds, __global ulong* out, uint baseNonce, uint mask) {
|
||||
uint gid = (uint)get_global_id(0);
|
||||
uint lid = (uint)get_local_id(0);
|
||||
uint nonce = baseNonce + gid;
|
||||
uint r0, r1, r2, r3, r4, r5, r6, r7;
|
||||
#if IGNEUM_EXCHANGE == 0
|
||||
IGNEUM_LOCAL_WORDS(xch, 2 * IGNEUM_GROUP);
|
||||
uint xk = 0u;
|
||||
#else
|
||||
(void)lid;
|
||||
#endif
|
||||
{ uint x = nonce ^ 0xbe8c5a0cu; x += 0x9e3779b9u; x = splitmix32(x); r0 = x ^ 0x7646e626u; } // SEEDW[0], 0x9e3779b9u * 1u, SEEDW[1]
|
||||
{ uint x = nonce ^ 0x7646e626u; x += 0x3c6ef372u; x = splitmix32(x); r1 = x ^ 0x508e29d0u; } // SEEDW[1], 0x9e3779b9u * 2u, SEEDW[2]
|
||||
{ uint x = nonce ^ 0x508e29d0u; x += 0xdaa66d2bu; x = splitmix32(x); r2 = x ^ 0xfb8a5b4au; } // SEEDW[2], 0x9e3779b9u * 3u, SEEDW[3]
|
||||
{ uint x = nonce ^ 0xfb8a5b4au; x += 0x78dde6e4u; x = splitmix32(x); r3 = x ^ 0x53c50955u; } // SEEDW[3], 0x9e3779b9u * 4u, SEEDW[4]
|
||||
{ uint x = nonce ^ 0x53c50955u; x += 0x1715609du; x = splitmix32(x); r4 = x ^ 0x685d62fcu; } // SEEDW[4], 0x9e3779b9u * 5u, SEEDW[5]
|
||||
{ uint x = nonce ^ 0x685d62fcu; x += 0xb54cda56u; x = splitmix32(x); r5 = x ^ 0x6065c013u; } // SEEDW[5], 0x9e3779b9u * 6u, SEEDW[6]
|
||||
{ uint x = nonce ^ 0x6065c013u; x += 0x5384540fu; x = splitmix32(x); r6 = x ^ 0x881096ebu; } // SEEDW[6], 0x9e3779b9u * 7u, SEEDW[7]
|
||||
{ uint x = nonce ^ 0x881096ebu; x += 0xf1bbcdc8u; x = splitmix32(x); r7 = x ^ 0xbe8c5a0cu; } // SEEDW[7], 0x9e3779b9u * 8u, SEEDW[0]
|
||||
|
||||
for (uint it = 0u; it < 8u; ++it) {
|
||||
uint sel = r0;
|
||||
r6 = r6 - r3; // 0 sub
|
||||
r2 = r2 | r6; // 1 or
|
||||
r4 = r4 + r0 + ((((sel >> 4u) & 1u) != 0u) ? 0x7731324bu : 0x1a579b38u); // 2 add
|
||||
r3 = r3 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 3 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 1u); r0 = r0 ^ t_; } // 4 shfl
|
||||
r4 = r4 ^ r1; // 5 xor
|
||||
r3 = r5 * r3 + r3; // 6 mad
|
||||
r1 = r1 * r6; // 7 mul
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x08000000u) & mask]; // 8 load
|
||||
r0 = r0 ^ r4; // 9 xor
|
||||
r0 = r0 + r7 + ((((sel >> 30u) & 1u) != 0u) ? 0xb10fcef8u : 0xee083919u); // 10 add
|
||||
r0 = mul_hi(r0, r2); // 11 mulhi
|
||||
r5 = r5 ^ r4; // 12 xor
|
||||
r7 = r3 * r0 + r7; // 13 mad
|
||||
r3 = r3 ^ ds[((rotl_imm(r4 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 14 load
|
||||
r2 = r2 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 15 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r7 = r7 ^ t_; } // 16 shfl
|
||||
r7 = r7 * r6; // 17 mul
|
||||
r2 = r3 * r1 + r2; // 18 mad
|
||||
r5 = rotr_var(r5, r2); // 19 rotr
|
||||
r5 = r5 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 20 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 16u); r6 = r6 ^ t_; } // 21 shfl
|
||||
r1 = mul_hi(r1, r2); // 22 mulhi
|
||||
r6 = r6 ^ r5; // 23 xor
|
||||
r5 = r2 * r0 + r5; // 24 mad
|
||||
r6 = r3 * r7 + r6; // 25 mad
|
||||
r6 = r6 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 26 load
|
||||
r1 = r5 * r5 + r1; // 27 mad
|
||||
r0 = r0 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 28 load
|
||||
r7 = r7 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0x0b1f02c7u : 0xd5e6c37cu); // 29 add
|
||||
r7 = r7 + r0 + ((((sel >> 0u) & 1u) != 0u) ? 0x2d6e8bd1u : 0x4e45ba51u); // 30 add
|
||||
r0 = r0 ^ r2; // 31 xor
|
||||
r6 = r6 + r2 + ((((sel >> 13u) & 1u) != 0u) ? 0x351422d2u : 0x1c91b2a1u); // 32 add
|
||||
r4 = mul_hi(r4, r5); // 33 mulhi
|
||||
r3 = rotr_var(r3, r5); // 34 rotr
|
||||
r3 = r3 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 35 load
|
||||
r4 = rotl_imm(r4, 13u); // 36 rotl
|
||||
r6 = r6 + r1 + ((((sel >> 7u) & 1u) != 0u) ? 0x64a28251u : 0x3f2970f7u); // 37 add
|
||||
r2 = r2 * r0; // 38 mul
|
||||
r3 = r3 + r4 + ((((sel >> 15u) & 1u) != 0u) ? 0x7378c955u : 0x3b7b3317u); // 39 add
|
||||
r0 = r0 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 40 load
|
||||
r6 = r6 ^ r5; // 41 xor
|
||||
r3 = r3 + r0 + ((((sel >> 2u) & 1u) != 0u) ? 0xb31f6a64u : 0x6678b059u); // 42 add
|
||||
r7 = r7 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 43 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 4u); r3 = r3 ^ t_; } // 44 shfl
|
||||
r1 = mul_hi(r1, r5); // 45 mulhi
|
||||
r0 = rotl_imm(r0, 6u); // 46 rotl
|
||||
r1 = r1 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 47 load
|
||||
r4 = r4 + r0 + ((((sel >> 17u) & 1u) != 0u) ? 0x5feccee6u : 0x43095946u); // 48 add
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 49 load
|
||||
r2 = r2 - r3; // 50 sub
|
||||
r7 = r7 + r2 + ((((sel >> 24u) & 1u) != 0u) ? 0x26e2b582u : 0xf45ecdf8u); // 51 add
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 52 load
|
||||
r6 = r6 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 53 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 4u); r7 = r7 ^ t_; } // 54 shfl
|
||||
r1 = rotr_var(r1, r2); // 55 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r0 = r0 ^ t_; } // 56 shfl
|
||||
r5 = rotl_imm(r5, 12u); // 57 rotl
|
||||
r5 = r5 + r3 + ((((sel >> 4u) & 1u) != 0u) ? 0x473b1718u : 0x6b4f35e8u); // 58 add
|
||||
r4 = rotr_var(r4, r1); // 59 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 1u); r4 = r4 ^ t_; } // 60 shfl
|
||||
r7 = r7 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 61 load
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 62 load
|
||||
r4 = r4 * r1; // 63 mul
|
||||
// latency-shadow block (Counter ASIC 3.0 item 8): 256 ALU instructions x 27 passes after instruction 63, no load
|
||||
for (uint sh = 0u; sh < 27u; ++sh) {
|
||||
r6 = r6 | r0; // s0 or
|
||||
r7 = r7 * r3; // s1 mul
|
||||
r3 = r5 * r5 + r3; // s2 mad
|
||||
r7 = r7 + r4 + ((((sel >> 6u) & 1u) != 0u) ? 0x38e58a06u : 0x97d3d105u); // s3 add
|
||||
r6 = r6 + r0 + ((((sel >> 20u) & 1u) != 0u) ? 0x9629673du : 0x29b87b9au); // s4 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 1u); r6 = r6 ^ t_; } // s5 shfl
|
||||
r5 = r3 * r5 + r5; // s6 mad
|
||||
r3 = rotl_imm(r3, 5u); // s7 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 12u) & 1u) != 0u) ? 0x4a5f55e1u : 0xa7c1fb0fu); // s8 add
|
||||
r5 = r5 * r7; // s9 mul
|
||||
r4 = r4 ^ r5; // s10 xor
|
||||
r6 = mul_hi(r6, r4); // s11 mulhi
|
||||
r5 = rotl_imm(r5, 13u); // s12 rotl
|
||||
r0 = r2 * r1 + r0; // s13 mad
|
||||
r2 = mul_hi(r2, r4); // s14 mulhi
|
||||
r3 = r3 | r1; // s15 or
|
||||
r0 = r0 - r5; // s16 sub
|
||||
r0 = r0 * r6; // s17 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 4u); r5 = r5 ^ t_; } // s18 shfl
|
||||
r3 = r3 | r4; // s19 or
|
||||
r6 = r3 * r2 + r6; // s20 mad
|
||||
r0 = rotl_imm(r0, 11u); // s21 rotl
|
||||
r3 = rotl_imm(r3, 25u); // s22 rotl
|
||||
r6 = r6 ^ r4; // s23 xor
|
||||
r4 = r4 | r2; // s24 or
|
||||
r1 = r1 + r6 + ((((sel >> 7u) & 1u) != 0u) ? 0xfc2f6d17u : 0x53e74799u); // s25 add
|
||||
r7 = rotr_var(r7, r6); // s26 rotr
|
||||
r7 = r7 + r5 + ((((sel >> 27u) & 1u) != 0u) ? 0x787048dcu : 0xb8fae90eu); // s27 add
|
||||
r4 = r4 ^ r1; // s28 xor
|
||||
r6 = r6 ^ r1; // s29 xor
|
||||
r2 = mul_hi(r2, r0); // s30 mulhi
|
||||
r1 = rotr_var(r1, r4); // s31 rotr
|
||||
r7 = r7 ^ r4; // s32 xor
|
||||
r3 = r2 * r6 + r3; // s33 mad
|
||||
r3 = r3 + r1 + ((((sel >> 12u) & 1u) != 0u) ? 0x9ffd510bu : 0xd2dc42ebu); // s34 add
|
||||
r4 = r4 + r7 + ((((sel >> 7u) & 1u) != 0u) ? 0xf871ba37u : 0xad8eda6fu); // s35 add
|
||||
r7 = r7 + r4 + ((((sel >> 3u) & 1u) != 0u) ? 0xd02d30cau : 0xb0cef8f4u); // s36 add
|
||||
r4 = r3 * r2 + r4; // s37 mad
|
||||
r5 = r7 * r7 + r5; // s38 mad
|
||||
r3 = r3 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x82e98a22u : 0x2bd506e2u); // s39 add
|
||||
r0 = rotl_imm(r0, 21u); // s40 rotl
|
||||
r1 = r1 | r0; // s41 or
|
||||
r7 = r7 + r5 + ((((sel >> 28u) & 1u) != 0u) ? 0x6435f6d8u : 0xc0386e49u); // s42 add
|
||||
r1 = mul_hi(r1, r0); // s43 mulhi
|
||||
r5 = r5 + r6 + ((((sel >> 12u) & 1u) != 0u) ? 0x1f682c68u : 0x3a8921a4u); // s44 add
|
||||
r4 = r4 + r5 + ((((sel >> 1u) & 1u) != 0u) ? 0xf246e180u : 0xeeca4334u); // s45 add
|
||||
r2 = r2 ^ r3; // s46 xor
|
||||
r2 = r2 ^ r5; // s47 xor
|
||||
r7 = r4 * r1 + r7; // s48 mad
|
||||
r2 = r6 * r1 + r2; // s49 mad
|
||||
r7 = r0 * r7 + r7; // s50 mad
|
||||
r5 = r5 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x41bd68ccu : 0x745776d8u); // s51 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 1u); r3 = r3 ^ t_; } // s52 shfl
|
||||
r4 = r4 | r0; // s53 or
|
||||
r0 = r2 * r4 + r0; // s54 mad
|
||||
r5 = rotl_imm(r5, 9u); // s55 rotl
|
||||
r5 = r5 * r7; // s56 mul
|
||||
r7 = rotl_imm(r7, 27u); // s57 rotl
|
||||
r7 = r7 - r6; // s58 sub
|
||||
r2 = rotl_imm(r2, 22u); // s59 rotl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 2u); r7 = r7 ^ t_; } // s60 shfl
|
||||
r4 = r4 + r0 + ((((sel >> 7u) & 1u) != 0u) ? 0xcba6a161u : 0x60d0ff55u); // s61 add
|
||||
r1 = r1 * r5; // s62 mul
|
||||
r7 = mul_hi(r7, r6); // s63 mulhi
|
||||
r5 = r5 * r0; // s64 mul
|
||||
r2 = rotr_var(r2, r5); // s65 rotr
|
||||
r0 = mul_hi(r0, r3); // s66 mulhi
|
||||
r0 = r0 | r7; // s67 or
|
||||
r0 = rotl_imm(r0, 19u); // s68 rotl
|
||||
r0 = rotl_imm(r0, 1u); // s69 rotl
|
||||
r2 = r2 | r3; // s70 or
|
||||
r3 = r7 * r5 + r3; // s71 mad
|
||||
r7 = r0 * r4 + r7; // s72 mad
|
||||
r4 = r4 | r3; // s73 or
|
||||
r3 = r3 ^ r1; // s74 xor
|
||||
r1 = r1 + r3 + ((((sel >> 15u) & 1u) != 0u) ? 0x8e35924eu : 0x3f4ffea2u); // s75 add
|
||||
r1 = r1 * r3; // s76 mul
|
||||
r0 = r3 * r2 + r0; // s77 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 16u); r2 = r2 ^ t_; } // s78 shfl
|
||||
r1 = mul_hi(r1, r0); // s79 mulhi
|
||||
r0 = mul_hi(r0, r4); // s80 mulhi
|
||||
r4 = r4 * r2; // s81 mul
|
||||
r5 = r5 - r7; // s82 sub
|
||||
r0 = r2 * r3 + r0; // s83 mad
|
||||
r7 = r7 - r2; // s84 sub
|
||||
r1 = r1 ^ r2; // s85 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r3, 16u); r6 = r6 ^ t_; } // s86 shfl
|
||||
r3 = r0 * r0 + r3; // s87 mad
|
||||
r0 = rotl_imm(r0, 20u); // s88 rotl
|
||||
r1 = r1 - r5; // s89 sub
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 16u); r6 = r6 ^ t_; } // s90 shfl
|
||||
r5 = r5 ^ r1; // s91 xor
|
||||
r5 = rotl_imm(r5, 12u); // s92 rotl
|
||||
r6 = r1 * r2 + r6; // s93 mad
|
||||
r2 = rotr_var(r2, r5); // s94 rotr
|
||||
r0 = r0 | r4; // s95 or
|
||||
r7 = r7 * r3; // s96 mul
|
||||
r5 = r5 | r0; // s97 or
|
||||
r3 = r3 + r1 + ((((sel >> 13u) & 1u) != 0u) ? 0x386bb642u : 0xbcfdd8a7u); // s98 add
|
||||
r0 = r0 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xa3962d03u : 0xc921a021u); // s99 add
|
||||
r1 = rotr_var(r1, r0); // s100 rotr
|
||||
r2 = r2 + r4 + ((((sel >> 18u) & 1u) != 0u) ? 0xc5990c5cu : 0x70da4067u); // s101 add
|
||||
r2 = rotl_imm(r2, 3u); // s102 rotl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 4u); r1 = r1 ^ t_; } // s103 shfl
|
||||
r0 = r0 ^ r2; // s104 xor
|
||||
r4 = r4 + r3 + ((((sel >> 28u) & 1u) != 0u) ? 0xb15aec75u : 0x1a5880cbu); // s105 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 2u); r4 = r4 ^ t_; } // s106 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 4u); r7 = r7 ^ t_; } // s107 shfl
|
||||
r1 = r7 * r3 + r1; // s108 mad
|
||||
r0 = r2 * r3 + r0; // s109 mad
|
||||
r2 = r2 | r4; // s110 or
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 1u); r6 = r6 ^ t_; } // s111 shfl
|
||||
r2 = r5 * r3 + r2; // s112 mad
|
||||
r4 = mul_hi(r4, r0); // s113 mulhi
|
||||
r7 = rotl_imm(r7, 29u); // s114 rotl
|
||||
r3 = r3 - r7; // s115 sub
|
||||
r0 = mul_hi(r0, r1); // s116 mulhi
|
||||
r4 = rotl_imm(r4, 12u); // s117 rotl
|
||||
r1 = r1 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0xc5e98ee5u : 0x24494ad3u); // s118 add
|
||||
r0 = r0 + r4 + ((((sel >> 4u) & 1u) != 0u) ? 0xa97bc949u : 0x1b2d1c83u); // s119 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 4u); r5 = r5 ^ t_; } // s120 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 8u); r3 = r3 ^ t_; } // s121 shfl
|
||||
r7 = rotl_imm(r7, 3u); // s122 rotl
|
||||
r4 = rotr_var(r4, r6); // s123 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 1u); r6 = r6 ^ t_; } // s124 shfl
|
||||
r5 = rotl_imm(r5, 21u); // s125 rotl
|
||||
r0 = r3 * r1 + r0; // s126 mad
|
||||
r0 = r0 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0xa7f6bb54u : 0xff14c08du); // s127 add
|
||||
r7 = r7 ^ r4; // s128 xor
|
||||
r3 = mul_hi(r3, r5); // s129 mulhi
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 8u); r5 = r5 ^ t_; } // s130 shfl
|
||||
r6 = mul_hi(r6, r4); // s131 mulhi
|
||||
r5 = r5 ^ r3; // s132 xor
|
||||
r0 = rotl_imm(r0, 19u); // s133 rotl
|
||||
r4 = rotr_var(r4, r3); // s134 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 16u); r1 = r1 ^ t_; } // s135 shfl
|
||||
r5 = r5 - r2; // s136 sub
|
||||
r3 = r3 ^ r5; // s137 xor
|
||||
r0 = r0 - r2; // s138 sub
|
||||
r3 = r3 * r7; // s139 mul
|
||||
r5 = r5 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0x2d465ca8u : 0x82ab98f4u); // s140 add
|
||||
r4 = r4 + r1 + ((((sel >> 11u) & 1u) != 0u) ? 0x237a9d8eu : 0x80594390u); // s141 add
|
||||
r3 = r7 * r0 + r3; // s142 mad
|
||||
r0 = r0 | r3; // s143 or
|
||||
r0 = mul_hi(r0, r1); // s144 mulhi
|
||||
r6 = r6 ^ r2; // s145 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 16u); r7 = r7 ^ t_; } // s146 shfl
|
||||
r0 = r0 + r1 + ((((sel >> 18u) & 1u) != 0u) ? 0xe68cf57eu : 0x07f15148u); // s147 add
|
||||
r7 = r7 | r3; // s148 or
|
||||
r4 = r4 ^ r2; // s149 xor
|
||||
r7 = r7 ^ r1; // s150 xor
|
||||
r5 = rotr_var(r5, r2); // s151 rotr
|
||||
r0 = r0 * r7; // s152 mul
|
||||
r4 = rotl_imm(r4, 16u); // s153 rotl
|
||||
r6 = rotr_var(r6, r4); // s154 rotr
|
||||
r6 = r6 ^ r1; // s155 xor
|
||||
r6 = r6 | r3; // s156 or
|
||||
r1 = r1 + r6 + ((((sel >> 19u) & 1u) != 0u) ? 0x84c2a09du : 0x14878c5au); // s157 add
|
||||
r4 = r4 - r1; // s158 sub
|
||||
r4 = r6 * r7 + r4; // s159 mad
|
||||
r1 = r1 * r5; // s160 mul
|
||||
r4 = r4 | r0; // s161 or
|
||||
r7 = rotl_imm(r7, 21u); // s162 rotl
|
||||
r0 = r0 - r1; // s163 sub
|
||||
r1 = r1 * r0; // s164 mul
|
||||
r2 = r2 - r1; // s165 sub
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 2u); r0 = r0 ^ t_; } // s166 shfl
|
||||
r0 = r0 - r4; // s167 sub
|
||||
r1 = rotr_var(r1, r6); // s168 rotr
|
||||
r7 = rotr_var(r7, r1); // s169 rotr
|
||||
r3 = r1 * r3 + r3; // s170 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 4u); r2 = r2 ^ t_; } // s171 shfl
|
||||
r2 = r2 - r1; // s172 sub
|
||||
r7 = r7 ^ r6; // s173 xor
|
||||
r4 = rotr_var(r4, r1); // s174 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r4 = r4 ^ t_; } // s175 shfl
|
||||
r6 = r6 * r2; // s176 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 4u); r0 = r0 ^ t_; } // s177 shfl
|
||||
r6 = r6 + r4 + ((((sel >> 21u) & 1u) != 0u) ? 0x509871e6u : 0x4fa43965u); // s178 add
|
||||
r0 = r0 ^ r6; // s179 xor
|
||||
r5 = r5 + r7 + ((((sel >> 0u) & 1u) != 0u) ? 0xb95b8a53u : 0xfb3c4bd8u); // s180 add
|
||||
r6 = r6 * r3; // s181 mul
|
||||
r7 = rotl_imm(r7, 1u); // s182 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 11u) & 1u) != 0u) ? 0x7047b7cfu : 0xc448a197u); // s183 add
|
||||
r3 = r3 * r5; // s184 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 1u); r5 = r5 ^ t_; } // s185 shfl
|
||||
r0 = r0 * r2; // s186 mul
|
||||
r4 = mul_hi(r4, r7); // s187 mulhi
|
||||
r2 = r7 * r2 + r2; // s188 mad
|
||||
r7 = r7 - r2; // s189 sub
|
||||
r2 = r2 + r5 + ((((sel >> 19u) & 1u) != 0u) ? 0x5caccc5du : 0x2fceaf49u); // s190 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 1u); r1 = r1 ^ t_; } // s191 shfl
|
||||
r7 = r7 + r1 + ((((sel >> 20u) & 1u) != 0u) ? 0x857bb5feu : 0xba0cee62u); // s192 add
|
||||
r6 = rotl_imm(r6, 22u); // s193 rotl
|
||||
r3 = r3 ^ r7; // s194 xor
|
||||
r7 = r7 * r0; // s195 mul
|
||||
r3 = r3 ^ r5; // s196 xor
|
||||
r5 = r5 + r1 + ((((sel >> 29u) & 1u) != 0u) ? 0xeae84577u : 0x8519428cu); // s197 add
|
||||
r7 = r4 * r4 + r7; // s198 mad
|
||||
r3 = r3 + r6 + ((((sel >> 8u) & 1u) != 0u) ? 0x0bfdbfa1u : 0x468639d3u); // s199 add
|
||||
r5 = r3 * r4 + r5; // s200 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 4u); r2 = r2 ^ t_; } // s201 shfl
|
||||
r7 = r7 + r4 + ((((sel >> 23u) & 1u) != 0u) ? 0x53ee8e50u : 0x18ec9a69u); // s202 add
|
||||
r6 = r6 ^ r0; // s203 xor
|
||||
r4 = r4 + r5 + ((((sel >> 4u) & 1u) != 0u) ? 0xac63376eu : 0x3e81485bu); // s204 add
|
||||
r7 = r1 * r4 + r7; // s205 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r1 = r1 ^ t_; } // s206 shfl
|
||||
r4 = r4 * r0; // s207 mul
|
||||
r1 = r1 | r6; // s208 or
|
||||
r6 = r6 + r4 + ((((sel >> 28u) & 1u) != 0u) ? 0x5b745519u : 0x66ccb75du); // s209 add
|
||||
r1 = rotr_var(r1, r3); // s210 rotr
|
||||
r5 = r5 - r4; // s211 sub
|
||||
r4 = r4 ^ r7; // s212 xor
|
||||
r1 = r1 + r7 + ((((sel >> 8u) & 1u) != 0u) ? 0x1c3dccf7u : 0x0ce0553du); // s213 add
|
||||
r0 = mul_hi(r0, r1); // s214 mulhi
|
||||
r3 = r3 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xfadc9205u : 0x96bfec89u); // s215 add
|
||||
r0 = rotl_imm(r0, 4u); // s216 rotl
|
||||
r6 = r6 ^ r4; // s217 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r6 = r6 ^ t_; } // s218 shfl
|
||||
r7 = r0 * r1 + r7; // s219 mad
|
||||
r4 = r4 * r0; // s220 mul
|
||||
r2 = r6 * r0 + r2; // s221 mad
|
||||
r5 = r5 ^ r4; // s222 xor
|
||||
r0 = r0 * r5; // s223 mul
|
||||
r2 = r2 + r0 + ((((sel >> 9u) & 1u) != 0u) ? 0x66b15c04u : 0x61495681u); // s224 add
|
||||
r3 = rotr_var(r3, r2); // s225 rotr
|
||||
r2 = r3 * r3 + r2; // s226 mad
|
||||
r6 = r6 ^ r0; // s227 xor
|
||||
r4 = rotl_imm(r4, 30u); // s228 rotl
|
||||
r2 = r2 + r3 + ((((sel >> 29u) & 1u) != 0u) ? 0x25bc17c2u : 0x70d05c34u); // s229 add
|
||||
r1 = rotr_var(r1, r5); // s230 rotr
|
||||
r1 = r1 + r4 + ((((sel >> 8u) & 1u) != 0u) ? 0x2d8728f3u : 0x32a28384u); // s231 add
|
||||
r1 = rotl_imm(r1, 6u); // s232 rotl
|
||||
r2 = r2 + r0 + ((((sel >> 21u) & 1u) != 0u) ? 0x61e81fe6u : 0xcf0949f1u); // s233 add
|
||||
r4 = rotl_imm(r4, 2u); // s234 rotl
|
||||
r2 = r2 - r6; // s235 sub
|
||||
r6 = r6 | r3; // s236 or
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 4u); r7 = r7 ^ t_; } // s237 shfl
|
||||
r1 = r1 ^ r3; // s238 xor
|
||||
r2 = rotl_imm(r2, 1u); // s239 rotl
|
||||
r5 = rotr_var(r5, r2); // s240 rotr
|
||||
r4 = rotl_imm(r4, 10u); // s241 rotl
|
||||
r6 = r6 + r2 + ((((sel >> 16u) & 1u) != 0u) ? 0x0a67565du : 0xe204fb50u); // s242 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r3, 2u); r1 = r1 ^ t_; } // s243 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 16u); r0 = r0 ^ t_; } // s244 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r5 = r5 ^ t_; } // s245 shfl
|
||||
r4 = r4 ^ r5; // s246 xor
|
||||
r7 = r7 ^ r2; // s247 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 16u); r0 = r0 ^ t_; } // s248 shfl
|
||||
r1 = mul_hi(r1, r3); // s249 mulhi
|
||||
r7 = r7 + r5 + ((((sel >> 31u) & 1u) != 0u) ? 0x5a79fd6fu : 0x3d6daffdu); // s250 add
|
||||
r6 = r6 + r0 + ((((sel >> 28u) & 1u) != 0u) ? 0x6beb28c2u : 0x84334aeau); // s251 add
|
||||
r7 = r7 - r1; // s252 sub
|
||||
r5 = rotr_var(r5, r2); // s253 rotr
|
||||
r2 = mul_hi(r2, r5); // s254 mulhi
|
||||
r6 = r6 * r3; // s255 mul
|
||||
}
|
||||
}
|
||||
uint lo = r0 ^ rotl_imm(r1, 7u) ^ rotl_imm(r2, 14u) ^ rotl_imm(r3, 21u);
|
||||
uint hi = r4 ^ rotl_imm(r5, 9u) ^ rotl_imm(r6, 18u) ^ rotl_imm(r7, 27u);
|
||||
out[gid] = ((ulong)hi << 32) | (ulong)lo;
|
||||
}
|
||||
|
||||
#if IGNEUM_EXCHANGE != 0
|
||||
// Reports the sub-group size this device uses for a work-group of IGNEUM_GROUP items. host.c runs it only when the
|
||||
// per-kernel query (clGetKernelSubGroupInfoKHR on igneum_hash) is unavailable; that query is preferred because a
|
||||
// compiler may pick a different wave width per kernel (RDNA: wave32 or wave64). See WAVEFRONT.md.
|
||||
IGNEUM_KERNEL_HASH void igneum_probe_subgroup(__global uint* out) {
|
||||
if (get_local_id(0) == 0u) { out[0] = get_sub_group_size(); out[1] = get_num_sub_groups(); }
|
||||
}
|
||||
#endif
|
||||
423
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel.cu
Normal file
423
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel.cu
Normal file
|
|
@ -0,0 +1,423 @@
|
|||
// Generated by igneum-pow export (generator v2) for seed "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000". Do not edit by hand.
|
||||
// Bit-exact twin of the Metal kernel for the same seed (see proto-cuda/CHECKLIST.md and program.metal).
|
||||
// Compiled ahead of time by nvcc together with proto-cuda/host.cu. No NVRTC.
|
||||
#include <cuda_runtime.h>
|
||||
#include <cstdint>
|
||||
#include "program.h"
|
||||
#include "memhard.h"
|
||||
|
||||
__device__ __forceinline__ uint32_t splitmix32(uint32_t x) {
|
||||
x ^= x >> 16; x *= 0x7feb352du;
|
||||
x ^= x >> 15; x *= 0x846ca68bu;
|
||||
x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
// n is a literal in 1..31 at every call site, so both shift amounts are in 1..31.
|
||||
__device__ __forceinline__ uint32_t rotl_imm(uint32_t x, uint32_t n) { return (x << n) | (x >> (32u - n)); }
|
||||
// n is masked to 0..31; the second shift amount is masked too, so n == 0 gives x.
|
||||
__device__ __forceinline__ uint32_t rotr_var(uint32_t x, uint32_t n) { n &= 31u; return (x >> n) | (x << ((32u - n) & 31u)); }
|
||||
__device__ __forceinline__ uint32_t ds_elem(uint32_t i, uint32_t d0, uint32_t d1) {
|
||||
uint32_t x = i ^ d0;
|
||||
x *= 0x9E3779B1u; x ^= x >> 15;
|
||||
x += d1;
|
||||
x *= 0x85EBCA77u; x ^= x >> 13;
|
||||
x *= 0xC2B2AE3Du; x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
|
||||
// Memory-hard dataset (MEMHARD.md). One thread per cache segment; one thread per 64-byte dataset item.
|
||||
// The core functions (mh_cache_segment, mh_item) are in memhard.h and are also compiled for the host.
|
||||
__global__ void igneum_cache_fill(uint32_t* cache, uint32_t nSegments) {
|
||||
uint32_t seg = blockIdx.x * blockDim.x + threadIdx.x;
|
||||
if (seg < nSegments) mh_cache_segment(cache, seg);
|
||||
}
|
||||
// Class v5: the window's leaves (leaves.bin, IGNEUM_STATE_LEAVES x 16 words) and their count.
|
||||
__global__ void igneum_build(uint32_t* ds, const uint32_t* cache, const uint32_t* leaves, uint32_t nLeaves, uint32_t nItems) {
|
||||
uint32_t t = blockIdx.x * blockDim.x + threadIdx.x;
|
||||
if (t < nItems) {
|
||||
uint32_t s[16];
|
||||
mh_item(cache, mh_leaf(leaves, nLeaves, t), t, s);
|
||||
for (uint32_t i = 0u; i < 16u; ++i) ds[(size_t)mh_addr(t, i)] = s[i];
|
||||
}
|
||||
}
|
||||
|
||||
// One hash per thread. blockDim.x is a multiple of 32; lane = threadIdx.x & 31 and every
|
||||
// __shfl_xor_sync stays inside the lane's own warp, exactly like simd_shuffle_xor inside a
|
||||
// 32-wide Metal SIMD group. Control flow is uniform, so the full 0xffffffff member mask is valid.
|
||||
__global__ void igneum_hash(const uint32_t* ds, uint64_t* out, uint32_t baseNonce, uint32_t mask) {
|
||||
uint32_t gid = blockIdx.x * blockDim.x + threadIdx.x;
|
||||
uint32_t nonce = baseNonce + gid;
|
||||
uint32_t r0, r1, r2, r3, r4, r5, r6, r7;
|
||||
{ uint32_t x = nonce ^ 0xbe8c5a0cu; x += 0x9e3779b9u; x = splitmix32(x); r0 = x ^ 0x7646e626u; } // SEEDW[0], 0x9e3779b9u * 1u, SEEDW[1]
|
||||
{ uint32_t x = nonce ^ 0x7646e626u; x += 0x3c6ef372u; x = splitmix32(x); r1 = x ^ 0x508e29d0u; } // SEEDW[1], 0x9e3779b9u * 2u, SEEDW[2]
|
||||
{ uint32_t x = nonce ^ 0x508e29d0u; x += 0xdaa66d2bu; x = splitmix32(x); r2 = x ^ 0xfb8a5b4au; } // SEEDW[2], 0x9e3779b9u * 3u, SEEDW[3]
|
||||
{ uint32_t x = nonce ^ 0xfb8a5b4au; x += 0x78dde6e4u; x = splitmix32(x); r3 = x ^ 0x53c50955u; } // SEEDW[3], 0x9e3779b9u * 4u, SEEDW[4]
|
||||
{ uint32_t x = nonce ^ 0x53c50955u; x += 0x1715609du; x = splitmix32(x); r4 = x ^ 0x685d62fcu; } // SEEDW[4], 0x9e3779b9u * 5u, SEEDW[5]
|
||||
{ uint32_t x = nonce ^ 0x685d62fcu; x += 0xb54cda56u; x = splitmix32(x); r5 = x ^ 0x6065c013u; } // SEEDW[5], 0x9e3779b9u * 6u, SEEDW[6]
|
||||
{ uint32_t x = nonce ^ 0x6065c013u; x += 0x5384540fu; x = splitmix32(x); r6 = x ^ 0x881096ebu; } // SEEDW[6], 0x9e3779b9u * 7u, SEEDW[7]
|
||||
{ uint32_t x = nonce ^ 0x881096ebu; x += 0xf1bbcdc8u; x = splitmix32(x); r7 = x ^ 0xbe8c5a0cu; } // SEEDW[7], 0x9e3779b9u * 8u, SEEDW[0]
|
||||
|
||||
for (uint32_t it = 0u; it < 8u; ++it) {
|
||||
uint32_t sel = r0;
|
||||
r6 = r6 - r3; // 0 sub
|
||||
r2 = r2 | r6; // 1 or
|
||||
r4 = r4 + r0 + ((((sel >> 4u) & 1u) != 0u) ? 0x7731324bu : 0x1a579b38u); // 2 add
|
||||
r3 = r3 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 3 load
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r6, 1); // 4 shfl
|
||||
r4 = r4 ^ r1; // 5 xor
|
||||
r3 = r5 * r3 + r3; // 6 mad
|
||||
r1 = r1 * r6; // 7 mul
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x08000000u) & mask]; // 8 load
|
||||
r0 = r0 ^ r4; // 9 xor
|
||||
r0 = r0 + r7 + ((((sel >> 30u) & 1u) != 0u) ? 0xb10fcef8u : 0xee083919u); // 10 add
|
||||
r0 = __umulhi(r0, r2); // 11 mulhi
|
||||
r5 = r5 ^ r4; // 12 xor
|
||||
r7 = r3 * r0 + r7; // 13 mad
|
||||
r3 = r3 ^ ds[((rotl_imm(r4 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 14 load
|
||||
r2 = r2 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 15 load
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r5, 16); // 16 shfl
|
||||
r7 = r7 * r6; // 17 mul
|
||||
r2 = r3 * r1 + r2; // 18 mad
|
||||
r5 = rotr_var(r5, r2); // 19 rotr
|
||||
r5 = r5 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 20 load
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r4, 16); // 21 shfl
|
||||
r1 = __umulhi(r1, r2); // 22 mulhi
|
||||
r6 = r6 ^ r5; // 23 xor
|
||||
r5 = r2 * r0 + r5; // 24 mad
|
||||
r6 = r3 * r7 + r6; // 25 mad
|
||||
r6 = r6 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 26 load
|
||||
r1 = r5 * r5 + r1; // 27 mad
|
||||
r0 = r0 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 28 load
|
||||
r7 = r7 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0x0b1f02c7u : 0xd5e6c37cu); // 29 add
|
||||
r7 = r7 + r0 + ((((sel >> 0u) & 1u) != 0u) ? 0x2d6e8bd1u : 0x4e45ba51u); // 30 add
|
||||
r0 = r0 ^ r2; // 31 xor
|
||||
r6 = r6 + r2 + ((((sel >> 13u) & 1u) != 0u) ? 0x351422d2u : 0x1c91b2a1u); // 32 add
|
||||
r4 = __umulhi(r4, r5); // 33 mulhi
|
||||
r3 = rotr_var(r3, r5); // 34 rotr
|
||||
r3 = r3 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 35 load
|
||||
r4 = rotl_imm(r4, 13u); // 36 rotl
|
||||
r6 = r6 + r1 + ((((sel >> 7u) & 1u) != 0u) ? 0x64a28251u : 0x3f2970f7u); // 37 add
|
||||
r2 = r2 * r0; // 38 mul
|
||||
r3 = r3 + r4 + ((((sel >> 15u) & 1u) != 0u) ? 0x7378c955u : 0x3b7b3317u); // 39 add
|
||||
r0 = r0 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 40 load
|
||||
r6 = r6 ^ r5; // 41 xor
|
||||
r3 = r3 + r0 + ((((sel >> 2u) & 1u) != 0u) ? 0xb31f6a64u : 0x6678b059u); // 42 add
|
||||
r7 = r7 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 43 load
|
||||
r3 = r3 ^ __shfl_xor_sync(0xffffffffu, r7, 4); // 44 shfl
|
||||
r1 = __umulhi(r1, r5); // 45 mulhi
|
||||
r0 = rotl_imm(r0, 6u); // 46 rotl
|
||||
r1 = r1 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 47 load
|
||||
r4 = r4 + r0 + ((((sel >> 17u) & 1u) != 0u) ? 0x5feccee6u : 0x43095946u); // 48 add
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 49 load
|
||||
r2 = r2 - r3; // 50 sub
|
||||
r7 = r7 + r2 + ((((sel >> 24u) & 1u) != 0u) ? 0x26e2b582u : 0xf45ecdf8u); // 51 add
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 52 load
|
||||
r6 = r6 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 53 load
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r5, 4); // 54 shfl
|
||||
r1 = rotr_var(r1, r2); // 55 rotr
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r5, 16); // 56 shfl
|
||||
r5 = rotl_imm(r5, 12u); // 57 rotl
|
||||
r5 = r5 + r3 + ((((sel >> 4u) & 1u) != 0u) ? 0x473b1718u : 0x6b4f35e8u); // 58 add
|
||||
r4 = rotr_var(r4, r1); // 59 rotr
|
||||
r4 = r4 ^ __shfl_xor_sync(0xffffffffu, r5, 1); // 60 shfl
|
||||
r7 = r7 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 61 load
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 62 load
|
||||
r4 = r4 * r1; // 63 mul
|
||||
// latency-shadow block (Counter ASIC 3.0 item 8): 256 ALU instructions x 27 passes after instruction 63, no load
|
||||
for (uint32_t sh = 0u; sh < 27u; ++sh) {
|
||||
r6 = r6 | r0; // s0 or
|
||||
r7 = r7 * r3; // s1 mul
|
||||
r3 = r5 * r5 + r3; // s2 mad
|
||||
r7 = r7 + r4 + ((((sel >> 6u) & 1u) != 0u) ? 0x38e58a06u : 0x97d3d105u); // s3 add
|
||||
r6 = r6 + r0 + ((((sel >> 20u) & 1u) != 0u) ? 0x9629673du : 0x29b87b9au); // s4 add
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r2, 1); // s5 shfl
|
||||
r5 = r3 * r5 + r5; // s6 mad
|
||||
r3 = rotl_imm(r3, 5u); // s7 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 12u) & 1u) != 0u) ? 0x4a5f55e1u : 0xa7c1fb0fu); // s8 add
|
||||
r5 = r5 * r7; // s9 mul
|
||||
r4 = r4 ^ r5; // s10 xor
|
||||
r6 = __umulhi(r6, r4); // s11 mulhi
|
||||
r5 = rotl_imm(r5, 13u); // s12 rotl
|
||||
r0 = r2 * r1 + r0; // s13 mad
|
||||
r2 = __umulhi(r2, r4); // s14 mulhi
|
||||
r3 = r3 | r1; // s15 or
|
||||
r0 = r0 - r5; // s16 sub
|
||||
r0 = r0 * r6; // s17 mul
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r1, 4); // s18 shfl
|
||||
r3 = r3 | r4; // s19 or
|
||||
r6 = r3 * r2 + r6; // s20 mad
|
||||
r0 = rotl_imm(r0, 11u); // s21 rotl
|
||||
r3 = rotl_imm(r3, 25u); // s22 rotl
|
||||
r6 = r6 ^ r4; // s23 xor
|
||||
r4 = r4 | r2; // s24 or
|
||||
r1 = r1 + r6 + ((((sel >> 7u) & 1u) != 0u) ? 0xfc2f6d17u : 0x53e74799u); // s25 add
|
||||
r7 = rotr_var(r7, r6); // s26 rotr
|
||||
r7 = r7 + r5 + ((((sel >> 27u) & 1u) != 0u) ? 0x787048dcu : 0xb8fae90eu); // s27 add
|
||||
r4 = r4 ^ r1; // s28 xor
|
||||
r6 = r6 ^ r1; // s29 xor
|
||||
r2 = __umulhi(r2, r0); // s30 mulhi
|
||||
r1 = rotr_var(r1, r4); // s31 rotr
|
||||
r7 = r7 ^ r4; // s32 xor
|
||||
r3 = r2 * r6 + r3; // s33 mad
|
||||
r3 = r3 + r1 + ((((sel >> 12u) & 1u) != 0u) ? 0x9ffd510bu : 0xd2dc42ebu); // s34 add
|
||||
r4 = r4 + r7 + ((((sel >> 7u) & 1u) != 0u) ? 0xf871ba37u : 0xad8eda6fu); // s35 add
|
||||
r7 = r7 + r4 + ((((sel >> 3u) & 1u) != 0u) ? 0xd02d30cau : 0xb0cef8f4u); // s36 add
|
||||
r4 = r3 * r2 + r4; // s37 mad
|
||||
r5 = r7 * r7 + r5; // s38 mad
|
||||
r3 = r3 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x82e98a22u : 0x2bd506e2u); // s39 add
|
||||
r0 = rotl_imm(r0, 21u); // s40 rotl
|
||||
r1 = r1 | r0; // s41 or
|
||||
r7 = r7 + r5 + ((((sel >> 28u) & 1u) != 0u) ? 0x6435f6d8u : 0xc0386e49u); // s42 add
|
||||
r1 = __umulhi(r1, r0); // s43 mulhi
|
||||
r5 = r5 + r6 + ((((sel >> 12u) & 1u) != 0u) ? 0x1f682c68u : 0x3a8921a4u); // s44 add
|
||||
r4 = r4 + r5 + ((((sel >> 1u) & 1u) != 0u) ? 0xf246e180u : 0xeeca4334u); // s45 add
|
||||
r2 = r2 ^ r3; // s46 xor
|
||||
r2 = r2 ^ r5; // s47 xor
|
||||
r7 = r4 * r1 + r7; // s48 mad
|
||||
r2 = r6 * r1 + r2; // s49 mad
|
||||
r7 = r0 * r7 + r7; // s50 mad
|
||||
r5 = r5 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x41bd68ccu : 0x745776d8u); // s51 add
|
||||
r3 = r3 ^ __shfl_xor_sync(0xffffffffu, r0, 1); // s52 shfl
|
||||
r4 = r4 | r0; // s53 or
|
||||
r0 = r2 * r4 + r0; // s54 mad
|
||||
r5 = rotl_imm(r5, 9u); // s55 rotl
|
||||
r5 = r5 * r7; // s56 mul
|
||||
r7 = rotl_imm(r7, 27u); // s57 rotl
|
||||
r7 = r7 - r6; // s58 sub
|
||||
r2 = rotl_imm(r2, 22u); // s59 rotl
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r1, 2); // s60 shfl
|
||||
r4 = r4 + r0 + ((((sel >> 7u) & 1u) != 0u) ? 0xcba6a161u : 0x60d0ff55u); // s61 add
|
||||
r1 = r1 * r5; // s62 mul
|
||||
r7 = __umulhi(r7, r6); // s63 mulhi
|
||||
r5 = r5 * r0; // s64 mul
|
||||
r2 = rotr_var(r2, r5); // s65 rotr
|
||||
r0 = __umulhi(r0, r3); // s66 mulhi
|
||||
r0 = r0 | r7; // s67 or
|
||||
r0 = rotl_imm(r0, 19u); // s68 rotl
|
||||
r0 = rotl_imm(r0, 1u); // s69 rotl
|
||||
r2 = r2 | r3; // s70 or
|
||||
r3 = r7 * r5 + r3; // s71 mad
|
||||
r7 = r0 * r4 + r7; // s72 mad
|
||||
r4 = r4 | r3; // s73 or
|
||||
r3 = r3 ^ r1; // s74 xor
|
||||
r1 = r1 + r3 + ((((sel >> 15u) & 1u) != 0u) ? 0x8e35924eu : 0x3f4ffea2u); // s75 add
|
||||
r1 = r1 * r3; // s76 mul
|
||||
r0 = r3 * r2 + r0; // s77 mad
|
||||
r2 = r2 ^ __shfl_xor_sync(0xffffffffu, r0, 16); // s78 shfl
|
||||
r1 = __umulhi(r1, r0); // s79 mulhi
|
||||
r0 = __umulhi(r0, r4); // s80 mulhi
|
||||
r4 = r4 * r2; // s81 mul
|
||||
r5 = r5 - r7; // s82 sub
|
||||
r0 = r2 * r3 + r0; // s83 mad
|
||||
r7 = r7 - r2; // s84 sub
|
||||
r1 = r1 ^ r2; // s85 xor
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r3, 16); // s86 shfl
|
||||
r3 = r0 * r0 + r3; // s87 mad
|
||||
r0 = rotl_imm(r0, 20u); // s88 rotl
|
||||
r1 = r1 - r5; // s89 sub
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r4, 16); // s90 shfl
|
||||
r5 = r5 ^ r1; // s91 xor
|
||||
r5 = rotl_imm(r5, 12u); // s92 rotl
|
||||
r6 = r1 * r2 + r6; // s93 mad
|
||||
r2 = rotr_var(r2, r5); // s94 rotr
|
||||
r0 = r0 | r4; // s95 or
|
||||
r7 = r7 * r3; // s96 mul
|
||||
r5 = r5 | r0; // s97 or
|
||||
r3 = r3 + r1 + ((((sel >> 13u) & 1u) != 0u) ? 0x386bb642u : 0xbcfdd8a7u); // s98 add
|
||||
r0 = r0 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xa3962d03u : 0xc921a021u); // s99 add
|
||||
r1 = rotr_var(r1, r0); // s100 rotr
|
||||
r2 = r2 + r4 + ((((sel >> 18u) & 1u) != 0u) ? 0xc5990c5cu : 0x70da4067u); // s101 add
|
||||
r2 = rotl_imm(r2, 3u); // s102 rotl
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r4, 4); // s103 shfl
|
||||
r0 = r0 ^ r2; // s104 xor
|
||||
r4 = r4 + r3 + ((((sel >> 28u) & 1u) != 0u) ? 0xb15aec75u : 0x1a5880cbu); // s105 add
|
||||
r4 = r4 ^ __shfl_xor_sync(0xffffffffu, r0, 2); // s106 shfl
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r0, 4); // s107 shfl
|
||||
r1 = r7 * r3 + r1; // s108 mad
|
||||
r0 = r2 * r3 + r0; // s109 mad
|
||||
r2 = r2 | r4; // s110 or
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r1, 1); // s111 shfl
|
||||
r2 = r5 * r3 + r2; // s112 mad
|
||||
r4 = __umulhi(r4, r0); // s113 mulhi
|
||||
r7 = rotl_imm(r7, 29u); // s114 rotl
|
||||
r3 = r3 - r7; // s115 sub
|
||||
r0 = __umulhi(r0, r1); // s116 mulhi
|
||||
r4 = rotl_imm(r4, 12u); // s117 rotl
|
||||
r1 = r1 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0xc5e98ee5u : 0x24494ad3u); // s118 add
|
||||
r0 = r0 + r4 + ((((sel >> 4u) & 1u) != 0u) ? 0xa97bc949u : 0x1b2d1c83u); // s119 add
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r4, 4); // s120 shfl
|
||||
r3 = r3 ^ __shfl_xor_sync(0xffffffffu, r2, 8); // s121 shfl
|
||||
r7 = rotl_imm(r7, 3u); // s122 rotl
|
||||
r4 = rotr_var(r4, r6); // s123 rotr
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r5, 1); // s124 shfl
|
||||
r5 = rotl_imm(r5, 21u); // s125 rotl
|
||||
r0 = r3 * r1 + r0; // s126 mad
|
||||
r0 = r0 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0xa7f6bb54u : 0xff14c08du); // s127 add
|
||||
r7 = r7 ^ r4; // s128 xor
|
||||
r3 = __umulhi(r3, r5); // s129 mulhi
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r7, 8); // s130 shfl
|
||||
r6 = __umulhi(r6, r4); // s131 mulhi
|
||||
r5 = r5 ^ r3; // s132 xor
|
||||
r0 = rotl_imm(r0, 19u); // s133 rotl
|
||||
r4 = rotr_var(r4, r3); // s134 rotr
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r2, 16); // s135 shfl
|
||||
r5 = r5 - r2; // s136 sub
|
||||
r3 = r3 ^ r5; // s137 xor
|
||||
r0 = r0 - r2; // s138 sub
|
||||
r3 = r3 * r7; // s139 mul
|
||||
r5 = r5 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0x2d465ca8u : 0x82ab98f4u); // s140 add
|
||||
r4 = r4 + r1 + ((((sel >> 11u) & 1u) != 0u) ? 0x237a9d8eu : 0x80594390u); // s141 add
|
||||
r3 = r7 * r0 + r3; // s142 mad
|
||||
r0 = r0 | r3; // s143 or
|
||||
r0 = __umulhi(r0, r1); // s144 mulhi
|
||||
r6 = r6 ^ r2; // s145 xor
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r1, 16); // s146 shfl
|
||||
r0 = r0 + r1 + ((((sel >> 18u) & 1u) != 0u) ? 0xe68cf57eu : 0x07f15148u); // s147 add
|
||||
r7 = r7 | r3; // s148 or
|
||||
r4 = r4 ^ r2; // s149 xor
|
||||
r7 = r7 ^ r1; // s150 xor
|
||||
r5 = rotr_var(r5, r2); // s151 rotr
|
||||
r0 = r0 * r7; // s152 mul
|
||||
r4 = rotl_imm(r4, 16u); // s153 rotl
|
||||
r6 = rotr_var(r6, r4); // s154 rotr
|
||||
r6 = r6 ^ r1; // s155 xor
|
||||
r6 = r6 | r3; // s156 or
|
||||
r1 = r1 + r6 + ((((sel >> 19u) & 1u) != 0u) ? 0x84c2a09du : 0x14878c5au); // s157 add
|
||||
r4 = r4 - r1; // s158 sub
|
||||
r4 = r6 * r7 + r4; // s159 mad
|
||||
r1 = r1 * r5; // s160 mul
|
||||
r4 = r4 | r0; // s161 or
|
||||
r7 = rotl_imm(r7, 21u); // s162 rotl
|
||||
r0 = r0 - r1; // s163 sub
|
||||
r1 = r1 * r0; // s164 mul
|
||||
r2 = r2 - r1; // s165 sub
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r6, 2); // s166 shfl
|
||||
r0 = r0 - r4; // s167 sub
|
||||
r1 = rotr_var(r1, r6); // s168 rotr
|
||||
r7 = rotr_var(r7, r1); // s169 rotr
|
||||
r3 = r1 * r3 + r3; // s170 mad
|
||||
r2 = r2 ^ __shfl_xor_sync(0xffffffffu, r0, 4); // s171 shfl
|
||||
r2 = r2 - r1; // s172 sub
|
||||
r7 = r7 ^ r6; // s173 xor
|
||||
r4 = rotr_var(r4, r1); // s174 rotr
|
||||
r4 = r4 ^ __shfl_xor_sync(0xffffffffu, r7, 2); // s175 shfl
|
||||
r6 = r6 * r2; // s176 mul
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r2, 4); // s177 shfl
|
||||
r6 = r6 + r4 + ((((sel >> 21u) & 1u) != 0u) ? 0x509871e6u : 0x4fa43965u); // s178 add
|
||||
r0 = r0 ^ r6; // s179 xor
|
||||
r5 = r5 + r7 + ((((sel >> 0u) & 1u) != 0u) ? 0xb95b8a53u : 0xfb3c4bd8u); // s180 add
|
||||
r6 = r6 * r3; // s181 mul
|
||||
r7 = rotl_imm(r7, 1u); // s182 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 11u) & 1u) != 0u) ? 0x7047b7cfu : 0xc448a197u); // s183 add
|
||||
r3 = r3 * r5; // s184 mul
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r7, 1); // s185 shfl
|
||||
r0 = r0 * r2; // s186 mul
|
||||
r4 = __umulhi(r4, r7); // s187 mulhi
|
||||
r2 = r7 * r2 + r2; // s188 mad
|
||||
r7 = r7 - r2; // s189 sub
|
||||
r2 = r2 + r5 + ((((sel >> 19u) & 1u) != 0u) ? 0x5caccc5du : 0x2fceaf49u); // s190 add
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r2, 1); // s191 shfl
|
||||
r7 = r7 + r1 + ((((sel >> 20u) & 1u) != 0u) ? 0x857bb5feu : 0xba0cee62u); // s192 add
|
||||
r6 = rotl_imm(r6, 22u); // s193 rotl
|
||||
r3 = r3 ^ r7; // s194 xor
|
||||
r7 = r7 * r0; // s195 mul
|
||||
r3 = r3 ^ r5; // s196 xor
|
||||
r5 = r5 + r1 + ((((sel >> 29u) & 1u) != 0u) ? 0xeae84577u : 0x8519428cu); // s197 add
|
||||
r7 = r4 * r4 + r7; // s198 mad
|
||||
r3 = r3 + r6 + ((((sel >> 8u) & 1u) != 0u) ? 0x0bfdbfa1u : 0x468639d3u); // s199 add
|
||||
r5 = r3 * r4 + r5; // s200 mad
|
||||
r2 = r2 ^ __shfl_xor_sync(0xffffffffu, r7, 4); // s201 shfl
|
||||
r7 = r7 + r4 + ((((sel >> 23u) & 1u) != 0u) ? 0x53ee8e50u : 0x18ec9a69u); // s202 add
|
||||
r6 = r6 ^ r0; // s203 xor
|
||||
r4 = r4 + r5 + ((((sel >> 4u) & 1u) != 0u) ? 0xac63376eu : 0x3e81485bu); // s204 add
|
||||
r7 = r1 * r4 + r7; // s205 mad
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r5, 16); // s206 shfl
|
||||
r4 = r4 * r0; // s207 mul
|
||||
r1 = r1 | r6; // s208 or
|
||||
r6 = r6 + r4 + ((((sel >> 28u) & 1u) != 0u) ? 0x5b745519u : 0x66ccb75du); // s209 add
|
||||
r1 = rotr_var(r1, r3); // s210 rotr
|
||||
r5 = r5 - r4; // s211 sub
|
||||
r4 = r4 ^ r7; // s212 xor
|
||||
r1 = r1 + r7 + ((((sel >> 8u) & 1u) != 0u) ? 0x1c3dccf7u : 0x0ce0553du); // s213 add
|
||||
r0 = __umulhi(r0, r1); // s214 mulhi
|
||||
r3 = r3 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xfadc9205u : 0x96bfec89u); // s215 add
|
||||
r0 = rotl_imm(r0, 4u); // s216 rotl
|
||||
r6 = r6 ^ r4; // s217 xor
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r7, 2); // s218 shfl
|
||||
r7 = r0 * r1 + r7; // s219 mad
|
||||
r4 = r4 * r0; // s220 mul
|
||||
r2 = r6 * r0 + r2; // s221 mad
|
||||
r5 = r5 ^ r4; // s222 xor
|
||||
r0 = r0 * r5; // s223 mul
|
||||
r2 = r2 + r0 + ((((sel >> 9u) & 1u) != 0u) ? 0x66b15c04u : 0x61495681u); // s224 add
|
||||
r3 = rotr_var(r3, r2); // s225 rotr
|
||||
r2 = r3 * r3 + r2; // s226 mad
|
||||
r6 = r6 ^ r0; // s227 xor
|
||||
r4 = rotl_imm(r4, 30u); // s228 rotl
|
||||
r2 = r2 + r3 + ((((sel >> 29u) & 1u) != 0u) ? 0x25bc17c2u : 0x70d05c34u); // s229 add
|
||||
r1 = rotr_var(r1, r5); // s230 rotr
|
||||
r1 = r1 + r4 + ((((sel >> 8u) & 1u) != 0u) ? 0x2d8728f3u : 0x32a28384u); // s231 add
|
||||
r1 = rotl_imm(r1, 6u); // s232 rotl
|
||||
r2 = r2 + r0 + ((((sel >> 21u) & 1u) != 0u) ? 0x61e81fe6u : 0xcf0949f1u); // s233 add
|
||||
r4 = rotl_imm(r4, 2u); // s234 rotl
|
||||
r2 = r2 - r6; // s235 sub
|
||||
r6 = r6 | r3; // s236 or
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r5, 4); // s237 shfl
|
||||
r1 = r1 ^ r3; // s238 xor
|
||||
r2 = rotl_imm(r2, 1u); // s239 rotl
|
||||
r5 = rotr_var(r5, r2); // s240 rotr
|
||||
r4 = rotl_imm(r4, 10u); // s241 rotl
|
||||
r6 = r6 + r2 + ((((sel >> 16u) & 1u) != 0u) ? 0x0a67565du : 0xe204fb50u); // s242 add
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r3, 2); // s243 shfl
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r1, 16); // s244 shfl
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r7, 2); // s245 shfl
|
||||
r4 = r4 ^ r5; // s246 xor
|
||||
r7 = r7 ^ r2; // s247 xor
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r6, 16); // s248 shfl
|
||||
r1 = __umulhi(r1, r3); // s249 mulhi
|
||||
r7 = r7 + r5 + ((((sel >> 31u) & 1u) != 0u) ? 0x5a79fd6fu : 0x3d6daffdu); // s250 add
|
||||
r6 = r6 + r0 + ((((sel >> 28u) & 1u) != 0u) ? 0x6beb28c2u : 0x84334aeau); // s251 add
|
||||
r7 = r7 - r1; // s252 sub
|
||||
r5 = rotr_var(r5, r2); // s253 rotr
|
||||
r2 = __umulhi(r2, r5); // s254 mulhi
|
||||
r6 = r6 * r3; // s255 mul
|
||||
}
|
||||
}
|
||||
uint32_t lo = r0 ^ rotl_imm(r1, 7u) ^ rotl_imm(r2, 14u) ^ rotl_imm(r3, 21u);
|
||||
uint32_t hi = r4 ^ rotl_imm(r5, 9u) ^ rotl_imm(r6, 18u) ^ rotl_imm(r7, 27u);
|
||||
out[gid] = ((uint64_t)hi << 32) | (uint64_t)lo;
|
||||
}
|
||||
|
||||
// Host-side launch wrappers. Declared in program.h, called from host.cu.
|
||||
cudaError_t igneum_launch_cache_fill(uint32_t* cache, uint32_t nSegments) {
|
||||
if (nSegments == 0u) return cudaErrorInvalidValue;
|
||||
uint32_t block = 256u;
|
||||
uint32_t grid = (nSegments + block - 1u) / block;
|
||||
igneum_cache_fill<<<grid, block>>>(cache, nSegments);
|
||||
return cudaGetLastError();
|
||||
}
|
||||
|
||||
cudaError_t igneum_launch_build(uint32_t* ds, const uint32_t* cache, const uint32_t* leaves, uint32_t nLeaves, uint32_t nItems) {
|
||||
if (nItems == 0u || nLeaves == 0u) return cudaErrorInvalidValue;
|
||||
uint32_t block = 256u;
|
||||
uint32_t grid = (nItems + block - 1u) / block;
|
||||
igneum_build<<<grid, block>>>(ds, cache, leaves, nLeaves, nItems);
|
||||
return cudaGetLastError();
|
||||
}
|
||||
|
||||
cudaError_t igneum_launch_hash(const uint32_t* ds, uint64_t* out, uint32_t baseNonce, uint32_t mask,
|
||||
uint32_t nonces, uint32_t blockWarps) {
|
||||
if (blockWarps == 0u || blockWarps > 32u) return cudaErrorInvalidValue;
|
||||
uint32_t block = 32u * blockWarps;
|
||||
if (nonces == 0u || (nonces % block) != 0u) return cudaErrorInvalidValue;
|
||||
igneum_hash<<<nonces / block, block>>>(ds, out, baseNonce, mask);
|
||||
return cudaGetLastError();
|
||||
}
|
||||
|
||||
cudaError_t igneum_hash_info(int* numRegs, int* blocksPerSM, uint32_t blockWarps) {
|
||||
cudaFuncAttributes attr;
|
||||
cudaError_t e = cudaFuncGetAttributes(&attr, igneum_hash);
|
||||
if (e != cudaSuccess) return e;
|
||||
*numRegs = attr.numRegs;
|
||||
return cudaOccupancyMaxActiveBlocksPerMultiprocessor(blocksPerSM, igneum_hash, (int)(32u * blockWarps), 0);
|
||||
}
|
||||
898
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel_bound.cl
Normal file
898
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel_bound.cl
Normal file
|
|
@ -0,0 +1,898 @@
|
|||
// Generated by igneum-pow export (generator v2) for seed "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000". Do not edit by hand.
|
||||
// OpenCL C twin of the Metal kernel for the same seed (see proto-opencl/README.md, WAVEFRONT.md and program.metal).
|
||||
// Built from source at runtime by proto-opencl/host.c, which passes these defines:
|
||||
// IGNEUM_GROUP work-group size of igneum_hash, a multiple of 32 (default 32: one work-group = one 32-lane unit)
|
||||
// IGNEUM_EXCHANGE 0 = local-memory exchange with a barrier (any device, any wave width; the default)
|
||||
// 1 = sub_group_shuffle_xor (cl_khr_subgroup_shuffle), only with IGNEUM_GROUP 32 and a sub-group size of exactly 32
|
||||
// 2 = intel_sub_group_shuffle_xor (cl_intel_subgroups), same condition
|
||||
// The verification unit is always 32 lanes. A 64-wide hardware wave (AMD GCN/CDNA, RDNA in wave64) runs two units;
|
||||
// the exchange masks are 1, 2, 4, 8, 16, so every partner lane lies inside the lane's own aligned run of 32.
|
||||
#ifndef IGNEUM_GROUP
|
||||
#define IGNEUM_GROUP 32
|
||||
#endif
|
||||
#ifndef IGNEUM_EXCHANGE
|
||||
#define IGNEUM_EXCHANGE 0
|
||||
#endif
|
||||
#ifdef __OPENCL_VERSION__
|
||||
#define IGNEUM_KERNEL_HASH __kernel __attribute__((reqd_work_group_size(IGNEUM_GROUP, 1, 1)))
|
||||
#define IGNEUM_LOCAL_WORDS(name, n) __local uint name[n]
|
||||
#if IGNEUM_EXCHANGE == 1
|
||||
#ifdef cl_khr_subgroups
|
||||
#pragma OPENCL EXTENSION cl_khr_subgroups : enable
|
||||
#endif
|
||||
#ifdef cl_khr_subgroup_shuffle
|
||||
#pragma OPENCL EXTENSION cl_khr_subgroup_shuffle : enable
|
||||
#endif
|
||||
#elif IGNEUM_EXCHANGE == 2
|
||||
#pragma OPENCL EXTENSION cl_intel_subgroups : enable
|
||||
#endif
|
||||
#else
|
||||
// Not an OpenCL compiler: proto-opencl/emu compiles this file as C++ and supplies the built-ins and these two macros.
|
||||
#include "emu_opencl.h"
|
||||
#endif
|
||||
|
||||
#if IGNEUM_EXCHANGE == 1
|
||||
#define IGNEUM_SHFL_XOR(dst, a, m) dst = sub_group_shuffle_xor((a), (uint)(m))
|
||||
#define IGNEUM_BCAST0(dst, a) dst = sub_group_broadcast((a), 0u)
|
||||
#elif IGNEUM_EXCHANGE == 2
|
||||
#define IGNEUM_SHFL_XOR(dst, a, m) dst = intel_sub_group_shuffle_xor((a), (uint)(m))
|
||||
#define IGNEUM_BCAST0(dst, a) dst = sub_group_broadcast((a), 0u)
|
||||
#else
|
||||
// Local-memory exchange. Two buffers of IGNEUM_GROUP words alternate (xk counts exchanges), so one barrier per
|
||||
// exchange is enough: a lane can only overwrite buffer b at exchange k+2 after passing barrier k+1, and every lane
|
||||
// reaches barrier k+1 only after its read of buffer b at exchange k. The partner lid ^ m stays inside the lane's
|
||||
// aligned run of 32 because m < 32. Control flow is uniform, so every work-item reaches every barrier.
|
||||
#define IGNEUM_SHFL_XOR(dst, a, m) { xch[(xk & 1u) * IGNEUM_GROUP + lid] = (a); barrier(CLK_LOCAL_MEM_FENCE); dst = xch[(xk & 1u) * IGNEUM_GROUP + (lid ^ (uint)(m))]; xk += 1u; }
|
||||
#define IGNEUM_BCAST0(dst, a) { xch[(xk & 1u) * IGNEUM_GROUP + lid] = (a); barrier(CLK_LOCAL_MEM_FENCE); dst = xch[(xk & 1u) * IGNEUM_GROUP + (lid & ~31u)]; xk += 1u; }
|
||||
#endif
|
||||
|
||||
static inline uint splitmix32(uint x) {
|
||||
x ^= x >> 16; x *= 0x7feb352du;
|
||||
x ^= x >> 15; x *= 0x846ca68bu;
|
||||
x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
// n is a literal in 1..31 at every call site. OpenCL rotate() rotates left by n modulo 32.
|
||||
static inline uint rotl_imm(uint x, uint n) { return rotate(x, n); }
|
||||
// Right rotation by n modulo 32 as a left rotation by (32 - n) modulo 32; n == 0 gives x.
|
||||
static inline uint rotr_var(uint x, uint n) { return rotate(x, (0u - n) & 31u); }
|
||||
static inline uint ds_elem(uint i, uint d0, uint d1) {
|
||||
uint x = i ^ d0;
|
||||
x *= 0x9E3779B1u; x ^= x >> 15;
|
||||
x += d1;
|
||||
x *= 0x85EBCA77u; x ^= x >> 13;
|
||||
x *= 0xC2B2AE3Du; x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
|
||||
// Memory-hard dataset core (MEMHARD.md). Cache: 2^26 words in 2^16 segments of 64 chained ChaCha12 lines.
|
||||
// Item: 8 rounds of 8 x seed-parameterised mixer + one 64-byte cache read, then 8 x final mixer (class v3, mixer multiplier 8,
|
||||
// docs/plans/mixer-x4.md: the round key of application j of round r is 0x9E3779B9 * (r * m + j + 1)). All parameters are literals.
|
||||
#define MH_CACHE_LINE_MASK 0x003fffffu
|
||||
#define MH_SEGMENT_LINES 64u
|
||||
#define MH_QR(a, b, c, d, r1, r2, r3, r4) { a += b; d ^= a; d = mh_rotl(d, r1); c += d; b ^= c; b = mh_rotl(b, r2); a += b; d ^= a; d = mh_rotl(d, r3); c += d; b ^= c; b = mh_rotl(b, r4); }
|
||||
static inline uint mh_rotl(uint x, uint n) { return (x << n) | (x >> (32u - n)); } // n in 1..31 at every call site
|
||||
|
||||
// y = ChaCha12 core(x) + x
|
||||
static inline void mh_chacha_block(const uint* x, uint* y) {
|
||||
for (uint i = 0u; i < 16u; ++i) y[i] = x[i];
|
||||
for (uint r = 0u; r < 6u; ++r) {
|
||||
MH_QR(y[0], y[4], y[8], y[12], 16u, 12u, 8u, 7u) MH_QR(y[1], y[5], y[9], y[13], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[6], y[10], y[14], 16u, 12u, 8u, 7u) MH_QR(y[3], y[7], y[11], y[15], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[0], y[5], y[10], y[15], 16u, 12u, 8u, 7u) MH_QR(y[1], y[6], y[11], y[12], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[7], y[8], y[13], 16u, 12u, 8u, 7u) MH_QR(y[3], y[4], y[9], y[14], 16u, 12u, 8u, 7u)
|
||||
}
|
||||
for (uint i = 0u; i < 16u; ++i) y[i] += x[i];
|
||||
}
|
||||
|
||||
// One cache segment: 64 chained lines written at cache[seg * 1024]. in_j = prev ^ (sigma || K || seg || j || tag), prev_0 = 0.
|
||||
static inline void mh_cache_segment(__global uint* cache, uint seg) {
|
||||
uint prev[16]; uint x[16]; uint y[16];
|
||||
for (uint i = 0u; i < 16u; ++i) prev[i] = 0u;
|
||||
for (uint j = 0u; j < MH_SEGMENT_LINES; ++j) {
|
||||
x[0] = 0x61707865u ^ prev[0]; x[1] = 0x3320646eu ^ prev[1]; x[2] = 0x79622d32u ^ prev[2]; x[3] = 0x6b206574u ^ prev[3];
|
||||
x[4] = 0xe23f9008u ^ prev[4];
|
||||
x[5] = 0xde33e763u ^ prev[5];
|
||||
x[6] = 0xc5ba415du ^ prev[6];
|
||||
x[7] = 0x8ddf6786u ^ prev[7];
|
||||
x[8] = 0x59f4a4beu ^ prev[8];
|
||||
x[9] = 0x8a3bc680u ^ prev[9];
|
||||
x[10] = 0x701b8e40u ^ prev[10];
|
||||
x[11] = 0x3b59025au ^ prev[11];
|
||||
x[12] = seg ^ prev[12]; x[13] = j ^ prev[13]; x[14] = 0x49676e65u ^ prev[14]; x[15] = 0x756d4d48u ^ prev[15];
|
||||
mh_chacha_block(x, y);
|
||||
__global uint* line = cache + ((seg * MH_SEGMENT_LINES + j) * 16u);
|
||||
for (uint i = 0u; i < 16u; ++i) { line[i] = y[i]; prev[i] = y[i]; }
|
||||
}
|
||||
}
|
||||
|
||||
// M_r: per word (s ^ (RC + rk)) * MUL, then a column round and a diagonal round with the seed-drawn rotations.
|
||||
static inline void mh_mixer(uint* s, uint rk) {
|
||||
s[0] = (s[0] ^ (0xe5bef5a3u + rk)) * 0xd1d79e7fu;
|
||||
s[1] = (s[1] ^ (0xdb2a7d90u + rk)) * 0xbcb61f8bu;
|
||||
s[2] = (s[2] ^ (0xcd1fa7e1u + rk)) * 0x970e91abu;
|
||||
s[3] = (s[3] ^ (0x30289419u + rk)) * 0xd749d96du;
|
||||
s[4] = (s[4] ^ (0x0a730d58u + rk)) * 0x143d7339u;
|
||||
s[5] = (s[5] ^ (0x432f8579u + rk)) * 0x2cde0d69u;
|
||||
s[6] = (s[6] ^ (0x6ab978a5u + rk)) * 0x12d0a8c1u;
|
||||
s[7] = (s[7] ^ (0x8c984f49u + rk)) * 0x2d335be9u;
|
||||
s[8] = (s[8] ^ (0x788c3c9eu + rk)) * 0x80a8aae9u;
|
||||
s[9] = (s[9] ^ (0x051eef02u + rk)) * 0x7ac896c7u;
|
||||
s[10] = (s[10] ^ (0x05e77db9u + rk)) * 0x9de23db7u;
|
||||
s[11] = (s[11] ^ (0x8b3cd4f3u + rk)) * 0xc362827du;
|
||||
s[12] = (s[12] ^ (0x6948d6cfu + rk)) * 0x5f4cdb5bu;
|
||||
s[13] = (s[13] ^ (0x0d8677f6u + rk)) * 0xfc6c5097u;
|
||||
s[14] = (s[14] ^ (0xf9505c8cu + rk)) * 0x6f547f83u;
|
||||
s[15] = (s[15] ^ (0x4513c163u + rk)) * 0x1ac31b47u;
|
||||
MH_QR(s[0], s[4], s[8], s[12], 28u, 11u, 18u, 18u) MH_QR(s[1], s[5], s[9], s[13], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[2], s[6], s[10], s[14], 28u, 11u, 18u, 18u) MH_QR(s[3], s[7], s[11], s[15], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[0], s[5], s[10], s[15], 13u, 28u, 5u, 13u) MH_QR(s[1], s[6], s[11], s[12], 13u, 28u, 5u, 13u)
|
||||
MH_QR(s[2], s[7], s[8], s[13], 13u, 28u, 5u, 13u) MH_QR(s[3], s[4], s[9], s[14], 13u, 28u, 5u, 13u)
|
||||
}
|
||||
|
||||
// Item t: 16 words. s = (K, t * MUL[i] + RC[i]); 8 rounds of 8 x mixer + cache line s[0] & mask; 8 x final mixer.
|
||||
static inline void mh_item(__global const uint* cache, __global const uint* leaf, uint t, uint* s) {
|
||||
s[0] = 0xe23f9008u;
|
||||
s[1] = 0xde33e763u;
|
||||
s[2] = 0xc5ba415du;
|
||||
s[3] = 0x8ddf6786u;
|
||||
s[4] = 0x59f4a4beu;
|
||||
s[5] = 0x8a3bc680u;
|
||||
s[6] = 0x701b8e40u;
|
||||
s[7] = 0x3b59025au;
|
||||
s[8] = t * 0xd1d79e7fu + 0xe5bef5a3u;
|
||||
s[9] = t * 0xbcb61f8bu + 0xdb2a7d90u;
|
||||
s[10] = t * 0x970e91abu + 0xcd1fa7e1u;
|
||||
s[11] = t * 0xd749d96du + 0x30289419u;
|
||||
s[12] = t * 0x143d7339u + 0x0a730d58u;
|
||||
s[13] = t * 0x2cde0d69u + 0x432f8579u;
|
||||
s[14] = t * 0x12d0a8c1u + 0x6ab978a5u;
|
||||
s[15] = t * 0x2d335be9u + 0x8c984f49u;
|
||||
for (uint i = 0u; i < 16u; ++i) s[i] ^= leaf[i];
|
||||
for (uint r = 0u; r < 8u; ++r) {
|
||||
for (uint j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (r * 8u + j + 1u));
|
||||
__global const uint* line = cache + ((s[0] & MH_CACHE_LINE_MASK) * 16u);
|
||||
for (uint i = 0u; i < 16u; ++i) s[i] ^= line[i];
|
||||
}
|
||||
for (uint j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (64u + j + 1u));
|
||||
}
|
||||
// Class v5 (docs/design/class-v5-stored-state.md): leaf(t) = leaves[t mod nLeaves], 16 words per leaf (leaves.bin).
|
||||
static inline __global const uint* mh_leaf(__global const uint* leaves, uint nLeaves, uint t) { return leaves + ((t % nLeaves) * 16u); }
|
||||
// Era layout (docs/plans/era-layout.md 1.2): dataset word w holds word mh_j(w) of item mh_t(w); j's bits sit at positions 1 9 13 14 of w.
|
||||
static inline uint mh_j(uint w) { return ((w >> 1u) & 1u) | (((w >> 9u) & 1u) << 1) | (((w >> 13u) & 1u) << 2) | (((w >> 14u) & 1u) << 3); }
|
||||
static inline uint mh_t(uint w) { w = (w & 0x00003fffu) | ((w >> 15u) << 14u); w = (w & 0x00001fffu) | ((w >> 14u) << 13u); w = (w & 0x000001ffu) | ((w >> 10u) << 9u); w = (w & 0x00000001u) | ((w >> 2u) << 1u); return w; }
|
||||
static inline uint mh_addr(uint t, uint j) { uint w = t; w = ((w >> 1u) << 2u) | (w & 0x00000001u) | (((j >> 0u) & 1u) << 1u); w = ((w >> 9u) << 10u) | (w & 0x000001ffu) | (((j >> 1u) & 1u) << 9u); w = ((w >> 13u) << 14u) | (w & 0x00001fffu) | (((j >> 2u) & 1u) << 13u); w = ((w >> 14u) << 15u) | (w & 0x00003fffu) | (((j >> 3u) & 1u) << 14u); return w; }
|
||||
// dataset[w] without the dataset: derive item mh_t(w) and take word mh_j(w).
|
||||
static inline uint mh_word(__global const uint* cache, __global const uint* leaves, uint nLeaves, uint w) { uint s[16]; mh_item(cache, mh_leaf(leaves, nLeaves, mh_t(w)), mh_t(w), s); return s[mh_j(w)]; }
|
||||
|
||||
// Memory-hard dataset (MEMHARD.md). One work-item per cache segment; one work-item per 64-byte dataset item.
|
||||
// The same constants as memhard.h in this pack (one emitter, three dialects).
|
||||
__kernel void igneum_cache_fill(__global uint* cache, uint nSegments) {
|
||||
uint seg = (uint)get_global_id(0);
|
||||
if (seg < nSegments) mh_cache_segment(cache, seg);
|
||||
}
|
||||
// Class v5: the window's leaves (leaves.bin, IGNEUM_STATE_LEAVES x 16 words) and their count.
|
||||
__kernel void igneum_build(__global uint* ds, __global const uint* cache, __global const uint* leaves, uint nLeaves, uint nItems) {
|
||||
uint t = (uint)get_global_id(0);
|
||||
if (t < nItems) {
|
||||
uint s[16];
|
||||
mh_item(cache, mh_leaf(leaves, nLeaves, t), t, s);
|
||||
for (uint i = 0u; i < 16u; ++i) ds[(ulong)mh_addr(t, i)] = s[i];
|
||||
}
|
||||
}
|
||||
|
||||
// One hash per work-item. IGNEUM_GROUP is a multiple of 32; lane = lid & 31 and every exchange stays inside the
|
||||
// lane's own aligned run of 32 work-items, exactly like simd_shuffle_xor inside a 32-wide Metal SIMD group and
|
||||
// __shfl_xor_sync inside a CUDA warp. Control flow is uniform (no branches at all).
|
||||
IGNEUM_KERNEL_HASH void igneum_hash(__global const uint* ds, __global ulong* out, uint baseNonce, uint mask) {
|
||||
uint gid = (uint)get_global_id(0);
|
||||
uint lid = (uint)get_local_id(0);
|
||||
uint nonce = baseNonce + gid;
|
||||
uint r0, r1, r2, r3, r4, r5, r6, r7;
|
||||
#if IGNEUM_EXCHANGE == 0
|
||||
IGNEUM_LOCAL_WORDS(xch, 2 * IGNEUM_GROUP);
|
||||
uint xk = 0u;
|
||||
#else
|
||||
(void)lid;
|
||||
#endif
|
||||
{ uint x = nonce ^ 0xbe8c5a0cu; x += 0x9e3779b9u; x = splitmix32(x); r0 = x ^ 0x7646e626u; } // SEEDW[0], 0x9e3779b9u * 1u, SEEDW[1]
|
||||
{ uint x = nonce ^ 0x7646e626u; x += 0x3c6ef372u; x = splitmix32(x); r1 = x ^ 0x508e29d0u; } // SEEDW[1], 0x9e3779b9u * 2u, SEEDW[2]
|
||||
{ uint x = nonce ^ 0x508e29d0u; x += 0xdaa66d2bu; x = splitmix32(x); r2 = x ^ 0xfb8a5b4au; } // SEEDW[2], 0x9e3779b9u * 3u, SEEDW[3]
|
||||
{ uint x = nonce ^ 0xfb8a5b4au; x += 0x78dde6e4u; x = splitmix32(x); r3 = x ^ 0x53c50955u; } // SEEDW[3], 0x9e3779b9u * 4u, SEEDW[4]
|
||||
{ uint x = nonce ^ 0x53c50955u; x += 0x1715609du; x = splitmix32(x); r4 = x ^ 0x685d62fcu; } // SEEDW[4], 0x9e3779b9u * 5u, SEEDW[5]
|
||||
{ uint x = nonce ^ 0x685d62fcu; x += 0xb54cda56u; x = splitmix32(x); r5 = x ^ 0x6065c013u; } // SEEDW[5], 0x9e3779b9u * 6u, SEEDW[6]
|
||||
{ uint x = nonce ^ 0x6065c013u; x += 0x5384540fu; x = splitmix32(x); r6 = x ^ 0x881096ebu; } // SEEDW[6], 0x9e3779b9u * 7u, SEEDW[7]
|
||||
{ uint x = nonce ^ 0x881096ebu; x += 0xf1bbcdc8u; x = splitmix32(x); r7 = x ^ 0xbe8c5a0cu; } // SEEDW[7], 0x9e3779b9u * 8u, SEEDW[0]
|
||||
|
||||
for (uint it = 0u; it < 8u; ++it) {
|
||||
uint sel = r0;
|
||||
r6 = r6 - r3; // 0 sub
|
||||
r2 = r2 | r6; // 1 or
|
||||
r4 = r4 + r0 + ((((sel >> 4u) & 1u) != 0u) ? 0x7731324bu : 0x1a579b38u); // 2 add
|
||||
r3 = r3 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 3 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 1u); r0 = r0 ^ t_; } // 4 shfl
|
||||
r4 = r4 ^ r1; // 5 xor
|
||||
r3 = r5 * r3 + r3; // 6 mad
|
||||
r1 = r1 * r6; // 7 mul
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x08000000u) & mask]; // 8 load
|
||||
r0 = r0 ^ r4; // 9 xor
|
||||
r0 = r0 + r7 + ((((sel >> 30u) & 1u) != 0u) ? 0xb10fcef8u : 0xee083919u); // 10 add
|
||||
r0 = mul_hi(r0, r2); // 11 mulhi
|
||||
r5 = r5 ^ r4; // 12 xor
|
||||
r7 = r3 * r0 + r7; // 13 mad
|
||||
r3 = r3 ^ ds[((rotl_imm(r4 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 14 load
|
||||
r2 = r2 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 15 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r7 = r7 ^ t_; } // 16 shfl
|
||||
r7 = r7 * r6; // 17 mul
|
||||
r2 = r3 * r1 + r2; // 18 mad
|
||||
r5 = rotr_var(r5, r2); // 19 rotr
|
||||
r5 = r5 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 20 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 16u); r6 = r6 ^ t_; } // 21 shfl
|
||||
r1 = mul_hi(r1, r2); // 22 mulhi
|
||||
r6 = r6 ^ r5; // 23 xor
|
||||
r5 = r2 * r0 + r5; // 24 mad
|
||||
r6 = r3 * r7 + r6; // 25 mad
|
||||
r6 = r6 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 26 load
|
||||
r1 = r5 * r5 + r1; // 27 mad
|
||||
r0 = r0 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 28 load
|
||||
r7 = r7 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0x0b1f02c7u : 0xd5e6c37cu); // 29 add
|
||||
r7 = r7 + r0 + ((((sel >> 0u) & 1u) != 0u) ? 0x2d6e8bd1u : 0x4e45ba51u); // 30 add
|
||||
r0 = r0 ^ r2; // 31 xor
|
||||
r6 = r6 + r2 + ((((sel >> 13u) & 1u) != 0u) ? 0x351422d2u : 0x1c91b2a1u); // 32 add
|
||||
r4 = mul_hi(r4, r5); // 33 mulhi
|
||||
r3 = rotr_var(r3, r5); // 34 rotr
|
||||
r3 = r3 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 35 load
|
||||
r4 = rotl_imm(r4, 13u); // 36 rotl
|
||||
r6 = r6 + r1 + ((((sel >> 7u) & 1u) != 0u) ? 0x64a28251u : 0x3f2970f7u); // 37 add
|
||||
r2 = r2 * r0; // 38 mul
|
||||
r3 = r3 + r4 + ((((sel >> 15u) & 1u) != 0u) ? 0x7378c955u : 0x3b7b3317u); // 39 add
|
||||
r0 = r0 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 40 load
|
||||
r6 = r6 ^ r5; // 41 xor
|
||||
r3 = r3 + r0 + ((((sel >> 2u) & 1u) != 0u) ? 0xb31f6a64u : 0x6678b059u); // 42 add
|
||||
r7 = r7 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 43 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 4u); r3 = r3 ^ t_; } // 44 shfl
|
||||
r1 = mul_hi(r1, r5); // 45 mulhi
|
||||
r0 = rotl_imm(r0, 6u); // 46 rotl
|
||||
r1 = r1 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 47 load
|
||||
r4 = r4 + r0 + ((((sel >> 17u) & 1u) != 0u) ? 0x5feccee6u : 0x43095946u); // 48 add
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 49 load
|
||||
r2 = r2 - r3; // 50 sub
|
||||
r7 = r7 + r2 + ((((sel >> 24u) & 1u) != 0u) ? 0x26e2b582u : 0xf45ecdf8u); // 51 add
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 52 load
|
||||
r6 = r6 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 53 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 4u); r7 = r7 ^ t_; } // 54 shfl
|
||||
r1 = rotr_var(r1, r2); // 55 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r0 = r0 ^ t_; } // 56 shfl
|
||||
r5 = rotl_imm(r5, 12u); // 57 rotl
|
||||
r5 = r5 + r3 + ((((sel >> 4u) & 1u) != 0u) ? 0x473b1718u : 0x6b4f35e8u); // 58 add
|
||||
r4 = rotr_var(r4, r1); // 59 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 1u); r4 = r4 ^ t_; } // 60 shfl
|
||||
r7 = r7 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 61 load
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 62 load
|
||||
r4 = r4 * r1; // 63 mul
|
||||
// latency-shadow block (Counter ASIC 3.0 item 8): 256 ALU instructions x 27 passes after instruction 63, no load
|
||||
for (uint sh = 0u; sh < 27u; ++sh) {
|
||||
r6 = r6 | r0; // s0 or
|
||||
r7 = r7 * r3; // s1 mul
|
||||
r3 = r5 * r5 + r3; // s2 mad
|
||||
r7 = r7 + r4 + ((((sel >> 6u) & 1u) != 0u) ? 0x38e58a06u : 0x97d3d105u); // s3 add
|
||||
r6 = r6 + r0 + ((((sel >> 20u) & 1u) != 0u) ? 0x9629673du : 0x29b87b9au); // s4 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 1u); r6 = r6 ^ t_; } // s5 shfl
|
||||
r5 = r3 * r5 + r5; // s6 mad
|
||||
r3 = rotl_imm(r3, 5u); // s7 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 12u) & 1u) != 0u) ? 0x4a5f55e1u : 0xa7c1fb0fu); // s8 add
|
||||
r5 = r5 * r7; // s9 mul
|
||||
r4 = r4 ^ r5; // s10 xor
|
||||
r6 = mul_hi(r6, r4); // s11 mulhi
|
||||
r5 = rotl_imm(r5, 13u); // s12 rotl
|
||||
r0 = r2 * r1 + r0; // s13 mad
|
||||
r2 = mul_hi(r2, r4); // s14 mulhi
|
||||
r3 = r3 | r1; // s15 or
|
||||
r0 = r0 - r5; // s16 sub
|
||||
r0 = r0 * r6; // s17 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 4u); r5 = r5 ^ t_; } // s18 shfl
|
||||
r3 = r3 | r4; // s19 or
|
||||
r6 = r3 * r2 + r6; // s20 mad
|
||||
r0 = rotl_imm(r0, 11u); // s21 rotl
|
||||
r3 = rotl_imm(r3, 25u); // s22 rotl
|
||||
r6 = r6 ^ r4; // s23 xor
|
||||
r4 = r4 | r2; // s24 or
|
||||
r1 = r1 + r6 + ((((sel >> 7u) & 1u) != 0u) ? 0xfc2f6d17u : 0x53e74799u); // s25 add
|
||||
r7 = rotr_var(r7, r6); // s26 rotr
|
||||
r7 = r7 + r5 + ((((sel >> 27u) & 1u) != 0u) ? 0x787048dcu : 0xb8fae90eu); // s27 add
|
||||
r4 = r4 ^ r1; // s28 xor
|
||||
r6 = r6 ^ r1; // s29 xor
|
||||
r2 = mul_hi(r2, r0); // s30 mulhi
|
||||
r1 = rotr_var(r1, r4); // s31 rotr
|
||||
r7 = r7 ^ r4; // s32 xor
|
||||
r3 = r2 * r6 + r3; // s33 mad
|
||||
r3 = r3 + r1 + ((((sel >> 12u) & 1u) != 0u) ? 0x9ffd510bu : 0xd2dc42ebu); // s34 add
|
||||
r4 = r4 + r7 + ((((sel >> 7u) & 1u) != 0u) ? 0xf871ba37u : 0xad8eda6fu); // s35 add
|
||||
r7 = r7 + r4 + ((((sel >> 3u) & 1u) != 0u) ? 0xd02d30cau : 0xb0cef8f4u); // s36 add
|
||||
r4 = r3 * r2 + r4; // s37 mad
|
||||
r5 = r7 * r7 + r5; // s38 mad
|
||||
r3 = r3 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x82e98a22u : 0x2bd506e2u); // s39 add
|
||||
r0 = rotl_imm(r0, 21u); // s40 rotl
|
||||
r1 = r1 | r0; // s41 or
|
||||
r7 = r7 + r5 + ((((sel >> 28u) & 1u) != 0u) ? 0x6435f6d8u : 0xc0386e49u); // s42 add
|
||||
r1 = mul_hi(r1, r0); // s43 mulhi
|
||||
r5 = r5 + r6 + ((((sel >> 12u) & 1u) != 0u) ? 0x1f682c68u : 0x3a8921a4u); // s44 add
|
||||
r4 = r4 + r5 + ((((sel >> 1u) & 1u) != 0u) ? 0xf246e180u : 0xeeca4334u); // s45 add
|
||||
r2 = r2 ^ r3; // s46 xor
|
||||
r2 = r2 ^ r5; // s47 xor
|
||||
r7 = r4 * r1 + r7; // s48 mad
|
||||
r2 = r6 * r1 + r2; // s49 mad
|
||||
r7 = r0 * r7 + r7; // s50 mad
|
||||
r5 = r5 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x41bd68ccu : 0x745776d8u); // s51 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 1u); r3 = r3 ^ t_; } // s52 shfl
|
||||
r4 = r4 | r0; // s53 or
|
||||
r0 = r2 * r4 + r0; // s54 mad
|
||||
r5 = rotl_imm(r5, 9u); // s55 rotl
|
||||
r5 = r5 * r7; // s56 mul
|
||||
r7 = rotl_imm(r7, 27u); // s57 rotl
|
||||
r7 = r7 - r6; // s58 sub
|
||||
r2 = rotl_imm(r2, 22u); // s59 rotl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 2u); r7 = r7 ^ t_; } // s60 shfl
|
||||
r4 = r4 + r0 + ((((sel >> 7u) & 1u) != 0u) ? 0xcba6a161u : 0x60d0ff55u); // s61 add
|
||||
r1 = r1 * r5; // s62 mul
|
||||
r7 = mul_hi(r7, r6); // s63 mulhi
|
||||
r5 = r5 * r0; // s64 mul
|
||||
r2 = rotr_var(r2, r5); // s65 rotr
|
||||
r0 = mul_hi(r0, r3); // s66 mulhi
|
||||
r0 = r0 | r7; // s67 or
|
||||
r0 = rotl_imm(r0, 19u); // s68 rotl
|
||||
r0 = rotl_imm(r0, 1u); // s69 rotl
|
||||
r2 = r2 | r3; // s70 or
|
||||
r3 = r7 * r5 + r3; // s71 mad
|
||||
r7 = r0 * r4 + r7; // s72 mad
|
||||
r4 = r4 | r3; // s73 or
|
||||
r3 = r3 ^ r1; // s74 xor
|
||||
r1 = r1 + r3 + ((((sel >> 15u) & 1u) != 0u) ? 0x8e35924eu : 0x3f4ffea2u); // s75 add
|
||||
r1 = r1 * r3; // s76 mul
|
||||
r0 = r3 * r2 + r0; // s77 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 16u); r2 = r2 ^ t_; } // s78 shfl
|
||||
r1 = mul_hi(r1, r0); // s79 mulhi
|
||||
r0 = mul_hi(r0, r4); // s80 mulhi
|
||||
r4 = r4 * r2; // s81 mul
|
||||
r5 = r5 - r7; // s82 sub
|
||||
r0 = r2 * r3 + r0; // s83 mad
|
||||
r7 = r7 - r2; // s84 sub
|
||||
r1 = r1 ^ r2; // s85 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r3, 16u); r6 = r6 ^ t_; } // s86 shfl
|
||||
r3 = r0 * r0 + r3; // s87 mad
|
||||
r0 = rotl_imm(r0, 20u); // s88 rotl
|
||||
r1 = r1 - r5; // s89 sub
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 16u); r6 = r6 ^ t_; } // s90 shfl
|
||||
r5 = r5 ^ r1; // s91 xor
|
||||
r5 = rotl_imm(r5, 12u); // s92 rotl
|
||||
r6 = r1 * r2 + r6; // s93 mad
|
||||
r2 = rotr_var(r2, r5); // s94 rotr
|
||||
r0 = r0 | r4; // s95 or
|
||||
r7 = r7 * r3; // s96 mul
|
||||
r5 = r5 | r0; // s97 or
|
||||
r3 = r3 + r1 + ((((sel >> 13u) & 1u) != 0u) ? 0x386bb642u : 0xbcfdd8a7u); // s98 add
|
||||
r0 = r0 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xa3962d03u : 0xc921a021u); // s99 add
|
||||
r1 = rotr_var(r1, r0); // s100 rotr
|
||||
r2 = r2 + r4 + ((((sel >> 18u) & 1u) != 0u) ? 0xc5990c5cu : 0x70da4067u); // s101 add
|
||||
r2 = rotl_imm(r2, 3u); // s102 rotl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 4u); r1 = r1 ^ t_; } // s103 shfl
|
||||
r0 = r0 ^ r2; // s104 xor
|
||||
r4 = r4 + r3 + ((((sel >> 28u) & 1u) != 0u) ? 0xb15aec75u : 0x1a5880cbu); // s105 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 2u); r4 = r4 ^ t_; } // s106 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 4u); r7 = r7 ^ t_; } // s107 shfl
|
||||
r1 = r7 * r3 + r1; // s108 mad
|
||||
r0 = r2 * r3 + r0; // s109 mad
|
||||
r2 = r2 | r4; // s110 or
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 1u); r6 = r6 ^ t_; } // s111 shfl
|
||||
r2 = r5 * r3 + r2; // s112 mad
|
||||
r4 = mul_hi(r4, r0); // s113 mulhi
|
||||
r7 = rotl_imm(r7, 29u); // s114 rotl
|
||||
r3 = r3 - r7; // s115 sub
|
||||
r0 = mul_hi(r0, r1); // s116 mulhi
|
||||
r4 = rotl_imm(r4, 12u); // s117 rotl
|
||||
r1 = r1 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0xc5e98ee5u : 0x24494ad3u); // s118 add
|
||||
r0 = r0 + r4 + ((((sel >> 4u) & 1u) != 0u) ? 0xa97bc949u : 0x1b2d1c83u); // s119 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 4u); r5 = r5 ^ t_; } // s120 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 8u); r3 = r3 ^ t_; } // s121 shfl
|
||||
r7 = rotl_imm(r7, 3u); // s122 rotl
|
||||
r4 = rotr_var(r4, r6); // s123 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 1u); r6 = r6 ^ t_; } // s124 shfl
|
||||
r5 = rotl_imm(r5, 21u); // s125 rotl
|
||||
r0 = r3 * r1 + r0; // s126 mad
|
||||
r0 = r0 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0xa7f6bb54u : 0xff14c08du); // s127 add
|
||||
r7 = r7 ^ r4; // s128 xor
|
||||
r3 = mul_hi(r3, r5); // s129 mulhi
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 8u); r5 = r5 ^ t_; } // s130 shfl
|
||||
r6 = mul_hi(r6, r4); // s131 mulhi
|
||||
r5 = r5 ^ r3; // s132 xor
|
||||
r0 = rotl_imm(r0, 19u); // s133 rotl
|
||||
r4 = rotr_var(r4, r3); // s134 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 16u); r1 = r1 ^ t_; } // s135 shfl
|
||||
r5 = r5 - r2; // s136 sub
|
||||
r3 = r3 ^ r5; // s137 xor
|
||||
r0 = r0 - r2; // s138 sub
|
||||
r3 = r3 * r7; // s139 mul
|
||||
r5 = r5 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0x2d465ca8u : 0x82ab98f4u); // s140 add
|
||||
r4 = r4 + r1 + ((((sel >> 11u) & 1u) != 0u) ? 0x237a9d8eu : 0x80594390u); // s141 add
|
||||
r3 = r7 * r0 + r3; // s142 mad
|
||||
r0 = r0 | r3; // s143 or
|
||||
r0 = mul_hi(r0, r1); // s144 mulhi
|
||||
r6 = r6 ^ r2; // s145 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 16u); r7 = r7 ^ t_; } // s146 shfl
|
||||
r0 = r0 + r1 + ((((sel >> 18u) & 1u) != 0u) ? 0xe68cf57eu : 0x07f15148u); // s147 add
|
||||
r7 = r7 | r3; // s148 or
|
||||
r4 = r4 ^ r2; // s149 xor
|
||||
r7 = r7 ^ r1; // s150 xor
|
||||
r5 = rotr_var(r5, r2); // s151 rotr
|
||||
r0 = r0 * r7; // s152 mul
|
||||
r4 = rotl_imm(r4, 16u); // s153 rotl
|
||||
r6 = rotr_var(r6, r4); // s154 rotr
|
||||
r6 = r6 ^ r1; // s155 xor
|
||||
r6 = r6 | r3; // s156 or
|
||||
r1 = r1 + r6 + ((((sel >> 19u) & 1u) != 0u) ? 0x84c2a09du : 0x14878c5au); // s157 add
|
||||
r4 = r4 - r1; // s158 sub
|
||||
r4 = r6 * r7 + r4; // s159 mad
|
||||
r1 = r1 * r5; // s160 mul
|
||||
r4 = r4 | r0; // s161 or
|
||||
r7 = rotl_imm(r7, 21u); // s162 rotl
|
||||
r0 = r0 - r1; // s163 sub
|
||||
r1 = r1 * r0; // s164 mul
|
||||
r2 = r2 - r1; // s165 sub
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 2u); r0 = r0 ^ t_; } // s166 shfl
|
||||
r0 = r0 - r4; // s167 sub
|
||||
r1 = rotr_var(r1, r6); // s168 rotr
|
||||
r7 = rotr_var(r7, r1); // s169 rotr
|
||||
r3 = r1 * r3 + r3; // s170 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 4u); r2 = r2 ^ t_; } // s171 shfl
|
||||
r2 = r2 - r1; // s172 sub
|
||||
r7 = r7 ^ r6; // s173 xor
|
||||
r4 = rotr_var(r4, r1); // s174 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r4 = r4 ^ t_; } // s175 shfl
|
||||
r6 = r6 * r2; // s176 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 4u); r0 = r0 ^ t_; } // s177 shfl
|
||||
r6 = r6 + r4 + ((((sel >> 21u) & 1u) != 0u) ? 0x509871e6u : 0x4fa43965u); // s178 add
|
||||
r0 = r0 ^ r6; // s179 xor
|
||||
r5 = r5 + r7 + ((((sel >> 0u) & 1u) != 0u) ? 0xb95b8a53u : 0xfb3c4bd8u); // s180 add
|
||||
r6 = r6 * r3; // s181 mul
|
||||
r7 = rotl_imm(r7, 1u); // s182 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 11u) & 1u) != 0u) ? 0x7047b7cfu : 0xc448a197u); // s183 add
|
||||
r3 = r3 * r5; // s184 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 1u); r5 = r5 ^ t_; } // s185 shfl
|
||||
r0 = r0 * r2; // s186 mul
|
||||
r4 = mul_hi(r4, r7); // s187 mulhi
|
||||
r2 = r7 * r2 + r2; // s188 mad
|
||||
r7 = r7 - r2; // s189 sub
|
||||
r2 = r2 + r5 + ((((sel >> 19u) & 1u) != 0u) ? 0x5caccc5du : 0x2fceaf49u); // s190 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 1u); r1 = r1 ^ t_; } // s191 shfl
|
||||
r7 = r7 + r1 + ((((sel >> 20u) & 1u) != 0u) ? 0x857bb5feu : 0xba0cee62u); // s192 add
|
||||
r6 = rotl_imm(r6, 22u); // s193 rotl
|
||||
r3 = r3 ^ r7; // s194 xor
|
||||
r7 = r7 * r0; // s195 mul
|
||||
r3 = r3 ^ r5; // s196 xor
|
||||
r5 = r5 + r1 + ((((sel >> 29u) & 1u) != 0u) ? 0xeae84577u : 0x8519428cu); // s197 add
|
||||
r7 = r4 * r4 + r7; // s198 mad
|
||||
r3 = r3 + r6 + ((((sel >> 8u) & 1u) != 0u) ? 0x0bfdbfa1u : 0x468639d3u); // s199 add
|
||||
r5 = r3 * r4 + r5; // s200 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 4u); r2 = r2 ^ t_; } // s201 shfl
|
||||
r7 = r7 + r4 + ((((sel >> 23u) & 1u) != 0u) ? 0x53ee8e50u : 0x18ec9a69u); // s202 add
|
||||
r6 = r6 ^ r0; // s203 xor
|
||||
r4 = r4 + r5 + ((((sel >> 4u) & 1u) != 0u) ? 0xac63376eu : 0x3e81485bu); // s204 add
|
||||
r7 = r1 * r4 + r7; // s205 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r1 = r1 ^ t_; } // s206 shfl
|
||||
r4 = r4 * r0; // s207 mul
|
||||
r1 = r1 | r6; // s208 or
|
||||
r6 = r6 + r4 + ((((sel >> 28u) & 1u) != 0u) ? 0x5b745519u : 0x66ccb75du); // s209 add
|
||||
r1 = rotr_var(r1, r3); // s210 rotr
|
||||
r5 = r5 - r4; // s211 sub
|
||||
r4 = r4 ^ r7; // s212 xor
|
||||
r1 = r1 + r7 + ((((sel >> 8u) & 1u) != 0u) ? 0x1c3dccf7u : 0x0ce0553du); // s213 add
|
||||
r0 = mul_hi(r0, r1); // s214 mulhi
|
||||
r3 = r3 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xfadc9205u : 0x96bfec89u); // s215 add
|
||||
r0 = rotl_imm(r0, 4u); // s216 rotl
|
||||
r6 = r6 ^ r4; // s217 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r6 = r6 ^ t_; } // s218 shfl
|
||||
r7 = r0 * r1 + r7; // s219 mad
|
||||
r4 = r4 * r0; // s220 mul
|
||||
r2 = r6 * r0 + r2; // s221 mad
|
||||
r5 = r5 ^ r4; // s222 xor
|
||||
r0 = r0 * r5; // s223 mul
|
||||
r2 = r2 + r0 + ((((sel >> 9u) & 1u) != 0u) ? 0x66b15c04u : 0x61495681u); // s224 add
|
||||
r3 = rotr_var(r3, r2); // s225 rotr
|
||||
r2 = r3 * r3 + r2; // s226 mad
|
||||
r6 = r6 ^ r0; // s227 xor
|
||||
r4 = rotl_imm(r4, 30u); // s228 rotl
|
||||
r2 = r2 + r3 + ((((sel >> 29u) & 1u) != 0u) ? 0x25bc17c2u : 0x70d05c34u); // s229 add
|
||||
r1 = rotr_var(r1, r5); // s230 rotr
|
||||
r1 = r1 + r4 + ((((sel >> 8u) & 1u) != 0u) ? 0x2d8728f3u : 0x32a28384u); // s231 add
|
||||
r1 = rotl_imm(r1, 6u); // s232 rotl
|
||||
r2 = r2 + r0 + ((((sel >> 21u) & 1u) != 0u) ? 0x61e81fe6u : 0xcf0949f1u); // s233 add
|
||||
r4 = rotl_imm(r4, 2u); // s234 rotl
|
||||
r2 = r2 - r6; // s235 sub
|
||||
r6 = r6 | r3; // s236 or
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 4u); r7 = r7 ^ t_; } // s237 shfl
|
||||
r1 = r1 ^ r3; // s238 xor
|
||||
r2 = rotl_imm(r2, 1u); // s239 rotl
|
||||
r5 = rotr_var(r5, r2); // s240 rotr
|
||||
r4 = rotl_imm(r4, 10u); // s241 rotl
|
||||
r6 = r6 + r2 + ((((sel >> 16u) & 1u) != 0u) ? 0x0a67565du : 0xe204fb50u); // s242 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r3, 2u); r1 = r1 ^ t_; } // s243 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 16u); r0 = r0 ^ t_; } // s244 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r5 = r5 ^ t_; } // s245 shfl
|
||||
r4 = r4 ^ r5; // s246 xor
|
||||
r7 = r7 ^ r2; // s247 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 16u); r0 = r0 ^ t_; } // s248 shfl
|
||||
r1 = mul_hi(r1, r3); // s249 mulhi
|
||||
r7 = r7 + r5 + ((((sel >> 31u) & 1u) != 0u) ? 0x5a79fd6fu : 0x3d6daffdu); // s250 add
|
||||
r6 = r6 + r0 + ((((sel >> 28u) & 1u) != 0u) ? 0x6beb28c2u : 0x84334aeau); // s251 add
|
||||
r7 = r7 - r1; // s252 sub
|
||||
r5 = rotr_var(r5, r2); // s253 rotr
|
||||
r2 = mul_hi(r2, r5); // s254 mulhi
|
||||
r6 = r6 * r3; // s255 mul
|
||||
}
|
||||
}
|
||||
uint lo = r0 ^ rotl_imm(r1, 7u) ^ rotl_imm(r2, 14u) ^ rotl_imm(r3, 21u);
|
||||
uint hi = r4 ^ rotl_imm(r5, 9u) ^ rotl_imm(r6, 18u) ^ rotl_imm(r7, 27u);
|
||||
out[gid] = ((ulong)hi << 32) | (ulong)lo;
|
||||
}
|
||||
|
||||
#if IGNEUM_EXCHANGE != 0
|
||||
// Reports the sub-group size this device uses for a work-group of IGNEUM_GROUP items. host.c runs it only when the
|
||||
// per-kernel query (clGetKernelSubGroupInfoKHR on igneum_hash) is unavailable; that query is preferred because a
|
||||
// compiler may pick a different wave width per kernel (RDNA: wave32 or wave64). See WAVEFRONT.md.
|
||||
IGNEUM_KERNEL_HASH void igneum_probe_subgroup(__global uint* out) {
|
||||
if (get_local_id(0) == 0u) { out[0] = get_sub_group_size(); out[1] = get_num_sub_groups(); }
|
||||
}
|
||||
#endif
|
||||
|
||||
// Header-bound variant (bind.rs): the init words come from initw, not SEEDW. Same body as igneum_hash.
|
||||
IGNEUM_KERNEL_HASH void igneum_hash_bound(__global const uint* ds, __global ulong* out, uint baseNonce, uint mask, __global const uint* initw) {
|
||||
uint gid = (uint)get_global_id(0);
|
||||
uint lid = (uint)get_local_id(0);
|
||||
uint nonce = baseNonce + gid;
|
||||
uint r0, r1, r2, r3, r4, r5, r6, r7;
|
||||
uint iw0 = initw[0], iw1 = initw[1], iw2 = initw[2], iw3 = initw[3], iw4 = initw[4], iw5 = initw[5], iw6 = initw[6], iw7 = initw[7];
|
||||
#if IGNEUM_EXCHANGE == 0
|
||||
IGNEUM_LOCAL_WORDS(xch, 2 * IGNEUM_GROUP);
|
||||
uint xk = 0u;
|
||||
#else
|
||||
(void)lid;
|
||||
#endif
|
||||
{ uint x = nonce ^ iw0; x += 0x9e3779b9u * 1u; x = splitmix32(x); r0 = x ^ iw1; }
|
||||
{ uint x = nonce ^ iw1; x += 0x9e3779b9u * 2u; x = splitmix32(x); r1 = x ^ iw2; }
|
||||
{ uint x = nonce ^ iw2; x += 0x9e3779b9u * 3u; x = splitmix32(x); r2 = x ^ iw3; }
|
||||
{ uint x = nonce ^ iw3; x += 0x9e3779b9u * 4u; x = splitmix32(x); r3 = x ^ iw4; }
|
||||
{ uint x = nonce ^ iw4; x += 0x9e3779b9u * 5u; x = splitmix32(x); r4 = x ^ iw5; }
|
||||
{ uint x = nonce ^ iw5; x += 0x9e3779b9u * 6u; x = splitmix32(x); r5 = x ^ iw6; }
|
||||
{ uint x = nonce ^ iw6; x += 0x9e3779b9u * 7u; x = splitmix32(x); r6 = x ^ iw7; }
|
||||
{ uint x = nonce ^ iw7; x += 0x9e3779b9u * 8u; x = splitmix32(x); r7 = x ^ iw0; }
|
||||
|
||||
for (uint it = 0u; it < 8u; ++it) {
|
||||
uint sel = r0;
|
||||
r6 = r6 - r3; // 0 sub
|
||||
r2 = r2 | r6; // 1 or
|
||||
r4 = r4 + r0 + ((((sel >> 4u) & 1u) != 0u) ? 0x7731324bu : 0x1a579b38u); // 2 add
|
||||
r3 = r3 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 3 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 1u); r0 = r0 ^ t_; } // 4 shfl
|
||||
r4 = r4 ^ r1; // 5 xor
|
||||
r3 = r5 * r3 + r3; // 6 mad
|
||||
r1 = r1 * r6; // 7 mul
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x08000000u) & mask]; // 8 load
|
||||
r0 = r0 ^ r4; // 9 xor
|
||||
r0 = r0 + r7 + ((((sel >> 30u) & 1u) != 0u) ? 0xb10fcef8u : 0xee083919u); // 10 add
|
||||
r0 = mul_hi(r0, r2); // 11 mulhi
|
||||
r5 = r5 ^ r4; // 12 xor
|
||||
r7 = r3 * r0 + r7; // 13 mad
|
||||
r3 = r3 ^ ds[((rotl_imm(r4 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 14 load
|
||||
r2 = r2 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 15 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r7 = r7 ^ t_; } // 16 shfl
|
||||
r7 = r7 * r6; // 17 mul
|
||||
r2 = r3 * r1 + r2; // 18 mad
|
||||
r5 = rotr_var(r5, r2); // 19 rotr
|
||||
r5 = r5 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 20 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 16u); r6 = r6 ^ t_; } // 21 shfl
|
||||
r1 = mul_hi(r1, r2); // 22 mulhi
|
||||
r6 = r6 ^ r5; // 23 xor
|
||||
r5 = r2 * r0 + r5; // 24 mad
|
||||
r6 = r3 * r7 + r6; // 25 mad
|
||||
r6 = r6 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 26 load
|
||||
r1 = r5 * r5 + r1; // 27 mad
|
||||
r0 = r0 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 28 load
|
||||
r7 = r7 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0x0b1f02c7u : 0xd5e6c37cu); // 29 add
|
||||
r7 = r7 + r0 + ((((sel >> 0u) & 1u) != 0u) ? 0x2d6e8bd1u : 0x4e45ba51u); // 30 add
|
||||
r0 = r0 ^ r2; // 31 xor
|
||||
r6 = r6 + r2 + ((((sel >> 13u) & 1u) != 0u) ? 0x351422d2u : 0x1c91b2a1u); // 32 add
|
||||
r4 = mul_hi(r4, r5); // 33 mulhi
|
||||
r3 = rotr_var(r3, r5); // 34 rotr
|
||||
r3 = r3 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 35 load
|
||||
r4 = rotl_imm(r4, 13u); // 36 rotl
|
||||
r6 = r6 + r1 + ((((sel >> 7u) & 1u) != 0u) ? 0x64a28251u : 0x3f2970f7u); // 37 add
|
||||
r2 = r2 * r0; // 38 mul
|
||||
r3 = r3 + r4 + ((((sel >> 15u) & 1u) != 0u) ? 0x7378c955u : 0x3b7b3317u); // 39 add
|
||||
r0 = r0 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 40 load
|
||||
r6 = r6 ^ r5; // 41 xor
|
||||
r3 = r3 + r0 + ((((sel >> 2u) & 1u) != 0u) ? 0xb31f6a64u : 0x6678b059u); // 42 add
|
||||
r7 = r7 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 43 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 4u); r3 = r3 ^ t_; } // 44 shfl
|
||||
r1 = mul_hi(r1, r5); // 45 mulhi
|
||||
r0 = rotl_imm(r0, 6u); // 46 rotl
|
||||
r1 = r1 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 47 load
|
||||
r4 = r4 + r0 + ((((sel >> 17u) & 1u) != 0u) ? 0x5feccee6u : 0x43095946u); // 48 add
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 49 load
|
||||
r2 = r2 - r3; // 50 sub
|
||||
r7 = r7 + r2 + ((((sel >> 24u) & 1u) != 0u) ? 0x26e2b582u : 0xf45ecdf8u); // 51 add
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 52 load
|
||||
r6 = r6 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 53 load
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 4u); r7 = r7 ^ t_; } // 54 shfl
|
||||
r1 = rotr_var(r1, r2); // 55 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r0 = r0 ^ t_; } // 56 shfl
|
||||
r5 = rotl_imm(r5, 12u); // 57 rotl
|
||||
r5 = r5 + r3 + ((((sel >> 4u) & 1u) != 0u) ? 0x473b1718u : 0x6b4f35e8u); // 58 add
|
||||
r4 = rotr_var(r4, r1); // 59 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 1u); r4 = r4 ^ t_; } // 60 shfl
|
||||
r7 = r7 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 61 load
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 62 load
|
||||
r4 = r4 * r1; // 63 mul
|
||||
// latency-shadow block (Counter ASIC 3.0 item 8): 256 ALU instructions x 27 passes after instruction 63, no load
|
||||
for (uint sh = 0u; sh < 27u; ++sh) {
|
||||
r6 = r6 | r0; // s0 or
|
||||
r7 = r7 * r3; // s1 mul
|
||||
r3 = r5 * r5 + r3; // s2 mad
|
||||
r7 = r7 + r4 + ((((sel >> 6u) & 1u) != 0u) ? 0x38e58a06u : 0x97d3d105u); // s3 add
|
||||
r6 = r6 + r0 + ((((sel >> 20u) & 1u) != 0u) ? 0x9629673du : 0x29b87b9au); // s4 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 1u); r6 = r6 ^ t_; } // s5 shfl
|
||||
r5 = r3 * r5 + r5; // s6 mad
|
||||
r3 = rotl_imm(r3, 5u); // s7 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 12u) & 1u) != 0u) ? 0x4a5f55e1u : 0xa7c1fb0fu); // s8 add
|
||||
r5 = r5 * r7; // s9 mul
|
||||
r4 = r4 ^ r5; // s10 xor
|
||||
r6 = mul_hi(r6, r4); // s11 mulhi
|
||||
r5 = rotl_imm(r5, 13u); // s12 rotl
|
||||
r0 = r2 * r1 + r0; // s13 mad
|
||||
r2 = mul_hi(r2, r4); // s14 mulhi
|
||||
r3 = r3 | r1; // s15 or
|
||||
r0 = r0 - r5; // s16 sub
|
||||
r0 = r0 * r6; // s17 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 4u); r5 = r5 ^ t_; } // s18 shfl
|
||||
r3 = r3 | r4; // s19 or
|
||||
r6 = r3 * r2 + r6; // s20 mad
|
||||
r0 = rotl_imm(r0, 11u); // s21 rotl
|
||||
r3 = rotl_imm(r3, 25u); // s22 rotl
|
||||
r6 = r6 ^ r4; // s23 xor
|
||||
r4 = r4 | r2; // s24 or
|
||||
r1 = r1 + r6 + ((((sel >> 7u) & 1u) != 0u) ? 0xfc2f6d17u : 0x53e74799u); // s25 add
|
||||
r7 = rotr_var(r7, r6); // s26 rotr
|
||||
r7 = r7 + r5 + ((((sel >> 27u) & 1u) != 0u) ? 0x787048dcu : 0xb8fae90eu); // s27 add
|
||||
r4 = r4 ^ r1; // s28 xor
|
||||
r6 = r6 ^ r1; // s29 xor
|
||||
r2 = mul_hi(r2, r0); // s30 mulhi
|
||||
r1 = rotr_var(r1, r4); // s31 rotr
|
||||
r7 = r7 ^ r4; // s32 xor
|
||||
r3 = r2 * r6 + r3; // s33 mad
|
||||
r3 = r3 + r1 + ((((sel >> 12u) & 1u) != 0u) ? 0x9ffd510bu : 0xd2dc42ebu); // s34 add
|
||||
r4 = r4 + r7 + ((((sel >> 7u) & 1u) != 0u) ? 0xf871ba37u : 0xad8eda6fu); // s35 add
|
||||
r7 = r7 + r4 + ((((sel >> 3u) & 1u) != 0u) ? 0xd02d30cau : 0xb0cef8f4u); // s36 add
|
||||
r4 = r3 * r2 + r4; // s37 mad
|
||||
r5 = r7 * r7 + r5; // s38 mad
|
||||
r3 = r3 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x82e98a22u : 0x2bd506e2u); // s39 add
|
||||
r0 = rotl_imm(r0, 21u); // s40 rotl
|
||||
r1 = r1 | r0; // s41 or
|
||||
r7 = r7 + r5 + ((((sel >> 28u) & 1u) != 0u) ? 0x6435f6d8u : 0xc0386e49u); // s42 add
|
||||
r1 = mul_hi(r1, r0); // s43 mulhi
|
||||
r5 = r5 + r6 + ((((sel >> 12u) & 1u) != 0u) ? 0x1f682c68u : 0x3a8921a4u); // s44 add
|
||||
r4 = r4 + r5 + ((((sel >> 1u) & 1u) != 0u) ? 0xf246e180u : 0xeeca4334u); // s45 add
|
||||
r2 = r2 ^ r3; // s46 xor
|
||||
r2 = r2 ^ r5; // s47 xor
|
||||
r7 = r4 * r1 + r7; // s48 mad
|
||||
r2 = r6 * r1 + r2; // s49 mad
|
||||
r7 = r0 * r7 + r7; // s50 mad
|
||||
r5 = r5 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x41bd68ccu : 0x745776d8u); // s51 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 1u); r3 = r3 ^ t_; } // s52 shfl
|
||||
r4 = r4 | r0; // s53 or
|
||||
r0 = r2 * r4 + r0; // s54 mad
|
||||
r5 = rotl_imm(r5, 9u); // s55 rotl
|
||||
r5 = r5 * r7; // s56 mul
|
||||
r7 = rotl_imm(r7, 27u); // s57 rotl
|
||||
r7 = r7 - r6; // s58 sub
|
||||
r2 = rotl_imm(r2, 22u); // s59 rotl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 2u); r7 = r7 ^ t_; } // s60 shfl
|
||||
r4 = r4 + r0 + ((((sel >> 7u) & 1u) != 0u) ? 0xcba6a161u : 0x60d0ff55u); // s61 add
|
||||
r1 = r1 * r5; // s62 mul
|
||||
r7 = mul_hi(r7, r6); // s63 mulhi
|
||||
r5 = r5 * r0; // s64 mul
|
||||
r2 = rotr_var(r2, r5); // s65 rotr
|
||||
r0 = mul_hi(r0, r3); // s66 mulhi
|
||||
r0 = r0 | r7; // s67 or
|
||||
r0 = rotl_imm(r0, 19u); // s68 rotl
|
||||
r0 = rotl_imm(r0, 1u); // s69 rotl
|
||||
r2 = r2 | r3; // s70 or
|
||||
r3 = r7 * r5 + r3; // s71 mad
|
||||
r7 = r0 * r4 + r7; // s72 mad
|
||||
r4 = r4 | r3; // s73 or
|
||||
r3 = r3 ^ r1; // s74 xor
|
||||
r1 = r1 + r3 + ((((sel >> 15u) & 1u) != 0u) ? 0x8e35924eu : 0x3f4ffea2u); // s75 add
|
||||
r1 = r1 * r3; // s76 mul
|
||||
r0 = r3 * r2 + r0; // s77 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 16u); r2 = r2 ^ t_; } // s78 shfl
|
||||
r1 = mul_hi(r1, r0); // s79 mulhi
|
||||
r0 = mul_hi(r0, r4); // s80 mulhi
|
||||
r4 = r4 * r2; // s81 mul
|
||||
r5 = r5 - r7; // s82 sub
|
||||
r0 = r2 * r3 + r0; // s83 mad
|
||||
r7 = r7 - r2; // s84 sub
|
||||
r1 = r1 ^ r2; // s85 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r3, 16u); r6 = r6 ^ t_; } // s86 shfl
|
||||
r3 = r0 * r0 + r3; // s87 mad
|
||||
r0 = rotl_imm(r0, 20u); // s88 rotl
|
||||
r1 = r1 - r5; // s89 sub
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 16u); r6 = r6 ^ t_; } // s90 shfl
|
||||
r5 = r5 ^ r1; // s91 xor
|
||||
r5 = rotl_imm(r5, 12u); // s92 rotl
|
||||
r6 = r1 * r2 + r6; // s93 mad
|
||||
r2 = rotr_var(r2, r5); // s94 rotr
|
||||
r0 = r0 | r4; // s95 or
|
||||
r7 = r7 * r3; // s96 mul
|
||||
r5 = r5 | r0; // s97 or
|
||||
r3 = r3 + r1 + ((((sel >> 13u) & 1u) != 0u) ? 0x386bb642u : 0xbcfdd8a7u); // s98 add
|
||||
r0 = r0 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xa3962d03u : 0xc921a021u); // s99 add
|
||||
r1 = rotr_var(r1, r0); // s100 rotr
|
||||
r2 = r2 + r4 + ((((sel >> 18u) & 1u) != 0u) ? 0xc5990c5cu : 0x70da4067u); // s101 add
|
||||
r2 = rotl_imm(r2, 3u); // s102 rotl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 4u); r1 = r1 ^ t_; } // s103 shfl
|
||||
r0 = r0 ^ r2; // s104 xor
|
||||
r4 = r4 + r3 + ((((sel >> 28u) & 1u) != 0u) ? 0xb15aec75u : 0x1a5880cbu); // s105 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 2u); r4 = r4 ^ t_; } // s106 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 4u); r7 = r7 ^ t_; } // s107 shfl
|
||||
r1 = r7 * r3 + r1; // s108 mad
|
||||
r0 = r2 * r3 + r0; // s109 mad
|
||||
r2 = r2 | r4; // s110 or
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 1u); r6 = r6 ^ t_; } // s111 shfl
|
||||
r2 = r5 * r3 + r2; // s112 mad
|
||||
r4 = mul_hi(r4, r0); // s113 mulhi
|
||||
r7 = rotl_imm(r7, 29u); // s114 rotl
|
||||
r3 = r3 - r7; // s115 sub
|
||||
r0 = mul_hi(r0, r1); // s116 mulhi
|
||||
r4 = rotl_imm(r4, 12u); // s117 rotl
|
||||
r1 = r1 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0xc5e98ee5u : 0x24494ad3u); // s118 add
|
||||
r0 = r0 + r4 + ((((sel >> 4u) & 1u) != 0u) ? 0xa97bc949u : 0x1b2d1c83u); // s119 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r4, 4u); r5 = r5 ^ t_; } // s120 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 8u); r3 = r3 ^ t_; } // s121 shfl
|
||||
r7 = rotl_imm(r7, 3u); // s122 rotl
|
||||
r4 = rotr_var(r4, r6); // s123 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 1u); r6 = r6 ^ t_; } // s124 shfl
|
||||
r5 = rotl_imm(r5, 21u); // s125 rotl
|
||||
r0 = r3 * r1 + r0; // s126 mad
|
||||
r0 = r0 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0xa7f6bb54u : 0xff14c08du); // s127 add
|
||||
r7 = r7 ^ r4; // s128 xor
|
||||
r3 = mul_hi(r3, r5); // s129 mulhi
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 8u); r5 = r5 ^ t_; } // s130 shfl
|
||||
r6 = mul_hi(r6, r4); // s131 mulhi
|
||||
r5 = r5 ^ r3; // s132 xor
|
||||
r0 = rotl_imm(r0, 19u); // s133 rotl
|
||||
r4 = rotr_var(r4, r3); // s134 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 16u); r1 = r1 ^ t_; } // s135 shfl
|
||||
r5 = r5 - r2; // s136 sub
|
||||
r3 = r3 ^ r5; // s137 xor
|
||||
r0 = r0 - r2; // s138 sub
|
||||
r3 = r3 * r7; // s139 mul
|
||||
r5 = r5 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0x2d465ca8u : 0x82ab98f4u); // s140 add
|
||||
r4 = r4 + r1 + ((((sel >> 11u) & 1u) != 0u) ? 0x237a9d8eu : 0x80594390u); // s141 add
|
||||
r3 = r7 * r0 + r3; // s142 mad
|
||||
r0 = r0 | r3; // s143 or
|
||||
r0 = mul_hi(r0, r1); // s144 mulhi
|
||||
r6 = r6 ^ r2; // s145 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 16u); r7 = r7 ^ t_; } // s146 shfl
|
||||
r0 = r0 + r1 + ((((sel >> 18u) & 1u) != 0u) ? 0xe68cf57eu : 0x07f15148u); // s147 add
|
||||
r7 = r7 | r3; // s148 or
|
||||
r4 = r4 ^ r2; // s149 xor
|
||||
r7 = r7 ^ r1; // s150 xor
|
||||
r5 = rotr_var(r5, r2); // s151 rotr
|
||||
r0 = r0 * r7; // s152 mul
|
||||
r4 = rotl_imm(r4, 16u); // s153 rotl
|
||||
r6 = rotr_var(r6, r4); // s154 rotr
|
||||
r6 = r6 ^ r1; // s155 xor
|
||||
r6 = r6 | r3; // s156 or
|
||||
r1 = r1 + r6 + ((((sel >> 19u) & 1u) != 0u) ? 0x84c2a09du : 0x14878c5au); // s157 add
|
||||
r4 = r4 - r1; // s158 sub
|
||||
r4 = r6 * r7 + r4; // s159 mad
|
||||
r1 = r1 * r5; // s160 mul
|
||||
r4 = r4 | r0; // s161 or
|
||||
r7 = rotl_imm(r7, 21u); // s162 rotl
|
||||
r0 = r0 - r1; // s163 sub
|
||||
r1 = r1 * r0; // s164 mul
|
||||
r2 = r2 - r1; // s165 sub
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 2u); r0 = r0 ^ t_; } // s166 shfl
|
||||
r0 = r0 - r4; // s167 sub
|
||||
r1 = rotr_var(r1, r6); // s168 rotr
|
||||
r7 = rotr_var(r7, r1); // s169 rotr
|
||||
r3 = r1 * r3 + r3; // s170 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r0, 4u); r2 = r2 ^ t_; } // s171 shfl
|
||||
r2 = r2 - r1; // s172 sub
|
||||
r7 = r7 ^ r6; // s173 xor
|
||||
r4 = rotr_var(r4, r1); // s174 rotr
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r4 = r4 ^ t_; } // s175 shfl
|
||||
r6 = r6 * r2; // s176 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 4u); r0 = r0 ^ t_; } // s177 shfl
|
||||
r6 = r6 + r4 + ((((sel >> 21u) & 1u) != 0u) ? 0x509871e6u : 0x4fa43965u); // s178 add
|
||||
r0 = r0 ^ r6; // s179 xor
|
||||
r5 = r5 + r7 + ((((sel >> 0u) & 1u) != 0u) ? 0xb95b8a53u : 0xfb3c4bd8u); // s180 add
|
||||
r6 = r6 * r3; // s181 mul
|
||||
r7 = rotl_imm(r7, 1u); // s182 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 11u) & 1u) != 0u) ? 0x7047b7cfu : 0xc448a197u); // s183 add
|
||||
r3 = r3 * r5; // s184 mul
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 1u); r5 = r5 ^ t_; } // s185 shfl
|
||||
r0 = r0 * r2; // s186 mul
|
||||
r4 = mul_hi(r4, r7); // s187 mulhi
|
||||
r2 = r7 * r2 + r2; // s188 mad
|
||||
r7 = r7 - r2; // s189 sub
|
||||
r2 = r2 + r5 + ((((sel >> 19u) & 1u) != 0u) ? 0x5caccc5du : 0x2fceaf49u); // s190 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r2, 1u); r1 = r1 ^ t_; } // s191 shfl
|
||||
r7 = r7 + r1 + ((((sel >> 20u) & 1u) != 0u) ? 0x857bb5feu : 0xba0cee62u); // s192 add
|
||||
r6 = rotl_imm(r6, 22u); // s193 rotl
|
||||
r3 = r3 ^ r7; // s194 xor
|
||||
r7 = r7 * r0; // s195 mul
|
||||
r3 = r3 ^ r5; // s196 xor
|
||||
r5 = r5 + r1 + ((((sel >> 29u) & 1u) != 0u) ? 0xeae84577u : 0x8519428cu); // s197 add
|
||||
r7 = r4 * r4 + r7; // s198 mad
|
||||
r3 = r3 + r6 + ((((sel >> 8u) & 1u) != 0u) ? 0x0bfdbfa1u : 0x468639d3u); // s199 add
|
||||
r5 = r3 * r4 + r5; // s200 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 4u); r2 = r2 ^ t_; } // s201 shfl
|
||||
r7 = r7 + r4 + ((((sel >> 23u) & 1u) != 0u) ? 0x53ee8e50u : 0x18ec9a69u); // s202 add
|
||||
r6 = r6 ^ r0; // s203 xor
|
||||
r4 = r4 + r5 + ((((sel >> 4u) & 1u) != 0u) ? 0xac63376eu : 0x3e81485bu); // s204 add
|
||||
r7 = r1 * r4 + r7; // s205 mad
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 16u); r1 = r1 ^ t_; } // s206 shfl
|
||||
r4 = r4 * r0; // s207 mul
|
||||
r1 = r1 | r6; // s208 or
|
||||
r6 = r6 + r4 + ((((sel >> 28u) & 1u) != 0u) ? 0x5b745519u : 0x66ccb75du); // s209 add
|
||||
r1 = rotr_var(r1, r3); // s210 rotr
|
||||
r5 = r5 - r4; // s211 sub
|
||||
r4 = r4 ^ r7; // s212 xor
|
||||
r1 = r1 + r7 + ((((sel >> 8u) & 1u) != 0u) ? 0x1c3dccf7u : 0x0ce0553du); // s213 add
|
||||
r0 = mul_hi(r0, r1); // s214 mulhi
|
||||
r3 = r3 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xfadc9205u : 0x96bfec89u); // s215 add
|
||||
r0 = rotl_imm(r0, 4u); // s216 rotl
|
||||
r6 = r6 ^ r4; // s217 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r6 = r6 ^ t_; } // s218 shfl
|
||||
r7 = r0 * r1 + r7; // s219 mad
|
||||
r4 = r4 * r0; // s220 mul
|
||||
r2 = r6 * r0 + r2; // s221 mad
|
||||
r5 = r5 ^ r4; // s222 xor
|
||||
r0 = r0 * r5; // s223 mul
|
||||
r2 = r2 + r0 + ((((sel >> 9u) & 1u) != 0u) ? 0x66b15c04u : 0x61495681u); // s224 add
|
||||
r3 = rotr_var(r3, r2); // s225 rotr
|
||||
r2 = r3 * r3 + r2; // s226 mad
|
||||
r6 = r6 ^ r0; // s227 xor
|
||||
r4 = rotl_imm(r4, 30u); // s228 rotl
|
||||
r2 = r2 + r3 + ((((sel >> 29u) & 1u) != 0u) ? 0x25bc17c2u : 0x70d05c34u); // s229 add
|
||||
r1 = rotr_var(r1, r5); // s230 rotr
|
||||
r1 = r1 + r4 + ((((sel >> 8u) & 1u) != 0u) ? 0x2d8728f3u : 0x32a28384u); // s231 add
|
||||
r1 = rotl_imm(r1, 6u); // s232 rotl
|
||||
r2 = r2 + r0 + ((((sel >> 21u) & 1u) != 0u) ? 0x61e81fe6u : 0xcf0949f1u); // s233 add
|
||||
r4 = rotl_imm(r4, 2u); // s234 rotl
|
||||
r2 = r2 - r6; // s235 sub
|
||||
r6 = r6 | r3; // s236 or
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r5, 4u); r7 = r7 ^ t_; } // s237 shfl
|
||||
r1 = r1 ^ r3; // s238 xor
|
||||
r2 = rotl_imm(r2, 1u); // s239 rotl
|
||||
r5 = rotr_var(r5, r2); // s240 rotr
|
||||
r4 = rotl_imm(r4, 10u); // s241 rotl
|
||||
r6 = r6 + r2 + ((((sel >> 16u) & 1u) != 0u) ? 0x0a67565du : 0xe204fb50u); // s242 add
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r3, 2u); r1 = r1 ^ t_; } // s243 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r1, 16u); r0 = r0 ^ t_; } // s244 shfl
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r7, 2u); r5 = r5 ^ t_; } // s245 shfl
|
||||
r4 = r4 ^ r5; // s246 xor
|
||||
r7 = r7 ^ r2; // s247 xor
|
||||
{ uint t_; IGNEUM_SHFL_XOR(t_, r6, 16u); r0 = r0 ^ t_; } // s248 shfl
|
||||
r1 = mul_hi(r1, r3); // s249 mulhi
|
||||
r7 = r7 + r5 + ((((sel >> 31u) & 1u) != 0u) ? 0x5a79fd6fu : 0x3d6daffdu); // s250 add
|
||||
r6 = r6 + r0 + ((((sel >> 28u) & 1u) != 0u) ? 0x6beb28c2u : 0x84334aeau); // s251 add
|
||||
r7 = r7 - r1; // s252 sub
|
||||
r5 = rotr_var(r5, r2); // s253 rotr
|
||||
r2 = mul_hi(r2, r5); // s254 mulhi
|
||||
r6 = r6 * r3; // s255 mul
|
||||
}
|
||||
}
|
||||
uint lo = r0 ^ rotl_imm(r1, 7u) ^ rotl_imm(r2, 14u) ^ rotl_imm(r3, 21u);
|
||||
uint hi = r4 ^ rotl_imm(r5, 9u) ^ rotl_imm(r6, 18u) ^ rotl_imm(r7, 27u);
|
||||
out[gid] = ((ulong)hi << 32) | (ulong)lo;
|
||||
}
|
||||
382
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel_bound.cu
Normal file
382
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/kernel_bound.cu
Normal file
|
|
@ -0,0 +1,382 @@
|
|||
// Generated by igneum-pow export (generator v2) for seed "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000". Do not edit by hand.
|
||||
// Header-bound twin of igneum_hash in kernel.cu: the init words come from a kernel argument, not SEEDW.
|
||||
// Host declarations (also in program_bound.h if present):
|
||||
// struct IgneumInitWords { uint32_t w[8]; };
|
||||
// cudaError_t igneum_launch_hash_bound(const uint32_t* ds, uint64_t* out, uint32_t baseNonce, uint32_t mask,
|
||||
// IgneumInitWords iw, uint32_t nonces, uint32_t blockWarps);
|
||||
// cudaError_t igneum_hash_bound_info(int* numRegs, int* blocksPerSM, uint32_t blockWarps);
|
||||
#include <cuda_runtime.h>
|
||||
#include <cstdint>
|
||||
#include "program.h"
|
||||
|
||||
struct IgneumInitWords { uint32_t w[8]; };
|
||||
|
||||
__device__ __forceinline__ uint32_t splitmix32(uint32_t x) {
|
||||
x ^= x >> 16; x *= 0x7feb352du;
|
||||
x ^= x >> 15; x *= 0x846ca68bu;
|
||||
x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
__device__ __forceinline__ uint32_t rotl_imm(uint32_t x, uint32_t n) { return (x << n) | (x >> (32u - n)); }
|
||||
__device__ __forceinline__ uint32_t rotr_var(uint32_t x, uint32_t n) { n &= 31u; return (x >> n) | (x << ((32u - n) & 31u)); }
|
||||
|
||||
__global__ void igneum_hash_bound(const uint32_t* ds, uint64_t* out, uint32_t baseNonce, uint32_t mask, IgneumInitWords iw) {
|
||||
uint32_t gid = blockIdx.x * blockDim.x + threadIdx.x;
|
||||
uint32_t nonce = baseNonce + gid;
|
||||
uint32_t r0, r1, r2, r3, r4, r5, r6, r7;
|
||||
{ uint32_t x = nonce ^ iw.w[0]; x += 0x9e3779b9u * 1u; x = splitmix32(x); r0 = x ^ iw.w[1]; }
|
||||
{ uint32_t x = nonce ^ iw.w[1]; x += 0x9e3779b9u * 2u; x = splitmix32(x); r1 = x ^ iw.w[2]; }
|
||||
{ uint32_t x = nonce ^ iw.w[2]; x += 0x9e3779b9u * 3u; x = splitmix32(x); r2 = x ^ iw.w[3]; }
|
||||
{ uint32_t x = nonce ^ iw.w[3]; x += 0x9e3779b9u * 4u; x = splitmix32(x); r3 = x ^ iw.w[4]; }
|
||||
{ uint32_t x = nonce ^ iw.w[4]; x += 0x9e3779b9u * 5u; x = splitmix32(x); r4 = x ^ iw.w[5]; }
|
||||
{ uint32_t x = nonce ^ iw.w[5]; x += 0x9e3779b9u * 6u; x = splitmix32(x); r5 = x ^ iw.w[6]; }
|
||||
{ uint32_t x = nonce ^ iw.w[6]; x += 0x9e3779b9u * 7u; x = splitmix32(x); r6 = x ^ iw.w[7]; }
|
||||
{ uint32_t x = nonce ^ iw.w[7]; x += 0x9e3779b9u * 8u; x = splitmix32(x); r7 = x ^ iw.w[0]; }
|
||||
|
||||
for (uint32_t it = 0u; it < 8u; ++it) {
|
||||
uint32_t sel = r0;
|
||||
r6 = r6 - r3; // 0 sub
|
||||
r2 = r2 | r6; // 1 or
|
||||
r4 = r4 + r0 + ((((sel >> 4u) & 1u) != 0u) ? 0x7731324bu : 0x1a579b38u); // 2 add
|
||||
r3 = r3 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 3 load
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r6, 1); // 4 shfl
|
||||
r4 = r4 ^ r1; // 5 xor
|
||||
r3 = r5 * r3 + r3; // 6 mad
|
||||
r1 = r1 * r6; // 7 mul
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x08000000u) & mask]; // 8 load
|
||||
r0 = r0 ^ r4; // 9 xor
|
||||
r0 = r0 + r7 + ((((sel >> 30u) & 1u) != 0u) ? 0xb10fcef8u : 0xee083919u); // 10 add
|
||||
r0 = __umulhi(r0, r2); // 11 mulhi
|
||||
r5 = r5 ^ r4; // 12 xor
|
||||
r7 = r3 * r0 + r7; // 13 mad
|
||||
r3 = r3 ^ ds[((rotl_imm(r4 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 14 load
|
||||
r2 = r2 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 15 load
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r5, 16); // 16 shfl
|
||||
r7 = r7 * r6; // 17 mul
|
||||
r2 = r3 * r1 + r2; // 18 mad
|
||||
r5 = rotr_var(r5, r2); // 19 rotr
|
||||
r5 = r5 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 20 load
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r4, 16); // 21 shfl
|
||||
r1 = __umulhi(r1, r2); // 22 mulhi
|
||||
r6 = r6 ^ r5; // 23 xor
|
||||
r5 = r2 * r0 + r5; // 24 mad
|
||||
r6 = r3 * r7 + r6; // 25 mad
|
||||
r6 = r6 ^ ds[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 26 load
|
||||
r1 = r5 * r5 + r1; // 27 mad
|
||||
r0 = r0 ^ ds[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 28 load
|
||||
r7 = r7 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0x0b1f02c7u : 0xd5e6c37cu); // 29 add
|
||||
r7 = r7 + r0 + ((((sel >> 0u) & 1u) != 0u) ? 0x2d6e8bd1u : 0x4e45ba51u); // 30 add
|
||||
r0 = r0 ^ r2; // 31 xor
|
||||
r6 = r6 + r2 + ((((sel >> 13u) & 1u) != 0u) ? 0x351422d2u : 0x1c91b2a1u); // 32 add
|
||||
r4 = __umulhi(r4, r5); // 33 mulhi
|
||||
r3 = rotr_var(r3, r5); // 34 rotr
|
||||
r3 = r3 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 35 load
|
||||
r4 = rotl_imm(r4, 13u); // 36 rotl
|
||||
r6 = r6 + r1 + ((((sel >> 7u) & 1u) != 0u) ? 0x64a28251u : 0x3f2970f7u); // 37 add
|
||||
r2 = r2 * r0; // 38 mul
|
||||
r3 = r3 + r4 + ((((sel >> 15u) & 1u) != 0u) ? 0x7378c955u : 0x3b7b3317u); // 39 add
|
||||
r0 = r0 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 40 load
|
||||
r6 = r6 ^ r5; // 41 xor
|
||||
r3 = r3 + r0 + ((((sel >> 2u) & 1u) != 0u) ? 0xb31f6a64u : 0x6678b059u); // 42 add
|
||||
r7 = r7 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & mask]; // 43 load
|
||||
r3 = r3 ^ __shfl_xor_sync(0xffffffffu, r7, 4); // 44 shfl
|
||||
r1 = __umulhi(r1, r5); // 45 mulhi
|
||||
r0 = rotl_imm(r0, 6u); // 46 rotl
|
||||
r1 = r1 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 47 load
|
||||
r4 = r4 + r0 + ((((sel >> 17u) & 1u) != 0u) ? 0x5feccee6u : 0x43095946u); // 48 add
|
||||
r2 = r2 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 49 load
|
||||
r2 = r2 - r3; // 50 sub
|
||||
r7 = r7 + r2 + ((((sel >> 24u) & 1u) != 0u) ? 0x26e2b582u : 0xf45ecdf8u); // 51 add
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & mask]; // 52 load
|
||||
r6 = r6 ^ ds[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & mask]; // 53 load
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r5, 4); // 54 shfl
|
||||
r1 = rotr_var(r1, r2); // 55 rotr
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r5, 16); // 56 shfl
|
||||
r5 = rotl_imm(r5, 12u); // 57 rotl
|
||||
r5 = r5 + r3 + ((((sel >> 4u) & 1u) != 0u) ? 0x473b1718u : 0x6b4f35e8u); // 58 add
|
||||
r4 = rotr_var(r4, r1); // 59 rotr
|
||||
r4 = r4 ^ __shfl_xor_sync(0xffffffffu, r5, 1); // 60 shfl
|
||||
r7 = r7 ^ ds[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 61 load
|
||||
r5 = r5 ^ ds[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & mask]; // 62 load
|
||||
r4 = r4 * r1; // 63 mul
|
||||
// latency-shadow block (Counter ASIC 3.0 item 8): 256 ALU instructions x 27 passes after instruction 63, no load
|
||||
for (uint32_t sh = 0u; sh < 27u; ++sh) {
|
||||
r6 = r6 | r0; // s0 or
|
||||
r7 = r7 * r3; // s1 mul
|
||||
r3 = r5 * r5 + r3; // s2 mad
|
||||
r7 = r7 + r4 + ((((sel >> 6u) & 1u) != 0u) ? 0x38e58a06u : 0x97d3d105u); // s3 add
|
||||
r6 = r6 + r0 + ((((sel >> 20u) & 1u) != 0u) ? 0x9629673du : 0x29b87b9au); // s4 add
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r2, 1); // s5 shfl
|
||||
r5 = r3 * r5 + r5; // s6 mad
|
||||
r3 = rotl_imm(r3, 5u); // s7 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 12u) & 1u) != 0u) ? 0x4a5f55e1u : 0xa7c1fb0fu); // s8 add
|
||||
r5 = r5 * r7; // s9 mul
|
||||
r4 = r4 ^ r5; // s10 xor
|
||||
r6 = __umulhi(r6, r4); // s11 mulhi
|
||||
r5 = rotl_imm(r5, 13u); // s12 rotl
|
||||
r0 = r2 * r1 + r0; // s13 mad
|
||||
r2 = __umulhi(r2, r4); // s14 mulhi
|
||||
r3 = r3 | r1; // s15 or
|
||||
r0 = r0 - r5; // s16 sub
|
||||
r0 = r0 * r6; // s17 mul
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r1, 4); // s18 shfl
|
||||
r3 = r3 | r4; // s19 or
|
||||
r6 = r3 * r2 + r6; // s20 mad
|
||||
r0 = rotl_imm(r0, 11u); // s21 rotl
|
||||
r3 = rotl_imm(r3, 25u); // s22 rotl
|
||||
r6 = r6 ^ r4; // s23 xor
|
||||
r4 = r4 | r2; // s24 or
|
||||
r1 = r1 + r6 + ((((sel >> 7u) & 1u) != 0u) ? 0xfc2f6d17u : 0x53e74799u); // s25 add
|
||||
r7 = rotr_var(r7, r6); // s26 rotr
|
||||
r7 = r7 + r5 + ((((sel >> 27u) & 1u) != 0u) ? 0x787048dcu : 0xb8fae90eu); // s27 add
|
||||
r4 = r4 ^ r1; // s28 xor
|
||||
r6 = r6 ^ r1; // s29 xor
|
||||
r2 = __umulhi(r2, r0); // s30 mulhi
|
||||
r1 = rotr_var(r1, r4); // s31 rotr
|
||||
r7 = r7 ^ r4; // s32 xor
|
||||
r3 = r2 * r6 + r3; // s33 mad
|
||||
r3 = r3 + r1 + ((((sel >> 12u) & 1u) != 0u) ? 0x9ffd510bu : 0xd2dc42ebu); // s34 add
|
||||
r4 = r4 + r7 + ((((sel >> 7u) & 1u) != 0u) ? 0xf871ba37u : 0xad8eda6fu); // s35 add
|
||||
r7 = r7 + r4 + ((((sel >> 3u) & 1u) != 0u) ? 0xd02d30cau : 0xb0cef8f4u); // s36 add
|
||||
r4 = r3 * r2 + r4; // s37 mad
|
||||
r5 = r7 * r7 + r5; // s38 mad
|
||||
r3 = r3 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x82e98a22u : 0x2bd506e2u); // s39 add
|
||||
r0 = rotl_imm(r0, 21u); // s40 rotl
|
||||
r1 = r1 | r0; // s41 or
|
||||
r7 = r7 + r5 + ((((sel >> 28u) & 1u) != 0u) ? 0x6435f6d8u : 0xc0386e49u); // s42 add
|
||||
r1 = __umulhi(r1, r0); // s43 mulhi
|
||||
r5 = r5 + r6 + ((((sel >> 12u) & 1u) != 0u) ? 0x1f682c68u : 0x3a8921a4u); // s44 add
|
||||
r4 = r4 + r5 + ((((sel >> 1u) & 1u) != 0u) ? 0xf246e180u : 0xeeca4334u); // s45 add
|
||||
r2 = r2 ^ r3; // s46 xor
|
||||
r2 = r2 ^ r5; // s47 xor
|
||||
r7 = r4 * r1 + r7; // s48 mad
|
||||
r2 = r6 * r1 + r2; // s49 mad
|
||||
r7 = r0 * r7 + r7; // s50 mad
|
||||
r5 = r5 + r6 + ((((sel >> 26u) & 1u) != 0u) ? 0x41bd68ccu : 0x745776d8u); // s51 add
|
||||
r3 = r3 ^ __shfl_xor_sync(0xffffffffu, r0, 1); // s52 shfl
|
||||
r4 = r4 | r0; // s53 or
|
||||
r0 = r2 * r4 + r0; // s54 mad
|
||||
r5 = rotl_imm(r5, 9u); // s55 rotl
|
||||
r5 = r5 * r7; // s56 mul
|
||||
r7 = rotl_imm(r7, 27u); // s57 rotl
|
||||
r7 = r7 - r6; // s58 sub
|
||||
r2 = rotl_imm(r2, 22u); // s59 rotl
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r1, 2); // s60 shfl
|
||||
r4 = r4 + r0 + ((((sel >> 7u) & 1u) != 0u) ? 0xcba6a161u : 0x60d0ff55u); // s61 add
|
||||
r1 = r1 * r5; // s62 mul
|
||||
r7 = __umulhi(r7, r6); // s63 mulhi
|
||||
r5 = r5 * r0; // s64 mul
|
||||
r2 = rotr_var(r2, r5); // s65 rotr
|
||||
r0 = __umulhi(r0, r3); // s66 mulhi
|
||||
r0 = r0 | r7; // s67 or
|
||||
r0 = rotl_imm(r0, 19u); // s68 rotl
|
||||
r0 = rotl_imm(r0, 1u); // s69 rotl
|
||||
r2 = r2 | r3; // s70 or
|
||||
r3 = r7 * r5 + r3; // s71 mad
|
||||
r7 = r0 * r4 + r7; // s72 mad
|
||||
r4 = r4 | r3; // s73 or
|
||||
r3 = r3 ^ r1; // s74 xor
|
||||
r1 = r1 + r3 + ((((sel >> 15u) & 1u) != 0u) ? 0x8e35924eu : 0x3f4ffea2u); // s75 add
|
||||
r1 = r1 * r3; // s76 mul
|
||||
r0 = r3 * r2 + r0; // s77 mad
|
||||
r2 = r2 ^ __shfl_xor_sync(0xffffffffu, r0, 16); // s78 shfl
|
||||
r1 = __umulhi(r1, r0); // s79 mulhi
|
||||
r0 = __umulhi(r0, r4); // s80 mulhi
|
||||
r4 = r4 * r2; // s81 mul
|
||||
r5 = r5 - r7; // s82 sub
|
||||
r0 = r2 * r3 + r0; // s83 mad
|
||||
r7 = r7 - r2; // s84 sub
|
||||
r1 = r1 ^ r2; // s85 xor
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r3, 16); // s86 shfl
|
||||
r3 = r0 * r0 + r3; // s87 mad
|
||||
r0 = rotl_imm(r0, 20u); // s88 rotl
|
||||
r1 = r1 - r5; // s89 sub
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r4, 16); // s90 shfl
|
||||
r5 = r5 ^ r1; // s91 xor
|
||||
r5 = rotl_imm(r5, 12u); // s92 rotl
|
||||
r6 = r1 * r2 + r6; // s93 mad
|
||||
r2 = rotr_var(r2, r5); // s94 rotr
|
||||
r0 = r0 | r4; // s95 or
|
||||
r7 = r7 * r3; // s96 mul
|
||||
r5 = r5 | r0; // s97 or
|
||||
r3 = r3 + r1 + ((((sel >> 13u) & 1u) != 0u) ? 0x386bb642u : 0xbcfdd8a7u); // s98 add
|
||||
r0 = r0 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xa3962d03u : 0xc921a021u); // s99 add
|
||||
r1 = rotr_var(r1, r0); // s100 rotr
|
||||
r2 = r2 + r4 + ((((sel >> 18u) & 1u) != 0u) ? 0xc5990c5cu : 0x70da4067u); // s101 add
|
||||
r2 = rotl_imm(r2, 3u); // s102 rotl
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r4, 4); // s103 shfl
|
||||
r0 = r0 ^ r2; // s104 xor
|
||||
r4 = r4 + r3 + ((((sel >> 28u) & 1u) != 0u) ? 0xb15aec75u : 0x1a5880cbu); // s105 add
|
||||
r4 = r4 ^ __shfl_xor_sync(0xffffffffu, r0, 2); // s106 shfl
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r0, 4); // s107 shfl
|
||||
r1 = r7 * r3 + r1; // s108 mad
|
||||
r0 = r2 * r3 + r0; // s109 mad
|
||||
r2 = r2 | r4; // s110 or
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r1, 1); // s111 shfl
|
||||
r2 = r5 * r3 + r2; // s112 mad
|
||||
r4 = __umulhi(r4, r0); // s113 mulhi
|
||||
r7 = rotl_imm(r7, 29u); // s114 rotl
|
||||
r3 = r3 - r7; // s115 sub
|
||||
r0 = __umulhi(r0, r1); // s116 mulhi
|
||||
r4 = rotl_imm(r4, 12u); // s117 rotl
|
||||
r1 = r1 + r4 + ((((sel >> 7u) & 1u) != 0u) ? 0xc5e98ee5u : 0x24494ad3u); // s118 add
|
||||
r0 = r0 + r4 + ((((sel >> 4u) & 1u) != 0u) ? 0xa97bc949u : 0x1b2d1c83u); // s119 add
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r4, 4); // s120 shfl
|
||||
r3 = r3 ^ __shfl_xor_sync(0xffffffffu, r2, 8); // s121 shfl
|
||||
r7 = rotl_imm(r7, 3u); // s122 rotl
|
||||
r4 = rotr_var(r4, r6); // s123 rotr
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r5, 1); // s124 shfl
|
||||
r5 = rotl_imm(r5, 21u); // s125 rotl
|
||||
r0 = r3 * r1 + r0; // s126 mad
|
||||
r0 = r0 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0xa7f6bb54u : 0xff14c08du); // s127 add
|
||||
r7 = r7 ^ r4; // s128 xor
|
||||
r3 = __umulhi(r3, r5); // s129 mulhi
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r7, 8); // s130 shfl
|
||||
r6 = __umulhi(r6, r4); // s131 mulhi
|
||||
r5 = r5 ^ r3; // s132 xor
|
||||
r0 = rotl_imm(r0, 19u); // s133 rotl
|
||||
r4 = rotr_var(r4, r3); // s134 rotr
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r2, 16); // s135 shfl
|
||||
r5 = r5 - r2; // s136 sub
|
||||
r3 = r3 ^ r5; // s137 xor
|
||||
r0 = r0 - r2; // s138 sub
|
||||
r3 = r3 * r7; // s139 mul
|
||||
r5 = r5 + r3 + ((((sel >> 24u) & 1u) != 0u) ? 0x2d465ca8u : 0x82ab98f4u); // s140 add
|
||||
r4 = r4 + r1 + ((((sel >> 11u) & 1u) != 0u) ? 0x237a9d8eu : 0x80594390u); // s141 add
|
||||
r3 = r7 * r0 + r3; // s142 mad
|
||||
r0 = r0 | r3; // s143 or
|
||||
r0 = __umulhi(r0, r1); // s144 mulhi
|
||||
r6 = r6 ^ r2; // s145 xor
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r1, 16); // s146 shfl
|
||||
r0 = r0 + r1 + ((((sel >> 18u) & 1u) != 0u) ? 0xe68cf57eu : 0x07f15148u); // s147 add
|
||||
r7 = r7 | r3; // s148 or
|
||||
r4 = r4 ^ r2; // s149 xor
|
||||
r7 = r7 ^ r1; // s150 xor
|
||||
r5 = rotr_var(r5, r2); // s151 rotr
|
||||
r0 = r0 * r7; // s152 mul
|
||||
r4 = rotl_imm(r4, 16u); // s153 rotl
|
||||
r6 = rotr_var(r6, r4); // s154 rotr
|
||||
r6 = r6 ^ r1; // s155 xor
|
||||
r6 = r6 | r3; // s156 or
|
||||
r1 = r1 + r6 + ((((sel >> 19u) & 1u) != 0u) ? 0x84c2a09du : 0x14878c5au); // s157 add
|
||||
r4 = r4 - r1; // s158 sub
|
||||
r4 = r6 * r7 + r4; // s159 mad
|
||||
r1 = r1 * r5; // s160 mul
|
||||
r4 = r4 | r0; // s161 or
|
||||
r7 = rotl_imm(r7, 21u); // s162 rotl
|
||||
r0 = r0 - r1; // s163 sub
|
||||
r1 = r1 * r0; // s164 mul
|
||||
r2 = r2 - r1; // s165 sub
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r6, 2); // s166 shfl
|
||||
r0 = r0 - r4; // s167 sub
|
||||
r1 = rotr_var(r1, r6); // s168 rotr
|
||||
r7 = rotr_var(r7, r1); // s169 rotr
|
||||
r3 = r1 * r3 + r3; // s170 mad
|
||||
r2 = r2 ^ __shfl_xor_sync(0xffffffffu, r0, 4); // s171 shfl
|
||||
r2 = r2 - r1; // s172 sub
|
||||
r7 = r7 ^ r6; // s173 xor
|
||||
r4 = rotr_var(r4, r1); // s174 rotr
|
||||
r4 = r4 ^ __shfl_xor_sync(0xffffffffu, r7, 2); // s175 shfl
|
||||
r6 = r6 * r2; // s176 mul
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r2, 4); // s177 shfl
|
||||
r6 = r6 + r4 + ((((sel >> 21u) & 1u) != 0u) ? 0x509871e6u : 0x4fa43965u); // s178 add
|
||||
r0 = r0 ^ r6; // s179 xor
|
||||
r5 = r5 + r7 + ((((sel >> 0u) & 1u) != 0u) ? 0xb95b8a53u : 0xfb3c4bd8u); // s180 add
|
||||
r6 = r6 * r3; // s181 mul
|
||||
r7 = rotl_imm(r7, 1u); // s182 rotl
|
||||
r0 = r0 + r7 + ((((sel >> 11u) & 1u) != 0u) ? 0x7047b7cfu : 0xc448a197u); // s183 add
|
||||
r3 = r3 * r5; // s184 mul
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r7, 1); // s185 shfl
|
||||
r0 = r0 * r2; // s186 mul
|
||||
r4 = __umulhi(r4, r7); // s187 mulhi
|
||||
r2 = r7 * r2 + r2; // s188 mad
|
||||
r7 = r7 - r2; // s189 sub
|
||||
r2 = r2 + r5 + ((((sel >> 19u) & 1u) != 0u) ? 0x5caccc5du : 0x2fceaf49u); // s190 add
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r2, 1); // s191 shfl
|
||||
r7 = r7 + r1 + ((((sel >> 20u) & 1u) != 0u) ? 0x857bb5feu : 0xba0cee62u); // s192 add
|
||||
r6 = rotl_imm(r6, 22u); // s193 rotl
|
||||
r3 = r3 ^ r7; // s194 xor
|
||||
r7 = r7 * r0; // s195 mul
|
||||
r3 = r3 ^ r5; // s196 xor
|
||||
r5 = r5 + r1 + ((((sel >> 29u) & 1u) != 0u) ? 0xeae84577u : 0x8519428cu); // s197 add
|
||||
r7 = r4 * r4 + r7; // s198 mad
|
||||
r3 = r3 + r6 + ((((sel >> 8u) & 1u) != 0u) ? 0x0bfdbfa1u : 0x468639d3u); // s199 add
|
||||
r5 = r3 * r4 + r5; // s200 mad
|
||||
r2 = r2 ^ __shfl_xor_sync(0xffffffffu, r7, 4); // s201 shfl
|
||||
r7 = r7 + r4 + ((((sel >> 23u) & 1u) != 0u) ? 0x53ee8e50u : 0x18ec9a69u); // s202 add
|
||||
r6 = r6 ^ r0; // s203 xor
|
||||
r4 = r4 + r5 + ((((sel >> 4u) & 1u) != 0u) ? 0xac63376eu : 0x3e81485bu); // s204 add
|
||||
r7 = r1 * r4 + r7; // s205 mad
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r5, 16); // s206 shfl
|
||||
r4 = r4 * r0; // s207 mul
|
||||
r1 = r1 | r6; // s208 or
|
||||
r6 = r6 + r4 + ((((sel >> 28u) & 1u) != 0u) ? 0x5b745519u : 0x66ccb75du); // s209 add
|
||||
r1 = rotr_var(r1, r3); // s210 rotr
|
||||
r5 = r5 - r4; // s211 sub
|
||||
r4 = r4 ^ r7; // s212 xor
|
||||
r1 = r1 + r7 + ((((sel >> 8u) & 1u) != 0u) ? 0x1c3dccf7u : 0x0ce0553du); // s213 add
|
||||
r0 = __umulhi(r0, r1); // s214 mulhi
|
||||
r3 = r3 + r6 + ((((sel >> 9u) & 1u) != 0u) ? 0xfadc9205u : 0x96bfec89u); // s215 add
|
||||
r0 = rotl_imm(r0, 4u); // s216 rotl
|
||||
r6 = r6 ^ r4; // s217 xor
|
||||
r6 = r6 ^ __shfl_xor_sync(0xffffffffu, r7, 2); // s218 shfl
|
||||
r7 = r0 * r1 + r7; // s219 mad
|
||||
r4 = r4 * r0; // s220 mul
|
||||
r2 = r6 * r0 + r2; // s221 mad
|
||||
r5 = r5 ^ r4; // s222 xor
|
||||
r0 = r0 * r5; // s223 mul
|
||||
r2 = r2 + r0 + ((((sel >> 9u) & 1u) != 0u) ? 0x66b15c04u : 0x61495681u); // s224 add
|
||||
r3 = rotr_var(r3, r2); // s225 rotr
|
||||
r2 = r3 * r3 + r2; // s226 mad
|
||||
r6 = r6 ^ r0; // s227 xor
|
||||
r4 = rotl_imm(r4, 30u); // s228 rotl
|
||||
r2 = r2 + r3 + ((((sel >> 29u) & 1u) != 0u) ? 0x25bc17c2u : 0x70d05c34u); // s229 add
|
||||
r1 = rotr_var(r1, r5); // s230 rotr
|
||||
r1 = r1 + r4 + ((((sel >> 8u) & 1u) != 0u) ? 0x2d8728f3u : 0x32a28384u); // s231 add
|
||||
r1 = rotl_imm(r1, 6u); // s232 rotl
|
||||
r2 = r2 + r0 + ((((sel >> 21u) & 1u) != 0u) ? 0x61e81fe6u : 0xcf0949f1u); // s233 add
|
||||
r4 = rotl_imm(r4, 2u); // s234 rotl
|
||||
r2 = r2 - r6; // s235 sub
|
||||
r6 = r6 | r3; // s236 or
|
||||
r7 = r7 ^ __shfl_xor_sync(0xffffffffu, r5, 4); // s237 shfl
|
||||
r1 = r1 ^ r3; // s238 xor
|
||||
r2 = rotl_imm(r2, 1u); // s239 rotl
|
||||
r5 = rotr_var(r5, r2); // s240 rotr
|
||||
r4 = rotl_imm(r4, 10u); // s241 rotl
|
||||
r6 = r6 + r2 + ((((sel >> 16u) & 1u) != 0u) ? 0x0a67565du : 0xe204fb50u); // s242 add
|
||||
r1 = r1 ^ __shfl_xor_sync(0xffffffffu, r3, 2); // s243 shfl
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r1, 16); // s244 shfl
|
||||
r5 = r5 ^ __shfl_xor_sync(0xffffffffu, r7, 2); // s245 shfl
|
||||
r4 = r4 ^ r5; // s246 xor
|
||||
r7 = r7 ^ r2; // s247 xor
|
||||
r0 = r0 ^ __shfl_xor_sync(0xffffffffu, r6, 16); // s248 shfl
|
||||
r1 = __umulhi(r1, r3); // s249 mulhi
|
||||
r7 = r7 + r5 + ((((sel >> 31u) & 1u) != 0u) ? 0x5a79fd6fu : 0x3d6daffdu); // s250 add
|
||||
r6 = r6 + r0 + ((((sel >> 28u) & 1u) != 0u) ? 0x6beb28c2u : 0x84334aeau); // s251 add
|
||||
r7 = r7 - r1; // s252 sub
|
||||
r5 = rotr_var(r5, r2); // s253 rotr
|
||||
r2 = __umulhi(r2, r5); // s254 mulhi
|
||||
r6 = r6 * r3; // s255 mul
|
||||
}
|
||||
}
|
||||
uint32_t lo = r0 ^ rotl_imm(r1, 7u) ^ rotl_imm(r2, 14u) ^ rotl_imm(r3, 21u);
|
||||
uint32_t hi = r4 ^ rotl_imm(r5, 9u) ^ rotl_imm(r6, 18u) ^ rotl_imm(r7, 27u);
|
||||
out[gid] = ((uint64_t)hi << 32) | (uint64_t)lo;
|
||||
}
|
||||
|
||||
cudaError_t igneum_launch_hash_bound(const uint32_t* ds, uint64_t* out, uint32_t baseNonce, uint32_t mask,
|
||||
IgneumInitWords iw, uint32_t nonces, uint32_t blockWarps) {
|
||||
if (blockWarps == 0u || blockWarps > 32u) return cudaErrorInvalidValue;
|
||||
uint32_t block = 32u * blockWarps;
|
||||
if (nonces == 0u || (nonces % block) != 0u) return cudaErrorInvalidValue;
|
||||
igneum_hash_bound<<<nonces / block, block>>>(ds, out, baseNonce, mask, iw);
|
||||
return cudaGetLastError();
|
||||
}
|
||||
|
||||
cudaError_t igneum_hash_bound_info(int* numRegs, int* blocksPerSM, uint32_t blockWarps) {
|
||||
cudaFuncAttributes attr;
|
||||
cudaError_t e = cudaFuncGetAttributes(&attr, igneum_hash_bound);
|
||||
if (e != cudaSuccess) return e;
|
||||
*numRegs = attr.numRegs;
|
||||
return cudaOccupancyMaxActiveBlocksPerMultiprocessor(blocksPerSM, igneum_hash_bound, (int)(32u * blockWarps), 0);
|
||||
}
|
||||
BIN
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/leaves.bin
Normal file
BIN
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/leaves.bin
Normal file
Binary file not shown.
116
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/memhard.h
Normal file
116
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/memhard.h
Normal file
|
|
@ -0,0 +1,116 @@
|
|||
// Generated by igneum-pow export (generator v2) for seed "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000". Do not edit by hand.
|
||||
// Memory-hard dataset core, the same text that the Mac's Metal kernels and CPU verifier were checked against.
|
||||
// Included by kernel.cu (device), host.cu (host reference) and proto-opencl/host.c (C99 host reference).
|
||||
// See proto-metal/MEMHARD.md for the construction. kernel.cl carries the same text in OpenCL C.
|
||||
#pragma once
|
||||
#ifdef __cplusplus
|
||||
#include <cstdint>
|
||||
#else
|
||||
#include <stdint.h>
|
||||
#endif
|
||||
#if defined(__CUDACC__)
|
||||
#define IGNEUM_HD __host__ __device__ __forceinline__
|
||||
#elif defined(_MSC_VER) && !defined(__cplusplus)
|
||||
#define IGNEUM_HD static __inline
|
||||
#else
|
||||
#define IGNEUM_HD static inline
|
||||
#endif
|
||||
// Memory-hard dataset core (MEMHARD.md). Cache: 2^26 words in 2^16 segments of 64 chained ChaCha12 lines.
|
||||
// Item: 8 rounds of 8 x seed-parameterised mixer + one 64-byte cache read, then 8 x final mixer (class v3, mixer multiplier 8,
|
||||
// docs/plans/mixer-x4.md: the round key of application j of round r is 0x9E3779B9 * (r * m + j + 1)). All parameters are literals.
|
||||
#define MH_CACHE_LINE_MASK 0x003fffffu
|
||||
#define MH_SEGMENT_LINES 64u
|
||||
#define MH_QR(a, b, c, d, r1, r2, r3, r4) { a += b; d ^= a; d = mh_rotl(d, r1); c += d; b ^= c; b = mh_rotl(b, r2); a += b; d ^= a; d = mh_rotl(d, r3); c += d; b ^= c; b = mh_rotl(b, r4); }
|
||||
IGNEUM_HD uint32_t mh_rotl(uint32_t x, uint32_t n) { return (x << n) | (x >> (32u - n)); } // n in 1..31 at every call site
|
||||
|
||||
// y = ChaCha12 core(x) + x
|
||||
IGNEUM_HD void mh_chacha_block(const uint32_t* x, uint32_t* y) {
|
||||
for (uint32_t i = 0u; i < 16u; ++i) y[i] = x[i];
|
||||
for (uint32_t r = 0u; r < 6u; ++r) {
|
||||
MH_QR(y[0], y[4], y[8], y[12], 16u, 12u, 8u, 7u) MH_QR(y[1], y[5], y[9], y[13], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[6], y[10], y[14], 16u, 12u, 8u, 7u) MH_QR(y[3], y[7], y[11], y[15], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[0], y[5], y[10], y[15], 16u, 12u, 8u, 7u) MH_QR(y[1], y[6], y[11], y[12], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[7], y[8], y[13], 16u, 12u, 8u, 7u) MH_QR(y[3], y[4], y[9], y[14], 16u, 12u, 8u, 7u)
|
||||
}
|
||||
for (uint32_t i = 0u; i < 16u; ++i) y[i] += x[i];
|
||||
}
|
||||
|
||||
// One cache segment: 64 chained lines written at cache[seg * 1024]. in_j = prev ^ (sigma || K || seg || j || tag), prev_0 = 0.
|
||||
IGNEUM_HD void mh_cache_segment(uint32_t* cache, uint32_t seg) {
|
||||
uint32_t prev[16]; uint32_t x[16]; uint32_t y[16];
|
||||
for (uint32_t i = 0u; i < 16u; ++i) prev[i] = 0u;
|
||||
for (uint32_t j = 0u; j < MH_SEGMENT_LINES; ++j) {
|
||||
x[0] = 0x61707865u ^ prev[0]; x[1] = 0x3320646eu ^ prev[1]; x[2] = 0x79622d32u ^ prev[2]; x[3] = 0x6b206574u ^ prev[3];
|
||||
x[4] = 0xe23f9008u ^ prev[4];
|
||||
x[5] = 0xde33e763u ^ prev[5];
|
||||
x[6] = 0xc5ba415du ^ prev[6];
|
||||
x[7] = 0x8ddf6786u ^ prev[7];
|
||||
x[8] = 0x59f4a4beu ^ prev[8];
|
||||
x[9] = 0x8a3bc680u ^ prev[9];
|
||||
x[10] = 0x701b8e40u ^ prev[10];
|
||||
x[11] = 0x3b59025au ^ prev[11];
|
||||
x[12] = seg ^ prev[12]; x[13] = j ^ prev[13]; x[14] = 0x49676e65u ^ prev[14]; x[15] = 0x756d4d48u ^ prev[15];
|
||||
mh_chacha_block(x, y);
|
||||
uint32_t* line = cache + ((seg * MH_SEGMENT_LINES + j) * 16u);
|
||||
for (uint32_t i = 0u; i < 16u; ++i) { line[i] = y[i]; prev[i] = y[i]; }
|
||||
}
|
||||
}
|
||||
|
||||
// M_r: per word (s ^ (RC + rk)) * MUL, then a column round and a diagonal round with the seed-drawn rotations.
|
||||
IGNEUM_HD void mh_mixer(uint32_t* s, uint32_t rk) {
|
||||
s[0] = (s[0] ^ (0xe5bef5a3u + rk)) * 0xd1d79e7fu;
|
||||
s[1] = (s[1] ^ (0xdb2a7d90u + rk)) * 0xbcb61f8bu;
|
||||
s[2] = (s[2] ^ (0xcd1fa7e1u + rk)) * 0x970e91abu;
|
||||
s[3] = (s[3] ^ (0x30289419u + rk)) * 0xd749d96du;
|
||||
s[4] = (s[4] ^ (0x0a730d58u + rk)) * 0x143d7339u;
|
||||
s[5] = (s[5] ^ (0x432f8579u + rk)) * 0x2cde0d69u;
|
||||
s[6] = (s[6] ^ (0x6ab978a5u + rk)) * 0x12d0a8c1u;
|
||||
s[7] = (s[7] ^ (0x8c984f49u + rk)) * 0x2d335be9u;
|
||||
s[8] = (s[8] ^ (0x788c3c9eu + rk)) * 0x80a8aae9u;
|
||||
s[9] = (s[9] ^ (0x051eef02u + rk)) * 0x7ac896c7u;
|
||||
s[10] = (s[10] ^ (0x05e77db9u + rk)) * 0x9de23db7u;
|
||||
s[11] = (s[11] ^ (0x8b3cd4f3u + rk)) * 0xc362827du;
|
||||
s[12] = (s[12] ^ (0x6948d6cfu + rk)) * 0x5f4cdb5bu;
|
||||
s[13] = (s[13] ^ (0x0d8677f6u + rk)) * 0xfc6c5097u;
|
||||
s[14] = (s[14] ^ (0xf9505c8cu + rk)) * 0x6f547f83u;
|
||||
s[15] = (s[15] ^ (0x4513c163u + rk)) * 0x1ac31b47u;
|
||||
MH_QR(s[0], s[4], s[8], s[12], 28u, 11u, 18u, 18u) MH_QR(s[1], s[5], s[9], s[13], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[2], s[6], s[10], s[14], 28u, 11u, 18u, 18u) MH_QR(s[3], s[7], s[11], s[15], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[0], s[5], s[10], s[15], 13u, 28u, 5u, 13u) MH_QR(s[1], s[6], s[11], s[12], 13u, 28u, 5u, 13u)
|
||||
MH_QR(s[2], s[7], s[8], s[13], 13u, 28u, 5u, 13u) MH_QR(s[3], s[4], s[9], s[14], 13u, 28u, 5u, 13u)
|
||||
}
|
||||
|
||||
// Item t: 16 words. s = (K, t * MUL[i] + RC[i]); 8 rounds of 8 x mixer + cache line s[0] & mask; 8 x final mixer.
|
||||
IGNEUM_HD void mh_item(const uint32_t* cache, const uint32_t* leaf, uint32_t t, uint32_t* s) {
|
||||
s[0] = 0xe23f9008u;
|
||||
s[1] = 0xde33e763u;
|
||||
s[2] = 0xc5ba415du;
|
||||
s[3] = 0x8ddf6786u;
|
||||
s[4] = 0x59f4a4beu;
|
||||
s[5] = 0x8a3bc680u;
|
||||
s[6] = 0x701b8e40u;
|
||||
s[7] = 0x3b59025au;
|
||||
s[8] = t * 0xd1d79e7fu + 0xe5bef5a3u;
|
||||
s[9] = t * 0xbcb61f8bu + 0xdb2a7d90u;
|
||||
s[10] = t * 0x970e91abu + 0xcd1fa7e1u;
|
||||
s[11] = t * 0xd749d96du + 0x30289419u;
|
||||
s[12] = t * 0x143d7339u + 0x0a730d58u;
|
||||
s[13] = t * 0x2cde0d69u + 0x432f8579u;
|
||||
s[14] = t * 0x12d0a8c1u + 0x6ab978a5u;
|
||||
s[15] = t * 0x2d335be9u + 0x8c984f49u;
|
||||
for (uint32_t i = 0u; i < 16u; ++i) s[i] ^= leaf[i];
|
||||
for (uint32_t r = 0u; r < 8u; ++r) {
|
||||
for (uint32_t j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (r * 8u + j + 1u));
|
||||
const uint32_t* line = cache + ((s[0] & MH_CACHE_LINE_MASK) * 16u);
|
||||
for (uint32_t i = 0u; i < 16u; ++i) s[i] ^= line[i];
|
||||
}
|
||||
for (uint32_t j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (64u + j + 1u));
|
||||
}
|
||||
// Class v5 (docs/design/class-v5-stored-state.md): leaf(t) = leaves[t mod nLeaves], 16 words per leaf (leaves.bin).
|
||||
IGNEUM_HD const uint32_t* mh_leaf(const uint32_t* leaves, uint32_t nLeaves, uint32_t t) { return leaves + ((t % nLeaves) * 16u); }
|
||||
// Era layout (docs/plans/era-layout.md 1.2): dataset word w holds word mh_j(w) of item mh_t(w); j's bits sit at positions 1 9 13 14 of w.
|
||||
IGNEUM_HD uint32_t mh_j(uint32_t w) { return ((w >> 1u) & 1u) | (((w >> 9u) & 1u) << 1) | (((w >> 13u) & 1u) << 2) | (((w >> 14u) & 1u) << 3); }
|
||||
IGNEUM_HD uint32_t mh_t(uint32_t w) { w = (w & 0x00003fffu) | ((w >> 15u) << 14u); w = (w & 0x00001fffu) | ((w >> 14u) << 13u); w = (w & 0x000001ffu) | ((w >> 10u) << 9u); w = (w & 0x00000001u) | ((w >> 2u) << 1u); return w; }
|
||||
IGNEUM_HD uint32_t mh_addr(uint32_t t, uint32_t j) { uint32_t w = t; w = ((w >> 1u) << 2u) | (w & 0x00000001u) | (((j >> 0u) & 1u) << 1u); w = ((w >> 9u) << 10u) | (w & 0x000001ffu) | (((j >> 1u) & 1u) << 9u); w = ((w >> 13u) << 14u) | (w & 0x00001fffu) | (((j >> 2u) & 1u) << 13u); w = ((w >> 14u) << 15u) | (w & 0x00003fffu) | (((j >> 3u) & 1u) << 14u); return w; }
|
||||
// dataset[w] without the dataset: derive item mh_t(w) and take word mh_j(w).
|
||||
IGNEUM_HD uint32_t mh_word(const uint32_t* cache, const uint32_t* leaves, uint32_t nLeaves, uint32_t w) { uint32_t s[16]; mh_item(cache, mh_leaf(leaves, nLeaves, mh_t(w)), mh_t(w), s); return s[mh_j(w)]; }
|
||||
115
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/memhard.metal
Normal file
115
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/memhard.metal
Normal file
|
|
@ -0,0 +1,115 @@
|
|||
#include <metal_stdlib>
|
||||
using namespace metal;
|
||||
// Memory-hard dataset core (MEMHARD.md). Cache: 2^26 words in 2^16 segments of 64 chained ChaCha12 lines.
|
||||
// Item: 8 rounds of 8 x seed-parameterised mixer + one 64-byte cache read, then 8 x final mixer (class v3, mixer multiplier 8,
|
||||
// docs/plans/mixer-x4.md: the round key of application j of round r is 0x9E3779B9 * (r * m + j + 1)). All parameters are literals.
|
||||
#define MH_CACHE_LINE_MASK 0x003fffffu
|
||||
#define MH_SEGMENT_LINES 64u
|
||||
#define MH_QR(a, b, c, d, r1, r2, r3, r4) { a += b; d ^= a; d = mh_rotl(d, r1); c += d; b ^= c; b = mh_rotl(b, r2); a += b; d ^= a; d = mh_rotl(d, r3); c += d; b ^= c; b = mh_rotl(b, r4); }
|
||||
inline uint mh_rotl(uint x, uint n) { return (x << n) | (x >> (32u - n)); } // n in 1..31 at every call site
|
||||
|
||||
// y = ChaCha12 core(x) + x
|
||||
inline void mh_chacha_block(const thread uint* x, thread uint* y) {
|
||||
for (uint i = 0u; i < 16u; ++i) y[i] = x[i];
|
||||
for (uint r = 0u; r < 6u; ++r) {
|
||||
MH_QR(y[0], y[4], y[8], y[12], 16u, 12u, 8u, 7u) MH_QR(y[1], y[5], y[9], y[13], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[6], y[10], y[14], 16u, 12u, 8u, 7u) MH_QR(y[3], y[7], y[11], y[15], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[0], y[5], y[10], y[15], 16u, 12u, 8u, 7u) MH_QR(y[1], y[6], y[11], y[12], 16u, 12u, 8u, 7u)
|
||||
MH_QR(y[2], y[7], y[8], y[13], 16u, 12u, 8u, 7u) MH_QR(y[3], y[4], y[9], y[14], 16u, 12u, 8u, 7u)
|
||||
}
|
||||
for (uint i = 0u; i < 16u; ++i) y[i] += x[i];
|
||||
}
|
||||
|
||||
// One cache segment: 64 chained lines written at cache[seg * 1024]. in_j = prev ^ (sigma || K || seg || j || tag), prev_0 = 0.
|
||||
inline void mh_cache_segment(device uint* cache, uint seg) {
|
||||
uint prev[16]; uint x[16]; uint y[16];
|
||||
for (uint i = 0u; i < 16u; ++i) prev[i] = 0u;
|
||||
for (uint j = 0u; j < MH_SEGMENT_LINES; ++j) {
|
||||
x[0] = 0x61707865u ^ prev[0]; x[1] = 0x3320646eu ^ prev[1]; x[2] = 0x79622d32u ^ prev[2]; x[3] = 0x6b206574u ^ prev[3];
|
||||
x[4] = 0xe23f9008u ^ prev[4];
|
||||
x[5] = 0xde33e763u ^ prev[5];
|
||||
x[6] = 0xc5ba415du ^ prev[6];
|
||||
x[7] = 0x8ddf6786u ^ prev[7];
|
||||
x[8] = 0x59f4a4beu ^ prev[8];
|
||||
x[9] = 0x8a3bc680u ^ prev[9];
|
||||
x[10] = 0x701b8e40u ^ prev[10];
|
||||
x[11] = 0x3b59025au ^ prev[11];
|
||||
x[12] = seg ^ prev[12]; x[13] = j ^ prev[13]; x[14] = 0x49676e65u ^ prev[14]; x[15] = 0x756d4d48u ^ prev[15];
|
||||
mh_chacha_block(x, y);
|
||||
device uint* line = cache + ((seg * MH_SEGMENT_LINES + j) * 16u);
|
||||
for (uint i = 0u; i < 16u; ++i) { line[i] = y[i]; prev[i] = y[i]; }
|
||||
}
|
||||
}
|
||||
|
||||
// M_r: per word (s ^ (RC + rk)) * MUL, then a column round and a diagonal round with the seed-drawn rotations.
|
||||
inline void mh_mixer(thread uint* s, uint rk) {
|
||||
s[0] = (s[0] ^ (0xe5bef5a3u + rk)) * 0xd1d79e7fu;
|
||||
s[1] = (s[1] ^ (0xdb2a7d90u + rk)) * 0xbcb61f8bu;
|
||||
s[2] = (s[2] ^ (0xcd1fa7e1u + rk)) * 0x970e91abu;
|
||||
s[3] = (s[3] ^ (0x30289419u + rk)) * 0xd749d96du;
|
||||
s[4] = (s[4] ^ (0x0a730d58u + rk)) * 0x143d7339u;
|
||||
s[5] = (s[5] ^ (0x432f8579u + rk)) * 0x2cde0d69u;
|
||||
s[6] = (s[6] ^ (0x6ab978a5u + rk)) * 0x12d0a8c1u;
|
||||
s[7] = (s[7] ^ (0x8c984f49u + rk)) * 0x2d335be9u;
|
||||
s[8] = (s[8] ^ (0x788c3c9eu + rk)) * 0x80a8aae9u;
|
||||
s[9] = (s[9] ^ (0x051eef02u + rk)) * 0x7ac896c7u;
|
||||
s[10] = (s[10] ^ (0x05e77db9u + rk)) * 0x9de23db7u;
|
||||
s[11] = (s[11] ^ (0x8b3cd4f3u + rk)) * 0xc362827du;
|
||||
s[12] = (s[12] ^ (0x6948d6cfu + rk)) * 0x5f4cdb5bu;
|
||||
s[13] = (s[13] ^ (0x0d8677f6u + rk)) * 0xfc6c5097u;
|
||||
s[14] = (s[14] ^ (0xf9505c8cu + rk)) * 0x6f547f83u;
|
||||
s[15] = (s[15] ^ (0x4513c163u + rk)) * 0x1ac31b47u;
|
||||
MH_QR(s[0], s[4], s[8], s[12], 28u, 11u, 18u, 18u) MH_QR(s[1], s[5], s[9], s[13], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[2], s[6], s[10], s[14], 28u, 11u, 18u, 18u) MH_QR(s[3], s[7], s[11], s[15], 28u, 11u, 18u, 18u)
|
||||
MH_QR(s[0], s[5], s[10], s[15], 13u, 28u, 5u, 13u) MH_QR(s[1], s[6], s[11], s[12], 13u, 28u, 5u, 13u)
|
||||
MH_QR(s[2], s[7], s[8], s[13], 13u, 28u, 5u, 13u) MH_QR(s[3], s[4], s[9], s[14], 13u, 28u, 5u, 13u)
|
||||
}
|
||||
|
||||
// Item t: 16 words. s = (K, t * MUL[i] + RC[i]); 8 rounds of 8 x mixer + cache line s[0] & mask; 8 x final mixer.
|
||||
inline void mh_item(device const uint* cache, device const uint* leaf, uint t, thread uint* s) {
|
||||
s[0] = 0xe23f9008u;
|
||||
s[1] = 0xde33e763u;
|
||||
s[2] = 0xc5ba415du;
|
||||
s[3] = 0x8ddf6786u;
|
||||
s[4] = 0x59f4a4beu;
|
||||
s[5] = 0x8a3bc680u;
|
||||
s[6] = 0x701b8e40u;
|
||||
s[7] = 0x3b59025au;
|
||||
s[8] = t * 0xd1d79e7fu + 0xe5bef5a3u;
|
||||
s[9] = t * 0xbcb61f8bu + 0xdb2a7d90u;
|
||||
s[10] = t * 0x970e91abu + 0xcd1fa7e1u;
|
||||
s[11] = t * 0xd749d96du + 0x30289419u;
|
||||
s[12] = t * 0x143d7339u + 0x0a730d58u;
|
||||
s[13] = t * 0x2cde0d69u + 0x432f8579u;
|
||||
s[14] = t * 0x12d0a8c1u + 0x6ab978a5u;
|
||||
s[15] = t * 0x2d335be9u + 0x8c984f49u;
|
||||
for (uint i = 0u; i < 16u; ++i) s[i] ^= leaf[i];
|
||||
for (uint r = 0u; r < 8u; ++r) {
|
||||
for (uint j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (r * 8u + j + 1u));
|
||||
device const uint* line = cache + ((s[0] & MH_CACHE_LINE_MASK) * 16u);
|
||||
for (uint i = 0u; i < 16u; ++i) s[i] ^= line[i];
|
||||
}
|
||||
for (uint j = 0u; j < 8u; ++j) mh_mixer(s, 0x9E3779B9u * (64u + j + 1u));
|
||||
}
|
||||
// Class v5 (docs/design/class-v5-stored-state.md): leaf(t) = leaves[t mod nLeaves], 16 words per leaf (leaves.bin).
|
||||
inline device const uint* mh_leaf(device const uint* leaves, uint nLeaves, uint t) { return leaves + ((t % nLeaves) * 16u); }
|
||||
// Era layout (docs/plans/era-layout.md 1.2): dataset word w holds word mh_j(w) of item mh_t(w); j's bits sit at positions 1 9 13 14 of w.
|
||||
inline uint mh_j(uint w) { return ((w >> 1u) & 1u) | (((w >> 9u) & 1u) << 1) | (((w >> 13u) & 1u) << 2) | (((w >> 14u) & 1u) << 3); }
|
||||
inline uint mh_t(uint w) { w = (w & 0x00003fffu) | ((w >> 15u) << 14u); w = (w & 0x00001fffu) | ((w >> 14u) << 13u); w = (w & 0x000001ffu) | ((w >> 10u) << 9u); w = (w & 0x00000001u) | ((w >> 2u) << 1u); return w; }
|
||||
inline uint mh_addr(uint t, uint j) { uint w = t; w = ((w >> 1u) << 2u) | (w & 0x00000001u) | (((j >> 0u) & 1u) << 1u); w = ((w >> 9u) << 10u) | (w & 0x000001ffu) | (((j >> 1u) & 1u) << 9u); w = ((w >> 13u) << 14u) | (w & 0x00001fffu) | (((j >> 2u) & 1u) << 13u); w = ((w >> 14u) << 15u) | (w & 0x00003fffu) | (((j >> 3u) & 1u) << 14u); return w; }
|
||||
// dataset[w] without the dataset: derive item mh_t(w) and take word mh_j(w).
|
||||
inline uint mh_word(device const uint* cache, device const uint* leaves, uint nLeaves, uint w) { uint s[16]; mh_item(cache, mh_leaf(leaves, nLeaves, mh_t(w)), mh_t(w), s); return s[mh_j(w)]; }
|
||||
|
||||
// One thread per segment (2^16 threads).
|
||||
kernel void igneum_cache_fill(device uint* cache [[buffer(0)]], uint gid [[thread_position_in_grid]]) {
|
||||
mh_cache_segment(cache, gid);
|
||||
}
|
||||
// One thread per 64-byte item (dataset words / 16 threads).
|
||||
// Class v5: the window's leaves (leaves.bin, IGNEUM_STATE_LEAVES x 16 words) in buffer 2, their count in buffer 3.
|
||||
kernel void igneum_build(device const uint* cache [[buffer(0)]], device uint* dataset [[buffer(1)]],
|
||||
device const uint* leaves [[buffer(2)]], constant uint& nLeaves [[buffer(3)]],
|
||||
uint gid [[thread_position_in_grid]]) {
|
||||
uint s[16];
|
||||
mh_item(cache, mh_leaf(leaves, nLeaves, gid), gid, s);
|
||||
for (uint i = 0u; i < 16u; ++i) dataset[mh_addr(gid, i)] = s[i];
|
||||
}
|
||||
97
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program.h
Normal file
97
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program.h
Normal file
|
|
@ -0,0 +1,97 @@
|
|||
// Generated by igneum-pow export (generator v2) for seed "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000". Do not edit by hand.
|
||||
// Program metadata for host.cu plus the launch wrappers defined in kernel.cu.
|
||||
// Also included by proto-opencl/host.c (C99), which defines IGNEUM_NO_CUDA first and reads only the macros.
|
||||
#pragma once
|
||||
#ifdef __cplusplus
|
||||
#include <cstdint>
|
||||
#else
|
||||
#include <stdint.h>
|
||||
#endif
|
||||
#ifndef IGNEUM_NO_CUDA
|
||||
#include <cuda_runtime.h>
|
||||
#endif
|
||||
|
||||
#define IGNEUM_SEED_STRING "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000"
|
||||
#define IGNEUM_SEED_BYTES_HEX "4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925"
|
||||
#define IGNEUM_GENERATOR 5
|
||||
#define IGNEUM_PROGRAM_ATTEMPT 0
|
||||
#define IGNEUM_PROGRAM_ID 0xe5a4ac5978462156ull
|
||||
#define IGNEUM_DAY_STRING "bytes:69676e65756d2d6461792ffd50000000000000"
|
||||
#define IGNEUM_DAY_BYTES_HEX "69676e65756d2d6461792ffd50000000000000"
|
||||
#define IGNEUM_DAY0 0xe23f9008u
|
||||
#define IGNEUM_DAY1 0xde33e763u
|
||||
#define IGNEUM_DATASET_LOG2 28
|
||||
#define IGNEUM_MASK 0x0fffffffu
|
||||
#define IGNEUM_LANES 32
|
||||
#define IGNEUM_ITERATIONS 8
|
||||
#define IGNEUM_INSTR_COUNT 64
|
||||
#define IGNEUM_LOADS_PER_HASH 128
|
||||
#define IGNEUM_WIDE_LOADS_PER_HASH 0
|
||||
#define IGNEUM_OP_MIX "load=16 add=11 shfl=7 mad=6 xor=6 mul=4 mulhi=4 rotr=4 rotl=3 sub=2 or=1"
|
||||
// Program class v5 (proof of stored state and of following, docs/design/class-v5-stored-state.md): generator version 5,
|
||||
// class v4 over a dataset whose every item is keyed by the window's execution state (IGNEUM_STATE_* below, leaves.bin);
|
||||
// a worker that runs another class refuses this pack, and a job line names the class it wants (class=v5 era=<hex>).
|
||||
#define IGNEUM_PROGRAM_CLASS "v5"
|
||||
#define IGNEUM_ERA_SEED_HEX "4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925"
|
||||
// Class v3 construction (Counter ASIC 2.0, 5 October 2026, docs/plans/mixer-x4.md): version 2 loads; the dataset item
|
||||
// derivation applies the mixer IGNEUM_MIXER_MULT times per round (memhard.h), and the cache follows the growth rule.
|
||||
#define IGNEUM_LOAD_CLASS "mx8-eraaf3a9139+sh256x27+state"
|
||||
#define IGNEUM_CLASS_MIXER_MULT 8
|
||||
#define IGNEUM_CACHE_GROWTH 1 // 1: cache words = 2^(26 + doublings(day)), doublings = floor(log2(1 + day / 1460))
|
||||
#define IGNEUM_LOAD_SLOTS 16
|
||||
#define IGNEUM_LOAD_MIX { 100, 0, 0 }
|
||||
#define IGNEUM_LOAD_WIDTH_COUNTS { 16, 0, 0 } // loads of 4, 16, 64 bytes per program
|
||||
#define IGNEUM_BYTES_PER_HASH 512
|
||||
#define IGNEUM_FOLD_ROT 11
|
||||
#define IGNEUM_FOLD_MUL 0x9e3779b1u
|
||||
// Class v5 state (docs/design/class-v5-stored-state.md): the window's reference chain block and the state root after it;
|
||||
// leaves.bin holds IGNEUM_STATE_LEAVES leaves of 16 little-endian words, leaf(t) = leaves[t mod IGNEUM_STATE_LEAVES].
|
||||
#define IGNEUM_STATE_BLOCK_HEX "4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925"
|
||||
#define IGNEUM_STATE_BLOCK_NUMBER 0
|
||||
#define IGNEUM_STATE_ROOT_HEX "7e37a9fb19b154d32daf5bf30a50d339a75029fbc9eec9ea20e95439dba5a311"
|
||||
#define IGNEUM_STATE_LEAVES 11
|
||||
#define IGNEUM_STATE_RECORDS 11
|
||||
#define IGNEUM_STATE_SAMPLED 0
|
||||
#define IGNEUM_STATE_LEAVES_FNV64 0xf141bfee2a8b11b0ull
|
||||
#define IGNEUM_STATE_LEAVES_FILE "leaves.bin"
|
||||
// Era layout (5 October 2026, docs/plans/era-layout.md): NOT the lottery hash. Every dataset load reads
|
||||
// idx = ((rotl(src * STRIDE_MUL, STRIDE_ROT) & window mask) | window offset) & MASK; the window of a load site is the
|
||||
// dataset, a half or a quarter of it (IGNEUM_ERA_WINDOWS: site:shrink:offset); dataset word w holds word j(w) of item
|
||||
// t(w) with j's bits at the INTERLEAVE positions (memhard.h: mh_t, mh_j, mh_addr).
|
||||
#define IGNEUM_ERA_LABEL "af3a9139"
|
||||
#define IGNEUM_ERA_SEED_WORDS { 0xaf3a9139u, 0x894db311u, 0xd6173573u, 0xc346bc0fu, 0x2d547cecu, 0xd5634ab8u, 0xb9584827u, 0x305d030eu }
|
||||
#define IGNEUM_ERA_ALLOWED_WIDTHS { 1, 0, 0 } // words, ascending, 0 = unused; one entry pins the width
|
||||
#define IGNEUM_ERA_WIDTH_WORDS 1
|
||||
#define IGNEUM_ERA_STRIDE_MUL 0x2cb18b35u
|
||||
#define IGNEUM_ERA_STRIDE_ROT 9
|
||||
#define IGNEUM_ERA_INTERLEAVE { 1, 9, 13, 14 }
|
||||
#define IGNEUM_ERA_WINDOWS "3:0:0 8:2:2 14:1:0 15:0:0 20:1:1 26:1:1 28:1:1 35:0:0 40:2:3 43:1:0 47:1:1 49:0:0 52:2:3 53:0:0 61:1:1 62:1:1"
|
||||
// Latency-shadow block (Counter ASIC 3.0 item 8, docs/analysis/latency-shadow-2026-10-06.md): NOT the lottery hash. A block of
|
||||
// IGNEUM_SHADOW_INSTRS ALU instructions (the ten non-load families) runs IGNEUM_SHADOW_REPS times at the end of every
|
||||
// iteration; the 16 loads, the acceptance rule and the base program are the class's without the shadow.
|
||||
#define IGNEUM_SHADOW_INSTRS 256
|
||||
#define IGNEUM_SHADOW_REPS 27
|
||||
#define IGNEUM_SHADOW_INSTRS_PER_HASH 55296
|
||||
#define IGNEUM_SHADOW_OP_MIX "add=45 mad=31 shfl=31 xor=30 rotl=28 mul=22 or=18 mulhi=17 rotr=17 sub=17"
|
||||
// 0 = closed-form dataset (ds_elem), 1 = memory-hard cache construction (MEMHARD.md, memhard.h)
|
||||
#define IGNEUM_DATASET_MODE 1
|
||||
|
||||
#define IGNEUM_SEEDW_INIT { 0xbe8c5a0cu, 0x7646e626u, 0x508e29d0u, 0xfb8a5b4au, 0x53c50955u, 0x685d62fcu, 0x6065c013u, 0x881096ebu }
|
||||
#define IGNEUM_KEY_INIT { 0xe23f9008u, 0xde33e763u, 0xc5ba415du, 0x8ddf6786u, 0x59f4a4beu, 0x8a3bc680u, 0x701b8e40u, 0x3b59025au }
|
||||
#define IGNEUM_CACHE_LOG2_WORDS 26
|
||||
#define IGNEUM_CACHE_SEGMENT_LOG2_LINES 6
|
||||
#define IGNEUM_CACHE_SEGMENTS 65536u
|
||||
#define IGNEUM_ITEM_ROUNDS 8
|
||||
#define IGNEUM_MIXER_MULT 8 // mixer applications per round and after the last read (class v3, docs/plans/mixer-x4.md)
|
||||
#define IGNEUM_MIX_ROT_INIT { 28u, 11u, 18u, 18u, 13u, 28u, 5u, 13u }
|
||||
#define IGNEUM_MIX_MUL_INIT { 0xd1d79e7fu, 0xbcb61f8bu, 0x970e91abu, 0xd749d96du, 0x143d7339u, 0x2cde0d69u, 0x12d0a8c1u, 0x2d335be9u, 0x80a8aae9u, 0x7ac896c7u, 0x9de23db7u, 0xc362827du, 0x5f4cdb5bu, 0xfc6c5097u, 0x6f547f83u, 0x1ac31b47u }
|
||||
#define IGNEUM_MIX_RC_INIT { 0xe5bef5a3u, 0xdb2a7d90u, 0xcd1fa7e1u, 0x30289419u, 0x0a730d58u, 0x432f8579u, 0x6ab978a5u, 0x8c984f49u, 0x788c3c9eu, 0x051eef02u, 0x05e77db9u, 0x8b3cd4f3u, 0x6948d6cfu, 0x0d8677f6u, 0xf9505c8cu, 0x4513c163u }
|
||||
|
||||
#ifndef IGNEUM_NO_CUDA
|
||||
// Defined in kernel.cu. All launch on the default stream and return cudaGetLastError().
|
||||
cudaError_t igneum_launch_cache_fill(uint32_t* cache, uint32_t nSegments);
|
||||
cudaError_t igneum_launch_build(uint32_t* ds, const uint32_t* cache, const uint32_t* leaves, uint32_t nLeaves, uint32_t nItems);
|
||||
cudaError_t igneum_launch_hash(const uint32_t* ds, uint64_t* out, uint32_t baseNonce, uint32_t mask,
|
||||
uint32_t nonces, uint32_t blockWarps);
|
||||
cudaError_t igneum_hash_info(int* numRegs, int* blocksPerSM, uint32_t blockWarps);
|
||||
#endif
|
||||
416
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program.json
Normal file
416
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program.json
Normal file
|
|
@ -0,0 +1,416 @@
|
|||
{
|
||||
"format": "igneum-program-pack-3",
|
||||
"generator": 5,
|
||||
"attempt": 0,
|
||||
"program_id": "0xe5a4ac5978462156",
|
||||
"program_id_derivation": "FNV-1a 64 over 'igneum-program/' || generator_le32 || seed_words as little-endian bytes || attempt_le32",
|
||||
"dataset_mode": "memory-hard",
|
||||
"seed": "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000",
|
||||
"seed_bytes": "4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925",
|
||||
"seed_words": ["0xbe8c5a0c", "0x7646e626", "0x508e29d0", "0xfb8a5b4a", "0x53c50955", "0x685d62fc", "0x6065c013", "0x881096eb"],
|
||||
"seed_derivation": "seed_words = FNV-1a 64 over seed_bytes (attempt 0) or seed_bytes || attempt_le32 (attempt k >= 1), basis ^ (salt * 0x9E3779B97F4A7C15) for salt 0..3, then h ^= h>>33; h *= 0xff51afd7ed558ccd; h ^= h>>33; words[2*salt] = low 32, words[2*salt+1] = high 32",
|
||||
"generator_rule": "version 2: exactly 16 load slots drawn first from instructions 1..63 (partial Fisher-Yates), the other 48 ops from the ten non-load weights (sum 75); a load's source is drawn from the registers other than dst written by an earlier instruction and not read by a load since; the candidate must pass the acceptance rule of spec 01 section 1.4.6 (static: no cyclically stale load source, every register has an injecting write; dynamic: 64 units on the seed-keyed closed-form dataset with no constant register bit, no lane-constant load site, under 164 saturated final values, every output bit within 136 of 1024, distinct addresses above 245760), else the next attempt of the seed is tried",
|
||||
"lanes": 32,
|
||||
"registers": 8,
|
||||
"iterations": 8,
|
||||
"instruction_count": 64,
|
||||
"loads_per_hash": 128,
|
||||
"program_class": "v5",
|
||||
"era_seed_bytes": "4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925",
|
||||
"state": {
|
||||
"block": "4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925",
|
||||
"block_number": 0,
|
||||
"root": "7e37a9fb19b154d32daf5bf30a50d339a75029fbc9eec9ea20e95439dba5a311",
|
||||
"leaves": 11,
|
||||
"records": 11,
|
||||
"sampled": false,
|
||||
"leaves_fnv1a64": "0xf141bfee2a8b11b0",
|
||||
"leaf_derivation": "leaves[i] = Blake2b-512('igneum-sd1/' || root || i_le32 || record_i) as 16 little-endian words; item t XORs leaves[t mod leaves] into its 16 initial words before the first mixer",
|
||||
"file": "leaves.bin"
|
||||
},
|
||||
"load_class": "mx8-eraaf3a9139+sh256x27+state",
|
||||
"mixer_mult": 8,
|
||||
"cache_growth": true,
|
||||
"mixer": "class v3 (Counter ASIC 2.0, 5 October 2026, docs/plans/mixer-x4.md): every mixer application of the item derivation is 8 applications with round keys (r * 8 + j + 1) * 0x9E3779B9, the 8 dependent cache reads per item unchanged; cache growth rule option C: cache words = 2^(26 + doublings(day)), dataset words = 2^(genesis_log2 + doublings(day)), doublings(day) = floor(log2(1 + day / 1460)) for day = days since genesis",
|
||||
"load_slots": 16,
|
||||
"load_mix_percent_4_16_64": [100, 0, 0],
|
||||
"load_width_counts_4_16_64": [16, 0, 0],
|
||||
"bytes_per_hash": 512,
|
||||
"wide_load": "read-width experiment (5 October 2026, docs/plans/read-width.md), NOT the lottery hash: a load of W words (width field, 4 or 16) reads dataset[b .. b + W) with b = (src & mask) & ~(W - 1) and folds every word into dst: x = dst ^ w[0]; for j in 1..W: x = (rotl(x, 11) * 0x9e3779b1) ^ w[j]; dst = x; width 1 is the plain load; the width is drawn per instruction from the class mix with one extra below(100) draw after the nine of version 2, and the program id is FNV-1a 64 over 'igneum-program-rw/' || generator_le32 || seed words || attempt_le32 || mix[3] || load_slots",
|
||||
"era": {
|
||||
"label": "af3a9139",
|
||||
"seed_words": ["0xaf3a9139", "0x894db311", "0xd6173573", "0xc346bc0f", "0x2d547cec", "0xd5634ab8", "0xb9584827", "0x305d030e"],
|
||||
"draw": "docs/plans/era-layout.md 1.1: SplitMix64 seeded with seed_words[0] | seed_words[1] << 32 of seed_words_from_bytes('igneum-era/' || n_le64 || E_n); width = allowed[below(|allowed|)], stride_mul = low32(next()) | 1, stride_rot = 1 + below(31), then four next() draws for a partial Fisher-Yates over positions log2(W)..15 of which 4 - log2(W) are used",
|
||||
"allowed_widths": [1],
|
||||
"width_words": 1,
|
||||
"stride_mul": "0x2cb18b35",
|
||||
"stride_rot": 9,
|
||||
"interleave": [1, 9, 13, 14],
|
||||
"address": "y = rotl(src * stride_mul, stride_rot); k = min(win, D - 26); idx = ((y & (mask >> k)) | ((off & (2^k - 1)) << (D - k))) & mask; a wide load aligns idx down to W words",
|
||||
"windows": "per instruction, after the width roll: win = below(3), off = low32(next()) & (2^win - 1); used on a load slot (the instruction's win and off fields)",
|
||||
"dataset_word": "dataset[w] = item(t(w))[j(w)]: j(w) gathers the bits of w at the interleave positions, t(w) is w with those bits removed",
|
||||
"program_id_suffix": "'era/' || allowed[3] || width_words || stride_mul_le32 || stride_rot_le32 || interleave[4]"
|
||||
},
|
||||
"op_mix": {"load": 16, "add": 11, "shfl": 7, "mad": 6, "xor": 6, "mul": 4, "mulhi": 4, "rotr": 4, "rotl": 3, "sub": 2, "or": 1},
|
||||
"register_init": "for i in 0..7: x = nonce ^ seed_words[i]; x += 0x9e3779b9 * (i+1) (mod 2^32); x = splitmix32(x); r[i] = x ^ seed_words[(i+1) & 7]",
|
||||
"splitmix32": "x ^= x>>16; x *= 0x7feb352d; x ^= x>>15; x *= 0x846ca68b; x ^= x>>16",
|
||||
"iteration": "sel = r0 sampled once at the top of each iteration, then all instructions in order",
|
||||
"output": "lo = r0 ^ rotl(r1,7) ^ rotl(r2,14) ^ rotl(r3,21); hi = r4 ^ rotl(r5,9) ^ rotl(r6,18) ^ rotl(r7,27); out = (hi << 32) | lo",
|
||||
"op_semantics": {
|
||||
"add": "dst = dst + src + (bit `bit` of sel ? imm2 : imm)",
|
||||
"sub": "dst = dst - src",
|
||||
"mul": "dst = dst * src (low 32)",
|
||||
"mulhi": "dst = high 32 bits of dst * src",
|
||||
"xor": "dst = dst ^ src",
|
||||
"or": "dst = dst | src",
|
||||
"rotl": "dst = rotl(dst, rot), rot in 1..31",
|
||||
"rotr": "dst = rotr(dst, src & 31)",
|
||||
"mad": "dst = src * src2 + dst",
|
||||
"shfl": "dst = dst ^ (src of lane (lane ^ mask)), mask in {1,2,4,8,16}, within the 32-lane warp",
|
||||
"load": "dst = dst ^ dataset[src & dataset.mask]",
|
||||
"wload": "base = (src of lane 0 & dataset.mask) & ~31; dst = dst ^ dataset[base + lane] (warp-coalesced 128-byte load, lever b, only when --wide-frac > 0)"
|
||||
},
|
||||
"dataset": {
|
||||
"log2_words": 28,
|
||||
"bytes": 1073741824,
|
||||
"mask": "0x0fffffff",
|
||||
"day": "bytes:69676e65756d2d6461792ffd50000000000000",
|
||||
"day_bytes": "69676e65756d2d6461792ffd50000000000000",
|
||||
"day_words_from": "seed_words_from_bytes(day_bytes)",
|
||||
"d0": "0xe23f9008",
|
||||
"d1": "0xde33e763",
|
||||
"mode": "memory-hard",
|
||||
"spec": "proto-metal/MEMHARD.md",
|
||||
"key": ["0xe23f9008", "0xde33e763", "0xc5ba415d", "0x8ddf6786", "0x59f4a4be", "0x8a3bc680", "0x701b8e40", "0x3b59025a"],
|
||||
"key_derivation": "the 8 words of seed_words_from_bytes(day_bytes); d0, d1 are key[0], key[1]",
|
||||
"cache": {"log2_words": 26, "bytes": 268435456, "line_words": 16, "segment_lines": 64, "segments": 65536, "block": "ChaCha12 core + feed-forward, rotations 16 12 8 7", "sigma": ["0x61707865", "0x3320646e", "0x79622d32", "0x6b206574"], "tag": ["0x49676e65", "0x756d4d48"], "chain": "in_j = prev_line ^ (sigma[0..3] || key[0..7] || seg || j || tag[0..1]); line_j = block(in_j); prev_0 = 0"},
|
||||
"mixer": {"draw": "SplitMix64 seeded with key[0] | key[1] << 32: rot[0..7] = 1 + next() % 31, mul[0..15] = low32(next()) | 1, rc[0..15] = low32(next())", "rot": [28, 11, 18, 18, 13, 28, 5, 13], "mul": ["0xd1d79e7f", "0xbcb61f8b", "0x970e91ab", "0xd749d96d", "0x143d7339", "0x2cde0d69", "0x12d0a8c1", "0x2d335be9", "0x80a8aae9", "0x7ac896c7", "0x9de23db7", "0xc362827d", "0x5f4cdb5b", "0xfc6c5097", "0x6f547f83", "0x1ac31b47"], "rc": ["0xe5bef5a3", "0xdb2a7d90", "0xcd1fa7e1", "0x30289419", "0x0a730d58", "0x432f8579", "0x6ab978a5", "0x8c984f49", "0x788c3c9e", "0x051eef02", "0x05e77db9", "0x8b3cd4f3", "0x6948d6cf", "0x0d8677f6", "0xf9505c8c", "0x4513c163"], "round": "for i in 0..15: s[i] = (s[i] ^ (rc[i] + (r+1) * 0x9E3779B9)) * mul[i]; then quarter rounds on columns (0,4,8,12) (1,5,9,13) (2,6,10,14) (3,7,11,15) with rot[0..3] and diagonals (0,5,10,15) (1,6,11,12) (2,7,8,13) (3,4,9,14) with rot[4..7]", "quarter_round": "a += b; d ^= a; d = rotl(d, r1); c += d; b ^= c; b = rotl(b, r2); a += b; d ^= a; d = rotl(d, r3); c += d; b ^= c; b = rotl(b, r4)"},
|
||||
"mixer_mult": 8,
|
||||
"item": "s[0..7] = key; s[8+i] = t * mul[i] + rc[i] for i in 0..7; for r in 0..7: for j in 0..7: s = M(s, rk = (r * 8 + j + 1) * 0x9E3779B9); line = s[0] & 0x003fffff; s[i] ^= cache[line * 16 + i]; then for j in 0..7: s = M(s, rk = (64 + j + 1) * 0x9E3779B9); item(t) = s",
|
||||
"word": "dataset[w] = item(w >> 4)[w & 15]"
|
||||
},
|
||||
"shadow": {"instrs": 256, "reps": 27, "instrs_per_hash": 55296, "op_mix": {"add": 45, "mad": 31, "shfl": 31, "xor": 30, "rotl": 28, "mul": 22, "or": 18, "mulhi": 17, "rotr": 17, "sub": 17}, "rule": "Counter ASIC 3.0 item 8 (docs/analysis/latency-shadow-2026-10-06.md): after the 64 base instructions the program stream draws instrs more ALU instructions (the non-load table, nine draws each, the source as on an ALU slot); the block runs reps times at the end of every iteration with the iteration's sel; the acceptance rule interprets the base program only", "program_id_suffix": "'shadow/' || instrs_le16 || reps_le16", "instructions": [
|
||||
{"i": 0, "op": "or", "dst": 6, "src": 0, "src2": 2, "imm": "0xc1eb1523", "imm2": "0x925958ea", "rot": 14, "bit": 14, "mask": 8},
|
||||
{"i": 1, "op": "mul", "dst": 7, "src": 3, "src2": 4, "imm": "0xc7044334", "imm2": "0x755566b6", "rot": 19, "bit": 17, "mask": 8},
|
||||
{"i": 2, "op": "mad", "dst": 3, "src": 5, "src2": 5, "imm": "0x33b52f17", "imm2": "0xcc459a7a", "rot": 5, "bit": 19, "mask": 16},
|
||||
{"i": 3, "op": "add", "dst": 7, "src": 4, "src2": 4, "imm": "0x97d3d105", "imm2": "0x38e58a06", "rot": 11, "bit": 6, "mask": 1},
|
||||
{"i": 4, "op": "add", "dst": 6, "src": 0, "src2": 6, "imm": "0x29b87b9a", "imm2": "0x9629673d", "rot": 16, "bit": 20, "mask": 8},
|
||||
{"i": 5, "op": "shfl", "dst": 6, "src": 2, "src2": 5, "imm": "0xb61f2e77", "imm2": "0x30e3807c", "rot": 1, "bit": 7, "mask": 1},
|
||||
{"i": 6, "op": "mad", "dst": 5, "src": 3, "src2": 5, "imm": "0x0de66253", "imm2": "0xfdfdd03e", "rot": 4, "bit": 27, "mask": 4},
|
||||
{"i": 7, "op": "rotl", "dst": 3, "src": 7, "src2": 1, "imm": "0x688514dd", "imm2": "0x139d08b5", "rot": 5, "bit": 1, "mask": 1},
|
||||
{"i": 8, "op": "add", "dst": 0, "src": 7, "src2": 3, "imm": "0xa7c1fb0f", "imm2": "0x4a5f55e1", "rot": 15, "bit": 12, "mask": 1},
|
||||
{"i": 9, "op": "mul", "dst": 5, "src": 7, "src2": 2, "imm": "0x1a64de5a", "imm2": "0xe37b0875", "rot": 5, "bit": 27, "mask": 16},
|
||||
{"i": 10, "op": "xor", "dst": 4, "src": 5, "src2": 3, "imm": "0x750d4f1a", "imm2": "0x2f3d9402", "rot": 3, "bit": 14, "mask": 2},
|
||||
{"i": 11, "op": "mulhi", "dst": 6, "src": 4, "src2": 5, "imm": "0x2eaa86c6", "imm2": "0x9fd5b22f", "rot": 30, "bit": 10, "mask": 1},
|
||||
{"i": 12, "op": "rotl", "dst": 5, "src": 2, "src2": 4, "imm": "0x8df50473", "imm2": "0x23d780ce", "rot": 13, "bit": 9, "mask": 1},
|
||||
{"i": 13, "op": "mad", "dst": 0, "src": 2, "src2": 1, "imm": "0xa956a7e6", "imm2": "0x58dbb356", "rot": 31, "bit": 11, "mask": 8},
|
||||
{"i": 14, "op": "mulhi", "dst": 2, "src": 4, "src2": 6, "imm": "0x6fbfb145", "imm2": "0x064eb015", "rot": 20, "bit": 2, "mask": 4},
|
||||
{"i": 15, "op": "or", "dst": 3, "src": 1, "src2": 5, "imm": "0xd313bc27", "imm2": "0x9f0fc235", "rot": 17, "bit": 16, "mask": 1},
|
||||
{"i": 16, "op": "sub", "dst": 0, "src": 5, "src2": 0, "imm": "0x86cbd336", "imm2": "0x94715167", "rot": 17, "bit": 31, "mask": 8},
|
||||
{"i": 17, "op": "mul", "dst": 0, "src": 6, "src2": 3, "imm": "0x95911c6b", "imm2": "0x816cea32", "rot": 23, "bit": 1, "mask": 2},
|
||||
{"i": 18, "op": "shfl", "dst": 5, "src": 1, "src2": 3, "imm": "0xf563dcad", "imm2": "0xb2889856", "rot": 14, "bit": 24, "mask": 4},
|
||||
{"i": 19, "op": "or", "dst": 3, "src": 4, "src2": 2, "imm": "0x7bb48d19", "imm2": "0xbf95ad6f", "rot": 8, "bit": 30, "mask": 16},
|
||||
{"i": 20, "op": "mad", "dst": 6, "src": 3, "src2": 2, "imm": "0xeb686a8d", "imm2": "0x8bda9f8a", "rot": 11, "bit": 18, "mask": 1},
|
||||
{"i": 21, "op": "rotl", "dst": 0, "src": 6, "src2": 0, "imm": "0x0915df51", "imm2": "0xa1365cda", "rot": 11, "bit": 25, "mask": 1},
|
||||
{"i": 22, "op": "rotl", "dst": 3, "src": 0, "src2": 3, "imm": "0xc09e4cdc", "imm2": "0x50c1369d", "rot": 25, "bit": 31, "mask": 2},
|
||||
{"i": 23, "op": "xor", "dst": 6, "src": 4, "src2": 1, "imm": "0xce01abec", "imm2": "0x5ff8c52e", "rot": 30, "bit": 19, "mask": 8},
|
||||
{"i": 24, "op": "or", "dst": 4, "src": 2, "src2": 4, "imm": "0xdacf9e4b", "imm2": "0x086a5684", "rot": 2, "bit": 10, "mask": 2},
|
||||
{"i": 25, "op": "add", "dst": 1, "src": 6, "src2": 3, "imm": "0x53e74799", "imm2": "0xfc2f6d17", "rot": 8, "bit": 7, "mask": 16},
|
||||
{"i": 26, "op": "rotr", "dst": 7, "src": 6, "src2": 7, "imm": "0xc89a75a0", "imm2": "0x67813253", "rot": 24, "bit": 13, "mask": 1},
|
||||
{"i": 27, "op": "add", "dst": 7, "src": 5, "src2": 4, "imm": "0xb8fae90e", "imm2": "0x787048dc", "rot": 26, "bit": 27, "mask": 1},
|
||||
{"i": 28, "op": "xor", "dst": 4, "src": 1, "src2": 0, "imm": "0x4c90c2ea", "imm2": "0x2ce7865b", "rot": 4, "bit": 7, "mask": 2},
|
||||
{"i": 29, "op": "xor", "dst": 6, "src": 1, "src2": 6, "imm": "0x6131ab08", "imm2": "0xd1825676", "rot": 4, "bit": 30, "mask": 2},
|
||||
{"i": 30, "op": "mulhi", "dst": 2, "src": 0, "src2": 1, "imm": "0xbe4a545f", "imm2": "0xdc29e823", "rot": 17, "bit": 5, "mask": 8},
|
||||
{"i": 31, "op": "rotr", "dst": 1, "src": 4, "src2": 1, "imm": "0x4d2f9cdc", "imm2": "0xdee1cc73", "rot": 5, "bit": 16, "mask": 4},
|
||||
{"i": 32, "op": "xor", "dst": 7, "src": 4, "src2": 5, "imm": "0x9d277aaa", "imm2": "0xa902fca5", "rot": 11, "bit": 3, "mask": 1},
|
||||
{"i": 33, "op": "mad", "dst": 3, "src": 2, "src2": 6, "imm": "0x7db09644", "imm2": "0x077bf3d9", "rot": 30, "bit": 19, "mask": 2},
|
||||
{"i": 34, "op": "add", "dst": 3, "src": 1, "src2": 1, "imm": "0xd2dc42eb", "imm2": "0x9ffd510b", "rot": 17, "bit": 12, "mask": 2},
|
||||
{"i": 35, "op": "add", "dst": 4, "src": 7, "src2": 0, "imm": "0xad8eda6f", "imm2": "0xf871ba37", "rot": 24, "bit": 7, "mask": 4},
|
||||
{"i": 36, "op": "add", "dst": 7, "src": 4, "src2": 3, "imm": "0xb0cef8f4", "imm2": "0xd02d30ca", "rot": 22, "bit": 3, "mask": 16},
|
||||
{"i": 37, "op": "mad", "dst": 4, "src": 3, "src2": 2, "imm": "0x6ed99a66", "imm2": "0x83dc97c0", "rot": 4, "bit": 1, "mask": 4},
|
||||
{"i": 38, "op": "mad", "dst": 5, "src": 7, "src2": 7, "imm": "0xe24fdd24", "imm2": "0x360f5c8b", "rot": 19, "bit": 9, "mask": 8},
|
||||
{"i": 39, "op": "add", "dst": 3, "src": 6, "src2": 4, "imm": "0x2bd506e2", "imm2": "0x82e98a22", "rot": 28, "bit": 26, "mask": 1},
|
||||
{"i": 40, "op": "rotl", "dst": 0, "src": 2, "src2": 1, "imm": "0xdb1faed3", "imm2": "0xf968a20d", "rot": 21, "bit": 30, "mask": 16},
|
||||
{"i": 41, "op": "or", "dst": 1, "src": 0, "src2": 1, "imm": "0xe6dfe522", "imm2": "0xb75813da", "rot": 10, "bit": 13, "mask": 1},
|
||||
{"i": 42, "op": "add", "dst": 7, "src": 5, "src2": 7, "imm": "0xc0386e49", "imm2": "0x6435f6d8", "rot": 26, "bit": 28, "mask": 16},
|
||||
{"i": 43, "op": "mulhi", "dst": 1, "src": 0, "src2": 4, "imm": "0xda497f7d", "imm2": "0xc4b748f4", "rot": 30, "bit": 10, "mask": 4},
|
||||
{"i": 44, "op": "add", "dst": 5, "src": 6, "src2": 0, "imm": "0x3a8921a4", "imm2": "0x1f682c68", "rot": 12, "bit": 12, "mask": 1},
|
||||
{"i": 45, "op": "add", "dst": 4, "src": 5, "src2": 6, "imm": "0xeeca4334", "imm2": "0xf246e180", "rot": 26, "bit": 1, "mask": 1},
|
||||
{"i": 46, "op": "xor", "dst": 2, "src": 3, "src2": 5, "imm": "0x052178c3", "imm2": "0xaab31f10", "rot": 28, "bit": 16, "mask": 8},
|
||||
{"i": 47, "op": "xor", "dst": 2, "src": 5, "src2": 0, "imm": "0x91fd7273", "imm2": "0xacc42b07", "rot": 22, "bit": 27, "mask": 4},
|
||||
{"i": 48, "op": "mad", "dst": 7, "src": 4, "src2": 1, "imm": "0x05a2970f", "imm2": "0x14876824", "rot": 10, "bit": 5, "mask": 1},
|
||||
{"i": 49, "op": "mad", "dst": 2, "src": 6, "src2": 1, "imm": "0x4440d6de", "imm2": "0x352023f9", "rot": 10, "bit": 25, "mask": 8},
|
||||
{"i": 50, "op": "mad", "dst": 7, "src": 0, "src2": 7, "imm": "0xa6f82633", "imm2": "0x78465772", "rot": 13, "bit": 8, "mask": 4},
|
||||
{"i": 51, "op": "add", "dst": 5, "src": 6, "src2": 5, "imm": "0x745776d8", "imm2": "0x41bd68cc", "rot": 29, "bit": 26, "mask": 1},
|
||||
{"i": 52, "op": "shfl", "dst": 3, "src": 0, "src2": 6, "imm": "0x8f94ebac", "imm2": "0xe4a208a0", "rot": 5, "bit": 16, "mask": 1},
|
||||
{"i": 53, "op": "or", "dst": 4, "src": 0, "src2": 5, "imm": "0x7b1235f3", "imm2": "0x408b4ea9", "rot": 27, "bit": 19, "mask": 8},
|
||||
{"i": 54, "op": "mad", "dst": 0, "src": 2, "src2": 4, "imm": "0x30114050", "imm2": "0xe71dff8f", "rot": 6, "bit": 30, "mask": 1},
|
||||
{"i": 55, "op": "rotl", "dst": 5, "src": 1, "src2": 7, "imm": "0x69f92198", "imm2": "0x65496f29", "rot": 9, "bit": 20, "mask": 4},
|
||||
{"i": 56, "op": "mul", "dst": 5, "src": 7, "src2": 6, "imm": "0x88b884d4", "imm2": "0x041b67f7", "rot": 25, "bit": 8, "mask": 16},
|
||||
{"i": 57, "op": "rotl", "dst": 7, "src": 5, "src2": 6, "imm": "0x7d1e18bc", "imm2": "0xf54ade27", "rot": 27, "bit": 20, "mask": 1},
|
||||
{"i": 58, "op": "sub", "dst": 7, "src": 6, "src2": 4, "imm": "0xa457bbab", "imm2": "0x4b05eb29", "rot": 6, "bit": 6, "mask": 4},
|
||||
{"i": 59, "op": "rotl", "dst": 2, "src": 7, "src2": 7, "imm": "0x4890bdda", "imm2": "0xadcb3d99", "rot": 22, "bit": 26, "mask": 16},
|
||||
{"i": 60, "op": "shfl", "dst": 7, "src": 1, "src2": 7, "imm": "0xe4da58eb", "imm2": "0x1d50080d", "rot": 22, "bit": 26, "mask": 2},
|
||||
{"i": 61, "op": "add", "dst": 4, "src": 0, "src2": 5, "imm": "0x60d0ff55", "imm2": "0xcba6a161", "rot": 6, "bit": 7, "mask": 8},
|
||||
{"i": 62, "op": "mul", "dst": 1, "src": 5, "src2": 1, "imm": "0x0fdc702e", "imm2": "0x6b1bd03a", "rot": 9, "bit": 23, "mask": 4},
|
||||
{"i": 63, "op": "mulhi", "dst": 7, "src": 6, "src2": 1, "imm": "0xcc44ed62", "imm2": "0x931dd006", "rot": 1, "bit": 0, "mask": 2},
|
||||
{"i": 64, "op": "mul", "dst": 5, "src": 0, "src2": 7, "imm": "0x448c3aba", "imm2": "0x1136818a", "rot": 23, "bit": 7, "mask": 1},
|
||||
{"i": 65, "op": "rotr", "dst": 2, "src": 5, "src2": 1, "imm": "0xb520dfb8", "imm2": "0xdce7d8f4", "rot": 22, "bit": 10, "mask": 8},
|
||||
{"i": 66, "op": "mulhi", "dst": 0, "src": 3, "src2": 4, "imm": "0x19fada48", "imm2": "0x3cc04e0e", "rot": 1, "bit": 24, "mask": 16},
|
||||
{"i": 67, "op": "or", "dst": 0, "src": 7, "src2": 3, "imm": "0x1c5905b4", "imm2": "0xf896dfb7", "rot": 31, "bit": 30, "mask": 8},
|
||||
{"i": 68, "op": "rotl", "dst": 0, "src": 1, "src2": 5, "imm": "0xa28511c1", "imm2": "0xbf649695", "rot": 19, "bit": 27, "mask": 1},
|
||||
{"i": 69, "op": "rotl", "dst": 0, "src": 7, "src2": 5, "imm": "0x22e2fef3", "imm2": "0x468bec9e", "rot": 1, "bit": 28, "mask": 1},
|
||||
{"i": 70, "op": "or", "dst": 2, "src": 3, "src2": 6, "imm": "0xcb542821", "imm2": "0xe084cf11", "rot": 15, "bit": 5, "mask": 8},
|
||||
{"i": 71, "op": "mad", "dst": 3, "src": 7, "src2": 5, "imm": "0xf06b8f6a", "imm2": "0x7382e15f", "rot": 6, "bit": 28, "mask": 16},
|
||||
{"i": 72, "op": "mad", "dst": 7, "src": 0, "src2": 4, "imm": "0x4f64c588", "imm2": "0x47b63ee9", "rot": 8, "bit": 10, "mask": 2},
|
||||
{"i": 73, "op": "or", "dst": 4, "src": 3, "src2": 2, "imm": "0xff138eea", "imm2": "0x6fd90eda", "rot": 23, "bit": 3, "mask": 16},
|
||||
{"i": 74, "op": "xor", "dst": 3, "src": 1, "src2": 0, "imm": "0x6011ece1", "imm2": "0x89579f15", "rot": 11, "bit": 2, "mask": 4},
|
||||
{"i": 75, "op": "add", "dst": 1, "src": 3, "src2": 4, "imm": "0x3f4ffea2", "imm2": "0x8e35924e", "rot": 12, "bit": 15, "mask": 2},
|
||||
{"i": 76, "op": "mul", "dst": 1, "src": 3, "src2": 7, "imm": "0x15efa846", "imm2": "0xa1c971f1", "rot": 14, "bit": 20, "mask": 1},
|
||||
{"i": 77, "op": "mad", "dst": 0, "src": 3, "src2": 2, "imm": "0x8660530a", "imm2": "0xa013524b", "rot": 14, "bit": 12, "mask": 8},
|
||||
{"i": 78, "op": "shfl", "dst": 2, "src": 0, "src2": 0, "imm": "0xcc2e96bd", "imm2": "0xe32c6f87", "rot": 10, "bit": 11, "mask": 16},
|
||||
{"i": 79, "op": "mulhi", "dst": 1, "src": 0, "src2": 6, "imm": "0x61931e3f", "imm2": "0x86383e7f", "rot": 17, "bit": 3, "mask": 1},
|
||||
{"i": 80, "op": "mulhi", "dst": 0, "src": 4, "src2": 7, "imm": "0xd9991afc", "imm2": "0xd00e8678", "rot": 24, "bit": 15, "mask": 16},
|
||||
{"i": 81, "op": "mul", "dst": 4, "src": 2, "src2": 3, "imm": "0x472f70f4", "imm2": "0x9fde4236", "rot": 10, "bit": 5, "mask": 4},
|
||||
{"i": 82, "op": "sub", "dst": 5, "src": 7, "src2": 6, "imm": "0x55a7acb1", "imm2": "0x9fde8df1", "rot": 30, "bit": 29, "mask": 16},
|
||||
{"i": 83, "op": "mad", "dst": 0, "src": 2, "src2": 3, "imm": "0x1c487c40", "imm2": "0xd9d9a1c3", "rot": 9, "bit": 5, "mask": 4},
|
||||
{"i": 84, "op": "sub", "dst": 7, "src": 2, "src2": 4, "imm": "0x515c0a72", "imm2": "0x6df5c83d", "rot": 28, "bit": 21, "mask": 4},
|
||||
{"i": 85, "op": "xor", "dst": 1, "src": 2, "src2": 4, "imm": "0x917f1174", "imm2": "0x8e72e519", "rot": 9, "bit": 18, "mask": 16},
|
||||
{"i": 86, "op": "shfl", "dst": 6, "src": 3, "src2": 3, "imm": "0xda856cc2", "imm2": "0x932ae8d8", "rot": 28, "bit": 1, "mask": 16},
|
||||
{"i": 87, "op": "mad", "dst": 3, "src": 0, "src2": 0, "imm": "0x4ef22321", "imm2": "0x5666ada3", "rot": 4, "bit": 4, "mask": 1},
|
||||
{"i": 88, "op": "rotl", "dst": 0, "src": 1, "src2": 3, "imm": "0xbb3ec00c", "imm2": "0xffe78937", "rot": 20, "bit": 19, "mask": 4},
|
||||
{"i": 89, "op": "sub", "dst": 1, "src": 5, "src2": 4, "imm": "0x8f295708", "imm2": "0x6608577f", "rot": 17, "bit": 18, "mask": 16},
|
||||
{"i": 90, "op": "shfl", "dst": 6, "src": 4, "src2": 5, "imm": "0xd953aa15", "imm2": "0x4ef9f062", "rot": 19, "bit": 9, "mask": 16},
|
||||
{"i": 91, "op": "xor", "dst": 5, "src": 1, "src2": 5, "imm": "0x6c02cde2", "imm2": "0x118dbd2b", "rot": 2, "bit": 20, "mask": 16},
|
||||
{"i": 92, "op": "rotl", "dst": 5, "src": 0, "src2": 5, "imm": "0x9f0b0c61", "imm2": "0xe26dbe7a", "rot": 12, "bit": 17, "mask": 2},
|
||||
{"i": 93, "op": "mad", "dst": 6, "src": 1, "src2": 2, "imm": "0x78a44a91", "imm2": "0x3ce118a9", "rot": 18, "bit": 16, "mask": 16},
|
||||
{"i": 94, "op": "rotr", "dst": 2, "src": 5, "src2": 3, "imm": "0x02295fcb", "imm2": "0xc3353c64", "rot": 5, "bit": 31, "mask": 16},
|
||||
{"i": 95, "op": "or", "dst": 0, "src": 4, "src2": 4, "imm": "0x727120c3", "imm2": "0x528f4416", "rot": 6, "bit": 20, "mask": 8},
|
||||
{"i": 96, "op": "mul", "dst": 7, "src": 3, "src2": 4, "imm": "0x919c42ea", "imm2": "0x87526aa5", "rot": 28, "bit": 1, "mask": 8},
|
||||
{"i": 97, "op": "or", "dst": 5, "src": 0, "src2": 7, "imm": "0x8a0a100c", "imm2": "0xb2c65cc3", "rot": 13, "bit": 7, "mask": 4},
|
||||
{"i": 98, "op": "add", "dst": 3, "src": 1, "src2": 1, "imm": "0xbcfdd8a7", "imm2": "0x386bb642", "rot": 8, "bit": 13, "mask": 8},
|
||||
{"i": 99, "op": "add", "dst": 0, "src": 6, "src2": 7, "imm": "0xc921a021", "imm2": "0xa3962d03", "rot": 31, "bit": 9, "mask": 1},
|
||||
{"i": 100, "op": "rotr", "dst": 1, "src": 0, "src2": 3, "imm": "0x918c6f69", "imm2": "0x1a125801", "rot": 18, "bit": 29, "mask": 16},
|
||||
{"i": 101, "op": "add", "dst": 2, "src": 4, "src2": 1, "imm": "0x70da4067", "imm2": "0xc5990c5c", "rot": 3, "bit": 18, "mask": 4},
|
||||
{"i": 102, "op": "rotl", "dst": 2, "src": 7, "src2": 0, "imm": "0x9048d081", "imm2": "0xc25a9f56", "rot": 3, "bit": 25, "mask": 16},
|
||||
{"i": 103, "op": "shfl", "dst": 1, "src": 4, "src2": 7, "imm": "0x70ca1427", "imm2": "0x6632093d", "rot": 18, "bit": 28, "mask": 4},
|
||||
{"i": 104, "op": "xor", "dst": 0, "src": 2, "src2": 0, "imm": "0x0637204c", "imm2": "0x3b1b7f94", "rot": 1, "bit": 12, "mask": 2},
|
||||
{"i": 105, "op": "add", "dst": 4, "src": 3, "src2": 5, "imm": "0x1a5880cb", "imm2": "0xb15aec75", "rot": 31, "bit": 28, "mask": 1},
|
||||
{"i": 106, "op": "shfl", "dst": 4, "src": 0, "src2": 4, "imm": "0x7c22cb00", "imm2": "0xc4df7949", "rot": 3, "bit": 15, "mask": 2},
|
||||
{"i": 107, "op": "shfl", "dst": 7, "src": 0, "src2": 0, "imm": "0xd43c76ca", "imm2": "0x8d5b9751", "rot": 31, "bit": 28, "mask": 4},
|
||||
{"i": 108, "op": "mad", "dst": 1, "src": 7, "src2": 3, "imm": "0x69879993", "imm2": "0x3c4e18b0", "rot": 19, "bit": 25, "mask": 4},
|
||||
{"i": 109, "op": "mad", "dst": 0, "src": 2, "src2": 3, "imm": "0x8ce3a4a6", "imm2": "0x1b3f0b74", "rot": 19, "bit": 8, "mask": 2},
|
||||
{"i": 110, "op": "or", "dst": 2, "src": 4, "src2": 5, "imm": "0xcdbdac98", "imm2": "0x8d0e9f92", "rot": 19, "bit": 13, "mask": 1},
|
||||
{"i": 111, "op": "shfl", "dst": 6, "src": 1, "src2": 1, "imm": "0xe74157f6", "imm2": "0xf92dead6", "rot": 4, "bit": 14, "mask": 1},
|
||||
{"i": 112, "op": "mad", "dst": 2, "src": 5, "src2": 3, "imm": "0x200831bc", "imm2": "0x1ee98048", "rot": 29, "bit": 14, "mask": 1},
|
||||
{"i": 113, "op": "mulhi", "dst": 4, "src": 0, "src2": 7, "imm": "0xd6827ead", "imm2": "0xc0ee1933", "rot": 26, "bit": 23, "mask": 16},
|
||||
{"i": 114, "op": "rotl", "dst": 7, "src": 2, "src2": 2, "imm": "0x6ae1be2b", "imm2": "0x83216a4e", "rot": 29, "bit": 25, "mask": 16},
|
||||
{"i": 115, "op": "sub", "dst": 3, "src": 7, "src2": 1, "imm": "0xa1d1646e", "imm2": "0x6fa998ce", "rot": 15, "bit": 29, "mask": 2},
|
||||
{"i": 116, "op": "mulhi", "dst": 0, "src": 1, "src2": 4, "imm": "0xd7921c15", "imm2": "0x8b2345a7", "rot": 10, "bit": 19, "mask": 4},
|
||||
{"i": 117, "op": "rotl", "dst": 4, "src": 0, "src2": 4, "imm": "0xc1b1b0ab", "imm2": "0xd19b3f39", "rot": 12, "bit": 16, "mask": 1},
|
||||
{"i": 118, "op": "add", "dst": 1, "src": 4, "src2": 5, "imm": "0x24494ad3", "imm2": "0xc5e98ee5", "rot": 22, "bit": 7, "mask": 16},
|
||||
{"i": 119, "op": "add", "dst": 0, "src": 4, "src2": 6, "imm": "0x1b2d1c83", "imm2": "0xa97bc949", "rot": 8, "bit": 4, "mask": 16},
|
||||
{"i": 120, "op": "shfl", "dst": 5, "src": 4, "src2": 7, "imm": "0xb93015ff", "imm2": "0xc22366cd", "rot": 23, "bit": 16, "mask": 4},
|
||||
{"i": 121, "op": "shfl", "dst": 3, "src": 2, "src2": 5, "imm": "0xc0aaacdb", "imm2": "0xaaa7d229", "rot": 31, "bit": 29, "mask": 8},
|
||||
{"i": 122, "op": "rotl", "dst": 7, "src": 1, "src2": 3, "imm": "0xcce6a73f", "imm2": "0x517b6b86", "rot": 3, "bit": 16, "mask": 8},
|
||||
{"i": 123, "op": "rotr", "dst": 4, "src": 6, "src2": 0, "imm": "0x8e156aca", "imm2": "0x0a894b3e", "rot": 2, "bit": 12, "mask": 2},
|
||||
{"i": 124, "op": "shfl", "dst": 6, "src": 5, "src2": 2, "imm": "0xd88c8446", "imm2": "0xf49fedf3", "rot": 3, "bit": 23, "mask": 1},
|
||||
{"i": 125, "op": "rotl", "dst": 5, "src": 2, "src2": 0, "imm": "0xd10b27cf", "imm2": "0x2dde1faa", "rot": 21, "bit": 7, "mask": 2},
|
||||
{"i": 126, "op": "mad", "dst": 0, "src": 3, "src2": 1, "imm": "0x619e1818", "imm2": "0x62add310", "rot": 26, "bit": 1, "mask": 4},
|
||||
{"i": 127, "op": "add", "dst": 0, "src": 3, "src2": 6, "imm": "0xff14c08d", "imm2": "0xa7f6bb54", "rot": 25, "bit": 24, "mask": 4},
|
||||
{"i": 128, "op": "xor", "dst": 7, "src": 4, "src2": 4, "imm": "0x34224086", "imm2": "0x5e9f858e", "rot": 4, "bit": 16, "mask": 8},
|
||||
{"i": 129, "op": "mulhi", "dst": 3, "src": 5, "src2": 6, "imm": "0x89c3cbdd", "imm2": "0x4e80334c", "rot": 29, "bit": 4, "mask": 1},
|
||||
{"i": 130, "op": "shfl", "dst": 5, "src": 7, "src2": 1, "imm": "0x37d4c3e2", "imm2": "0xbd54378c", "rot": 14, "bit": 26, "mask": 8},
|
||||
{"i": 131, "op": "mulhi", "dst": 6, "src": 4, "src2": 3, "imm": "0x4f5fac7c", "imm2": "0x74070028", "rot": 5, "bit": 25, "mask": 8},
|
||||
{"i": 132, "op": "xor", "dst": 5, "src": 3, "src2": 0, "imm": "0x690eb003", "imm2": "0x8516a581", "rot": 20, "bit": 21, "mask": 2},
|
||||
{"i": 133, "op": "rotl", "dst": 0, "src": 5, "src2": 7, "imm": "0x42b31885", "imm2": "0x08249acb", "rot": 19, "bit": 6, "mask": 8},
|
||||
{"i": 134, "op": "rotr", "dst": 4, "src": 3, "src2": 3, "imm": "0x07d729ed", "imm2": "0xf564da8a", "rot": 17, "bit": 9, "mask": 16},
|
||||
{"i": 135, "op": "shfl", "dst": 1, "src": 2, "src2": 0, "imm": "0x19130ef2", "imm2": "0xf59815e5", "rot": 17, "bit": 12, "mask": 16},
|
||||
{"i": 136, "op": "sub", "dst": 5, "src": 2, "src2": 5, "imm": "0xdeb56168", "imm2": "0x871104c4", "rot": 13, "bit": 17, "mask": 16},
|
||||
{"i": 137, "op": "xor", "dst": 3, "src": 5, "src2": 4, "imm": "0x884fc7da", "imm2": "0x83a4a5bf", "rot": 23, "bit": 12, "mask": 4},
|
||||
{"i": 138, "op": "sub", "dst": 0, "src": 2, "src2": 6, "imm": "0x3e67a6fd", "imm2": "0x844d2039", "rot": 3, "bit": 24, "mask": 4},
|
||||
{"i": 139, "op": "mul", "dst": 3, "src": 7, "src2": 7, "imm": "0x965933b4", "imm2": "0xf37ef93d", "rot": 6, "bit": 9, "mask": 4},
|
||||
{"i": 140, "op": "add", "dst": 5, "src": 3, "src2": 3, "imm": "0x82ab98f4", "imm2": "0x2d465ca8", "rot": 11, "bit": 24, "mask": 1},
|
||||
{"i": 141, "op": "add", "dst": 4, "src": 1, "src2": 7, "imm": "0x80594390", "imm2": "0x237a9d8e", "rot": 14, "bit": 11, "mask": 8},
|
||||
{"i": 142, "op": "mad", "dst": 3, "src": 7, "src2": 0, "imm": "0xd37861a1", "imm2": "0x2fe70eae", "rot": 2, "bit": 11, "mask": 16},
|
||||
{"i": 143, "op": "or", "dst": 0, "src": 3, "src2": 5, "imm": "0x1a018e2c", "imm2": "0x301aa5f6", "rot": 15, "bit": 26, "mask": 1},
|
||||
{"i": 144, "op": "mulhi", "dst": 0, "src": 1, "src2": 5, "imm": "0x69d22dfc", "imm2": "0x4a90cb1e", "rot": 21, "bit": 9, "mask": 16},
|
||||
{"i": 145, "op": "xor", "dst": 6, "src": 2, "src2": 0, "imm": "0xf3f41f47", "imm2": "0xdf8d33e9", "rot": 22, "bit": 0, "mask": 8},
|
||||
{"i": 146, "op": "shfl", "dst": 7, "src": 1, "src2": 1, "imm": "0xf8a52377", "imm2": "0x0613ef41", "rot": 24, "bit": 12, "mask": 16},
|
||||
{"i": 147, "op": "add", "dst": 0, "src": 1, "src2": 3, "imm": "0x07f15148", "imm2": "0xe68cf57e", "rot": 2, "bit": 18, "mask": 4},
|
||||
{"i": 148, "op": "or", "dst": 7, "src": 3, "src2": 0, "imm": "0xbfd39b49", "imm2": "0xc7f6e97c", "rot": 13, "bit": 27, "mask": 2},
|
||||
{"i": 149, "op": "xor", "dst": 4, "src": 2, "src2": 7, "imm": "0x47409e43", "imm2": "0x04ef7db9", "rot": 8, "bit": 6, "mask": 4},
|
||||
{"i": 150, "op": "xor", "dst": 7, "src": 1, "src2": 1, "imm": "0x5d42a156", "imm2": "0xc2f77c48", "rot": 27, "bit": 17, "mask": 4},
|
||||
{"i": 151, "op": "rotr", "dst": 5, "src": 2, "src2": 7, "imm": "0x7c170030", "imm2": "0xa3102916", "rot": 5, "bit": 6, "mask": 1},
|
||||
{"i": 152, "op": "mul", "dst": 0, "src": 7, "src2": 5, "imm": "0x16d009af", "imm2": "0x2a89533b", "rot": 26, "bit": 8, "mask": 4},
|
||||
{"i": 153, "op": "rotl", "dst": 4, "src": 5, "src2": 0, "imm": "0x6e54eb71", "imm2": "0xe2d27256", "rot": 16, "bit": 18, "mask": 1},
|
||||
{"i": 154, "op": "rotr", "dst": 6, "src": 4, "src2": 5, "imm": "0x128f24a4", "imm2": "0x59888463", "rot": 15, "bit": 17, "mask": 16},
|
||||
{"i": 155, "op": "xor", "dst": 6, "src": 1, "src2": 2, "imm": "0x34f276ff", "imm2": "0xef5086b4", "rot": 2, "bit": 10, "mask": 4},
|
||||
{"i": 156, "op": "or", "dst": 6, "src": 3, "src2": 4, "imm": "0x16cc2c2f", "imm2": "0xfae5a09b", "rot": 12, "bit": 31, "mask": 4},
|
||||
{"i": 157, "op": "add", "dst": 1, "src": 6, "src2": 7, "imm": "0x14878c5a", "imm2": "0x84c2a09d", "rot": 30, "bit": 19, "mask": 8},
|
||||
{"i": 158, "op": "sub", "dst": 4, "src": 1, "src2": 0, "imm": "0xfe18e4b0", "imm2": "0x239df0cc", "rot": 14, "bit": 20, "mask": 2},
|
||||
{"i": 159, "op": "mad", "dst": 4, "src": 6, "src2": 7, "imm": "0xbdfcc4e0", "imm2": "0x0870dee8", "rot": 1, "bit": 7, "mask": 8},
|
||||
{"i": 160, "op": "mul", "dst": 1, "src": 5, "src2": 1, "imm": "0xc837b43d", "imm2": "0x027ac950", "rot": 4, "bit": 22, "mask": 16},
|
||||
{"i": 161, "op": "or", "dst": 4, "src": 0, "src2": 2, "imm": "0x9c28f8f6", "imm2": "0x3db71fe6", "rot": 11, "bit": 6, "mask": 16},
|
||||
{"i": 162, "op": "rotl", "dst": 7, "src": 3, "src2": 6, "imm": "0xb6e56f55", "imm2": "0x5fcbd27c", "rot": 21, "bit": 10, "mask": 1},
|
||||
{"i": 163, "op": "sub", "dst": 0, "src": 1, "src2": 4, "imm": "0x8d95ae70", "imm2": "0xd4f10284", "rot": 15, "bit": 12, "mask": 8},
|
||||
{"i": 164, "op": "mul", "dst": 1, "src": 0, "src2": 6, "imm": "0xa9bdf99e", "imm2": "0x1f3dceb6", "rot": 21, "bit": 20, "mask": 16},
|
||||
{"i": 165, "op": "sub", "dst": 2, "src": 1, "src2": 4, "imm": "0xda245058", "imm2": "0x2038e7f6", "rot": 30, "bit": 11, "mask": 2},
|
||||
{"i": 166, "op": "shfl", "dst": 0, "src": 6, "src2": 3, "imm": "0xbbc8eee8", "imm2": "0xa73472fb", "rot": 23, "bit": 4, "mask": 2},
|
||||
{"i": 167, "op": "sub", "dst": 0, "src": 4, "src2": 2, "imm": "0x43540d69", "imm2": "0xb3f116ca", "rot": 28, "bit": 2, "mask": 2},
|
||||
{"i": 168, "op": "rotr", "dst": 1, "src": 6, "src2": 6, "imm": "0x57cea752", "imm2": "0xc1136e71", "rot": 25, "bit": 3, "mask": 8},
|
||||
{"i": 169, "op": "rotr", "dst": 7, "src": 1, "src2": 5, "imm": "0x4550a465", "imm2": "0x14362935", "rot": 25, "bit": 9, "mask": 16},
|
||||
{"i": 170, "op": "mad", "dst": 3, "src": 1, "src2": 3, "imm": "0x8c961d80", "imm2": "0x7de2e1db", "rot": 10, "bit": 6, "mask": 4},
|
||||
{"i": 171, "op": "shfl", "dst": 2, "src": 0, "src2": 0, "imm": "0xda047fca", "imm2": "0x459e17ab", "rot": 22, "bit": 26, "mask": 4},
|
||||
{"i": 172, "op": "sub", "dst": 2, "src": 1, "src2": 2, "imm": "0xbcc341fd", "imm2": "0x17a3b229", "rot": 8, "bit": 25, "mask": 8},
|
||||
{"i": 173, "op": "xor", "dst": 7, "src": 6, "src2": 7, "imm": "0x2302b424", "imm2": "0x78780982", "rot": 23, "bit": 7, "mask": 8},
|
||||
{"i": 174, "op": "rotr", "dst": 4, "src": 1, "src2": 2, "imm": "0xcad6fb5f", "imm2": "0x1807628e", "rot": 8, "bit": 21, "mask": 4},
|
||||
{"i": 175, "op": "shfl", "dst": 4, "src": 7, "src2": 7, "imm": "0xf5057e8e", "imm2": "0xd14d6cc5", "rot": 3, "bit": 12, "mask": 2},
|
||||
{"i": 176, "op": "mul", "dst": 6, "src": 2, "src2": 1, "imm": "0xc2926356", "imm2": "0xae74dc32", "rot": 1, "bit": 21, "mask": 8},
|
||||
{"i": 177, "op": "shfl", "dst": 0, "src": 2, "src2": 7, "imm": "0x8d94d98e", "imm2": "0x7bd6f4dc", "rot": 27, "bit": 14, "mask": 4},
|
||||
{"i": 178, "op": "add", "dst": 6, "src": 4, "src2": 6, "imm": "0x4fa43965", "imm2": "0x509871e6", "rot": 28, "bit": 21, "mask": 4},
|
||||
{"i": 179, "op": "xor", "dst": 0, "src": 6, "src2": 6, "imm": "0xe278fe6c", "imm2": "0x75d8a63f", "rot": 21, "bit": 15, "mask": 2},
|
||||
{"i": 180, "op": "add", "dst": 5, "src": 7, "src2": 7, "imm": "0xfb3c4bd8", "imm2": "0xb95b8a53", "rot": 22, "bit": 0, "mask": 4},
|
||||
{"i": 181, "op": "mul", "dst": 6, "src": 3, "src2": 2, "imm": "0x5a1347ec", "imm2": "0x3ccc5389", "rot": 27, "bit": 26, "mask": 4},
|
||||
{"i": 182, "op": "rotl", "dst": 7, "src": 4, "src2": 7, "imm": "0xdc6db2f7", "imm2": "0x7cdff0c6", "rot": 1, "bit": 2, "mask": 4},
|
||||
{"i": 183, "op": "add", "dst": 0, "src": 7, "src2": 7, "imm": "0xc448a197", "imm2": "0x7047b7cf", "rot": 24, "bit": 11, "mask": 8},
|
||||
{"i": 184, "op": "mul", "dst": 3, "src": 5, "src2": 0, "imm": "0xa5c0e492", "imm2": "0xaf7d8a86", "rot": 11, "bit": 16, "mask": 8},
|
||||
{"i": 185, "op": "shfl", "dst": 5, "src": 7, "src2": 1, "imm": "0x5965794b", "imm2": "0x36ddca6b", "rot": 7, "bit": 25, "mask": 1},
|
||||
{"i": 186, "op": "mul", "dst": 0, "src": 2, "src2": 5, "imm": "0xdfdbcc0f", "imm2": "0xac7b7091", "rot": 11, "bit": 31, "mask": 8},
|
||||
{"i": 187, "op": "mulhi", "dst": 4, "src": 7, "src2": 5, "imm": "0x3d8999e9", "imm2": "0xf1848db8", "rot": 15, "bit": 21, "mask": 8},
|
||||
{"i": 188, "op": "mad", "dst": 2, "src": 7, "src2": 2, "imm": "0xda3af48c", "imm2": "0x8e33a97c", "rot": 6, "bit": 30, "mask": 4},
|
||||
{"i": 189, "op": "sub", "dst": 7, "src": 2, "src2": 4, "imm": "0xa32da8c1", "imm2": "0xf7486d06", "rot": 25, "bit": 16, "mask": 4},
|
||||
{"i": 190, "op": "add", "dst": 2, "src": 5, "src2": 1, "imm": "0x2fceaf49", "imm2": "0x5caccc5d", "rot": 24, "bit": 19, "mask": 2},
|
||||
{"i": 191, "op": "shfl", "dst": 1, "src": 2, "src2": 2, "imm": "0x2a71f5e6", "imm2": "0xb9b51e89", "rot": 31, "bit": 25, "mask": 1},
|
||||
{"i": 192, "op": "add", "dst": 7, "src": 1, "src2": 7, "imm": "0xba0cee62", "imm2": "0x857bb5fe", "rot": 4, "bit": 20, "mask": 2},
|
||||
{"i": 193, "op": "rotl", "dst": 6, "src": 3, "src2": 3, "imm": "0xc7fe35cf", "imm2": "0x515c1201", "rot": 22, "bit": 7, "mask": 16},
|
||||
{"i": 194, "op": "xor", "dst": 3, "src": 7, "src2": 4, "imm": "0xe496d1f1", "imm2": "0x11046e3e", "rot": 16, "bit": 9, "mask": 4},
|
||||
{"i": 195, "op": "mul", "dst": 7, "src": 0, "src2": 2, "imm": "0xc7d5c4c2", "imm2": "0x077e940b", "rot": 23, "bit": 29, "mask": 4},
|
||||
{"i": 196, "op": "xor", "dst": 3, "src": 5, "src2": 3, "imm": "0xbe6db14a", "imm2": "0xf736da5a", "rot": 19, "bit": 6, "mask": 2},
|
||||
{"i": 197, "op": "add", "dst": 5, "src": 1, "src2": 5, "imm": "0x8519428c", "imm2": "0xeae84577", "rot": 18, "bit": 29, "mask": 1},
|
||||
{"i": 198, "op": "mad", "dst": 7, "src": 4, "src2": 4, "imm": "0xb272a029", "imm2": "0x071f6a5c", "rot": 5, "bit": 28, "mask": 8},
|
||||
{"i": 199, "op": "add", "dst": 3, "src": 6, "src2": 3, "imm": "0x468639d3", "imm2": "0x0bfdbfa1", "rot": 20, "bit": 8, "mask": 16},
|
||||
{"i": 200, "op": "mad", "dst": 5, "src": 3, "src2": 4, "imm": "0xffba18b7", "imm2": "0xbdcf9e81", "rot": 13, "bit": 31, "mask": 8},
|
||||
{"i": 201, "op": "shfl", "dst": 2, "src": 7, "src2": 7, "imm": "0xb7533191", "imm2": "0x8d58cc8c", "rot": 7, "bit": 29, "mask": 4},
|
||||
{"i": 202, "op": "add", "dst": 7, "src": 4, "src2": 6, "imm": "0x18ec9a69", "imm2": "0x53ee8e50", "rot": 12, "bit": 23, "mask": 16},
|
||||
{"i": 203, "op": "xor", "dst": 6, "src": 0, "src2": 5, "imm": "0x12011ff1", "imm2": "0xfbb57ac6", "rot": 16, "bit": 25, "mask": 16},
|
||||
{"i": 204, "op": "add", "dst": 4, "src": 5, "src2": 3, "imm": "0x3e81485b", "imm2": "0xac63376e", "rot": 1, "bit": 4, "mask": 1},
|
||||
{"i": 205, "op": "mad", "dst": 7, "src": 1, "src2": 4, "imm": "0xeb9eee5c", "imm2": "0x3f1daecd", "rot": 19, "bit": 29, "mask": 4},
|
||||
{"i": 206, "op": "shfl", "dst": 1, "src": 5, "src2": 2, "imm": "0x4fc50d06", "imm2": "0xb5533b31", "rot": 22, "bit": 27, "mask": 16},
|
||||
{"i": 207, "op": "mul", "dst": 4, "src": 0, "src2": 5, "imm": "0x202d6001", "imm2": "0x50388f85", "rot": 5, "bit": 18, "mask": 16},
|
||||
{"i": 208, "op": "or", "dst": 1, "src": 6, "src2": 7, "imm": "0x8743c0d4", "imm2": "0x9df69539", "rot": 14, "bit": 16, "mask": 1},
|
||||
{"i": 209, "op": "add", "dst": 6, "src": 4, "src2": 6, "imm": "0x66ccb75d", "imm2": "0x5b745519", "rot": 28, "bit": 28, "mask": 8},
|
||||
{"i": 210, "op": "rotr", "dst": 1, "src": 3, "src2": 1, "imm": "0xe9933132", "imm2": "0x6702fe85", "rot": 16, "bit": 19, "mask": 1},
|
||||
{"i": 211, "op": "sub", "dst": 5, "src": 4, "src2": 4, "imm": "0xfe04b942", "imm2": "0x40dd2ce8", "rot": 24, "bit": 8, "mask": 4},
|
||||
{"i": 212, "op": "xor", "dst": 4, "src": 7, "src2": 7, "imm": "0x10827878", "imm2": "0xef0de8fc", "rot": 17, "bit": 11, "mask": 8},
|
||||
{"i": 213, "op": "add", "dst": 1, "src": 7, "src2": 1, "imm": "0x0ce0553d", "imm2": "0x1c3dccf7", "rot": 26, "bit": 8, "mask": 1},
|
||||
{"i": 214, "op": "mulhi", "dst": 0, "src": 1, "src2": 1, "imm": "0xa71d7581", "imm2": "0xe567c74c", "rot": 2, "bit": 29, "mask": 2},
|
||||
{"i": 215, "op": "add", "dst": 3, "src": 6, "src2": 2, "imm": "0x96bfec89", "imm2": "0xfadc9205", "rot": 30, "bit": 9, "mask": 16},
|
||||
{"i": 216, "op": "rotl", "dst": 0, "src": 2, "src2": 5, "imm": "0x947ca474", "imm2": "0xd2a1b650", "rot": 4, "bit": 1, "mask": 1},
|
||||
{"i": 217, "op": "xor", "dst": 6, "src": 4, "src2": 5, "imm": "0x0b7617de", "imm2": "0xc6c0a9a2", "rot": 31, "bit": 15, "mask": 2},
|
||||
{"i": 218, "op": "shfl", "dst": 6, "src": 7, "src2": 1, "imm": "0x738f0a8f", "imm2": "0xfc868560", "rot": 2, "bit": 20, "mask": 2},
|
||||
{"i": 219, "op": "mad", "dst": 7, "src": 0, "src2": 1, "imm": "0x2960f2bc", "imm2": "0x7e9ef9b7", "rot": 21, "bit": 4, "mask": 16},
|
||||
{"i": 220, "op": "mul", "dst": 4, "src": 0, "src2": 2, "imm": "0x6014d800", "imm2": "0xd19ac7d1", "rot": 1, "bit": 13, "mask": 1},
|
||||
{"i": 221, "op": "mad", "dst": 2, "src": 6, "src2": 0, "imm": "0x4b558fc2", "imm2": "0x2d48bd3c", "rot": 18, "bit": 31, "mask": 2},
|
||||
{"i": 222, "op": "xor", "dst": 5, "src": 4, "src2": 6, "imm": "0x0260f3ab", "imm2": "0x8fce22f1", "rot": 21, "bit": 26, "mask": 16},
|
||||
{"i": 223, "op": "mul", "dst": 0, "src": 5, "src2": 0, "imm": "0xe07ab58f", "imm2": "0x5349a41d", "rot": 20, "bit": 3, "mask": 1},
|
||||
{"i": 224, "op": "add", "dst": 2, "src": 0, "src2": 5, "imm": "0x61495681", "imm2": "0x66b15c04", "rot": 24, "bit": 9, "mask": 16},
|
||||
{"i": 225, "op": "rotr", "dst": 3, "src": 2, "src2": 5, "imm": "0x291268bd", "imm2": "0xf046aa79", "rot": 27, "bit": 15, "mask": 4},
|
||||
{"i": 226, "op": "mad", "dst": 2, "src": 3, "src2": 3, "imm": "0xfd6dfa6c", "imm2": "0x2c0db93d", "rot": 29, "bit": 2, "mask": 16},
|
||||
{"i": 227, "op": "xor", "dst": 6, "src": 0, "src2": 4, "imm": "0xd4ae1ee2", "imm2": "0xe6a67972", "rot": 26, "bit": 1, "mask": 4},
|
||||
{"i": 228, "op": "rotl", "dst": 4, "src": 7, "src2": 3, "imm": "0xf6627a9d", "imm2": "0x99078da3", "rot": 30, "bit": 5, "mask": 16},
|
||||
{"i": 229, "op": "add", "dst": 2, "src": 3, "src2": 7, "imm": "0x70d05c34", "imm2": "0x25bc17c2", "rot": 20, "bit": 29, "mask": 4},
|
||||
{"i": 230, "op": "rotr", "dst": 1, "src": 5, "src2": 2, "imm": "0xef874bea", "imm2": "0x3bfbfecc", "rot": 17, "bit": 2, "mask": 2},
|
||||
{"i": 231, "op": "add", "dst": 1, "src": 4, "src2": 5, "imm": "0x32a28384", "imm2": "0x2d8728f3", "rot": 1, "bit": 8, "mask": 8},
|
||||
{"i": 232, "op": "rotl", "dst": 1, "src": 6, "src2": 3, "imm": "0x877c54ab", "imm2": "0x1d5e5014", "rot": 6, "bit": 24, "mask": 2},
|
||||
{"i": 233, "op": "add", "dst": 2, "src": 0, "src2": 5, "imm": "0xcf0949f1", "imm2": "0x61e81fe6", "rot": 2, "bit": 21, "mask": 16},
|
||||
{"i": 234, "op": "rotl", "dst": 4, "src": 5, "src2": 5, "imm": "0x5017c32c", "imm2": "0x1ab34dfe", "rot": 2, "bit": 1, "mask": 2},
|
||||
{"i": 235, "op": "sub", "dst": 2, "src": 6, "src2": 2, "imm": "0xadbdb811", "imm2": "0x2c2d4f2b", "rot": 30, "bit": 29, "mask": 8},
|
||||
{"i": 236, "op": "or", "dst": 6, "src": 3, "src2": 4, "imm": "0x700e204f", "imm2": "0x27294e7b", "rot": 13, "bit": 0, "mask": 4},
|
||||
{"i": 237, "op": "shfl", "dst": 7, "src": 5, "src2": 4, "imm": "0xc39f71ba", "imm2": "0x9a90263d", "rot": 28, "bit": 9, "mask": 4},
|
||||
{"i": 238, "op": "xor", "dst": 1, "src": 3, "src2": 6, "imm": "0x9f84a5b3", "imm2": "0x95691d15", "rot": 14, "bit": 22, "mask": 1},
|
||||
{"i": 239, "op": "rotl", "dst": 2, "src": 4, "src2": 0, "imm": "0xa3b79ed5", "imm2": "0x05f568bc", "rot": 1, "bit": 15, "mask": 4},
|
||||
{"i": 240, "op": "rotr", "dst": 5, "src": 2, "src2": 7, "imm": "0xaaabbb0d", "imm2": "0x71ce4cd6", "rot": 2, "bit": 26, "mask": 4},
|
||||
{"i": 241, "op": "rotl", "dst": 4, "src": 1, "src2": 7, "imm": "0x8146fbd3", "imm2": "0xde65fc31", "rot": 10, "bit": 9, "mask": 4},
|
||||
{"i": 242, "op": "add", "dst": 6, "src": 2, "src2": 7, "imm": "0xe204fb50", "imm2": "0x0a67565d", "rot": 5, "bit": 16, "mask": 1},
|
||||
{"i": 243, "op": "shfl", "dst": 1, "src": 3, "src2": 7, "imm": "0xcb7a3c1b", "imm2": "0x863f002c", "rot": 18, "bit": 1, "mask": 2},
|
||||
{"i": 244, "op": "shfl", "dst": 0, "src": 1, "src2": 3, "imm": "0x92eb030c", "imm2": "0x3cff4c6d", "rot": 1, "bit": 22, "mask": 16},
|
||||
{"i": 245, "op": "shfl", "dst": 5, "src": 7, "src2": 6, "imm": "0x046c969c", "imm2": "0xd21d7e76", "rot": 5, "bit": 31, "mask": 2},
|
||||
{"i": 246, "op": "xor", "dst": 4, "src": 5, "src2": 1, "imm": "0x2e3015f0", "imm2": "0x1b305a5c", "rot": 22, "bit": 16, "mask": 8},
|
||||
{"i": 247, "op": "xor", "dst": 7, "src": 2, "src2": 0, "imm": "0x8238cf71", "imm2": "0xb7874d94", "rot": 9, "bit": 1, "mask": 2},
|
||||
{"i": 248, "op": "shfl", "dst": 0, "src": 6, "src2": 7, "imm": "0x437cbc99", "imm2": "0x7fb41fed", "rot": 17, "bit": 1, "mask": 16},
|
||||
{"i": 249, "op": "mulhi", "dst": 1, "src": 3, "src2": 2, "imm": "0xeb293d88", "imm2": "0xb92f5968", "rot": 3, "bit": 20, "mask": 2},
|
||||
{"i": 250, "op": "add", "dst": 7, "src": 5, "src2": 3, "imm": "0x3d6daffd", "imm2": "0x5a79fd6f", "rot": 6, "bit": 31, "mask": 2},
|
||||
{"i": 251, "op": "add", "dst": 6, "src": 0, "src2": 3, "imm": "0x84334aea", "imm2": "0x6beb28c2", "rot": 27, "bit": 28, "mask": 2},
|
||||
{"i": 252, "op": "sub", "dst": 7, "src": 1, "src2": 6, "imm": "0xe6177638", "imm2": "0x814d3fc5", "rot": 25, "bit": 2, "mask": 2},
|
||||
{"i": 253, "op": "rotr", "dst": 5, "src": 2, "src2": 7, "imm": "0x95560bac", "imm2": "0xbf8b01a9", "rot": 14, "bit": 5, "mask": 16},
|
||||
{"i": 254, "op": "mulhi", "dst": 2, "src": 5, "src2": 1, "imm": "0x318956e5", "imm2": "0x7c20b373", "rot": 13, "bit": 30, "mask": 2},
|
||||
{"i": 255, "op": "mul", "dst": 6, "src": 3, "src2": 3, "imm": "0xe1998e49", "imm2": "0x59353e5a", "rot": 15, "bit": 2, "mask": 4}
|
||||
]},
|
||||
"instructions": [
|
||||
{"i": 0, "op": "sub", "dst": 6, "src": 3, "src2": 3, "imm": "0x77dfbe60", "imm2": "0x404c8b9c", "rot": 24, "bit": 8, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 1, "op": "or", "dst": 2, "src": 6, "src2": 7, "imm": "0xb2e79058", "imm2": "0xc2485073", "rot": 21, "bit": 23, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 2, "op": "add", "dst": 4, "src": 0, "src2": 6, "imm": "0x1a579b38", "imm2": "0x7731324b", "rot": 6, "bit": 4, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 3, "op": "load", "dst": 3, "src": 6, "src2": 0, "imm": "0x683f5bc5", "imm2": "0xc8a218fe", "rot": 25, "bit": 4, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 4, "op": "shfl", "dst": 0, "src": 6, "src2": 4, "imm": "0x4d7cdcbb", "imm2": "0x1a5bf1a4", "rot": 8, "bit": 28, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 5, "op": "xor", "dst": 4, "src": 1, "src2": 0, "imm": "0xa2b86827", "imm2": "0x94a44439", "rot": 21, "bit": 22, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 6, "op": "mad", "dst": 3, "src": 5, "src2": 3, "imm": "0x58a4ea3f", "imm2": "0x467879fa", "rot": 6, "bit": 31, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 7, "op": "mul", "dst": 1, "src": 6, "src2": 1, "imm": "0x31fcc19c", "imm2": "0x176cb88f", "rot": 1, "bit": 11, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 8, "op": "load", "dst": 2, "src": 0, "src2": 3, "imm": "0x89b12386", "imm2": "0x661ed5e5", "rot": 11, "bit": 13, "mask": 2, "width": 1, "win": 2, "off": 2},
|
||||
{"i": 9, "op": "xor", "dst": 0, "src": 4, "src2": 1, "imm": "0x69150105", "imm2": "0x8e1f04b9", "rot": 3, "bit": 30, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 10, "op": "add", "dst": 0, "src": 7, "src2": 5, "imm": "0xee083919", "imm2": "0xb10fcef8", "rot": 9, "bit": 30, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 11, "op": "mulhi", "dst": 0, "src": 2, "src2": 3, "imm": "0x99ac7c83", "imm2": "0x34eb2845", "rot": 18, "bit": 12, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 12, "op": "xor", "dst": 5, "src": 4, "src2": 0, "imm": "0x97cfc887", "imm2": "0x0458f053", "rot": 28, "bit": 24, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 13, "op": "mad", "dst": 7, "src": 3, "src2": 0, "imm": "0x503e87b4", "imm2": "0x2c5d2617", "rot": 3, "bit": 15, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 14, "op": "load", "dst": 3, "src": 4, "src2": 4, "imm": "0xfae98035", "imm2": "0x6c127281", "rot": 19, "bit": 29, "mask": 4, "width": 1, "win": 1, "off": 0},
|
||||
{"i": 15, "op": "load", "dst": 2, "src": 5, "src2": 2, "imm": "0x67601ed5", "imm2": "0x3c546ed1", "rot": 28, "bit": 28, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 16, "op": "shfl", "dst": 7, "src": 5, "src2": 6, "imm": "0x7789be79", "imm2": "0x87496b8e", "rot": 22, "bit": 2, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 17, "op": "mul", "dst": 7, "src": 6, "src2": 2, "imm": "0x559099bb", "imm2": "0x6a72e1a6", "rot": 1, "bit": 11, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 18, "op": "mad", "dst": 2, "src": 3, "src2": 1, "imm": "0x4f5bae5e", "imm2": "0x4c8c5762", "rot": 27, "bit": 31, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 19, "op": "rotr", "dst": 5, "src": 2, "src2": 4, "imm": "0x5dc5e6bf", "imm2": "0x36bc207d", "rot": 9, "bit": 4, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 20, "op": "load", "dst": 5, "src": 3, "src2": 7, "imm": "0x4a278885", "imm2": "0xcd1044b0", "rot": 30, "bit": 0, "mask": 4, "width": 1, "win": 1, "off": 1},
|
||||
{"i": 21, "op": "shfl", "dst": 6, "src": 4, "src2": 3, "imm": "0x48b94ffc", "imm2": "0xd8e14d80", "rot": 2, "bit": 31, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 22, "op": "mulhi", "dst": 1, "src": 2, "src2": 7, "imm": "0x47bd8492", "imm2": "0x28fa8d0b", "rot": 19, "bit": 26, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 23, "op": "xor", "dst": 6, "src": 5, "src2": 1, "imm": "0x4ed6baf4", "imm2": "0xa8c31f74", "rot": 30, "bit": 2, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 24, "op": "mad", "dst": 5, "src": 2, "src2": 0, "imm": "0x6174747d", "imm2": "0xfc912524", "rot": 18, "bit": 29, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 25, "op": "mad", "dst": 6, "src": 3, "src2": 7, "imm": "0x113a6442", "imm2": "0xbe52ec57", "rot": 15, "bit": 15, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 26, "op": "load", "dst": 6, "src": 5, "src2": 1, "imm": "0x9d3eb1cc", "imm2": "0x375c73bb", "rot": 29, "bit": 7, "mask": 8, "width": 1, "win": 1, "off": 1},
|
||||
{"i": 27, "op": "mad", "dst": 1, "src": 5, "src2": 5, "imm": "0x150088c3", "imm2": "0x7f17e789", "rot": 8, "bit": 25, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 28, "op": "load", "dst": 0, "src": 6, "src2": 3, "imm": "0x4b412223", "imm2": "0x9192ceeb", "rot": 16, "bit": 28, "mask": 1, "width": 1, "win": 1, "off": 1},
|
||||
{"i": 29, "op": "add", "dst": 7, "src": 4, "src2": 6, "imm": "0xd5e6c37c", "imm2": "0x0b1f02c7", "rot": 2, "bit": 7, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 30, "op": "add", "dst": 7, "src": 0, "src2": 6, "imm": "0x4e45ba51", "imm2": "0x2d6e8bd1", "rot": 7, "bit": 0, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 31, "op": "xor", "dst": 0, "src": 2, "src2": 7, "imm": "0xef601d89", "imm2": "0x805402db", "rot": 7, "bit": 28, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 32, "op": "add", "dst": 6, "src": 2, "src2": 3, "imm": "0x1c91b2a1", "imm2": "0x351422d2", "rot": 1, "bit": 13, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 33, "op": "mulhi", "dst": 4, "src": 5, "src2": 7, "imm": "0x7cf21567", "imm2": "0x1075ee0d", "rot": 26, "bit": 17, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 34, "op": "rotr", "dst": 3, "src": 5, "src2": 6, "imm": "0x60d8f191", "imm2": "0x40523707", "rot": 3, "bit": 10, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 35, "op": "load", "dst": 3, "src": 0, "src2": 5, "imm": "0xb1dc3f6e", "imm2": "0x74440531", "rot": 21, "bit": 10, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 36, "op": "rotl", "dst": 4, "src": 6, "src2": 4, "imm": "0xf0d17569", "imm2": "0xd281fd01", "rot": 13, "bit": 4, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 37, "op": "add", "dst": 6, "src": 1, "src2": 6, "imm": "0x3f2970f7", "imm2": "0x64a28251", "rot": 4, "bit": 7, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 38, "op": "mul", "dst": 2, "src": 0, "src2": 3, "imm": "0x6463770c", "imm2": "0xb85d603f", "rot": 27, "bit": 1, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 39, "op": "add", "dst": 3, "src": 4, "src2": 3, "imm": "0x3b7b3317", "imm2": "0x7378c955", "rot": 28, "bit": 15, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 40, "op": "load", "dst": 0, "src": 7, "src2": 3, "imm": "0xd0fb5982", "imm2": "0x2e4d31a8", "rot": 25, "bit": 27, "mask": 4, "width": 1, "win": 2, "off": 3},
|
||||
{"i": 41, "op": "xor", "dst": 6, "src": 5, "src2": 0, "imm": "0xe96a9e6e", "imm2": "0x8d73cc3e", "rot": 17, "bit": 11, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 42, "op": "add", "dst": 3, "src": 0, "src2": 6, "imm": "0x6678b059", "imm2": "0xb31f6a64", "rot": 11, "bit": 2, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 43, "op": "load", "dst": 7, "src": 3, "src2": 2, "imm": "0x65a841c5", "imm2": "0x6aae1403", "rot": 9, "bit": 9, "mask": 1, "width": 1, "win": 1, "off": 0},
|
||||
{"i": 44, "op": "shfl", "dst": 3, "src": 7, "src2": 6, "imm": "0x650ba38d", "imm2": "0x07218508", "rot": 12, "bit": 13, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 45, "op": "mulhi", "dst": 1, "src": 5, "src2": 7, "imm": "0x18736788", "imm2": "0x9f820eed", "rot": 13, "bit": 29, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 46, "op": "rotl", "dst": 0, "src": 1, "src2": 5, "imm": "0x48deba75", "imm2": "0xefc508fa", "rot": 6, "bit": 6, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 47, "op": "load", "dst": 1, "src": 7, "src2": 4, "imm": "0x28a04384", "imm2": "0xc67cb0ac", "rot": 5, "bit": 5, "mask": 2, "width": 1, "win": 1, "off": 1},
|
||||
{"i": 48, "op": "add", "dst": 4, "src": 0, "src2": 5, "imm": "0x43095946", "imm2": "0x5feccee6", "rot": 22, "bit": 17, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 49, "op": "load", "dst": 2, "src": 0, "src2": 2, "imm": "0x3e78cdd8", "imm2": "0x7c170311", "rot": 10, "bit": 5, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 50, "op": "sub", "dst": 2, "src": 3, "src2": 7, "imm": "0xaff0cdb2", "imm2": "0xde0b7ee0", "rot": 21, "bit": 11, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 51, "op": "add", "dst": 7, "src": 2, "src2": 1, "imm": "0xf45ecdf8", "imm2": "0x26e2b582", "rot": 28, "bit": 24, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 52, "op": "load", "dst": 5, "src": 7, "src2": 7, "imm": "0x67353e69", "imm2": "0xb2d56082", "rot": 27, "bit": 14, "mask": 2, "width": 1, "win": 2, "off": 3},
|
||||
{"i": 53, "op": "load", "dst": 6, "src": 3, "src2": 1, "imm": "0x5a58c787", "imm2": "0x53a0467b", "rot": 6, "bit": 16, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 54, "op": "shfl", "dst": 7, "src": 5, "src2": 4, "imm": "0x3c237ad5", "imm2": "0xf2681187", "rot": 14, "bit": 26, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 55, "op": "rotr", "dst": 1, "src": 2, "src2": 3, "imm": "0x1c25c077", "imm2": "0x125de1ff", "rot": 13, "bit": 13, "mask": 2, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 56, "op": "shfl", "dst": 0, "src": 5, "src2": 0, "imm": "0x6465afe2", "imm2": "0x23b41d56", "rot": 31, "bit": 28, "mask": 16, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 57, "op": "rotl", "dst": 5, "src": 0, "src2": 2, "imm": "0xa427ff6e", "imm2": "0x014a0deb", "rot": 12, "bit": 28, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 58, "op": "add", "dst": 5, "src": 3, "src2": 7, "imm": "0x6b4f35e8", "imm2": "0x473b1718", "rot": 25, "bit": 4, "mask": 8, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 59, "op": "rotr", "dst": 4, "src": 1, "src2": 3, "imm": "0x66cc96ef", "imm2": "0xcb8c8e56", "rot": 7, "bit": 6, "mask": 4, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 60, "op": "shfl", "dst": 4, "src": 5, "src2": 5, "imm": "0x461bbf9f", "imm2": "0xc1ffd350", "rot": 19, "bit": 2, "mask": 1, "width": 1, "win": 0, "off": 0},
|
||||
{"i": 61, "op": "load", "dst": 7, "src": 0, "src2": 1, "imm": "0x4d893571", "imm2": "0x5583e644", "rot": 19, "bit": 19, "mask": 8, "width": 1, "win": 1, "off": 1},
|
||||
{"i": 62, "op": "load", "dst": 5, "src": 7, "src2": 3, "imm": "0x7a9049b4", "imm2": "0xbb7ebdf8", "rot": 15, "bit": 11, "mask": 8, "width": 1, "win": 1, "off": 1},
|
||||
{"i": 63, "op": "mul", "dst": 4, "src": 1, "src2": 7, "imm": "0xe7ae8a83", "imm2": "0x1ded64e8", "rot": 21, "bit": 1, "mask": 8, "width": 1, "win": 0, "off": 0}
|
||||
]
|
||||
}
|
||||
368
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program.metal
Normal file
368
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program.metal
Normal file
|
|
@ -0,0 +1,368 @@
|
|||
#include <metal_stdlib>
|
||||
using namespace metal;
|
||||
|
||||
#define MASK 0x0fffffffu
|
||||
constant uint SEEDW[8] = { 0xbe8c5a0cu, 0x7646e626u, 0x508e29d0u, 0xfb8a5b4au, 0x53c50955u, 0x685d62fcu, 0x6065c013u, 0x881096ebu };
|
||||
|
||||
inline uint splitmix32(uint x) {
|
||||
x ^= x >> 16; x *= 0x7feb352du;
|
||||
x ^= x >> 15; x *= 0x846ca68bu;
|
||||
x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
inline uint rotl_imm(uint x, uint n) { return (x << n) | (x >> (32u - n)); } // n in 1..31
|
||||
inline uint rotr_var(uint x, uint n) { n &= 31u; return (x >> n) | (x << ((32u - n) & 31u)); }
|
||||
inline uint ds_elem(uint i, uint d0, uint d1) {
|
||||
uint x = i ^ d0;
|
||||
x *= 0x9E3779B1u; x ^= x >> 15;
|
||||
x += d1;
|
||||
x *= 0x85EBCA77u; x ^= x >> 13;
|
||||
x *= 0xC2B2AE3Du; x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
|
||||
kernel void igneum_hash(device const uint* dataset [[buffer(0)]],
|
||||
device ulong* out [[buffer(1)]],
|
||||
constant uint& baseNonce [[buffer(2)]],
|
||||
uint gid [[thread_position_in_grid]]) {
|
||||
uint nonce = baseNonce + gid;
|
||||
uint r0, r1, r2, r3, r4, r5, r6, r7;
|
||||
{ uint x = nonce ^ SEEDW[0]; x += 0x9e3779b9u * 1u; x = splitmix32(x); r0 = x ^ SEEDW[1]; }
|
||||
{ uint x = nonce ^ SEEDW[1]; x += 0x9e3779b9u * 2u; x = splitmix32(x); r1 = x ^ SEEDW[2]; }
|
||||
{ uint x = nonce ^ SEEDW[2]; x += 0x9e3779b9u * 3u; x = splitmix32(x); r2 = x ^ SEEDW[3]; }
|
||||
{ uint x = nonce ^ SEEDW[3]; x += 0x9e3779b9u * 4u; x = splitmix32(x); r3 = x ^ SEEDW[4]; }
|
||||
{ uint x = nonce ^ SEEDW[4]; x += 0x9e3779b9u * 5u; x = splitmix32(x); r4 = x ^ SEEDW[5]; }
|
||||
{ uint x = nonce ^ SEEDW[5]; x += 0x9e3779b9u * 6u; x = splitmix32(x); r5 = x ^ SEEDW[6]; }
|
||||
{ uint x = nonce ^ SEEDW[6]; x += 0x9e3779b9u * 7u; x = splitmix32(x); r6 = x ^ SEEDW[7]; }
|
||||
{ uint x = nonce ^ SEEDW[7]; x += 0x9e3779b9u * 8u; x = splitmix32(x); r7 = x ^ SEEDW[0]; }
|
||||
|
||||
for (uint it = 0u; it < 8u; ++it) {
|
||||
uint sel = r0;
|
||||
r6 = r6 - r3; // 0
|
||||
r2 = r2 | r6; // 1
|
||||
r4 = r4 + r0 + select(0x1a579b38u, 0x7731324bu, ((sel >> 4u) & 1u) != 0u); // 2
|
||||
r3 = r3 ^ dataset[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 3
|
||||
r0 = r0 ^ simd_shuffle_xor(r6, (ushort)1); // 4
|
||||
r4 = r4 ^ r1; // 5
|
||||
r3 = r5 * r3 + r3; // 6
|
||||
r1 = r1 * r6; // 7
|
||||
r2 = r2 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x08000000u) & MASK]; // 8
|
||||
r0 = r0 ^ r4; // 9
|
||||
r0 = r0 + r7 + select(0xee083919u, 0xb10fcef8u, ((sel >> 30u) & 1u) != 0u); // 10
|
||||
r0 = mulhi(r0, r2); // 11
|
||||
r5 = r5 ^ r4; // 12
|
||||
r7 = r3 * r0 + r7; // 13
|
||||
r3 = r3 ^ dataset[((rotl_imm(r4 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & MASK]; // 14
|
||||
r2 = r2 ^ dataset[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 15
|
||||
r7 = r7 ^ simd_shuffle_xor(r5, (ushort)16); // 16
|
||||
r7 = r7 * r6; // 17
|
||||
r2 = r3 * r1 + r2; // 18
|
||||
r5 = rotr_var(r5, r2); // 19
|
||||
r5 = r5 ^ dataset[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 20
|
||||
r6 = r6 ^ simd_shuffle_xor(r4, (ushort)16); // 21
|
||||
r1 = mulhi(r1, r2); // 22
|
||||
r6 = r6 ^ r5; // 23
|
||||
r5 = r2 * r0 + r5; // 24
|
||||
r6 = r3 * r7 + r6; // 25
|
||||
r6 = r6 ^ dataset[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 26
|
||||
r1 = r5 * r5 + r1; // 27
|
||||
r0 = r0 ^ dataset[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 28
|
||||
r7 = r7 + r4 + select(0xd5e6c37cu, 0x0b1f02c7u, ((sel >> 7u) & 1u) != 0u); // 29
|
||||
r7 = r7 + r0 + select(0x4e45ba51u, 0x2d6e8bd1u, ((sel >> 0u) & 1u) != 0u); // 30
|
||||
r0 = r0 ^ r2; // 31
|
||||
r6 = r6 + r2 + select(0x1c91b2a1u, 0x351422d2u, ((sel >> 13u) & 1u) != 0u); // 32
|
||||
r4 = mulhi(r4, r5); // 33
|
||||
r3 = rotr_var(r3, r5); // 34
|
||||
r3 = r3 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 35
|
||||
r4 = rotl_imm(r4, 13u); // 36
|
||||
r6 = r6 + r1 + select(0x3f2970f7u, 0x64a28251u, ((sel >> 7u) & 1u) != 0u); // 37
|
||||
r2 = r2 * r0; // 38
|
||||
r3 = r3 + r4 + select(0x3b7b3317u, 0x7378c955u, ((sel >> 15u) & 1u) != 0u); // 39
|
||||
r0 = r0 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & MASK]; // 40
|
||||
r6 = r6 ^ r5; // 41
|
||||
r3 = r3 + r0 + select(0x6678b059u, 0xb31f6a64u, ((sel >> 2u) & 1u) != 0u); // 42
|
||||
r7 = r7 ^ dataset[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & MASK]; // 43
|
||||
r3 = r3 ^ simd_shuffle_xor(r7, (ushort)4); // 44
|
||||
r1 = mulhi(r1, r5); // 45
|
||||
r0 = rotl_imm(r0, 6u); // 46
|
||||
r1 = r1 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 47
|
||||
r4 = r4 + r0 + select(0x43095946u, 0x5feccee6u, ((sel >> 17u) & 1u) != 0u); // 48
|
||||
r2 = r2 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 49
|
||||
r2 = r2 - r3; // 50
|
||||
r7 = r7 + r2 + select(0xf45ecdf8u, 0x26e2b582u, ((sel >> 24u) & 1u) != 0u); // 51
|
||||
r5 = r5 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & MASK]; // 52
|
||||
r6 = r6 ^ dataset[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 53
|
||||
r7 = r7 ^ simd_shuffle_xor(r5, (ushort)4); // 54
|
||||
r1 = rotr_var(r1, r2); // 55
|
||||
r0 = r0 ^ simd_shuffle_xor(r5, (ushort)16); // 56
|
||||
r5 = rotl_imm(r5, 12u); // 57
|
||||
r5 = r5 + r3 + select(0x6b4f35e8u, 0x473b1718u, ((sel >> 4u) & 1u) != 0u); // 58
|
||||
r4 = rotr_var(r4, r1); // 59
|
||||
r4 = r4 ^ simd_shuffle_xor(r5, (ushort)1); // 60
|
||||
r7 = r7 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 61
|
||||
r5 = r5 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 62
|
||||
r4 = r4 * r1; // 63
|
||||
// latency-shadow block (Counter ASIC 3.0 item 8): 256 ALU instructions x 27 passes after instruction 63, no load
|
||||
for (uint sh = 0u; sh < 27u; ++sh) {
|
||||
r6 = r6 | r0; // s0 or
|
||||
r7 = r7 * r3; // s1 mul
|
||||
r3 = r5 * r5 + r3; // s2 mad
|
||||
r7 = r7 + r4 + select(0x97d3d105u, 0x38e58a06u, ((sel >> 6u) & 1u) != 0u); // s3 add
|
||||
r6 = r6 + r0 + select(0x29b87b9au, 0x9629673du, ((sel >> 20u) & 1u) != 0u); // s4 add
|
||||
r6 = r6 ^ simd_shuffle_xor(r2, (ushort)1); // s5 shfl
|
||||
r5 = r3 * r5 + r5; // s6 mad
|
||||
r3 = rotl_imm(r3, 5u); // s7 rotl
|
||||
r0 = r0 + r7 + select(0xa7c1fb0fu, 0x4a5f55e1u, ((sel >> 12u) & 1u) != 0u); // s8 add
|
||||
r5 = r5 * r7; // s9 mul
|
||||
r4 = r4 ^ r5; // s10 xor
|
||||
r6 = mulhi(r6, r4); // s11 mulhi
|
||||
r5 = rotl_imm(r5, 13u); // s12 rotl
|
||||
r0 = r2 * r1 + r0; // s13 mad
|
||||
r2 = mulhi(r2, r4); // s14 mulhi
|
||||
r3 = r3 | r1; // s15 or
|
||||
r0 = r0 - r5; // s16 sub
|
||||
r0 = r0 * r6; // s17 mul
|
||||
r5 = r5 ^ simd_shuffle_xor(r1, (ushort)4); // s18 shfl
|
||||
r3 = r3 | r4; // s19 or
|
||||
r6 = r3 * r2 + r6; // s20 mad
|
||||
r0 = rotl_imm(r0, 11u); // s21 rotl
|
||||
r3 = rotl_imm(r3, 25u); // s22 rotl
|
||||
r6 = r6 ^ r4; // s23 xor
|
||||
r4 = r4 | r2; // s24 or
|
||||
r1 = r1 + r6 + select(0x53e74799u, 0xfc2f6d17u, ((sel >> 7u) & 1u) != 0u); // s25 add
|
||||
r7 = rotr_var(r7, r6); // s26 rotr
|
||||
r7 = r7 + r5 + select(0xb8fae90eu, 0x787048dcu, ((sel >> 27u) & 1u) != 0u); // s27 add
|
||||
r4 = r4 ^ r1; // s28 xor
|
||||
r6 = r6 ^ r1; // s29 xor
|
||||
r2 = mulhi(r2, r0); // s30 mulhi
|
||||
r1 = rotr_var(r1, r4); // s31 rotr
|
||||
r7 = r7 ^ r4; // s32 xor
|
||||
r3 = r2 * r6 + r3; // s33 mad
|
||||
r3 = r3 + r1 + select(0xd2dc42ebu, 0x9ffd510bu, ((sel >> 12u) & 1u) != 0u); // s34 add
|
||||
r4 = r4 + r7 + select(0xad8eda6fu, 0xf871ba37u, ((sel >> 7u) & 1u) != 0u); // s35 add
|
||||
r7 = r7 + r4 + select(0xb0cef8f4u, 0xd02d30cau, ((sel >> 3u) & 1u) != 0u); // s36 add
|
||||
r4 = r3 * r2 + r4; // s37 mad
|
||||
r5 = r7 * r7 + r5; // s38 mad
|
||||
r3 = r3 + r6 + select(0x2bd506e2u, 0x82e98a22u, ((sel >> 26u) & 1u) != 0u); // s39 add
|
||||
r0 = rotl_imm(r0, 21u); // s40 rotl
|
||||
r1 = r1 | r0; // s41 or
|
||||
r7 = r7 + r5 + select(0xc0386e49u, 0x6435f6d8u, ((sel >> 28u) & 1u) != 0u); // s42 add
|
||||
r1 = mulhi(r1, r0); // s43 mulhi
|
||||
r5 = r5 + r6 + select(0x3a8921a4u, 0x1f682c68u, ((sel >> 12u) & 1u) != 0u); // s44 add
|
||||
r4 = r4 + r5 + select(0xeeca4334u, 0xf246e180u, ((sel >> 1u) & 1u) != 0u); // s45 add
|
||||
r2 = r2 ^ r3; // s46 xor
|
||||
r2 = r2 ^ r5; // s47 xor
|
||||
r7 = r4 * r1 + r7; // s48 mad
|
||||
r2 = r6 * r1 + r2; // s49 mad
|
||||
r7 = r0 * r7 + r7; // s50 mad
|
||||
r5 = r5 + r6 + select(0x745776d8u, 0x41bd68ccu, ((sel >> 26u) & 1u) != 0u); // s51 add
|
||||
r3 = r3 ^ simd_shuffle_xor(r0, (ushort)1); // s52 shfl
|
||||
r4 = r4 | r0; // s53 or
|
||||
r0 = r2 * r4 + r0; // s54 mad
|
||||
r5 = rotl_imm(r5, 9u); // s55 rotl
|
||||
r5 = r5 * r7; // s56 mul
|
||||
r7 = rotl_imm(r7, 27u); // s57 rotl
|
||||
r7 = r7 - r6; // s58 sub
|
||||
r2 = rotl_imm(r2, 22u); // s59 rotl
|
||||
r7 = r7 ^ simd_shuffle_xor(r1, (ushort)2); // s60 shfl
|
||||
r4 = r4 + r0 + select(0x60d0ff55u, 0xcba6a161u, ((sel >> 7u) & 1u) != 0u); // s61 add
|
||||
r1 = r1 * r5; // s62 mul
|
||||
r7 = mulhi(r7, r6); // s63 mulhi
|
||||
r5 = r5 * r0; // s64 mul
|
||||
r2 = rotr_var(r2, r5); // s65 rotr
|
||||
r0 = mulhi(r0, r3); // s66 mulhi
|
||||
r0 = r0 | r7; // s67 or
|
||||
r0 = rotl_imm(r0, 19u); // s68 rotl
|
||||
r0 = rotl_imm(r0, 1u); // s69 rotl
|
||||
r2 = r2 | r3; // s70 or
|
||||
r3 = r7 * r5 + r3; // s71 mad
|
||||
r7 = r0 * r4 + r7; // s72 mad
|
||||
r4 = r4 | r3; // s73 or
|
||||
r3 = r3 ^ r1; // s74 xor
|
||||
r1 = r1 + r3 + select(0x3f4ffea2u, 0x8e35924eu, ((sel >> 15u) & 1u) != 0u); // s75 add
|
||||
r1 = r1 * r3; // s76 mul
|
||||
r0 = r3 * r2 + r0; // s77 mad
|
||||
r2 = r2 ^ simd_shuffle_xor(r0, (ushort)16); // s78 shfl
|
||||
r1 = mulhi(r1, r0); // s79 mulhi
|
||||
r0 = mulhi(r0, r4); // s80 mulhi
|
||||
r4 = r4 * r2; // s81 mul
|
||||
r5 = r5 - r7; // s82 sub
|
||||
r0 = r2 * r3 + r0; // s83 mad
|
||||
r7 = r7 - r2; // s84 sub
|
||||
r1 = r1 ^ r2; // s85 xor
|
||||
r6 = r6 ^ simd_shuffle_xor(r3, (ushort)16); // s86 shfl
|
||||
r3 = r0 * r0 + r3; // s87 mad
|
||||
r0 = rotl_imm(r0, 20u); // s88 rotl
|
||||
r1 = r1 - r5; // s89 sub
|
||||
r6 = r6 ^ simd_shuffle_xor(r4, (ushort)16); // s90 shfl
|
||||
r5 = r5 ^ r1; // s91 xor
|
||||
r5 = rotl_imm(r5, 12u); // s92 rotl
|
||||
r6 = r1 * r2 + r6; // s93 mad
|
||||
r2 = rotr_var(r2, r5); // s94 rotr
|
||||
r0 = r0 | r4; // s95 or
|
||||
r7 = r7 * r3; // s96 mul
|
||||
r5 = r5 | r0; // s97 or
|
||||
r3 = r3 + r1 + select(0xbcfdd8a7u, 0x386bb642u, ((sel >> 13u) & 1u) != 0u); // s98 add
|
||||
r0 = r0 + r6 + select(0xc921a021u, 0xa3962d03u, ((sel >> 9u) & 1u) != 0u); // s99 add
|
||||
r1 = rotr_var(r1, r0); // s100 rotr
|
||||
r2 = r2 + r4 + select(0x70da4067u, 0xc5990c5cu, ((sel >> 18u) & 1u) != 0u); // s101 add
|
||||
r2 = rotl_imm(r2, 3u); // s102 rotl
|
||||
r1 = r1 ^ simd_shuffle_xor(r4, (ushort)4); // s103 shfl
|
||||
r0 = r0 ^ r2; // s104 xor
|
||||
r4 = r4 + r3 + select(0x1a5880cbu, 0xb15aec75u, ((sel >> 28u) & 1u) != 0u); // s105 add
|
||||
r4 = r4 ^ simd_shuffle_xor(r0, (ushort)2); // s106 shfl
|
||||
r7 = r7 ^ simd_shuffle_xor(r0, (ushort)4); // s107 shfl
|
||||
r1 = r7 * r3 + r1; // s108 mad
|
||||
r0 = r2 * r3 + r0; // s109 mad
|
||||
r2 = r2 | r4; // s110 or
|
||||
r6 = r6 ^ simd_shuffle_xor(r1, (ushort)1); // s111 shfl
|
||||
r2 = r5 * r3 + r2; // s112 mad
|
||||
r4 = mulhi(r4, r0); // s113 mulhi
|
||||
r7 = rotl_imm(r7, 29u); // s114 rotl
|
||||
r3 = r3 - r7; // s115 sub
|
||||
r0 = mulhi(r0, r1); // s116 mulhi
|
||||
r4 = rotl_imm(r4, 12u); // s117 rotl
|
||||
r1 = r1 + r4 + select(0x24494ad3u, 0xc5e98ee5u, ((sel >> 7u) & 1u) != 0u); // s118 add
|
||||
r0 = r0 + r4 + select(0x1b2d1c83u, 0xa97bc949u, ((sel >> 4u) & 1u) != 0u); // s119 add
|
||||
r5 = r5 ^ simd_shuffle_xor(r4, (ushort)4); // s120 shfl
|
||||
r3 = r3 ^ simd_shuffle_xor(r2, (ushort)8); // s121 shfl
|
||||
r7 = rotl_imm(r7, 3u); // s122 rotl
|
||||
r4 = rotr_var(r4, r6); // s123 rotr
|
||||
r6 = r6 ^ simd_shuffle_xor(r5, (ushort)1); // s124 shfl
|
||||
r5 = rotl_imm(r5, 21u); // s125 rotl
|
||||
r0 = r3 * r1 + r0; // s126 mad
|
||||
r0 = r0 + r3 + select(0xff14c08du, 0xa7f6bb54u, ((sel >> 24u) & 1u) != 0u); // s127 add
|
||||
r7 = r7 ^ r4; // s128 xor
|
||||
r3 = mulhi(r3, r5); // s129 mulhi
|
||||
r5 = r5 ^ simd_shuffle_xor(r7, (ushort)8); // s130 shfl
|
||||
r6 = mulhi(r6, r4); // s131 mulhi
|
||||
r5 = r5 ^ r3; // s132 xor
|
||||
r0 = rotl_imm(r0, 19u); // s133 rotl
|
||||
r4 = rotr_var(r4, r3); // s134 rotr
|
||||
r1 = r1 ^ simd_shuffle_xor(r2, (ushort)16); // s135 shfl
|
||||
r5 = r5 - r2; // s136 sub
|
||||
r3 = r3 ^ r5; // s137 xor
|
||||
r0 = r0 - r2; // s138 sub
|
||||
r3 = r3 * r7; // s139 mul
|
||||
r5 = r5 + r3 + select(0x82ab98f4u, 0x2d465ca8u, ((sel >> 24u) & 1u) != 0u); // s140 add
|
||||
r4 = r4 + r1 + select(0x80594390u, 0x237a9d8eu, ((sel >> 11u) & 1u) != 0u); // s141 add
|
||||
r3 = r7 * r0 + r3; // s142 mad
|
||||
r0 = r0 | r3; // s143 or
|
||||
r0 = mulhi(r0, r1); // s144 mulhi
|
||||
r6 = r6 ^ r2; // s145 xor
|
||||
r7 = r7 ^ simd_shuffle_xor(r1, (ushort)16); // s146 shfl
|
||||
r0 = r0 + r1 + select(0x07f15148u, 0xe68cf57eu, ((sel >> 18u) & 1u) != 0u); // s147 add
|
||||
r7 = r7 | r3; // s148 or
|
||||
r4 = r4 ^ r2; // s149 xor
|
||||
r7 = r7 ^ r1; // s150 xor
|
||||
r5 = rotr_var(r5, r2); // s151 rotr
|
||||
r0 = r0 * r7; // s152 mul
|
||||
r4 = rotl_imm(r4, 16u); // s153 rotl
|
||||
r6 = rotr_var(r6, r4); // s154 rotr
|
||||
r6 = r6 ^ r1; // s155 xor
|
||||
r6 = r6 | r3; // s156 or
|
||||
r1 = r1 + r6 + select(0x14878c5au, 0x84c2a09du, ((sel >> 19u) & 1u) != 0u); // s157 add
|
||||
r4 = r4 - r1; // s158 sub
|
||||
r4 = r6 * r7 + r4; // s159 mad
|
||||
r1 = r1 * r5; // s160 mul
|
||||
r4 = r4 | r0; // s161 or
|
||||
r7 = rotl_imm(r7, 21u); // s162 rotl
|
||||
r0 = r0 - r1; // s163 sub
|
||||
r1 = r1 * r0; // s164 mul
|
||||
r2 = r2 - r1; // s165 sub
|
||||
r0 = r0 ^ simd_shuffle_xor(r6, (ushort)2); // s166 shfl
|
||||
r0 = r0 - r4; // s167 sub
|
||||
r1 = rotr_var(r1, r6); // s168 rotr
|
||||
r7 = rotr_var(r7, r1); // s169 rotr
|
||||
r3 = r1 * r3 + r3; // s170 mad
|
||||
r2 = r2 ^ simd_shuffle_xor(r0, (ushort)4); // s171 shfl
|
||||
r2 = r2 - r1; // s172 sub
|
||||
r7 = r7 ^ r6; // s173 xor
|
||||
r4 = rotr_var(r4, r1); // s174 rotr
|
||||
r4 = r4 ^ simd_shuffle_xor(r7, (ushort)2); // s175 shfl
|
||||
r6 = r6 * r2; // s176 mul
|
||||
r0 = r0 ^ simd_shuffle_xor(r2, (ushort)4); // s177 shfl
|
||||
r6 = r6 + r4 + select(0x4fa43965u, 0x509871e6u, ((sel >> 21u) & 1u) != 0u); // s178 add
|
||||
r0 = r0 ^ r6; // s179 xor
|
||||
r5 = r5 + r7 + select(0xfb3c4bd8u, 0xb95b8a53u, ((sel >> 0u) & 1u) != 0u); // s180 add
|
||||
r6 = r6 * r3; // s181 mul
|
||||
r7 = rotl_imm(r7, 1u); // s182 rotl
|
||||
r0 = r0 + r7 + select(0xc448a197u, 0x7047b7cfu, ((sel >> 11u) & 1u) != 0u); // s183 add
|
||||
r3 = r3 * r5; // s184 mul
|
||||
r5 = r5 ^ simd_shuffle_xor(r7, (ushort)1); // s185 shfl
|
||||
r0 = r0 * r2; // s186 mul
|
||||
r4 = mulhi(r4, r7); // s187 mulhi
|
||||
r2 = r7 * r2 + r2; // s188 mad
|
||||
r7 = r7 - r2; // s189 sub
|
||||
r2 = r2 + r5 + select(0x2fceaf49u, 0x5caccc5du, ((sel >> 19u) & 1u) != 0u); // s190 add
|
||||
r1 = r1 ^ simd_shuffle_xor(r2, (ushort)1); // s191 shfl
|
||||
r7 = r7 + r1 + select(0xba0cee62u, 0x857bb5feu, ((sel >> 20u) & 1u) != 0u); // s192 add
|
||||
r6 = rotl_imm(r6, 22u); // s193 rotl
|
||||
r3 = r3 ^ r7; // s194 xor
|
||||
r7 = r7 * r0; // s195 mul
|
||||
r3 = r3 ^ r5; // s196 xor
|
||||
r5 = r5 + r1 + select(0x8519428cu, 0xeae84577u, ((sel >> 29u) & 1u) != 0u); // s197 add
|
||||
r7 = r4 * r4 + r7; // s198 mad
|
||||
r3 = r3 + r6 + select(0x468639d3u, 0x0bfdbfa1u, ((sel >> 8u) & 1u) != 0u); // s199 add
|
||||
r5 = r3 * r4 + r5; // s200 mad
|
||||
r2 = r2 ^ simd_shuffle_xor(r7, (ushort)4); // s201 shfl
|
||||
r7 = r7 + r4 + select(0x18ec9a69u, 0x53ee8e50u, ((sel >> 23u) & 1u) != 0u); // s202 add
|
||||
r6 = r6 ^ r0; // s203 xor
|
||||
r4 = r4 + r5 + select(0x3e81485bu, 0xac63376eu, ((sel >> 4u) & 1u) != 0u); // s204 add
|
||||
r7 = r1 * r4 + r7; // s205 mad
|
||||
r1 = r1 ^ simd_shuffle_xor(r5, (ushort)16); // s206 shfl
|
||||
r4 = r4 * r0; // s207 mul
|
||||
r1 = r1 | r6; // s208 or
|
||||
r6 = r6 + r4 + select(0x66ccb75du, 0x5b745519u, ((sel >> 28u) & 1u) != 0u); // s209 add
|
||||
r1 = rotr_var(r1, r3); // s210 rotr
|
||||
r5 = r5 - r4; // s211 sub
|
||||
r4 = r4 ^ r7; // s212 xor
|
||||
r1 = r1 + r7 + select(0x0ce0553du, 0x1c3dccf7u, ((sel >> 8u) & 1u) != 0u); // s213 add
|
||||
r0 = mulhi(r0, r1); // s214 mulhi
|
||||
r3 = r3 + r6 + select(0x96bfec89u, 0xfadc9205u, ((sel >> 9u) & 1u) != 0u); // s215 add
|
||||
r0 = rotl_imm(r0, 4u); // s216 rotl
|
||||
r6 = r6 ^ r4; // s217 xor
|
||||
r6 = r6 ^ simd_shuffle_xor(r7, (ushort)2); // s218 shfl
|
||||
r7 = r0 * r1 + r7; // s219 mad
|
||||
r4 = r4 * r0; // s220 mul
|
||||
r2 = r6 * r0 + r2; // s221 mad
|
||||
r5 = r5 ^ r4; // s222 xor
|
||||
r0 = r0 * r5; // s223 mul
|
||||
r2 = r2 + r0 + select(0x61495681u, 0x66b15c04u, ((sel >> 9u) & 1u) != 0u); // s224 add
|
||||
r3 = rotr_var(r3, r2); // s225 rotr
|
||||
r2 = r3 * r3 + r2; // s226 mad
|
||||
r6 = r6 ^ r0; // s227 xor
|
||||
r4 = rotl_imm(r4, 30u); // s228 rotl
|
||||
r2 = r2 + r3 + select(0x70d05c34u, 0x25bc17c2u, ((sel >> 29u) & 1u) != 0u); // s229 add
|
||||
r1 = rotr_var(r1, r5); // s230 rotr
|
||||
r1 = r1 + r4 + select(0x32a28384u, 0x2d8728f3u, ((sel >> 8u) & 1u) != 0u); // s231 add
|
||||
r1 = rotl_imm(r1, 6u); // s232 rotl
|
||||
r2 = r2 + r0 + select(0xcf0949f1u, 0x61e81fe6u, ((sel >> 21u) & 1u) != 0u); // s233 add
|
||||
r4 = rotl_imm(r4, 2u); // s234 rotl
|
||||
r2 = r2 - r6; // s235 sub
|
||||
r6 = r6 | r3; // s236 or
|
||||
r7 = r7 ^ simd_shuffle_xor(r5, (ushort)4); // s237 shfl
|
||||
r1 = r1 ^ r3; // s238 xor
|
||||
r2 = rotl_imm(r2, 1u); // s239 rotl
|
||||
r5 = rotr_var(r5, r2); // s240 rotr
|
||||
r4 = rotl_imm(r4, 10u); // s241 rotl
|
||||
r6 = r6 + r2 + select(0xe204fb50u, 0x0a67565du, ((sel >> 16u) & 1u) != 0u); // s242 add
|
||||
r1 = r1 ^ simd_shuffle_xor(r3, (ushort)2); // s243 shfl
|
||||
r0 = r0 ^ simd_shuffle_xor(r1, (ushort)16); // s244 shfl
|
||||
r5 = r5 ^ simd_shuffle_xor(r7, (ushort)2); // s245 shfl
|
||||
r4 = r4 ^ r5; // s246 xor
|
||||
r7 = r7 ^ r2; // s247 xor
|
||||
r0 = r0 ^ simd_shuffle_xor(r6, (ushort)16); // s248 shfl
|
||||
r1 = mulhi(r1, r3); // s249 mulhi
|
||||
r7 = r7 + r5 + select(0x3d6daffdu, 0x5a79fd6fu, ((sel >> 31u) & 1u) != 0u); // s250 add
|
||||
r6 = r6 + r0 + select(0x84334aeau, 0x6beb28c2u, ((sel >> 28u) & 1u) != 0u); // s251 add
|
||||
r7 = r7 - r1; // s252 sub
|
||||
r5 = rotr_var(r5, r2); // s253 rotr
|
||||
r2 = mulhi(r2, r5); // s254 mulhi
|
||||
r6 = r6 * r3; // s255 mul
|
||||
}
|
||||
}
|
||||
uint lo = r0 ^ rotl_imm(r1, 7u) ^ rotl_imm(r2, 14u) ^ rotl_imm(r3, 21u);
|
||||
uint hi = r4 ^ rotl_imm(r5, 9u) ^ rotl_imm(r6, 18u) ^ rotl_imm(r7, 27u);
|
||||
out[gid] = ((ulong)hi << 32) | (ulong)lo;
|
||||
}
|
||||
370
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program_bound.metal
Normal file
370
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/program_bound.metal
Normal file
|
|
@ -0,0 +1,370 @@
|
|||
#include <metal_stdlib>
|
||||
using namespace metal;
|
||||
|
||||
#define MASK 0x0fffffffu
|
||||
constant uint SEEDW[8] = { 0xbe8c5a0cu, 0x7646e626u, 0x508e29d0u, 0xfb8a5b4au, 0x53c50955u, 0x685d62fcu, 0x6065c013u, 0x881096ebu };
|
||||
|
||||
inline uint splitmix32(uint x) {
|
||||
x ^= x >> 16; x *= 0x7feb352du;
|
||||
x ^= x >> 15; x *= 0x846ca68bu;
|
||||
x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
inline uint rotl_imm(uint x, uint n) { return (x << n) | (x >> (32u - n)); } // n in 1..31
|
||||
inline uint rotr_var(uint x, uint n) { n &= 31u; return (x >> n) | (x << ((32u - n) & 31u)); }
|
||||
inline uint ds_elem(uint i, uint d0, uint d1) {
|
||||
uint x = i ^ d0;
|
||||
x *= 0x9E3779B1u; x ^= x >> 15;
|
||||
x += d1;
|
||||
x *= 0x85EBCA77u; x ^= x >> 13;
|
||||
x *= 0xC2B2AE3Du; x ^= x >> 16;
|
||||
return x;
|
||||
}
|
||||
|
||||
// Header-bound variant: the init words come from buffer 3 (bind.rs), not from SEEDW.
|
||||
kernel void igneum_hash_bound(device const uint* dataset [[buffer(0)]],
|
||||
device ulong* out [[buffer(1)]],
|
||||
constant uint& baseNonce [[buffer(2)]],
|
||||
constant uint* initw [[buffer(3)]],
|
||||
uint gid [[thread_position_in_grid]]) {
|
||||
uint nonce = baseNonce + gid;
|
||||
uint r0, r1, r2, r3, r4, r5, r6, r7;
|
||||
{ uint x = nonce ^ initw[0]; x += 0x9e3779b9u * 1u; x = splitmix32(x); r0 = x ^ initw[1]; }
|
||||
{ uint x = nonce ^ initw[1]; x += 0x9e3779b9u * 2u; x = splitmix32(x); r1 = x ^ initw[2]; }
|
||||
{ uint x = nonce ^ initw[2]; x += 0x9e3779b9u * 3u; x = splitmix32(x); r2 = x ^ initw[3]; }
|
||||
{ uint x = nonce ^ initw[3]; x += 0x9e3779b9u * 4u; x = splitmix32(x); r3 = x ^ initw[4]; }
|
||||
{ uint x = nonce ^ initw[4]; x += 0x9e3779b9u * 5u; x = splitmix32(x); r4 = x ^ initw[5]; }
|
||||
{ uint x = nonce ^ initw[5]; x += 0x9e3779b9u * 6u; x = splitmix32(x); r5 = x ^ initw[6]; }
|
||||
{ uint x = nonce ^ initw[6]; x += 0x9e3779b9u * 7u; x = splitmix32(x); r6 = x ^ initw[7]; }
|
||||
{ uint x = nonce ^ initw[7]; x += 0x9e3779b9u * 8u; x = splitmix32(x); r7 = x ^ initw[0]; }
|
||||
|
||||
for (uint it = 0u; it < 8u; ++it) {
|
||||
uint sel = r0;
|
||||
r6 = r6 - r3; // 0
|
||||
r2 = r2 | r6; // 1
|
||||
r4 = r4 + r0 + select(0x1a579b38u, 0x7731324bu, ((sel >> 4u) & 1u) != 0u); // 2
|
||||
r3 = r3 ^ dataset[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 3
|
||||
r0 = r0 ^ simd_shuffle_xor(r6, (ushort)1); // 4
|
||||
r4 = r4 ^ r1; // 5
|
||||
r3 = r5 * r3 + r3; // 6
|
||||
r1 = r1 * r6; // 7
|
||||
r2 = r2 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x08000000u) & MASK]; // 8
|
||||
r0 = r0 ^ r4; // 9
|
||||
r0 = r0 + r7 + select(0xee083919u, 0xb10fcef8u, ((sel >> 30u) & 1u) != 0u); // 10
|
||||
r0 = mulhi(r0, r2); // 11
|
||||
r5 = r5 ^ r4; // 12
|
||||
r7 = r3 * r0 + r7; // 13
|
||||
r3 = r3 ^ dataset[((rotl_imm(r4 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & MASK]; // 14
|
||||
r2 = r2 ^ dataset[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 15
|
||||
r7 = r7 ^ simd_shuffle_xor(r5, (ushort)16); // 16
|
||||
r7 = r7 * r6; // 17
|
||||
r2 = r3 * r1 + r2; // 18
|
||||
r5 = rotr_var(r5, r2); // 19
|
||||
r5 = r5 ^ dataset[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 20
|
||||
r6 = r6 ^ simd_shuffle_xor(r4, (ushort)16); // 21
|
||||
r1 = mulhi(r1, r2); // 22
|
||||
r6 = r6 ^ r5; // 23
|
||||
r5 = r2 * r0 + r5; // 24
|
||||
r6 = r3 * r7 + r6; // 25
|
||||
r6 = r6 ^ dataset[((rotl_imm(r5 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 26
|
||||
r1 = r5 * r5 + r1; // 27
|
||||
r0 = r0 ^ dataset[((rotl_imm(r6 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 28
|
||||
r7 = r7 + r4 + select(0xd5e6c37cu, 0x0b1f02c7u, ((sel >> 7u) & 1u) != 0u); // 29
|
||||
r7 = r7 + r0 + select(0x4e45ba51u, 0x2d6e8bd1u, ((sel >> 0u) & 1u) != 0u); // 30
|
||||
r0 = r0 ^ r2; // 31
|
||||
r6 = r6 + r2 + select(0x1c91b2a1u, 0x351422d2u, ((sel >> 13u) & 1u) != 0u); // 32
|
||||
r4 = mulhi(r4, r5); // 33
|
||||
r3 = rotr_var(r3, r5); // 34
|
||||
r3 = r3 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 35
|
||||
r4 = rotl_imm(r4, 13u); // 36
|
||||
r6 = r6 + r1 + select(0x3f2970f7u, 0x64a28251u, ((sel >> 7u) & 1u) != 0u); // 37
|
||||
r2 = r2 * r0; // 38
|
||||
r3 = r3 + r4 + select(0x3b7b3317u, 0x7378c955u, ((sel >> 15u) & 1u) != 0u); // 39
|
||||
r0 = r0 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & MASK]; // 40
|
||||
r6 = r6 ^ r5; // 41
|
||||
r3 = r3 + r0 + select(0x6678b059u, 0xb31f6a64u, ((sel >> 2u) & 1u) != 0u); // 42
|
||||
r7 = r7 ^ dataset[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x00000000u) & MASK]; // 43
|
||||
r3 = r3 ^ simd_shuffle_xor(r7, (ushort)4); // 44
|
||||
r1 = mulhi(r1, r5); // 45
|
||||
r0 = rotl_imm(r0, 6u); // 46
|
||||
r1 = r1 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 47
|
||||
r4 = r4 + r0 + select(0x43095946u, 0x5feccee6u, ((sel >> 17u) & 1u) != 0u); // 48
|
||||
r2 = r2 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 49
|
||||
r2 = r2 - r3; // 50
|
||||
r7 = r7 + r2 + select(0xf45ecdf8u, 0x26e2b582u, ((sel >> 24u) & 1u) != 0u); // 51
|
||||
r5 = r5 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x03ffffffu) | 0x0c000000u) & MASK]; // 52
|
||||
r6 = r6 ^ dataset[((rotl_imm(r3 * 0x2cb18b35u, 9u) & 0x0fffffffu) | 0x00000000u) & MASK]; // 53
|
||||
r7 = r7 ^ simd_shuffle_xor(r5, (ushort)4); // 54
|
||||
r1 = rotr_var(r1, r2); // 55
|
||||
r0 = r0 ^ simd_shuffle_xor(r5, (ushort)16); // 56
|
||||
r5 = rotl_imm(r5, 12u); // 57
|
||||
r5 = r5 + r3 + select(0x6b4f35e8u, 0x473b1718u, ((sel >> 4u) & 1u) != 0u); // 58
|
||||
r4 = rotr_var(r4, r1); // 59
|
||||
r4 = r4 ^ simd_shuffle_xor(r5, (ushort)1); // 60
|
||||
r7 = r7 ^ dataset[((rotl_imm(r0 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 61
|
||||
r5 = r5 ^ dataset[((rotl_imm(r7 * 0x2cb18b35u, 9u) & 0x07ffffffu) | 0x08000000u) & MASK]; // 62
|
||||
r4 = r4 * r1; // 63
|
||||
// latency-shadow block (Counter ASIC 3.0 item 8): 256 ALU instructions x 27 passes after instruction 63, no load
|
||||
for (uint sh = 0u; sh < 27u; ++sh) {
|
||||
r6 = r6 | r0; // s0 or
|
||||
r7 = r7 * r3; // s1 mul
|
||||
r3 = r5 * r5 + r3; // s2 mad
|
||||
r7 = r7 + r4 + select(0x97d3d105u, 0x38e58a06u, ((sel >> 6u) & 1u) != 0u); // s3 add
|
||||
r6 = r6 + r0 + select(0x29b87b9au, 0x9629673du, ((sel >> 20u) & 1u) != 0u); // s4 add
|
||||
r6 = r6 ^ simd_shuffle_xor(r2, (ushort)1); // s5 shfl
|
||||
r5 = r3 * r5 + r5; // s6 mad
|
||||
r3 = rotl_imm(r3, 5u); // s7 rotl
|
||||
r0 = r0 + r7 + select(0xa7c1fb0fu, 0x4a5f55e1u, ((sel >> 12u) & 1u) != 0u); // s8 add
|
||||
r5 = r5 * r7; // s9 mul
|
||||
r4 = r4 ^ r5; // s10 xor
|
||||
r6 = mulhi(r6, r4); // s11 mulhi
|
||||
r5 = rotl_imm(r5, 13u); // s12 rotl
|
||||
r0 = r2 * r1 + r0; // s13 mad
|
||||
r2 = mulhi(r2, r4); // s14 mulhi
|
||||
r3 = r3 | r1; // s15 or
|
||||
r0 = r0 - r5; // s16 sub
|
||||
r0 = r0 * r6; // s17 mul
|
||||
r5 = r5 ^ simd_shuffle_xor(r1, (ushort)4); // s18 shfl
|
||||
r3 = r3 | r4; // s19 or
|
||||
r6 = r3 * r2 + r6; // s20 mad
|
||||
r0 = rotl_imm(r0, 11u); // s21 rotl
|
||||
r3 = rotl_imm(r3, 25u); // s22 rotl
|
||||
r6 = r6 ^ r4; // s23 xor
|
||||
r4 = r4 | r2; // s24 or
|
||||
r1 = r1 + r6 + select(0x53e74799u, 0xfc2f6d17u, ((sel >> 7u) & 1u) != 0u); // s25 add
|
||||
r7 = rotr_var(r7, r6); // s26 rotr
|
||||
r7 = r7 + r5 + select(0xb8fae90eu, 0x787048dcu, ((sel >> 27u) & 1u) != 0u); // s27 add
|
||||
r4 = r4 ^ r1; // s28 xor
|
||||
r6 = r6 ^ r1; // s29 xor
|
||||
r2 = mulhi(r2, r0); // s30 mulhi
|
||||
r1 = rotr_var(r1, r4); // s31 rotr
|
||||
r7 = r7 ^ r4; // s32 xor
|
||||
r3 = r2 * r6 + r3; // s33 mad
|
||||
r3 = r3 + r1 + select(0xd2dc42ebu, 0x9ffd510bu, ((sel >> 12u) & 1u) != 0u); // s34 add
|
||||
r4 = r4 + r7 + select(0xad8eda6fu, 0xf871ba37u, ((sel >> 7u) & 1u) != 0u); // s35 add
|
||||
r7 = r7 + r4 + select(0xb0cef8f4u, 0xd02d30cau, ((sel >> 3u) & 1u) != 0u); // s36 add
|
||||
r4 = r3 * r2 + r4; // s37 mad
|
||||
r5 = r7 * r7 + r5; // s38 mad
|
||||
r3 = r3 + r6 + select(0x2bd506e2u, 0x82e98a22u, ((sel >> 26u) & 1u) != 0u); // s39 add
|
||||
r0 = rotl_imm(r0, 21u); // s40 rotl
|
||||
r1 = r1 | r0; // s41 or
|
||||
r7 = r7 + r5 + select(0xc0386e49u, 0x6435f6d8u, ((sel >> 28u) & 1u) != 0u); // s42 add
|
||||
r1 = mulhi(r1, r0); // s43 mulhi
|
||||
r5 = r5 + r6 + select(0x3a8921a4u, 0x1f682c68u, ((sel >> 12u) & 1u) != 0u); // s44 add
|
||||
r4 = r4 + r5 + select(0xeeca4334u, 0xf246e180u, ((sel >> 1u) & 1u) != 0u); // s45 add
|
||||
r2 = r2 ^ r3; // s46 xor
|
||||
r2 = r2 ^ r5; // s47 xor
|
||||
r7 = r4 * r1 + r7; // s48 mad
|
||||
r2 = r6 * r1 + r2; // s49 mad
|
||||
r7 = r0 * r7 + r7; // s50 mad
|
||||
r5 = r5 + r6 + select(0x745776d8u, 0x41bd68ccu, ((sel >> 26u) & 1u) != 0u); // s51 add
|
||||
r3 = r3 ^ simd_shuffle_xor(r0, (ushort)1); // s52 shfl
|
||||
r4 = r4 | r0; // s53 or
|
||||
r0 = r2 * r4 + r0; // s54 mad
|
||||
r5 = rotl_imm(r5, 9u); // s55 rotl
|
||||
r5 = r5 * r7; // s56 mul
|
||||
r7 = rotl_imm(r7, 27u); // s57 rotl
|
||||
r7 = r7 - r6; // s58 sub
|
||||
r2 = rotl_imm(r2, 22u); // s59 rotl
|
||||
r7 = r7 ^ simd_shuffle_xor(r1, (ushort)2); // s60 shfl
|
||||
r4 = r4 + r0 + select(0x60d0ff55u, 0xcba6a161u, ((sel >> 7u) & 1u) != 0u); // s61 add
|
||||
r1 = r1 * r5; // s62 mul
|
||||
r7 = mulhi(r7, r6); // s63 mulhi
|
||||
r5 = r5 * r0; // s64 mul
|
||||
r2 = rotr_var(r2, r5); // s65 rotr
|
||||
r0 = mulhi(r0, r3); // s66 mulhi
|
||||
r0 = r0 | r7; // s67 or
|
||||
r0 = rotl_imm(r0, 19u); // s68 rotl
|
||||
r0 = rotl_imm(r0, 1u); // s69 rotl
|
||||
r2 = r2 | r3; // s70 or
|
||||
r3 = r7 * r5 + r3; // s71 mad
|
||||
r7 = r0 * r4 + r7; // s72 mad
|
||||
r4 = r4 | r3; // s73 or
|
||||
r3 = r3 ^ r1; // s74 xor
|
||||
r1 = r1 + r3 + select(0x3f4ffea2u, 0x8e35924eu, ((sel >> 15u) & 1u) != 0u); // s75 add
|
||||
r1 = r1 * r3; // s76 mul
|
||||
r0 = r3 * r2 + r0; // s77 mad
|
||||
r2 = r2 ^ simd_shuffle_xor(r0, (ushort)16); // s78 shfl
|
||||
r1 = mulhi(r1, r0); // s79 mulhi
|
||||
r0 = mulhi(r0, r4); // s80 mulhi
|
||||
r4 = r4 * r2; // s81 mul
|
||||
r5 = r5 - r7; // s82 sub
|
||||
r0 = r2 * r3 + r0; // s83 mad
|
||||
r7 = r7 - r2; // s84 sub
|
||||
r1 = r1 ^ r2; // s85 xor
|
||||
r6 = r6 ^ simd_shuffle_xor(r3, (ushort)16); // s86 shfl
|
||||
r3 = r0 * r0 + r3; // s87 mad
|
||||
r0 = rotl_imm(r0, 20u); // s88 rotl
|
||||
r1 = r1 - r5; // s89 sub
|
||||
r6 = r6 ^ simd_shuffle_xor(r4, (ushort)16); // s90 shfl
|
||||
r5 = r5 ^ r1; // s91 xor
|
||||
r5 = rotl_imm(r5, 12u); // s92 rotl
|
||||
r6 = r1 * r2 + r6; // s93 mad
|
||||
r2 = rotr_var(r2, r5); // s94 rotr
|
||||
r0 = r0 | r4; // s95 or
|
||||
r7 = r7 * r3; // s96 mul
|
||||
r5 = r5 | r0; // s97 or
|
||||
r3 = r3 + r1 + select(0xbcfdd8a7u, 0x386bb642u, ((sel >> 13u) & 1u) != 0u); // s98 add
|
||||
r0 = r0 + r6 + select(0xc921a021u, 0xa3962d03u, ((sel >> 9u) & 1u) != 0u); // s99 add
|
||||
r1 = rotr_var(r1, r0); // s100 rotr
|
||||
r2 = r2 + r4 + select(0x70da4067u, 0xc5990c5cu, ((sel >> 18u) & 1u) != 0u); // s101 add
|
||||
r2 = rotl_imm(r2, 3u); // s102 rotl
|
||||
r1 = r1 ^ simd_shuffle_xor(r4, (ushort)4); // s103 shfl
|
||||
r0 = r0 ^ r2; // s104 xor
|
||||
r4 = r4 + r3 + select(0x1a5880cbu, 0xb15aec75u, ((sel >> 28u) & 1u) != 0u); // s105 add
|
||||
r4 = r4 ^ simd_shuffle_xor(r0, (ushort)2); // s106 shfl
|
||||
r7 = r7 ^ simd_shuffle_xor(r0, (ushort)4); // s107 shfl
|
||||
r1 = r7 * r3 + r1; // s108 mad
|
||||
r0 = r2 * r3 + r0; // s109 mad
|
||||
r2 = r2 | r4; // s110 or
|
||||
r6 = r6 ^ simd_shuffle_xor(r1, (ushort)1); // s111 shfl
|
||||
r2 = r5 * r3 + r2; // s112 mad
|
||||
r4 = mulhi(r4, r0); // s113 mulhi
|
||||
r7 = rotl_imm(r7, 29u); // s114 rotl
|
||||
r3 = r3 - r7; // s115 sub
|
||||
r0 = mulhi(r0, r1); // s116 mulhi
|
||||
r4 = rotl_imm(r4, 12u); // s117 rotl
|
||||
r1 = r1 + r4 + select(0x24494ad3u, 0xc5e98ee5u, ((sel >> 7u) & 1u) != 0u); // s118 add
|
||||
r0 = r0 + r4 + select(0x1b2d1c83u, 0xa97bc949u, ((sel >> 4u) & 1u) != 0u); // s119 add
|
||||
r5 = r5 ^ simd_shuffle_xor(r4, (ushort)4); // s120 shfl
|
||||
r3 = r3 ^ simd_shuffle_xor(r2, (ushort)8); // s121 shfl
|
||||
r7 = rotl_imm(r7, 3u); // s122 rotl
|
||||
r4 = rotr_var(r4, r6); // s123 rotr
|
||||
r6 = r6 ^ simd_shuffle_xor(r5, (ushort)1); // s124 shfl
|
||||
r5 = rotl_imm(r5, 21u); // s125 rotl
|
||||
r0 = r3 * r1 + r0; // s126 mad
|
||||
r0 = r0 + r3 + select(0xff14c08du, 0xa7f6bb54u, ((sel >> 24u) & 1u) != 0u); // s127 add
|
||||
r7 = r7 ^ r4; // s128 xor
|
||||
r3 = mulhi(r3, r5); // s129 mulhi
|
||||
r5 = r5 ^ simd_shuffle_xor(r7, (ushort)8); // s130 shfl
|
||||
r6 = mulhi(r6, r4); // s131 mulhi
|
||||
r5 = r5 ^ r3; // s132 xor
|
||||
r0 = rotl_imm(r0, 19u); // s133 rotl
|
||||
r4 = rotr_var(r4, r3); // s134 rotr
|
||||
r1 = r1 ^ simd_shuffle_xor(r2, (ushort)16); // s135 shfl
|
||||
r5 = r5 - r2; // s136 sub
|
||||
r3 = r3 ^ r5; // s137 xor
|
||||
r0 = r0 - r2; // s138 sub
|
||||
r3 = r3 * r7; // s139 mul
|
||||
r5 = r5 + r3 + select(0x82ab98f4u, 0x2d465ca8u, ((sel >> 24u) & 1u) != 0u); // s140 add
|
||||
r4 = r4 + r1 + select(0x80594390u, 0x237a9d8eu, ((sel >> 11u) & 1u) != 0u); // s141 add
|
||||
r3 = r7 * r0 + r3; // s142 mad
|
||||
r0 = r0 | r3; // s143 or
|
||||
r0 = mulhi(r0, r1); // s144 mulhi
|
||||
r6 = r6 ^ r2; // s145 xor
|
||||
r7 = r7 ^ simd_shuffle_xor(r1, (ushort)16); // s146 shfl
|
||||
r0 = r0 + r1 + select(0x07f15148u, 0xe68cf57eu, ((sel >> 18u) & 1u) != 0u); // s147 add
|
||||
r7 = r7 | r3; // s148 or
|
||||
r4 = r4 ^ r2; // s149 xor
|
||||
r7 = r7 ^ r1; // s150 xor
|
||||
r5 = rotr_var(r5, r2); // s151 rotr
|
||||
r0 = r0 * r7; // s152 mul
|
||||
r4 = rotl_imm(r4, 16u); // s153 rotl
|
||||
r6 = rotr_var(r6, r4); // s154 rotr
|
||||
r6 = r6 ^ r1; // s155 xor
|
||||
r6 = r6 | r3; // s156 or
|
||||
r1 = r1 + r6 + select(0x14878c5au, 0x84c2a09du, ((sel >> 19u) & 1u) != 0u); // s157 add
|
||||
r4 = r4 - r1; // s158 sub
|
||||
r4 = r6 * r7 + r4; // s159 mad
|
||||
r1 = r1 * r5; // s160 mul
|
||||
r4 = r4 | r0; // s161 or
|
||||
r7 = rotl_imm(r7, 21u); // s162 rotl
|
||||
r0 = r0 - r1; // s163 sub
|
||||
r1 = r1 * r0; // s164 mul
|
||||
r2 = r2 - r1; // s165 sub
|
||||
r0 = r0 ^ simd_shuffle_xor(r6, (ushort)2); // s166 shfl
|
||||
r0 = r0 - r4; // s167 sub
|
||||
r1 = rotr_var(r1, r6); // s168 rotr
|
||||
r7 = rotr_var(r7, r1); // s169 rotr
|
||||
r3 = r1 * r3 + r3; // s170 mad
|
||||
r2 = r2 ^ simd_shuffle_xor(r0, (ushort)4); // s171 shfl
|
||||
r2 = r2 - r1; // s172 sub
|
||||
r7 = r7 ^ r6; // s173 xor
|
||||
r4 = rotr_var(r4, r1); // s174 rotr
|
||||
r4 = r4 ^ simd_shuffle_xor(r7, (ushort)2); // s175 shfl
|
||||
r6 = r6 * r2; // s176 mul
|
||||
r0 = r0 ^ simd_shuffle_xor(r2, (ushort)4); // s177 shfl
|
||||
r6 = r6 + r4 + select(0x4fa43965u, 0x509871e6u, ((sel >> 21u) & 1u) != 0u); // s178 add
|
||||
r0 = r0 ^ r6; // s179 xor
|
||||
r5 = r5 + r7 + select(0xfb3c4bd8u, 0xb95b8a53u, ((sel >> 0u) & 1u) != 0u); // s180 add
|
||||
r6 = r6 * r3; // s181 mul
|
||||
r7 = rotl_imm(r7, 1u); // s182 rotl
|
||||
r0 = r0 + r7 + select(0xc448a197u, 0x7047b7cfu, ((sel >> 11u) & 1u) != 0u); // s183 add
|
||||
r3 = r3 * r5; // s184 mul
|
||||
r5 = r5 ^ simd_shuffle_xor(r7, (ushort)1); // s185 shfl
|
||||
r0 = r0 * r2; // s186 mul
|
||||
r4 = mulhi(r4, r7); // s187 mulhi
|
||||
r2 = r7 * r2 + r2; // s188 mad
|
||||
r7 = r7 - r2; // s189 sub
|
||||
r2 = r2 + r5 + select(0x2fceaf49u, 0x5caccc5du, ((sel >> 19u) & 1u) != 0u); // s190 add
|
||||
r1 = r1 ^ simd_shuffle_xor(r2, (ushort)1); // s191 shfl
|
||||
r7 = r7 + r1 + select(0xba0cee62u, 0x857bb5feu, ((sel >> 20u) & 1u) != 0u); // s192 add
|
||||
r6 = rotl_imm(r6, 22u); // s193 rotl
|
||||
r3 = r3 ^ r7; // s194 xor
|
||||
r7 = r7 * r0; // s195 mul
|
||||
r3 = r3 ^ r5; // s196 xor
|
||||
r5 = r5 + r1 + select(0x8519428cu, 0xeae84577u, ((sel >> 29u) & 1u) != 0u); // s197 add
|
||||
r7 = r4 * r4 + r7; // s198 mad
|
||||
r3 = r3 + r6 + select(0x468639d3u, 0x0bfdbfa1u, ((sel >> 8u) & 1u) != 0u); // s199 add
|
||||
r5 = r3 * r4 + r5; // s200 mad
|
||||
r2 = r2 ^ simd_shuffle_xor(r7, (ushort)4); // s201 shfl
|
||||
r7 = r7 + r4 + select(0x18ec9a69u, 0x53ee8e50u, ((sel >> 23u) & 1u) != 0u); // s202 add
|
||||
r6 = r6 ^ r0; // s203 xor
|
||||
r4 = r4 + r5 + select(0x3e81485bu, 0xac63376eu, ((sel >> 4u) & 1u) != 0u); // s204 add
|
||||
r7 = r1 * r4 + r7; // s205 mad
|
||||
r1 = r1 ^ simd_shuffle_xor(r5, (ushort)16); // s206 shfl
|
||||
r4 = r4 * r0; // s207 mul
|
||||
r1 = r1 | r6; // s208 or
|
||||
r6 = r6 + r4 + select(0x66ccb75du, 0x5b745519u, ((sel >> 28u) & 1u) != 0u); // s209 add
|
||||
r1 = rotr_var(r1, r3); // s210 rotr
|
||||
r5 = r5 - r4; // s211 sub
|
||||
r4 = r4 ^ r7; // s212 xor
|
||||
r1 = r1 + r7 + select(0x0ce0553du, 0x1c3dccf7u, ((sel >> 8u) & 1u) != 0u); // s213 add
|
||||
r0 = mulhi(r0, r1); // s214 mulhi
|
||||
r3 = r3 + r6 + select(0x96bfec89u, 0xfadc9205u, ((sel >> 9u) & 1u) != 0u); // s215 add
|
||||
r0 = rotl_imm(r0, 4u); // s216 rotl
|
||||
r6 = r6 ^ r4; // s217 xor
|
||||
r6 = r6 ^ simd_shuffle_xor(r7, (ushort)2); // s218 shfl
|
||||
r7 = r0 * r1 + r7; // s219 mad
|
||||
r4 = r4 * r0; // s220 mul
|
||||
r2 = r6 * r0 + r2; // s221 mad
|
||||
r5 = r5 ^ r4; // s222 xor
|
||||
r0 = r0 * r5; // s223 mul
|
||||
r2 = r2 + r0 + select(0x61495681u, 0x66b15c04u, ((sel >> 9u) & 1u) != 0u); // s224 add
|
||||
r3 = rotr_var(r3, r2); // s225 rotr
|
||||
r2 = r3 * r3 + r2; // s226 mad
|
||||
r6 = r6 ^ r0; // s227 xor
|
||||
r4 = rotl_imm(r4, 30u); // s228 rotl
|
||||
r2 = r2 + r3 + select(0x70d05c34u, 0x25bc17c2u, ((sel >> 29u) & 1u) != 0u); // s229 add
|
||||
r1 = rotr_var(r1, r5); // s230 rotr
|
||||
r1 = r1 + r4 + select(0x32a28384u, 0x2d8728f3u, ((sel >> 8u) & 1u) != 0u); // s231 add
|
||||
r1 = rotl_imm(r1, 6u); // s232 rotl
|
||||
r2 = r2 + r0 + select(0xcf0949f1u, 0x61e81fe6u, ((sel >> 21u) & 1u) != 0u); // s233 add
|
||||
r4 = rotl_imm(r4, 2u); // s234 rotl
|
||||
r2 = r2 - r6; // s235 sub
|
||||
r6 = r6 | r3; // s236 or
|
||||
r7 = r7 ^ simd_shuffle_xor(r5, (ushort)4); // s237 shfl
|
||||
r1 = r1 ^ r3; // s238 xor
|
||||
r2 = rotl_imm(r2, 1u); // s239 rotl
|
||||
r5 = rotr_var(r5, r2); // s240 rotr
|
||||
r4 = rotl_imm(r4, 10u); // s241 rotl
|
||||
r6 = r6 + r2 + select(0xe204fb50u, 0x0a67565du, ((sel >> 16u) & 1u) != 0u); // s242 add
|
||||
r1 = r1 ^ simd_shuffle_xor(r3, (ushort)2); // s243 shfl
|
||||
r0 = r0 ^ simd_shuffle_xor(r1, (ushort)16); // s244 shfl
|
||||
r5 = r5 ^ simd_shuffle_xor(r7, (ushort)2); // s245 shfl
|
||||
r4 = r4 ^ r5; // s246 xor
|
||||
r7 = r7 ^ r2; // s247 xor
|
||||
r0 = r0 ^ simd_shuffle_xor(r6, (ushort)16); // s248 shfl
|
||||
r1 = mulhi(r1, r3); // s249 mulhi
|
||||
r7 = r7 + r5 + select(0x3d6daffdu, 0x5a79fd6fu, ((sel >> 31u) & 1u) != 0u); // s250 add
|
||||
r6 = r6 + r0 + select(0x84334aeau, 0x6beb28c2u, ((sel >> 28u) & 1u) != 0u); // s251 add
|
||||
r7 = r7 - r1; // s252 sub
|
||||
r5 = rotr_var(r5, r2); // s253 rotr
|
||||
r2 = mulhi(r2, r5); // s254 mulhi
|
||||
r6 = r6 * r3; // s255 mul
|
||||
}
|
||||
}
|
||||
uint lo = r0 ^ rotl_imm(r1, 7u) ^ rotl_imm(r2, 14u) ^ rotl_imm(r3, 21u);
|
||||
uint hi = r4 ^ rotl_imm(r5, 9u) ^ rotl_imm(r6, 18u) ^ rotl_imm(r7, 27u);
|
||||
out[gid] = ((ulong)hi << 32) | (ulong)lo;
|
||||
}
|
||||
BIN
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/state.igsd1
Normal file
BIN
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/state.igsd1
Normal file
Binary file not shown.
57
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/vectors.h
Normal file
57
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/vectors.h
Normal file
|
|
@ -0,0 +1,57 @@
|
|||
// Generated by igneum-pow export (generator v2) for seed "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000". Do not edit by hand.
|
||||
// Expected outputs: igneum-pow (Rust) CPU interpreter, generator v5, memory-hard dataset
|
||||
#pragma once
|
||||
#ifdef __cplusplus
|
||||
#include <cstdint>
|
||||
#else
|
||||
#include <stdint.h>
|
||||
#endif
|
||||
|
||||
#define IGNEUM_VEC_WARPS 3
|
||||
static const uint32_t IGNEUM_VEC_BASE[IGNEUM_VEC_WARPS] = { 0u, 4096u, 1000000u };
|
||||
static const uint64_t IGNEUM_VEC_OUT[IGNEUM_VEC_WARPS][32] = {
|
||||
{ // base nonce 0
|
||||
0xe552166a03298f7full, 0xf526f1c5bbcce3feull, 0xd2599307c7dff26dull, 0x292f1d7c62935cefull, 0x176dd4b8dc247c8eull, 0x916c618e3f222f3cull, 0xc4b4cd32e2b17869ull, 0x083985b4514fdb52ull,
|
||||
0x471c78fc23623557ull, 0x3ef558f8f91fef48ull, 0xc3a23b4f40a9b42eull, 0xe0cad6d14d62a59eull, 0x23be7de4ce7cde95ull, 0x3eb342c0a40d848aull, 0x93bd93cc7f35be92ull, 0xd468fe59ee140d1aull,
|
||||
0x242f02aefcdd4907ull, 0x4a0af1c5c1807a38ull, 0x40cc94e1ba07667dull, 0xb20da649934a63f1ull, 0x172560c509523166ull, 0x7608ac11340f39efull, 0x4b401d953c834cbaull, 0x66017b835f9d80e2ull,
|
||||
0x9150f03bfeac28d3ull, 0xdf54e85028a3591aull, 0x637a68f8ad0a322full, 0x48f87298734fdb52ull, 0x65f3200db1282d59ull, 0x6c48d31c96caa0e4ull, 0x3b06a9d989aa3eb7ull, 0x316f5a49394af8abull
|
||||
},
|
||||
{ // base nonce 4096
|
||||
0xd2d4d57cca870a92ull, 0xc2a46c89e9e8d2e8ull, 0xe9d5a1048e9e357full, 0xa482ea1a68fb0fdfull, 0x9d4c01c50ebafb81ull, 0xe844ac46bf40758aull, 0x4fccaf252c2cbbf7ull, 0x90f96462f44878a4ull,
|
||||
0x07bad307b27ea868ull, 0xf93fba205fd5a9ccull, 0xc08d0da873abfff4ull, 0x5fcb36b12eabd4b3ull, 0x48cbb9465226c503ull, 0x7281def60f0c6982ull, 0x609c4a01b5ff6b09ull, 0x505e0d5083c5ebfaull,
|
||||
0x2e6370a4045ee3daull, 0x16df206ad3197bf6ull, 0xb63552c2e683f34aull, 0xb48d70f0f4fe8c3cull, 0xe4f941f4d5cac0bdull, 0xfe43ac9df2145859ull, 0x5964c115d0d9616full, 0xc2fb078d60b1b21cull,
|
||||
0x663a7905f56d4585ull, 0xfad4000340b1e774ull, 0x0e9cf62f4e5b5fe9ull, 0xa284122fdd4037e3ull, 0x0e24488e1fe86f6eull, 0x563f0ca5d9d78390ull, 0x3b9c6c5bb8ca2771ull, 0xe2dafada40c22022ull
|
||||
},
|
||||
{ // base nonce 1000000
|
||||
0x46d0145d98c681e2ull, 0x2909904cf57e5364ull, 0xb3686f5347a3920dull, 0x48c80e2427513b60ull, 0xae8b429006c2876aull, 0xdfeeddc0b4714cbbull, 0xa839ddb421d4b06bull, 0x77265ae1bde25c0bull,
|
||||
0x666fda9b8a119724ull, 0x2fba7b9e71ddd3b3ull, 0xc81c4e05c94a23feull, 0x49c52ee7e025f9bbull, 0x7951fad6f47e130dull, 0x9769f514ba0c5527ull, 0xe8cf1f573d573d41ull, 0x5f6c19433ba166d0ull,
|
||||
0x70806160896dc78bull, 0xbcbd57d0462c2b4full, 0xce880c91f263b526ull, 0x4bd3f39d60e5b6f3ull, 0x065159ea440706f5ull, 0xe4cbee1294943b62ull, 0xa2ce61093e845bdcull, 0xc6195798c0f2cdd0ull,
|
||||
0x1c4b63a0aea5fd36ull, 0xe5254c6a049e115aull, 0x9bdf1a4ca31ecf17ull, 0x64f8159b9c84ddd4ull, 0x78c6caf1a64a7293ull, 0xe360359048ecdc56ull, 0xf55b82c8ed94c781ull, 0x423ef7066859a0b0ull
|
||||
}
|
||||
};
|
||||
|
||||
// Dataset self-test: dataset[0..15] and dataset[IGNEUM_MASK] (268435455).
|
||||
static const uint32_t IGNEUM_DS_HEAD[16] = {
|
||||
0x0790a3c8u, 0xf061f4c1u, 0xc93a88c3u, 0x0b6b0fb8u, 0x4dc6214au, 0x7e397ebau, 0x62bdca7cu, 0x2cd12c0cu,
|
||||
0xb8a05ca1u, 0xe4e4e28fu, 0xb8623289u, 0x8244cdd4u, 0x12bb8880u, 0xd8cb0b94u, 0x6c8c05b9u, 0x26fd923bu
|
||||
};
|
||||
static const uint32_t IGNEUM_DS_LAST_INDEX = 268435455u;
|
||||
static const uint32_t IGNEUM_DS_LAST = 0xc5f96c2eu;
|
||||
// 64 sampled dataset words (index, value) computed on the Mac.
|
||||
#define IGNEUM_DS_SAMPLES 64
|
||||
static const uint32_t IGNEUM_DS_SAMPLE_INDEX[IGNEUM_DS_SAMPLES] = {
|
||||
59471966u, 217795994u, 208353206u, 42483309u, 172547758u, 148076330u, 183853158u, 214389424u, 267488061u, 169781097u, 184093494u, 153880993u, 84977930u, 46426879u, 3093825u, 225364072u, 44593546u, 260713159u, 168250303u, 52384140u, 223401610u, 45554030u, 95410555u, 175039924u, 79171087u, 267580473u, 24168642u, 37981670u, 171551130u, 195559979u, 204611762u, 140997658u, 138925853u, 86637313u, 20736778u, 219665210u, 160430336u, 264654675u, 8013395u, 228945585u, 213884386u, 104419827u, 44185464u, 142737231u, 99284897u, 132475900u, 61861762u, 132056166u, 262388043u, 91878046u, 117353561u, 124768597u, 71352993u, 190698941u, 46055428u, 55281366u, 165145231u, 106810753u, 171985651u, 232085256u, 159510492u, 40072060u, 209107596u, 39023794u
|
||||
};
|
||||
static const uint32_t IGNEUM_DS_SAMPLE_VALUE[IGNEUM_DS_SAMPLES] = {
|
||||
0x9dead642u, 0x94ef5e4au, 0x74892075u, 0xbc8466d3u, 0x5da3db7au, 0x5f7d336cu, 0xdb6e2d4bu, 0x707871e4u, 0xab1a8c26u, 0x095645e2u, 0x009e1047u, 0x0e043163u, 0x7f7a83f7u, 0xc80d8368u, 0x5c82379du, 0x6f040f0au, 0x9c15d89bu, 0x2e8158f1u, 0x899eb1dbu, 0x6b7b78ceu, 0x87c98f2du, 0x9bf58c2cu, 0xfe7c3758u, 0x135eb1c6u, 0x2241ea1cu, 0x58a5f34bu, 0x96891cc6u, 0xf3e92116u, 0xca2b5299u, 0xc48760fbu, 0xbfc8e6ecu, 0xf10dbadfu, 0x19672a34u, 0x2b38c7f0u, 0x2cc23ef9u, 0x732c2e9au, 0xa733b519u, 0xb4f99cc9u, 0xe0f2ceddu, 0xe405baa6u, 0x2515e05eu, 0xb5fe054eu, 0x62e77ca0u, 0x18a85e70u, 0x8c0deaf1u, 0x42a2f9c8u, 0x06e2ad43u, 0x14a20999u, 0x1d817d84u, 0x3493d381u, 0xc60d470bu, 0x6366a26bu, 0xfa46d099u, 0xa3d3c90eu, 0xcc73ca26u, 0x3ed2d84eu, 0xfca19a4fu, 0xa43b015eu, 0x1f5203b5u, 0x3bfb700cu, 0x1cb4554cu, 0x4c45d8afu, 0x271554b7u, 0xfb3d3fb8u
|
||||
};
|
||||
// Cache self-test (memory-hard mode): cache[0..15], the last 16 words, and FNV-1a 64 over all 2^26 words.
|
||||
static const uint32_t IGNEUM_CACHE_HEAD[16] = {
|
||||
0x47e15959u, 0xf138e163u, 0x4fc0aa19u, 0xbe50897cu, 0xddcc43a9u, 0xfbecd8bau, 0x3be5e048u, 0xc330cb77u,
|
||||
0xc00ff902u, 0x62eb64d7u, 0x825ef196u, 0x9bd34544u, 0x84c52834u, 0x6f7303eau, 0x5546e71fu, 0x3f0b835cu
|
||||
};
|
||||
static const uint32_t IGNEUM_CACHE_LAST[16] = {
|
||||
0xfe146921u, 0x0d972fc0u, 0xa5383d3eu, 0xf751aaf9u, 0x52509092u, 0x447ad742u, 0x18253e5cu, 0x15dae1fbu,
|
||||
0xff5b8c7du, 0x8914179bu, 0xdfec91fau, 0x80d3d462u, 0xec0a5caeu, 0xa8af9a8bu, 0x1b7935a6u, 0x1026962fu
|
||||
};
|
||||
static const uint64_t IGNEUM_CACHE_FNV64 = 0x7334fa46e5d972ebull;
|
||||
36
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/vectors.json
Normal file
36
proto-cuda/packs-ca3-v5/v5-dn3-epoch0/vectors.json
Normal file
|
|
@ -0,0 +1,36 @@
|
|||
{
|
||||
"seed": "igneum-epoch/4020cb4382e3fe4b281c817c02582e147d8f851f566ae9172b28912b8e68b925/day/69676e65756d2d6461792ffd50000000000000",
|
||||
"day": "bytes:69676e65756d2d6461792ffd50000000000000",
|
||||
"dataset_mode": "memory-hard",
|
||||
"dataset_log2_words": 28,
|
||||
"mask": "0x0fffffff",
|
||||
"lanes": 32,
|
||||
"source": "igneum-pow (Rust) CPU interpreter, generator v5, memory-hard dataset",
|
||||
"warps": [
|
||||
{"base_nonce": 0, "expected": [
|
||||
"0xe552166a03298f7f", "0xf526f1c5bbcce3fe", "0xd2599307c7dff26d", "0x292f1d7c62935cef", "0x176dd4b8dc247c8e", "0x916c618e3f222f3c", "0xc4b4cd32e2b17869", "0x083985b4514fdb52",
|
||||
"0x471c78fc23623557", "0x3ef558f8f91fef48", "0xc3a23b4f40a9b42e", "0xe0cad6d14d62a59e", "0x23be7de4ce7cde95", "0x3eb342c0a40d848a", "0x93bd93cc7f35be92", "0xd468fe59ee140d1a",
|
||||
"0x242f02aefcdd4907", "0x4a0af1c5c1807a38", "0x40cc94e1ba07667d", "0xb20da649934a63f1", "0x172560c509523166", "0x7608ac11340f39ef", "0x4b401d953c834cba", "0x66017b835f9d80e2",
|
||||
"0x9150f03bfeac28d3", "0xdf54e85028a3591a", "0x637a68f8ad0a322f", "0x48f87298734fdb52", "0x65f3200db1282d59", "0x6c48d31c96caa0e4", "0x3b06a9d989aa3eb7", "0x316f5a49394af8ab"
|
||||
]},
|
||||
{"base_nonce": 4096, "expected": [
|
||||
"0xd2d4d57cca870a92", "0xc2a46c89e9e8d2e8", "0xe9d5a1048e9e357f", "0xa482ea1a68fb0fdf", "0x9d4c01c50ebafb81", "0xe844ac46bf40758a", "0x4fccaf252c2cbbf7", "0x90f96462f44878a4",
|
||||
"0x07bad307b27ea868", "0xf93fba205fd5a9cc", "0xc08d0da873abfff4", "0x5fcb36b12eabd4b3", "0x48cbb9465226c503", "0x7281def60f0c6982", "0x609c4a01b5ff6b09", "0x505e0d5083c5ebfa",
|
||||
"0x2e6370a4045ee3da", "0x16df206ad3197bf6", "0xb63552c2e683f34a", "0xb48d70f0f4fe8c3c", "0xe4f941f4d5cac0bd", "0xfe43ac9df2145859", "0x5964c115d0d9616f", "0xc2fb078d60b1b21c",
|
||||
"0x663a7905f56d4585", "0xfad4000340b1e774", "0x0e9cf62f4e5b5fe9", "0xa284122fdd4037e3", "0x0e24488e1fe86f6e", "0x563f0ca5d9d78390", "0x3b9c6c5bb8ca2771", "0xe2dafada40c22022"
|
||||
]},
|
||||
{"base_nonce": 1000000, "expected": [
|
||||
"0x46d0145d98c681e2", "0x2909904cf57e5364", "0xb3686f5347a3920d", "0x48c80e2427513b60", "0xae8b429006c2876a", "0xdfeeddc0b4714cbb", "0xa839ddb421d4b06b", "0x77265ae1bde25c0b",
|
||||
"0x666fda9b8a119724", "0x2fba7b9e71ddd3b3", "0xc81c4e05c94a23fe", "0x49c52ee7e025f9bb", "0x7951fad6f47e130d", "0x9769f514ba0c5527", "0xe8cf1f573d573d41", "0x5f6c19433ba166d0",
|
||||
"0x70806160896dc78b", "0xbcbd57d0462c2b4f", "0xce880c91f263b526", "0x4bd3f39d60e5b6f3", "0x065159ea440706f5", "0xe4cbee1294943b62", "0xa2ce61093e845bdc", "0xc6195798c0f2cdd0",
|
||||
"0x1c4b63a0aea5fd36", "0xe5254c6a049e115a", "0x9bdf1a4ca31ecf17", "0x64f8159b9c84ddd4", "0x78c6caf1a64a7293", "0xe360359048ecdc56", "0xf55b82c8ed94c781", "0x423ef7066859a0b0"
|
||||
]}
|
||||
],
|
||||
"dataset_head": ["0x0790a3c8", "0xf061f4c1", "0xc93a88c3", "0x0b6b0fb8", "0x4dc6214a", "0x7e397eba", "0x62bdca7c", "0x2cd12c0c", "0xb8a05ca1", "0xe4e4e28f", "0xb8623289", "0x8244cdd4", "0x12bb8880", "0xd8cb0b94", "0x6c8c05b9", "0x26fd923b"],
|
||||
"dataset_last_index": 268435455,
|
||||
"dataset_last": "0xc5f96c2e",
|
||||
"dataset_samples": [{"index": 59471966, "value": "0x9dead642"}, {"index": 217795994, "value": "0x94ef5e4a"}, {"index": 208353206, "value": "0x74892075"}, {"index": 42483309, "value": "0xbc8466d3"}, {"index": 172547758, "value": "0x5da3db7a"}, {"index": 148076330, "value": "0x5f7d336c"}, {"index": 183853158, "value": "0xdb6e2d4b"}, {"index": 214389424, "value": "0x707871e4"}, {"index": 267488061, "value": "0xab1a8c26"}, {"index": 169781097, "value": "0x095645e2"}, {"index": 184093494, "value": "0x009e1047"}, {"index": 153880993, "value": "0x0e043163"}, {"index": 84977930, "value": "0x7f7a83f7"}, {"index": 46426879, "value": "0xc80d8368"}, {"index": 3093825, "value": "0x5c82379d"}, {"index": 225364072, "value": "0x6f040f0a"}, {"index": 44593546, "value": "0x9c15d89b"}, {"index": 260713159, "value": "0x2e8158f1"}, {"index": 168250303, "value": "0x899eb1db"}, {"index": 52384140, "value": "0x6b7b78ce"}, {"index": 223401610, "value": "0x87c98f2d"}, {"index": 45554030, "value": "0x9bf58c2c"}, {"index": 95410555, "value": "0xfe7c3758"}, {"index": 175039924, "value": "0x135eb1c6"}, {"index": 79171087, "value": "0x2241ea1c"}, {"index": 267580473, "value": "0x58a5f34b"}, {"index": 24168642, "value": "0x96891cc6"}, {"index": 37981670, "value": "0xf3e92116"}, {"index": 171551130, "value": "0xca2b5299"}, {"index": 195559979, "value": "0xc48760fb"}, {"index": 204611762, "value": "0xbfc8e6ec"}, {"index": 140997658, "value": "0xf10dbadf"}, {"index": 138925853, "value": "0x19672a34"}, {"index": 86637313, "value": "0x2b38c7f0"}, {"index": 20736778, "value": "0x2cc23ef9"}, {"index": 219665210, "value": "0x732c2e9a"}, {"index": 160430336, "value": "0xa733b519"}, {"index": 264654675, "value": "0xb4f99cc9"}, {"index": 8013395, "value": "0xe0f2cedd"}, {"index": 228945585, "value": "0xe405baa6"}, {"index": 213884386, "value": "0x2515e05e"}, {"index": 104419827, "value": "0xb5fe054e"}, {"index": 44185464, "value": "0x62e77ca0"}, {"index": 142737231, "value": "0x18a85e70"}, {"index": 99284897, "value": "0x8c0deaf1"}, {"index": 132475900, "value": "0x42a2f9c8"}, {"index": 61861762, "value": "0x06e2ad43"}, {"index": 132056166, "value": "0x14a20999"}, {"index": 262388043, "value": "0x1d817d84"}, {"index": 91878046, "value": "0x3493d381"}, {"index": 117353561, "value": "0xc60d470b"}, {"index": 124768597, "value": "0x6366a26b"}, {"index": 71352993, "value": "0xfa46d099"}, {"index": 190698941, "value": "0xa3d3c90e"}, {"index": 46055428, "value": "0xcc73ca26"}, {"index": 55281366, "value": "0x3ed2d84e"}, {"index": 165145231, "value": "0xfca19a4f"}, {"index": 106810753, "value": "0xa43b015e"}, {"index": 171985651, "value": "0x1f5203b5"}, {"index": 232085256, "value": "0x3bfb700c"}, {"index": 159510492, "value": "0x1cb4554c"}, {"index": 40072060, "value": "0x4c45d8af"}, {"index": 209107596, "value": "0x271554b7"}, {"index": 39023794, "value": "0xfb3d3fb8"}],
|
||||
"cache_head": ["0x47e15959", "0xf138e163", "0x4fc0aa19", "0xbe50897c", "0xddcc43a9", "0xfbecd8ba", "0x3be5e048", "0xc330cb77", "0xc00ff902", "0x62eb64d7", "0x825ef196", "0x9bd34544", "0x84c52834", "0x6f7303ea", "0x5546e71f", "0x3f0b835c"],
|
||||
"cache_last_line": ["0xfe146921", "0x0d972fc0", "0xa5383d3e", "0xf751aaf9", "0x52509092", "0x447ad742", "0x18253e5c", "0x15dae1fb", "0xff5b8c7d", "0x8914179b", "0xdfec91fa", "0x80d3d462", "0xec0a5cae", "0xa8af9a8b", "0x1b7935a6", "0x1026962f"],
|
||||
"cache_fnv1a64": "0x7334fa46e5d972eb"
|
||||
}
|
||||
Loading…
Reference in a new issue