Skip to content

[GEMM] Fix typo for streamk 3072x4096x3072 case #2898

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
whitneywhtsang merged 3 commits into from
Dec 3, 2024
Merged

[GEMM] Fix typo for streamk 3072x4096x3072 case #2898

Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
3f5c597
[GEMM] Fix typo for streamk 3072x4096x3072 case
whitneywhtsang Dec 2, 2024
7f56c6d
initialize results tensor
whitneywhtsang Dec 2, 2024
215783a
address review comment
whitneywhtsang Dec 3, 2024
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
19 changes: 10 additions & 9 deletions benchmarks/triton_kernels_benchmark/gemm_benchmark.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -268,6 +268,7 @@ def get_shapes(B, M, N, K, transpose_a, transpose_b):
def benchmark(B, M, N, K, provider):
a_shape, b_shape = get_shapes(B, M, N, K, transpose_a=TRANSPOSE_A, transpose_b=TRANSPOSE_B)

torch.manual_seed(0)
a = torch.rand(a_shape, device='xpu', dtype=torch.bfloat16)
b = torch.rand(b_shape, device='xpu', dtype=torch.bfloat16)

Expand All @@ -291,10 +292,10 @@ def benchmark(B, M, N, K, provider):
elif provider == 'triton':
assert len(a.shape) == len(b.shape), 'Incompatible sizes'
if len(a.shape) == 3:
c = torch.empty((B, M, N), device='xpu', dtype=torch.float32)
c = torch.zeros((B, M, N), device='xpu', dtype=torch.float32)
else:
assert len(a.shape) == 2, 'Expecting shape of length 2'
c = torch.empty((M, N), device='xpu', dtype=torch.float32)
c = torch.zeros((M, N), device='xpu', dtype=torch.float32)
triton_fn = lambda: matmul(a, b, c, transpose_a=TRANSPOSE_A, transpose_b=TRANSPOSE_B)
torch_fn = lambda: torch.matmul(torch_a, torch_b).to(torch.float32)
rtol = 1e-2 if a.dtype == torch.bfloat16 else 1e-3
Expand All @@ -304,17 +305,17 @@ def benchmark(B, M, N, K, provider):
kernel_name='matmul_kernel_with_block_pointers')
elif provider == 'xetla':
if B == 1:
c = torch.empty((M, N), device='xpu', dtype=torch.float32)
acc = torch.empty((M, N), device='xpu', dtype=torch.float32)
cnt = torch.empty((M, N), device='xpu', dtype=torch.int32)
c = torch.zeros((M, N), device='xpu', dtype=torch.float32)
acc = torch.zeros((M, N), device='xpu', dtype=torch.float32)
cnt = torch.zeros((M, N), device='xpu', dtype=torch.int32)
else:
c = torch.empty((B, M, N), device='xpu', dtype=torch.float32)
acc = torch.empty((B, M, N), device='xpu', dtype=torch.float32)
cnt = torch.empty((B, M, N), device='xpu', dtype=torch.int32)
c = torch.zeros((B, M, N), device='xpu', dtype=torch.float32)
acc = torch.zeros((B, M, N), device='xpu', dtype=torch.float32)
cnt = torch.zeros((B, M, N), device='xpu', dtype=torch.int32)
name = f'gemm_shape_{B}_{M}_{K}_{N}'
# FIXME: Use gemm_streamk_benchmark.py when Triton streamk can get
# better performance.
if (B, M, N, K) == (1, 3072, 4096, 3072):
if (B, M, K, N) == (1, 3072, 4096, 3072):
name = 'gemm_streamk_shape_3072_4096_3072'
func = getattr(xetla_kernel, name)
xetla_fn = lambda: func(a, b, c, acc, cnt)
Expand Down