Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

src:cpu:aarch64 restricts the conv operation to just one post-op when… #1689

Closed
wants to merge 2 commits into from
Closed

src:cpu:aarch64 restricts the conv operation to just one post-op when… #1689

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
746fb35
src:cpu:aarch64 restricts the conv operation to just one post-op when…
renato-arantes Jun 20, 2023
5e4e68c
Merge branch 'oneapi-src:master' into fp16_postops
renato-arantes Sep 22, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions src/cpu/aarch64/acl_convolution_utils.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -317,6 +317,11 @@ status_t init_conf_indirect_gemm(acl_conv_conf_t &acp, memory_desc_t &src_md,
if (dnnl_get_max_threads() < 28 && !acp.fast_math)
return status::unimplemented;

// Restricts the conv operation to just one post-op when the src/wei/dst
// data type is fp16
if (attr.post_ops_.len() > 1 && dst_md.data_type == data_type::f16)
return status::unimplemented;

// If we do not need to pad input channels for fast math mode then it would
// be faster to run convolution with im2row instead of using indirect kernel
int block_by = arm_compute::block_by(acp.weights_info.weight_format());
Expand Down
5 changes: 4 additions & 1 deletion src/cpu/aarch64/acl_gemm_convolution.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,7 +80,10 @@ struct acl_gemm_convolution_fwd_t : public primitive_t {
src_type, wei_type, bia_type, dst_type, undef)
&& !has_zero_dim_memory()
&& attr()->has_default_values(smask_t::post_ops, dst_type)
&& output_scales_mask_ok() && zero_points_ok();
&& output_scales_mask_ok() && zero_points_ok()
&& !(attr_.post_ops_.len() > 1
&& dst_md_.data_type == data_type::f16);

if (!ok) return status::unimplemented;

CHECK(acl_convolution_utils::init_conf_gemm(acp_, src_md_,
Expand Down