Prover floor: the miner-phase copy of the real-card playbook; route 2 design (what changes on chain: nothing; the pool rows)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
igneum-labs
parent
d1d7d19a06
commit
87ba8c6159
1 changed files with 100 additions and 0 deletions
100
tools/prover-floor/pc-12gb-card-miner.ps1
Normal file
100
tools/prover-floor/pc-12gb-card-miner.ps1
Normal file
|
|
@ -0,0 +1,100 @@
|
|||
# Prover floor, route 1 (6 October 2026, the project lead: a real 12 GB card lands today; can it mine AND prove?). The same
|
||||
# fixtures as sweeps 3 and 4 on the card itself: `--mode shard` (the core proof, its verify, then the compressed
|
||||
# proof, in one run; the 1-s sampler split at the core RESULT gives the core-only peak) at thresholds 2^26 and
|
||||
# 2^27, the v1 shard and an empty block. Two phases, each its own job: FLOOR_PHASE=alone (publish WITH
|
||||
# --stop-miners; measures the card's own idle first) and FLOOR_PHASE=miner (publish WITHOUT --stop-miners; measures
|
||||
# the miner's working set on the card first). Works on PC 1 or PC 2: the card is found by its memory (under
|
||||
# 13,000 MiB; FLOOR_CARD_INDEX overrides) and every proof runs on it through IGNEUM_CUDA_DEVICE (the host passes
|
||||
# the index to the SDK, which starts the server with CUDA_VISIBLE_DEVICES=<index>; CUDA_DEVICE_ORDER=PCI_BUS_ID
|
||||
# keeps nvidia-smi's and CUDA's numbering the same). The kit: /opt/igneum-floor/home/.sp1/bin/sp1-gpu-server (the
|
||||
# v3 build, pc2-build-server.ps1 on this PC) and the host built from the fetched package igneum-prove-wsl2-floor
|
||||
# (jobs\floor-kit\, a fetch job with --extract); each is tested first and named if missing. Never touches the
|
||||
# app's /opt/igneum host or /root/.sp1 server; leaves the prover ON; the runner restores the miners.
|
||||
$ErrorActionPreference = 'Continue'
|
||||
$urlFile = if ($env:IGNEUM_APP_DIR) { Join-Path $env:IGNEUM_APP_DIR 'app.url' } else { Join-Path $env:LOCALAPPDATA 'igneum\app\app.url' }
|
||||
if (-not (Test-Path $urlFile)) { $urlFile = Join-Path $env:LOCALAPPDATA 'igneum\app\app.url' }
|
||||
$base = (Get-Content $urlFile -Raw).Trim().TrimEnd('/')
|
||||
function Stamp { (Get-Date).ToUniversalTime().ToString('yyyy-MM-ddTHH:mm:ssZ') }
|
||||
function Prove($on) { try { (Invoke-RestMethod -Method Post -Uri "$base/api/prove" -ContentType 'application/json' -Body (@{on=$on} | ConvertTo-Json -Compress) -TimeoutSec 10) | ConvertTo-Json -Compress } catch { "error: $_" } }
|
||||
$phase = 'miner' # the beside-the-miner phase: publish WITHOUT --stop-miners
|
||||
"RESULT start $(Stamp) phase=$phase machine=$env:COMPUTERNAME"
|
||||
"RESULT cards $((& nvidia-smi --query-gpu=index,name,memory.total,memory.used,pci.bus_id --format=csv,noheader,nounits 2>$null) -join ' | ')"
|
||||
# the 12 GB card: the first index whose memory.total is under 13,000 MiB (a 3060 12 GB reads 12,288; a 4070 12,282)
|
||||
$cards = & nvidia-smi --query-gpu=index,memory.total --format=csv,noheader,nounits 2>$null | ForEach-Object { $p = $_ -split ',\s*'; [pscustomobject]@{ index = [int]$p[0]; total = [int]$p[1] } }
|
||||
$card = if ($env:FLOOR_CARD_INDEX) { [int]$env:FLOOR_CARD_INDEX } else { ($cards | Where-Object { $_.total -lt 13000 } | Select-Object -First 1).index }
|
||||
if ($null -eq $card) { "RESULT measure_failed no card under 13,000 MiB on this PC (set FLOOR_CARD_INDEX to force one)"; "RESULT end $(Stamp)"; exit 2 }
|
||||
"RESULT card index=$card total_mib=$(($cards | Where-Object { $_.index -eq $card }).total)"
|
||||
"RESULT prover_off $(Stamp) $(Prove $false)"
|
||||
Start-Sleep -Seconds 30
|
||||
$job = $env:IGNEUM_JOB_DIR; if (-not $job) { $job = Join-Path $env:TEMP 'igneum-floor-card' }; New-Item -ItemType Directory -Force -Path $job | Out-Null
|
||||
function WslPath($p) { $w = (& wsl.exe -d Ubuntu-24.04 -u root -- wslpath -a ($p -replace '\\', '/') 2>$null); if ($w) { ($w -replace "`0", '').Trim() } else { '/mnt/c' + ($p.Substring(2) -replace '\\', '/') } }
|
||||
$jobW = WslPath $job
|
||||
# the kit (the wiped-jobs-folder class): the fetched package under the jobs folder, tested before use
|
||||
$jobs = Split-Path $env:IGNEUM_JOB_DIR
|
||||
$kit = Join-Path $jobs 'floor-kit'
|
||||
$kitOk = Test-Path (Join-Path $kit 'igneum-prove-wsl2-floor\package\proving\igneum-prove\Cargo.toml')
|
||||
"RESULT kit $(if ($kitOk) { "present $kit" } else { "MISSING: publish the fetch job of igneum-prove-wsl2-floor.zip to this machine (--dir jobs --to floor-kit --extract) first" })"
|
||||
$kitW = if ($kitOk) { WslPath (Join-Path $kit 'igneum-prove-wsl2-floor\package') } else { '/nonexistent' }
|
||||
$bash = @'
|
||||
set -uo pipefail
|
||||
export PATH="$HOME/.cargo/bin:$PATH" CUDA_DEVICE_ORDER=PCI_BUS_ID
|
||||
CUDA_DIR="$(ls -d /usr/local/cuda-12.* 2>/dev/null | sort -V | tail -1 || true)"; [ -n "$CUDA_DIR" ] && export PATH="$CUDA_DIR/bin:$PATH" && export LD_LIBRARY_PATH="$CUDA_DIR/lib64:/usr/lib/wsl/lib:${LD_LIBRARY_PATH:-}"
|
||||
stamp() { date -u +%Y-%m-%dT%H:%M:%SZ; }
|
||||
JOB='JOBW_PLACEHOLDER'; KIT='KITW_PLACEHOLDER'; CARD=CARD_PLACEHOLDER; PHASE='PHASE_PLACEHOLDER'
|
||||
FLOORHOME=/opt/igneum-floor/home; SRV=$FLOORHOME/.sp1/bin/sp1-gpu-server; H=/opt/igneum-floor/host/igneum-prove-host
|
||||
echo "RESULT wsl_cards $(nvidia-smi --query-gpu=index,name,memory.total,pci.bus_id --format=csv,noheader,nounits 2>/dev/null | tr '\n' ';')"
|
||||
[ -x "$SRV" ] || { echo "RESULT measure_failed no patched server at $SRV: run the build job (tools/prover-floor/pc2-build-server.ps1) on this PC first (25 to 45 min cold)"; exit 2; }
|
||||
echo "RESULT patched_server sha256=$(sha256sum $SRV | cut -c1-16) version=$($SRV --version 2>/dev/null)"
|
||||
if [ ! -x "$H" ]; then
|
||||
[ -f "$KIT/proving/igneum-prove/Cargo.toml" ] || { echo "RESULT measure_failed no host and no kit: fetch igneum-prove-wsl2-floor.zip first"; exit 2; }
|
||||
echo "STAGE host build $(stamp)"
|
||||
mkdir -p /opt/igneum-floor/host-src && rsync -a "$KIT/" /opt/igneum-floor/host-src/ && find /opt/igneum-floor/host-src -type f -exec touch {} +
|
||||
TD=/root/igneum-prove/proving/igneum-prove/target; [ -d "$TD" ] || TD=/opt/igneum-floor/host-target
|
||||
( cd /opt/igneum-floor/host-src/proving/igneum-prove && CARGO_TARGET_DIR=$TD nice -n 19 cargo build --release -p igneum-prove-host --features igneum-prove-host/cuda > $JOB/host-build.log 2>&1 ) || { echo "RESULT measure_failed host build; tail:"; tail -n 30 $JOB/host-build.log; exit 2; }
|
||||
mkdir -p /opt/igneum-floor/host && cp $TD/release/igneum-prove-host /opt/igneum-floor/host/ && echo "RESULT host built sha256=$(sha256sum $H | cut -c1-16)"
|
||||
fi
|
||||
FX="/opt/igneum-floor/host-src/proving/fixtures"; [ -d "$FX" ] || FX="$KIT/proving/fixtures"
|
||||
V1="$FX/fees-v1-shards2.json"; EMPTY="$FX/block-72854-empty-block-first.json"
|
||||
echo "RESULT host_ids $($H --mode id 2>/dev/null | tr '\n' ' ' | cut -c1-200)"
|
||||
pkill -f sp1-gpu-server 2>/dev/null; sleep 2; rm -f /tmp/sp1-cuda-*.sock
|
||||
smi() { nvidia-smi -i $CARD --query-gpu=memory.used --format=csv,noheader,nounits | head -1; }
|
||||
echo "RESULT card_before phase=$PHASE used_mib=$(smi) (alone: the card's own idle; miner: idle + the miner's working set on this card)"
|
||||
runshard() { # name fixture env...
|
||||
local name="$1" fx="$2"; shift 2
|
||||
local tag="$name-$(basename $fx .json)"
|
||||
pkill -f sp1-gpu-server 2>/dev/null; sleep 2; rm -f /tmp/sp1-cuda-*.sock
|
||||
local csv="$JOB/smi-$tag.csv" log="$JOB/log-$tag.txt"
|
||||
nvidia-smi -i $CARD --query-gpu=timestamp,memory.used,utilization.gpu --format=csv,noheader,nounits -l 1 > "$csv" 2>/dev/null &
|
||||
local SMI=$!
|
||||
local t0=$(date +%s)
|
||||
env HOME=$FLOORHOME SP1_PROVER=cuda IGNEUM_CUDA_DEVICE=$CARD RUST_LOG=off SP1_GPU_FLOOR_LOG=1 "$@" $H "$fx" --mode shard --shard 0 --out "$JOB/res-$tag.json" > "$log" 2>&1
|
||||
local rc=$?
|
||||
local wall=$(( $(date +%s) - t0 ))
|
||||
pkill -f sp1-gpu-server 2>/dev/null; sleep 1; kill $SMI 2>/dev/null; sleep 1; rm -f /tmp/sp1-cuda-*.sock
|
||||
local n=$(wc -l < "$csv")
|
||||
local core=$(grep -E "^RESULT core shard" "$log" | tail -1); local comp=$(grep -E "^RESULT compressed shard" "$log" | tail -1)
|
||||
local cyc=$(grep -E "^RESULT execute shard" "$log" | tail -1 | sed -E 's/.*: ([0-9]+) cycles.*/\1/')
|
||||
local core_at=$(echo "$core" | sed -E 's/.* at ([0-9T:.-]+Z?)$/\1/'); local core_epoch=$(date -u -d "${core_at}" +%s 2>/dev/null || echo 0)
|
||||
local split=$(( core_epoch - t0 )); [ $split -lt 1 ] && split=$n
|
||||
local corepeak=$(awk -F', *' -v s=$split 'NR<=s+1 { if ($2+0 > m) m=$2+0 } END { print m+0 }' "$csv")
|
||||
local peak=$(awk -F', *' '{ if ($2+0 > m) m=$2+0 } END { print m+0 }' "$csv")
|
||||
local cl=$(echo "$core" | sed -E 's/.*prove ([0-9.]+) s, proof ([0-9]+) bytes, verify ([0-9.]+) s, ([A-Z ]+);.*/core_s=\1 core_bytes=\2 core_verify_s=\3 core_\4/')
|
||||
local pl=$(echo "$comp" | sed -E 's/.*prove ([0-9.]+) s, proof ([0-9]+) bytes, verify ([0-9.]+) s, ([A-Z ]+);.*/compressed_s=\1 compressed_bytes=\2 compressed_verify_s=\3 compressed_\4/')
|
||||
local err=$(grep -iE "error|panick|out of memory|OOM|unsupported" "$log" | grep -v "^FLOOR" | head -1 | cut -c1-200)
|
||||
echo "RESULT card12 phase=$PHASE cfg=$name fixture=$(basename $fx .json) card=$CARD core_peak_mib=$corepeak full_peak_mib=$peak split_s=$split samples=$n wall_s=$wall cycles=${cyc:-na} ${cl:-no_core_result} ${pl:-no_compressed_result} exit=$rc env='$*' ${err:+err=$err}"
|
||||
grep -E "^FLOOR (memory|opts|grow)|^RESULT cuda device" "$log" | sed "s/^/RESULT floorline phase=$PHASE cfg=$name fixture=$(basename $fx .json) /" | head -10
|
||||
}
|
||||
runshard e26 "$V1" SP1_GPU_ELEMENT_THRESHOLD=67108864
|
||||
runshard e26 "$EMPTY" SP1_GPU_ELEMENT_THRESHOLD=67108864
|
||||
runshard e27 "$V1" SP1_GPU_ELEMENT_THRESHOLD=134217728
|
||||
runshard e27 "$EMPTY" SP1_GPU_ELEMENT_THRESHOLD=134217728
|
||||
pkill -f sp1-gpu-server 2>/dev/null; sleep 1; rm -f /tmp/sp1-cuda-*.sock
|
||||
echo "RESULT card_after phase=$PHASE used_mib=$(smi)"
|
||||
echo "RESULT measure_end $(stamp)"
|
||||
'@
|
||||
$bash = $bash.Replace('JOBW_PLACEHOLDER', $jobW).Replace('KITW_PLACEHOLDER', $kitW).Replace('CARD_PLACEHOLDER', "$card").Replace('PHASE_PLACEHOLDER', $phase)
|
||||
$bashFile = Join-Path $job 'card12.sh'
|
||||
[IO.File]::WriteAllText($bashFile, ($bash -replace "`r`n", "`n"), (New-Object System.Text.UTF8Encoding $false))
|
||||
& wsl.exe -d Ubuntu-24.04 -u root -- bash (WslPath $bashFile) 2>&1 | ForEach-Object { ($_ -replace "`0", '') }
|
||||
"RESULT prover_on $(Stamp) $(Prove $true)"
|
||||
"RESULT end $(Stamp)"
|
||||
Loading…
Reference in a new issue