mirror of
https://github.com/luxfi/gpu.git
synced 2026-10-30 16:23:49 +00:00
zeekayandHanzo Dev
b1e39595f7
Standalone microbenchmarks for the deterministic CLOB matcher (one thread per book, sequential price-time priority, parity-checked vs OpenMP CPU) across all three GPU makers: CUDA, HIP, Metal. Optimization: order-major SoA layout coalesces warp reads (adjacent books read contiguous words instead of DEPTH-strided). Pure memory-access change; identical deterministic fills (parity MATCH at every N). Measured, tuned per backend (taker orders/sec): - AMD Radeon 8060S (HIP, SoA preload TPB=64): 12.76B (1.94B baseline, 6.6x) - NVIDIA GB10 (CUDA, SoA lowreg TPB=32): 9.13B (1.54B baseline, 5.9x) - Apple M4 Max (Metal, AoS TPB=256): 5.60B (SoA neutral on unified mem) - Apple M1 Max (Metal): 2.80B Findings: coalescing is a discrete/APU-memory win (neutral-to-negative on Apple unified memory); NVIDIA is occupancy-bound (prefers lowreg/small blocks) while AMD is bandwidth-rich (prefers register-resident preload) — opposite optima. Includes dex_tune.cu (TPB x variant sweep) + per-backend build/tune scripts. Co-authored-by: Hanzo Dev <[email protected]>
20 lines
1.1 KiB
Bash
Executable File
20 lines
1.1 KiB
Bash
Executable File
#!/bin/bash
|
|
# build_metal.sh — build the deterministic per-book DEX matcher for Apple Silicon (Metal).
|
|
# Verified on ra (Apple M1 Max, 64GB) and dbc (Apple M4 Max, 128GB).
|
|
# NOTE: on Apple unified memory the coalesced-SoA rewrite is NEUTRAL-to-NEGATIVE (the large
|
|
# shared cache already absorbs the strided AoS access); AoS is the tuned default here. Both are
|
|
# built so you can confirm on your own silicon.
|
|
# Usage: ./build_metal.sh [N] (N = number of books/markets, default 2000000)
|
|
set -e
|
|
LIBOMP=${LIBOMP:-$( (command -v brew >/dev/null && brew --prefix libomp) 2>/dev/null || echo /opt/homebrew/opt/libomp)}
|
|
SDK=$(xcrun --show-sdk-path)
|
|
here=$(cd "$(dirname "$0")" && pwd)
|
|
N=${1:-2000000}
|
|
cc() { xcrun clang++ -O3 -std=c++17 -isysroot "$SDK" -I"$LIBOMP/include" -L"$LIBOMP/lib" \
|
|
-Xpreprocessor -fopenmp -lomp -framework Metal -framework Foundation -o "$1" "$2"; }
|
|
cc /tmp/dex_books "$here/dex_books.mm"
|
|
cc /tmp/dex_books_opt "$here/dex_books_opt.mm"
|
|
export DYLD_LIBRARY_PATH=$LIBOMP/lib
|
|
echo "--- baseline (AoS) — tuned default on Apple ---"; /tmp/dex_books "$N"
|
|
echo "--- coalesced SoA (discrete-GPU optimization) ---"; /tmp/dex_books_opt "$N"
|