* The GCC used in the GCN-GPU images was increase from version 8 to
version 10. This was necessary due to PR #1145 which made GCC require
GCC >=10. This patch was previously part of #1161 but has been merged
into
this PR.
* A patch has been applied to ROCm-OpenCL-Runtime to fix a linking error
in which there were multiple definitions of `ret_val`. This issue is
highlighted here:
https://github.com/ROCm/ROCm-OpenCL-Runtime/issues/113.
This was previously part #1161 but has been moved into this PR.
* The Dockerfile's `RUN` command (built to layers in the Docker image)
have been refactored so sources and built objects are deleted in the
same RUN command as where they were built and installed. This reduces
the size of the image substantially: from 16.3GB down to 6.6GB.
* The `apt upgrade` has been removed. This step (previously at the start
of the file) did nothing of importance. Removing it saves both time
building the image and reduces the size of the image by a small amount.
* `--depth=1` is used when cloning repositories so the entire commit
tree
tree is not pulled each time. This saves some time when building the
image.
* `apt -y update` has been added where `apt -y install` is used so
CACHED image layers do not become an issue in the future if the image
were to be rebuilt.
This removes the gfx803 and gfx801 targets for building applications as
GCN3 will be removed from gem5. It also removes the copy/paste bug from
the HACC docker which is clobbering the HCC targets and removing gfx902.
Change-Id: I9a0d7fda437e797baf0f743a0a450948b9260b07
Co-authored-by: Harshil Patel <hpppatel@ucdavis.edu>
This adds the setting up of environment variables to the gcn-gpu
Dockerfile from the halofinder Dockerfile in gem5-resources
so that we don't need to use a seperate Docker image in the
gpu tests.
Change-Id: Ifcc7a4c6bbcd5289ce9561923366e9ed193f170c
gdb was originally part of the ROCm 1.6 Dockerfile a few years ago. It
got removed when updating to ROCm 4.0. This adds it back as being able
to debug things is quite useful.
Change-Id: I3f8148cde79e6cc5233fa3c8c830b64817f01d3a
Prior to this patch those building from these Dockerfiles could
encounter caching issues where the `apt -y update` RUN was loaded from a
cached layer prior to running the `install` command. Typically this was
trying to obtain a package from a wrong IP address. The fix for this is
to run this all in one Docker RUN to avoid loading a broken cache.
Change-Id: If309c5c1d4a0240fed670abe980772d90f7d2172
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/59350
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
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>
In the gcn-gpu image, rocBLAS wasn't able to be installed. This was
due to us not installing rocm-cmake, as rocBLAS is dependent on it
and will download the most recent version of rocm-cmake if it
isn't installed. The most recent version of rocm-cmake wasn't
compatible with the version of ROCm we're using.
This patch installs rocm-cmake before building and installing
rocBLAS instead of after.
Change-Id: Iaaa34d5e0d6594fddd0d1a7d147f43405163ca89
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/50847
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
Previously, we were using ROCm 1.6.2 as there were issues with some of
the machine learning applications that weren't present on 1.6.2.
However, after re-running them we've found that they, and all other
applications previously tested, run to completion.
Additionally, there have been patches to enable BLIT kernels which made
it so we no longer need to build HIP and MIOpen differently for APU and
DGPU code. This allows us to install HIP directly from the .deb packages
instead of from source. Installing from the .deb packages also avoid the
hipDeviceSynchronize() bug. Finally, this makes it so most GPU programs
can be run as-is without modifications to remove hipMalloc/hipMemcpy
calls as was done previously.
Change-Id: Ic61b09ed200b19f759d891487cde874abd607537
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/37675
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
The Python version installed in the Dockerfile for GCN3 by apt-get is
too old to build gem5. This bumps the version to the most recent Python
to avoid needing to update this file too much.
Python 3.9 is install via PPA since it is not available in the official
Ubuntu 16.04 repository. Likewise, pip is installed from "source" as it
is not available for Python 3.9 in from neither the PPA nor Ubuntu.
Change-Id: Ia919f31cf9c9063e1df091cea15590526715739b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/37219
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Daniel Gerzhoy <daniel.gerzhoy@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This dockerfile creates an image that installs the software stack needed
to run both machine learning and non-machine learning applications using
the GCN3 gpu model, while also applying patches to the software stack to
optimize machine learning applications, as well as APUs, which is the
current type of GPU in the GCN3 GPU model.
Change-Id: If36c2df1c00c895e27e9d741027fd10c17bf224e
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29192
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
