gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[PYTHON] Added benchmarking code

Browse Source
This commit is contained in:
Philippe Tillet
2021-01-31 12:59:18 -05:00
parent 0b025db2ee
commit 9b31244897
1 changed files with 7 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
python/tests/test_matmul.py
View File

@@ -83,14 +83,16 @@ def do_bench(fn, flops = 0, warmup = 10, rep = 50):
def perf_op(dtype=th.float16, warmup=10, rep=50):
AT, BT = False, False
configs = [(N, N, N) for N in [128, 8192]]
for M, N, K in configs:
import pandas as pd
df = pd.DataFrame(columns=['AT', 'BT', 'N', 'TRITON', 'TORCH'])
Ns = [128, 256, 512, 1024, 1536, 2048, 3072, 4096, 6144, 8192]
configs = [(AT, BT, N, N, N) for AT in [False, True] for BT in [False, True] for N in Ns]
for AT, BT, M, N, K in configs:
a = th.randn((K, M) if AT else (M, K), device='cuda', dtype=dtype) / K**.5
b = th.randn((N, K) if BT else (K, N), device='cuda', dtype=dtype) / K**.5
if AT: a = a.t()
if BT: b = b.t()
a = a[::,::]
b = b[::,::]
TH_MS, TH_TFLOPS, _ = do_bench(lambda: th.matmul(a, b), flops = M*N*K*2, warmup = warmup, rep = rep)
TT_MS, TT_TFLOPS, _ = do_bench(lambda: tt.ops.matmul(a, b), flops = M*N*K*2, warmup = warmup, rep = rep)
print((M, N, K), TH_MS, TT_MS)
df = df.append({'AT': AT, 'BT': BT, 'N': N, 'TRITON': TT_TFLOPS, 'TORCH': TH_TFLOPS}, ignore_index=True)
print(df)