github/ROCm
1
1
Fork 0
mirror of https://github.com/ROCm/ROCm.git synced 2026-04-05 03:01:17 -04:00
Code Issues Packages Projects Releases Wiki Activity

Enable usage of block pointer semantics for AMD gpus (#301)

Browse Source 
* Enable usage of block pointer semantics for AMD gpus

This commit enables usage of block pointer semantics by enabling
rewrite_tensor_pointer_pass that rewrites block pointer loads/stores
to legacy loads/stores.

* Update FA fwd in tutorial to use the block pointers

* use 90 compute capability for amd gpus in python/triton/compiler/compiler.py

Co-authored-by: Alexander Efimov <efimov.alexander@gmail.com>

---------

Co-authored-by: Ognjen Plavsic <ognjen.plavsic@dxc.com>
Co-authored-by: Lixun Zhang <lixun.zhang@amd.com>
Co-authored-by: Aleksandr Efimov <130555951+alefimov-amd@users.noreply.github.com>
Co-authored-by: Alexander Efimov <efimov.alexander@gmail.com>
This commit is contained in:
jayfurmanek
2023-08-24 13:05:12 -05:00
committed by GitHub
parent fa429316d4
commit ff7e707f87
6 changed files with 171 additions and 123 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
python/test/unit/language/test_block_pointer.py
View File

@@ -23,7 +23,7 @@ def block_copy_kernel(a_ptr, b_ptr, N, BLOCK_SIZE: tl.constexpr, padding_option:
for padding in ("zero", "nan")])
def test_block_copy(dtype_str, n, padding_option):
capability = torch.cuda.get_device_capability()
if capability[0] >= 9:
if torch.version.hip is None and capability[0] >= 9:
pytest.skip("Hopper support is working in progress")
dtype = getattr(torch, dtype_str)
@@ -82,7 +82,7 @@ def matmul_no_scf_with_advance_kernel(
])
def test_block_ptr_matmul_no_scf(shape, num_warps):
capability = torch.cuda.get_device_capability()
if capability[0] >= 9:
if torch.version.hip is None and capability[0] >= 9:
pytest.skip("Hopper support is working in progress")
m, n, k = shape