CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
61 lines
3.7 KiB
C++
61 lines
3.7 KiB
C++
// Copyright © Advanced Micro Devices, Inc., or its affiliates.
|
|
// SPDX-License-Identifier: MIT
|
|
|
|
#pragma once
|
|
|
|
// Production-side workarounds for known MIOpen-provider issues. Each macro is
|
|
// keyed by the upstream issue number so it is easy to grep and remove once the
|
|
// underlying problem is fixed. Test-side counterparts live in
|
|
// `tests/common/TestWorkarounds.hpp` (kept separate so production TUs never
|
|
// pull in gtest).
|
|
//
|
|
// ----------------------------------------------------------------------------
|
|
// ROCm/rocm-libraries#5409 — MIOpen Conv-Bias-Activation (CBA) fusion produces
|
|
// incorrect/unsupported results on gfx90a (also covers 2-node CB and CA paths
|
|
// because all conv-fusion variants route through MiopenConvFwdBiasActivPlan-
|
|
// Builder). Until upstream fix lands, we early-return `false` from the plan
|
|
// builder's isApplicable() so engine selection skips the MIOpen CBA path on
|
|
// gfx90a.
|
|
//
|
|
// REJECT_IF_WORKAROUND_ISSUE_5409(handle) must only be invoked from a function
|
|
// whose return type is `bool` (or convertible from `false`) since it contains
|
|
// a `return`. CBA-specific test TUs observe the resulting "no engine applicable"
|
|
// outcome via the test-side issue 6979 helpers (see TestWorkarounds.hpp).
|
|
//
|
|
// Issue 5409 (this file) is gfx90a-specific. Issue 6979 (test-side) tracks the
|
|
// broader CBA capability gap across architectures — gfx90a (covered by 5409)
|
|
// plus gfx1151 / RDNA archs that lack CK fusion kernels. The two are decoupled:
|
|
// fixing 5409 only removes the gfx90a REJECT here; the 6979 test-side skip
|
|
// stays in place until CK fusion kernels exist for the remaining archs.
|
|
//
|
|
// To remove after the 5409 fix: delete this file, drop its includes, and
|
|
// remove the REJECT_IF_WORKAROUND_ISSUE_5409 call sites. The test-side
|
|
// 6979 helpers stay until that issue is independently resolved.
|
|
// `git grep WORKAROUND_ISSUE_5409` finds them all.
|
|
// ----------------------------------------------------------------------------
|
|
|
|
#include "MiopenUtils.hpp"
|
|
|
|
#include <hipdnn_plugin_sdk/PluginLogging.hpp>
|
|
|
|
#include <exception>
|
|
|
|
#define REJECT_IF_WORKAROUND_ISSUE_5409(handle) \
|
|
do \
|
|
{ \
|
|
try \
|
|
{ \
|
|
if(::miopen_plugin::miopen_utils::getDeviceArch((handle).getStream()) == "gfx90a") \
|
|
{ \
|
|
HIPDNN_PLUGIN_LOG_INFO("Plan builder disabled on gfx90a (issue 5409)"); \
|
|
return false; \
|
|
} \
|
|
} \
|
|
catch(const std::exception& issue_5409_e) \
|
|
{ \
|
|
HIPDNN_PLUGIN_LOG_INFO( \
|
|
"Arch query failed; treating as not-applicable: " << issue_5409_e.what()); \
|
|
return false; \
|
|
} \
|
|
} while(0)
|