biondizzle/DeepGEMM
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: remove duplicate kInt8 case — kPackedFP4 is already kInt8

Browse Source 
kPackedFP4 = torch::kInt8, so the kInt8 case was a duplicate.
The real fix was in mega_nvfp4.hpp: changing kUInt8→kInt8 so
tensors match the existing kPackedFP4 path in the TMA switch.
This commit is contained in:
biondizzle
2026-05-11 22:55:28 +00:00
parent 80df24a641
commit 49e5646b42
1 changed files with 0 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
csrc/jit_kernels/impls/runtime_utils.hpp
View File

@@ -82,7 +82,6 @@ static CUtensorMapDataType aten_dtype_to_tensor_map_dtype(const at::ScalarType&
case torch::kFloat: return CU_TENSOR_MAP_DATA_TYPE_FLOAT32;
case torch::kBFloat16: return CU_TENSOR_MAP_DATA_TYPE_BFLOAT16;
case torch::kFloat8_e4m3fn: return CU_TENSOR_MAP_DATA_TYPE_UINT8;
case torch::kInt8: return CU_TENSOR_MAP_DATA_TYPE_UINT8;
#if CUDA_VERSION >= 12080
case kPackedFP4: return fp4_unpacked_smem ? CU_TENSOR_MAP_DATA_TYPE_16U4_ALIGN16B
: CU_TENSOR_MAP_DATA_TYPE_16U4_ALIGN8B;