Skip to content

[Benchmark] Add matmul FP16 benchmark results #1067

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Oct 19, 2025
Merged

[Benchmark] Add matmul FP16 benchmark results #1067

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 36 additions & 0 deletions benchmark/matmul/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,36 @@
# FP16 Matmul Benchmark (8192×8192)

This document records the throughput achieved by `benchmark_matmul.py` when multiplying FP16 matrices sized `M = N = 8192` across different `K` dimensions using the default autotuning search space.

## Environment

- Repository commit: `17bd0a6c651f599bec1397e0b91830c3ddc93076`
- GPUs: `NVIDIA H800 SXM` on driver `560.35.05`

## How to Reproduce

```bash
cd benchmark/matmul
python - <<'PY'
from benchmark_matmul import matmul

M = 8192
N = 8192
for K in [256, 512, 1024, 2048, 4096, 8192, 16384]:
res = matmul(M, N, K, False)
tflops = 2 * M * N * K / res.latency * 1e-12
print(f"K={K:5d} latency={res.latency:.6f}s TFlops={tflops:.3f}")
PY
```

## Results

| K | Latency (s) | Throughput (TFLOPs) |
|-------|-------------|---------------------|
| 256 | 0.089056 | 386 |
| 512 | 0.132064 | 520 |
| 1024 | 0.218816 | 628 |
| 2048 | 0.390112 | 705 |
| 4096 | 0.746752 | 736 |
| 8192 | 1.449888 | 758 |
| 16384 | 2.871168 | 766 |
3 changes: 3 additions & 0 deletions benchmark/matmul/benchmark_matmul.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
import itertools
import logging

import tilelang
import tilelang.language as T
from tilelang.autotuner import autotune
from tilelang import jit
Expand Down Expand Up @@ -187,6 +188,8 @@ def main(

# Enable (or disable) swizzling optimization
T.use_swizzle(panel_size=10, enable=enable_rasteration)
# to utilize swizzle tma layout
T.annotate_layout({C_shared: tilelang.layout.make_swizzled_layout(C_shared)})

# Clear out the accumulation buffer
T.clear(C_local)
Expand Down
14 changes: 7 additions & 7 deletions benchmark/matmul_fp8/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,10 +27,10 @@ PY

| K | Latency (s) | Throughput (TFLOPs) |
|-------|-------------|---------------------|
| 256 | 0.091488 | 376 |
| 512 | 0.110496 | 622 |
| 1024 | 0.148256 | 927 |
| 2048 | 0.234080 | 1174 |
| 4096 | 0.398944 | 1378 |
| 8192 | 0.752416 | 1461 |
| 16384 | 1.443808 | 1523 |
| 256 | 0.060352 | 569 |
| 512 | 0.080096 | 858 |
| 1024 | 0.121696 | 1129 |
| 2048 | 0.204672 | 1343 |
| 4096 | 0.374816 | 1467 |
| 8192 | 0.729664 | 1507 |
| 16384 | 1.427264 | 1541 |
4 changes: 3 additions & 1 deletion benchmark/matmul_fp8/benchmark_matmul.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
import argparse
import itertools
import logging

import tilelang
import tilelang.language as T
from tilelang.autotuner import autotune
from tilelang import jit
Expand Down Expand Up @@ -190,6 +190,8 @@ def main(

# Enable (or disable) swizzling optimization
T.use_swizzle(panel_size=10, enable=enable_rasteration)
# to utilize swizzle tma layout
T.annotate_layout({C_shared: tilelang.layout.make_swizzled_layout(C_shared)})

# Clear out the accumulation buffer
T.clear(C_local)
Expand Down