Skip to content

[Issue]: ATOM nightly docker can not run on gfx942 #69

@sunway513

Description

@sunway513

Problem Description

I've seen the following error message when trying to execute the rocm/atom:latest docker on gfx942 system:

:3:hip_fatbin.cpp           :537 : 2500313198678 us:  Device name: amdgcn-amd-amdhsa--gfx942:sramecc+:xnack- Generic name: amdgcn-amd-amdhsa--gfx9-4-generic:sramecc+:xnack-
:1:hip_fatbin.cpp           :694 : 2500313198684 us:  No compatible code objects found for: gfx942:sramecc+:xnack-, value of HIP_FORCE_SPIRV_CODEOBJECT: 0
[aiter] import [module_rmsnorm] under /aiter-test/aiter/jit/module_rmsnorm.so
:3:hip_device_runtime.cpp   :687 : 2500313199379 us:   hipGetDevice ( 0x7ffeb5815144 )
:3:hip_device_runtime.cpp   :699 : 2500313199387 us:  hipGetDevice: Returned hipSuccess : 0
:3:hip_device_runtime.cpp   :687 : 2500313199395 us:   hipGetDevice ( 0x7ffeb5815004 )

If I reinstall AITER there, the issue can be fixed. That implies the AITER package was not correctly built there.
Recommend to update the prebuilt AITER package with both gfx942 and gfx950 targets, and refresh.

Operating System

ubuntu

CPU

epyc

GPU

gfx942

ROCm Version

rocm7.1.1

ROCm Component

No response

Steps to Reproduce

No response

(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support

No response

Additional Information

No response

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions