util-docker: Update how rocBLAS is built in gcn-gpu Dockerfile
rocBLAS builds for all targets by default, but it requires we set the HCC_AMDGPU_TARGET env variable to all the targets we want to build for. This change explicitly sets all the gfx versions we intend to support in gem5, and removes the extraneous "-a all" from the install command. Additionally, doing this allowed for DNNMark to run on gfx902. Change-Id: Id1a00433beaa23f6935c12073d5bddc38431886b Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/54623 Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com> Maintainer: Matt Sinclair <mattdsinclair@gmail.com> Reviewed-by: Matthew Poremba <matthew.poremba@amd.com> Tested-by: kokoro <noreply+kokoro@google.com>
This commit is contained in:
committed by
Bobby Bruce
parent
398d4f9d07
commit
515c89b860
@@ -105,11 +105,11 @@ RUN apt install rocm-cmake
|
||||
RUN git clone -b rocm-4.0.0 \
|
||||
https://github.com/ROCmSoftwarePlatform/rocBLAS.git && mkdir rocBLAS/build
|
||||
|
||||
ENV HCC_AMDGPU_TARGET=gfx801
|
||||
ENV HCC_AMDGPU_TARGET=gfx801,gfx803,gfx900,gfx902
|
||||
WORKDIR rocBLAS
|
||||
# rocBLAS needs to be built from source otherwise gfx801 gets an error in HIP
|
||||
# rocBLAS needs to be built from source otherwise certain gfx versions get errors in HIP
|
||||
# about there being no GPU binary available
|
||||
RUN ./install.sh -d -a all -i
|
||||
RUN ./install.sh -d -i
|
||||
WORKDIR /
|
||||
|
||||
# MIOpen dependencies + MIOpen
|
||||
|
||||
Reference in New Issue
Block a user