DeepSeek/DeepGEMM
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepGEMM synced 2025-06-26 23:15:49 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Compatible with padding

Browse Source
This commit is contained in:
Chenggang Zhao
2025-04-11 17:38:22 +08:00
parent 4c111418a2
commit 23d4289365
2 changed files with 9 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
deep_gemm/jit/compiler.py
View File

@@ -101,7 +101,7 @@ def build(name: str, arg_defs: tuple, code: str) -> Runtime:
'-gencode=arch=compute_90a,code=sm_90a',
'--ptxas-options=--register-usage-level=10' + (',--verbose' if 'DG_PTXAS_VERBOSE' in os.environ else ''),
# Suppress some unnecessary warnings, such as unused variables for certain `constexpr` branch cases
'--diag-suppress=177,174,940']
'--diag-suppress=39,174,177,940']
cxx_flags = ['-fPIC', '-O3', '-Wno-deprecated-declarations', '-Wno-abi', '-fconcepts']
flags = [*nvcc_flags, f'--compiler-options={",".join(cxx_flags)}']
include_dirs = [get_jit_include_dir()]