From c9cb48203cbd3a1728d427e780570d9bb9692926 Mon Sep 17 00:00:00 2001
From: Joe Schoonover <joe@fluidnumerics.com>
Date: Wed, 10 Jun 2026 17:49:16 -0400
Subject: [PATCH] Add single-precision and MI300X variants to docker-publish
 workflow

Extend the GitHub Actions docker-publish matrix so each published image
is available in both single and double precision, and add MI300X
(gfx942) as a new published target.

- docker/x86_sm100/Dockerfile, docker/x86_gfx942/Dockerfile: add a
  SELF_DOUBLE_PRECISION build-arg (default ON) wired to the
  SELF_ENABLE_DOUBLE_PRECISION CMake option, so the same Dockerfile
  builds either precision.
- docker/x86_gfx942/Dockerfile: new, modeled on the gfx90a Dockerfile,
  targeting gfx942 via the latest-x86-rocm643-gfx942 selfish base image.
- docker-publish.yml: pass SELF_DOUBLE_PRECISION through build-args and
  add matrix entries. Published tags:
    latest-x86-cuda130-sm100        (double, unchanged)
    latest-x86-cuda130-sm100-sp     (single, new)
    latest-x86-rocm643-gfx942       (double, new)
    latest-x86-rocm643-gfx942-sp    (single, new)

Existing tags are unchanged for backward compatibility; single-precision
images use the -sp suffix. Each variant keeps a separate buildcache tag.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---
 .github/workflows/docker-publish.yml | 17 +++++++++++++++
 docker/x86_gfx942/Dockerfile         | 31 ++++++++++++++++++++++++++++
 docker/x86_sm100/Dockerfile          |  5 +++++
 3 files changed, 53 insertions(+)
 create mode 100644 docker/x86_gfx942/Dockerfile

diff --git a/.github/workflows/docker-publish.yml b/.github/workflows/docker-publish.yml
index 71f711a2e..e675dfe60 100644
--- a/.github/workflows/docker-publish.yml
+++ b/.github/workflows/docker-publish.yml
@@ -30,6 +30,22 @@ jobs:
             context: docker/x86_sm100
             selfish_image: higherordermethods/selfish:latest-x86-cuda130-sm100
             tag_suffix: x86-cuda130-sm100
+            double_precision: "ON"
+          - variant: x86-cuda130-sm100-sp
+            context: docker/x86_sm100
+            selfish_image: higherordermethods/selfish:latest-x86-cuda130-sm100
+            tag_suffix: x86-cuda130-sm100-sp
+            double_precision: "OFF"
+          - variant: x86-rocm643-gfx942
+            context: docker/x86_gfx942
+            selfish_image: higherordermethods/selfish:latest-x86-rocm643-gfx942
+            tag_suffix: x86-rocm643-gfx942
+            double_precision: "ON"
+          - variant: x86-rocm643-gfx942-sp
+            context: docker/x86_gfx942
+            selfish_image: higherordermethods/selfish:latest-x86-rocm643-gfx942
+            tag_suffix: x86-rocm643-gfx942-sp
+            double_precision: "OFF"
 
     steps:
       - name: Checkout repository
@@ -70,5 +86,6 @@ jobs:
           build-args: |
             SELFISH_IMAGE=${{ matrix.selfish_image }}
             SELFISH_SHA=${{ steps.selfish.outputs.sha }}
+            SELF_DOUBLE_PRECISION=${{ matrix.double_precision }}
           cache-from: type=registry,ref=${{ env.REGISTRY_IMAGE }}:buildcache-${{ matrix.tag_suffix }}
           cache-to: type=registry,ref=${{ env.REGISTRY_IMAGE }}:buildcache-${{ matrix.tag_suffix }},mode=max
diff --git a/docker/x86_gfx942/Dockerfile b/docker/x86_gfx942/Dockerfile
new file mode 100644
index 000000000..f88d2a4fa
--- /dev/null
+++ b/docker/x86_gfx942/Dockerfile
@@ -0,0 +1,31 @@
+ARG SELFISH_IMAGE=higherordermethods/selfish:latest-x86-rocm643-gfx942
+FROM ${SELFISH_IMAGE}
+
+ARG SELFISH_SHA=""
+LABEL org.opencontainers.image.base.name="${SELFISH_IMAGE}"
+LABEL org.opencontainers.image.base.digest="${SELFISH_SHA}"
+
+# Floating-point precision of the build. ON -> double precision (default),
+# OFF -> single precision. Maps to the SELF_ENABLE_DOUBLE_PRECISION CMake option.
+ARG SELF_DOUBLE_PRECISION=ON
+
+COPY . /opt/self/src
+
+RUN source /opt/spack-environment/activate.sh && \
+    mkdir -p /opt/self/build && \
+    cd /opt/self/build && \
+    FC=gfortran cmake \
+      -DCMAKE_INSTALL_PREFIX=/opt/self/install \
+      -DCMAKE_BUILD_TYPE=Release \
+      -DSELF_ENABLE_GPU=ON \
+      -DSELF_GPU_BACKEND=HIP \
+      -DSELF_ENABLE_TESTING=ON \
+      -DSELF_ENABLE_DOUBLE_PRECISION=${SELF_DOUBLE_PRECISION} \
+      -DCMAKE_HIP_ARCHITECTURES="gfx942" \
+      -DGPU_TARGETS="gfx942" \
+      -DSELF_ENABLE_EXAMPLES=ON \
+      /opt/self/src && \
+    make -j$(nproc) && \
+    make install
+
+WORKDIR /opt/self/build
diff --git a/docker/x86_sm100/Dockerfile b/docker/x86_sm100/Dockerfile
index 6d23cedaa..49363b38a 100644
--- a/docker/x86_sm100/Dockerfile
+++ b/docker/x86_sm100/Dockerfile
@@ -5,6 +5,10 @@ ARG SELFISH_SHA=""
 LABEL org.opencontainers.image.base.name="${SELFISH_IMAGE}"
 LABEL org.opencontainers.image.base.digest="${SELFISH_SHA}"
 
+# Floating-point precision of the build. ON -> double precision (default),
+# OFF -> single precision. Maps to the SELF_ENABLE_DOUBLE_PRECISION CMake option.
+ARG SELF_DOUBLE_PRECISION=ON
+
 COPY . /opt/self/src
 
 RUN source /opt/spack-environment/activate.sh && \
@@ -16,6 +20,7 @@ RUN source /opt/spack-environment/activate.sh && \
       -DSELF_ENABLE_GPU=ON \
       -DSELF_GPU_BACKEND=CUDA \
       -DSELF_ENABLE_TESTING=ON \
+      -DSELF_ENABLE_DOUBLE_PRECISION=${SELF_DOUBLE_PRECISION} \
       -DCMAKE_CUDA_ARCHITECTURES="100" \
       -DSELF_ENABLE_EXAMPLES=ON \
       /opt/self/src && \