leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

Adding more Threadblock Tiles for Mixed-input TensorOp (BF16 * S8) in cutlass_library (#1132)

Browse Source 
* Adding more tiles in the cutlass_library for mixed-input support.

* fix rebase issue

* more tiles to upcast a
This commit is contained in:
Manish Gupta
2023-10-13 11:33:15 -04:00
committed by GitHub
parent fa8dfe631f
commit 757275f279
14 changed files with 830 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/unit/gemm/device/CMakeLists.txt
+2
View File
@@ -350,10 +350,12 @@ cutlass_test_unit_add_executable(
# Upcast on Operand A
gemm_universal_u8t_f16n_f16t_mixed_input_tensor_op_f16_sm80.cu
gemm_universal_s8t_f16n_f16t_mixed_input_tensor_op_f16_sm80.cu
gemm_universal_s8t_bf16n_bf16t_mixed_input_tensor_op_f32_sm80.cu
# Upcast on Operand B
gemm_universal_f16t_u8n_f16t_mixed_input_tensor_op_f16_sm80.cu
gemm_universal_f16t_s8n_f16t_mixed_input_tensor_op_f16_sm80.cu
gemm_universal_bf16t_s8n_bf16t_mixed_input_tensor_op_f32_sm80.cu
)
cutlass_test_unit_add_executable(