Linux RDMA and InfiniBand development
 help / color / mirror / Atom feed
From: Stephen Bates <sbates@raithlin.com>
To: linux-rdma@vger.kernel.org
Cc: Stephen Bates <sbates@raithlin.com>,
	sshaul@nvidia.com, stephen.bates@amd.com,
	abhijit.gangurde@amd.com, nikhil.agarwal@amd.com
Subject: [PATCH v1 2/2] perftest: extend ROCm CI to compile device code per target arch
Date: Mon,  9 Nov 2026 15:42:14 -0600	[thread overview]
Message-ID: <20261109214215.3152974-2-sbates@raithlin.com> (raw)
In-Reply-To: <20261109214215.3152974-1-sbates@raithlin.com>

The ROCm CI job added previously builds everything but passes no
--with-rocm-arch flag, so the hip kernel is compiled for the host
only. Now that --with-rocm-arch exists and the 7.2.4 slim image ships
device support for gfx950 and gfx1250, fan the job into one leg per
architecture.

fail-fast: false keeps both legs running so a device-specific failure
in one does not hide the other leg's result.

Signed-off-by: Stephen Bates <sbates@raithlin.com>
---
 .github/workflows/ubuntu24_04_rocm7_2.yaml | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/ubuntu24_04_rocm7_2.yaml b/.github/workflows/ubuntu24_04_rocm7_2.yaml
index e8f3ea5..3a3ce03 100644
--- a/.github/workflows/ubuntu24_04_rocm7_2.yaml
+++ b/.github/workflows/ubuntu24_04_rocm7_2.yaml
@@ -15,11 +15,19 @@ jobs:
 
     # AMD's slim ROCm image. The -full and -complete tags carry the whole ML
     # stack -- Composable Kernel alone is over 2 GB -- which a compile check
-    # never touches. This one is about 1.1 GB and still ships hipcc and the
-    # HIP and HSA headers needed to compile the ROCm backend.
+    # never touches. This one is about 1.1 GB and still ships hipcc, the HIP
+    # and HSA headers, and device support for gfx950 and gfx1250.
     container: rocm/dev-ubuntu-24.04:7.2.4
 
-    name: ROCm
+    strategy:
+      # Report every architecture's result. A failure in generated device code
+      # is far more likely to be specific to one target than common to both,
+      # and cancelling the other leg hides exactly the comparison that matters.
+      fail-fast: false
+      matrix:
+        rocm_arch: [gfx950, gfx1250]
+
+    name: ROCm (${{ matrix.rocm_arch }})
 
     steps:
       - name: Checkout repository
@@ -63,7 +71,8 @@ jobs:
       - name: Configure the build
         run: |
           ./configure --enable-rocm --enable-rocm-dmabuf \
-                      --with-rocm="$ROCM_PATH"
+                      --with-rocm="$ROCM_PATH" \
+                      --with-rocm-arch=${{ matrix.rocm_arch }}
 
       - name: Build perftest
         run: make
-- 
2.53.0



      reply	other threads:[~2026-11-09 21:42 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-09 21:42 [PATCH v1 1/2] perftest: add data validation for ROCm device memory Stephen Bates
2026-11-09 21:42 ` Stephen Bates [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261109214215.3152974-2-sbates@raithlin.com \
    --to=sbates@raithlin.com \
    --cc=abhijit.gangurde@amd.com \
    --cc=linux-rdma@vger.kernel.org \
    --cc=nikhil.agarwal@amd.com \
    --cc=sshaul@nvidia.com \
    --cc=stephen.bates@amd.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox