Skip to content

Commit

Permalink
hotfix: amd gemv was flaky
Browse files Browse the repository at this point in the history
  • Loading branch information
geohot committed Dec 2, 2024
1 parent 4e46c67 commit 6c1efb9
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion test/external/speed_v_theoretical.py
Original file line number Diff line number Diff line change
Expand Up @@ -91,7 +91,7 @@ def test_gemm_4096(self): self._test_matmul(4096, nv_tflops=95, amd_tflops=70)
def test_gemm_8192(self): self._test_matmul(8192, nv_tflops=130, amd_tflops=70)

def test_gemv_16384_4096(self): self._test_matmul(16384, 4096, 1, nv_gbs=430, amd_gbs=400)
def test_gemv_4096_16384(self): self._test_matmul(4096, 16384, 1, nv_gbs=430, amd_gbs=400)
def test_gemv_4096_16384(self): self._test_matmul(4096, 16384, 1, nv_gbs=430, amd_gbs=380) # AMD was flaky at 400

# TODO: tiny7 is slower than tiny12
def test_conv_3x3_256_32_32_256_256(self): self._test_conv_3x3(256, 32, 32, 256, 256, nv_tflops=27, amd_tflops=18)
Expand Down

0 comments on commit 6c1efb9

Please sign in to comment.