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

style(examples): typo (#1080)

Browse Source 
* Update ampere_tensorop_conv2dfprop.cu

learning cutlass, PR a typo.

* Update ampere_gemm_operand_reduction_fusion.cu
This commit is contained in:
tpoisonooo 2023-09-11 22:13:22 +08:00 committed by GitHub
parent 34bbadd3ff
commit a77b2c9cb8
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
examples/16_ampere_tensorop_conv2dfprop/ampere_tensorop_conv2dfprop.cu
View File

@ -53,7 +53,7 @@ can be used to form warp tiles (the tile shape each warp computes),
and multiple warp tiles can be used to compute threadblock tiles and multiple warp tiles can be used to compute threadblock tiles
(the tile shape computed by a threadblock). (the tile shape computed by a threadblock).
In thie example, we split variable initialization into two parts. In this example, we split variable initialization into two parts.
1. Setting up data properties: describes how tensors are laid out in the memory 1. Setting up data properties: describes how tensors are laid out in the memory
and how the kernel can view them (logical to physical mapping) and how the kernel can view them (logical to physical mapping)

2
examples/23_ampere_gemm_operand_reduction_fusion/ampere_gemm_operand_reduction_fusion.cu
View File

@ -30,7 +30,7 @@
**************************************************************************************************/ **************************************************************************************************/
/** /**
The example demenstrates how to reduce one of the operands of the GEMM along the k-dimension when The example demonstrates how to reduce one of the operands of the GEMM along the k-dimension when
computing GEMM. So the output also contains either a Mx1 or 1XN vector. It only works with Ampere computing GEMM. So the output also contains either a Mx1 or 1XN vector. It only works with Ampere
16x8x16 FP16/BF16 tensor cores, though it is not difficult to apply to other Turing/Ampere tensor 16x8x16 FP16/BF16 tensor cores, though it is not difficult to apply to other Turing/Ampere tensor
core instructions. core instructions.