[AMD] Remove deprecated macro from COnvUtils (#124158)

Summary:
This is not great, but our ATen-cpu is not completely GPU agnostic. Previously we have worked on D54453492 (https://github.com/pytorch/pytorch/pull/121082) and D54528255, but there are a few things we haven't resolved, and it's exploding here. So we'll continue to fix them until all are gone.

This ROCm block is for 4.3 which is very old. I don't think it should be supported any more. So let's just kill this macro

Test Plan: CI

Differential Revision: D56172660

Pull Request resolved: https://github.com/pytorch/pytorch/pull/124158
Approved by: https://github.com/jeffdaily, https://github.com/nmacchioni
This commit is contained in:
Xiaodong Wang
2024-04-19 19:00:31 +00:00
committed by PyTorch MergeBot
parent b0d83726bd
commit 8869b543e8
2 changed files with 0 additions and 3 deletions

View File

@ -2269,7 +2269,6 @@ exclude_patterns = [
"aten/src/ATen/Context.cpp",
"aten/src/ATen/DLConvertor.cpp",
"aten/src/ATen/core/Array.h",
"aten/src/ATen/native/ConvUtils.h",
"aten/src/ATen/native/quantized/ConvUtils.h",
"aten/src/ATen/native/sparse/SparseBlasImpl.cpp", # triton implementation
"aten/src/ATen/native/transformers/attention.cpp",

View File

@ -358,7 +358,6 @@ static inline bool miopen_conv_use_channels_last(const at::Tensor& input, const
}
bool can_use_miopen_channels_last_2d = false;
#if defined(USE_ROCM) && (ROCM_VERSION >= 40300)
// TODO: Remove PYTORCH_MIOPEN_SUGGEST_NHWC once ROCm officially supports NHWC in MIOpen
// See #64427
static c10::optional<bool> PYTORCH_MIOPEN_SUGGEST_NHWC = c10::utils::check_env("PYTORCH_MIOPEN_SUGGEST_NHWC");
@ -370,7 +369,6 @@ static inline bool miopen_conv_use_channels_last(const at::Tensor& input, const
( (input_memory_format == at::MemoryFormat::ChannelsLast) ||
(weight_memory_format == at::MemoryFormat::ChannelsLast) )
);
#endif
bool can_use_miopen_channels_last_3d = false;