ViperEkura/AstrAI
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

chore: relocate kernel benchmarks to csrc/bench

Browse Source 
- Move benchmark_gemv.py, benchmark_swiglu.py, and benchmark_gemv_common.py from scripts/tools/ to csrc/bench/ so kernel benchmarks live next to the kernels they measure
- Update reproduction commands in decode_linear_benchmark.md, swiglu_benchmark.md, and cuda_kernels.md
- Codify the placement convention in AGENTS.md: kernel benchmarks in csrc/bench/, pure-CUDA harnesses in csrc/tests/*.cu, engine and evaluation benchmarks in scripts/
This commit is contained in:
ViperEkura
2026-09-03 21:06:33 +08:00
parent 76f1c10feb
commit 28d11f1610
6 changed files with 6 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/tools/benchmark_gemv.py → csrc/bench/benchmark_gemv.py
View File
scripts/tools/benchmark_gemv_common.py → csrc/bench/benchmark_gemv_common.py
View File
scripts/tools/benchmark_swiglu.py → csrc/bench/benchmark_swiglu.py
View File
docs/developer/cuda_kernels.md
+1 -1
View File
@@ -84,7 +84,7 @@ OPT 1.3B M=1 is +4.54%. Qwen2 and LLaMA 3 70B M=1, and all three new
families at M=8, remain exact PyTorch fallbacks. families at M=8, remain exact PyTorch fallbacks.
These are synthetic projection-chain measurements, not whole-model throughput These are synthetic projection-chain measurements, not whole-model throughput
claims. Reproduce them with `scripts/tools/benchmark_gemv_common.py`. claims. Reproduce them with `csrc/bench/benchmark_gemv_common.py`.
The AstrAI 1B A→B→B→A results use the real `InferenceEngine`, including scheduler, The AstrAI 1B A→B→B→A results use the real `InferenceEngine`, including scheduler,
sampling, and CUDA Graph. M=8 stays on PyTorch because its remaining sampling, and CUDA Graph. M=8 stays on PyTorch because its remaining
docs/developer/decode_linear_benchmark.md
+3 -3
View File
@@ -1,6 +1,6 @@
# Decode linear shape benchmark # Decode linear shape benchmark
`scripts/tools/benchmark_gemv.py` records the `F.linear` baseline used to decide `csrc/bench/benchmark_gemv.py` records the `F.linear` baseline used to decide
whether a BF16 GEMV or small-M kernel should enter automatic inference dispatch. whether a BF16 GEMV or small-M kernel should enter automatic inference dispatch.
It does not change model execution or select a custom kernel. It does not change model execution or select a custom kernel.
@@ -10,7 +10,7 @@ modes. Results include device-event latency samples, p50/p90/p99, estimated
effective IO bandwidth, and CUDA kernel launches per call. effective IO bandwidth, and CUDA kernel launches per call.
```bash ```bash
CUDA_VISIBLE_DEVICES=0 python scripts/tools/benchmark_gemv.py \ CUDA_VISIBLE_DEVICES=0 python csrc/bench/benchmark_gemv.py \
--output results/decode_linear.json \ --output results/decode_linear.json \
--markdown-output results/decode_linear.md --markdown-output results/decode_linear.md
``` ```
@@ -25,7 +25,7 @@ For direct A/B coverage of the custom kernel and guarded dispatcher across
traditional LLaMA and GPT-NeoX decode shapes, use: traditional LLaMA and GPT-NeoX decode shapes, use:
```bash ```bash
CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python scripts/tools/benchmark_gemv_common.py \ CUDA_VISIBLE_DEVICES=0 PYTHONPATH=. python csrc/bench/benchmark_gemv_common.py \
--suite all --family traditional --m 2 4 \ --suite all --family traditional --m 2 4 \
--output results/gemv_common.json --output results/gemv_common.json
``` ```
docs/developer/swiglu_benchmark.md
+2 -2
View File
@@ -1,12 +1,12 @@
# Fused SwiGLU benchmark # Fused SwiGLU benchmark
`scripts/tools/benchmark_swiglu.py` compares the directly callable fused BF16 `csrc/bench/benchmark_swiglu.py` compares the directly callable fused BF16
SwiGLU primitive with both `F.linear` and the existing two-GEMV chain. It covers SwiGLU primitive with both `F.linear` and the existing two-GEMV chain. It covers
the native AstrAI 1B MLP plus LLaMA 2 7B/13B, LLaMA 3 8B, and GPT-NeoX 20B the native AstrAI 1B MLP plus LLaMA 2 7B/13B, LLaMA 3 8B, and GPT-NeoX 20B
up/gate shapes at M=1/2/4/8 in eager and CUDA Graph modes. up/gate shapes at M=1/2/4/8 in eager and CUDA Graph modes.
```bash ```bash
CUDA_VISIBLE_DEVICES=0 python scripts/tools/benchmark_swiglu.py \ CUDA_VISIBLE_DEVICES=0 python csrc/bench/benchmark_swiglu.py \
--output results/swiglu.json \ --output results/swiglu.json \
--markdown-output results/swiglu.md \ --markdown-output results/swiglu.md \
--m-values 1,2,4,8 --mode both \ --m-values 1,2,4,8 --mode both \