Sync with main, enforce no cache modifiers on remote stores

.devcontainer/devcontainer.json

@@ -9,7 +9,7 @@
     // Creates a stable agent socket at ~/.ssh/ssh-agent.sock and optionally loads ~/.ssh/id_rsa.
     "initializeCommand": "bash -lc \"bash '${localWorkspaceFolder}/.devcontainer/ensure-ssh-agent.sh'\"",
     "runArgs": [
-        "--name=${localEnv:USER}-iris-dev",
+        "--name=${localEnv:USER}-${localWorkspaceFolderBasename}-dev",
         "--network=host",
         "--device=/dev/kfd",
         "--device=/dev/dri",

.devcontainer/ensure-ssh-agent.sh

@@ -14,13 +14,29 @@ SOCK="${HOME}/.ssh/ssh-agent.sock"
 
 mkdir -p "${HOME}/.ssh"
 
+# Check if socket exists AND has keys loaded
 if [[ -S "${SOCK}" ]]; then
-  exit 0
+  if SSH_AUTH_SOCK="${SOCK}" ssh-add -l >/dev/null 2>&1; then
+    # Agent is running and has keys loaded
+    exit 0
+  fi
+
+  # Check if agent is alive but just has no keys
+  if SSH_AUTH_SOCK="${SOCK}" ssh-add -l 2>&1 | grep -q "no identities"; then
+    # Agent is alive, just needs keys loaded - continue to key loading below
+    :
+  else
+    # Agent is dead or socket is stale, remove it
+    rm -f "${SOCK}" 2>/dev/null || true
+  fi
 fi
 
-rm -f "${SOCK}"
-ssh-agent -a "${SOCK}" -t 8h >/dev/null
+# Start agent if socket doesn't exist
+if [[ ! -S "${SOCK}" ]]; then
+  ssh-agent -a "${SOCK}" -t 8h >/dev/null || true
+fi
 
+# Load SSH key if it exists
 if [[ -f "${HOME}/.ssh/id_rsa" ]]; then
   SSH_AUTH_SOCK="${SOCK}" ssh-add "${HOME}/.ssh/id_rsa" >/dev/null 2>&1 || true
 fi

.github/scripts/acquire_gpus.sh

@@ -0,0 +1,40 @@
+#!/bin/bash
+# SPDX-License-Identifier: MIT
+# Copyright (c) 2025 Advanced Micro Devices, Inc. All rights reserved.
+#
+# Acquire GPUs for CI workflows - to be called as a workflow step
+# Usage: acquire_gpus.sh <num_gpus>
+#
+# Exports GPU_DEVICES environment variable to $GITHUB_ENV for use in subsequent steps
+
+set -e
+
+NUM_GPUS=$1
+
+if [ -z "$NUM_GPUS" ]; then
+    echo "[ERROR] Missing required argument"
+    echo "Usage: $0 <num_gpus>"
+    exit 1
+fi
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+
+echo "[ACQUIRE-GPUS] Acquiring $NUM_GPUS GPU(s)"
+source "$SCRIPT_DIR/gpu_allocator.sh"
+acquire_gpus "$NUM_GPUS"
+
+echo "[ACQUIRE-GPUS] Allocated GPUs: $GPU_DEVICES"
+echo "[ACQUIRE-GPUS] GPU allocation details:"
+echo "  GPU_DEVICES=$GPU_DEVICES"
+echo "  ALLOCATED_GPU_BITMAP=$ALLOCATED_GPU_BITMAP"
+
+# Export to GITHUB_ENV so subsequent steps can use these variables
+if [ -n "$GITHUB_ENV" ]; then
+    {
+        echo "GPU_DEVICES=$GPU_DEVICES"
+        echo "ALLOCATED_GPU_BITMAP=$ALLOCATED_GPU_BITMAP"
+    } >> "$GITHUB_ENV"
+    echo "[ACQUIRE-GPUS] Exported variables to GITHUB_ENV"
+else
+    echo "[ACQUIRE-GPUS] WARNING: GITHUB_ENV not set, variables not exported"
+fi

.github/scripts/container_build.sh

@@ -32,34 +32,34 @@ echo "✅ /dev/shm size OK (${shm_size_gb}GB)"
 if [ "$CONTAINER_RUNTIME" = "apptainer" ]; then
     echo "[INFO] Building with Apptainer..."
 
-    # Create persistent Apptainer directory
-    mkdir -p ~/apptainer
-
-    # Define paths
-    IMAGE_PATH=~/apptainer/iris-dev.sif
-    DEF_FILE=apptainer/iris.def
-    CHECKSUM_FILE=~/apptainer/iris-dev.sif.checksum
-
     # Verify def file exists
+    DEF_FILE=apptainer/iris.def
     if [ ! -f "$DEF_FILE" ]; then
         echo "[ERROR] Definition file $DEF_FILE not found"
         exit 1
     fi
 
-    # Calculate checksum of the def file
-    NEW_CHECKSUM=$(sha256sum "$DEF_FILE" | awk '{print $1}')
+    # Calculate checksum of the def file to use as subdirectory name
+    DEF_CHECKSUM=$(sha256sum "$DEF_FILE" | awk '{print $1}')
+
+    # Create persistent Apptainer directory with checksum subdirectory
+    mkdir -p "${HOME}/iris-apptainer-images/${DEF_CHECKSUM}"
+
+    # Define paths
+    IMAGE_PATH="${HOME}/iris-apptainer-images/${DEF_CHECKSUM}/iris-dev.sif"
+    CHECKSUM_FILE="${HOME}/iris-apptainer-images/${DEF_CHECKSUM}/iris-dev.sif.checksum"
 
     # Check if image exists and has a valid checksum
     REBUILD_NEEDED=true
     if [ -f "$IMAGE_PATH" ] && [ -f "$CHECKSUM_FILE" ]; then
         OLD_CHECKSUM=$(head -n1 "$CHECKSUM_FILE" 2>/dev/null)
         # Validate checksum format (64 hex characters for SHA256)
-        if [[ "$OLD_CHECKSUM" =~ ^[a-f0-9]{64}$ ]] && [ "$OLD_CHECKSUM" = "$NEW_CHECKSUM" ]; then
-            echo "[INFO] Def file unchanged (checksum: $NEW_CHECKSUM)"
+        if [[ "$OLD_CHECKSUM" =~ ^[a-f0-9]{64}$ ]] && [ "$OLD_CHECKSUM" = "$DEF_CHECKSUM" ]; then
+            echo "[INFO] Def file unchanged (checksum: $DEF_CHECKSUM)"
             echo "[INFO] Skipping rebuild, using existing image at $IMAGE_PATH"
             REBUILD_NEEDED=false
         else
-            echo "[INFO] Def file changed (old: ${OLD_CHECKSUM:-<invalid>}, new: $NEW_CHECKSUM)"
+            echo "[INFO] Def file changed (old: ${OLD_CHECKSUM:-<invalid>}, new: $DEF_CHECKSUM)"
             echo "[INFO] Rebuilding Apptainer image..."
         fi
     else
@@ -70,9 +70,9 @@ if [ "$CONTAINER_RUNTIME" = "apptainer" ]; then
     if [ "$REBUILD_NEEDED" = true ]; then
         if apptainer build --force "$IMAGE_PATH" "$DEF_FILE"; then
             # Store the checksum only if build succeeded
-            echo "$NEW_CHECKSUM" > "$CHECKSUM_FILE"
+            echo "$DEF_CHECKSUM" > "$CHECKSUM_FILE"
             echo "[INFO] Built image: $IMAGE_PATH"
-            echo "[INFO] Checksum saved: $NEW_CHECKSUM"
+            echo "[INFO] Checksum saved: $DEF_CHECKSUM"
         else
             echo "[ERROR] Apptainer build failed"
             exit 1
@@ -83,14 +83,19 @@ elif [ "$CONTAINER_RUNTIME" = "docker" ]; then
     echo "[INFO] Checking Docker images..."
     # Use GitHub variable if set, otherwise default to iris-dev
     IMAGE_NAME=${DOCKER_IMAGE_NAME:-"iris-dev"}
-    
+
     # Check if the image exists
     if docker image inspect "$IMAGE_NAME" &> /dev/null; then
         echo "[INFO] Using existing Docker image: $IMAGE_NAME"
     else
-        echo "[WARNING] Docker image $IMAGE_NAME not found"
-        echo "[INFO] Please build it using: ./build_triton_image.sh"
-        echo "[INFO] Or pull it if available from registry"
+        echo "[INFO] Docker image $IMAGE_NAME not found, building..."
+        DOCKER_DIR="$(dirname "$(realpath "$0")")/../../docker"
+        if docker build -t "$IMAGE_NAME" "$DOCKER_DIR"; then
+            echo "[INFO] Built Docker image: $IMAGE_NAME"
+        else
+            echo "[ERROR] Docker build failed"
+            exit 1
+        fi
     fi
 fi
 

.github/scripts/container_exec.sh

@@ -51,13 +51,21 @@ if [ "$CONTAINER_RUNTIME" = "apptainer" ]; then
     # Find image
     if [ -n "$CUSTOM_IMAGE" ]; then
         IMAGE="$CUSTOM_IMAGE"
-    elif [ -f ~/apptainer/iris-dev.sif ]; then
-        IMAGE=~/apptainer/iris-dev.sif
-    elif [ -f apptainer/images/iris.sif ]; then
-        IMAGE="apptainer/images/iris.sif"
     else
-        echo "[ERROR] Apptainer image not found" >&2
-        exit 1
+        # Calculate checksum of def file to find the correct subdirectory
+        DEF_FILE=apptainer/iris.def
+        if [ ! -f "$DEF_FILE" ]; then
+            echo "[ERROR] Definition file $DEF_FILE not found" >&2
+            exit 1
+        fi
+        DEF_CHECKSUM=$(sha256sum "$DEF_FILE" | awk '{print $1}')
+
+        if [ -f "${HOME}/iris-apptainer-images/${DEF_CHECKSUM}/iris-dev.sif" ]; then
+            IMAGE="${HOME}/iris-apptainer-images/${DEF_CHECKSUM}/iris-dev.sif"
+        else
+            echo "[ERROR] Apptainer image not found" >&2
+            exit 1
+        fi
     fi
 
     # Create temporary overlay in workspace with unique name based on PID and timestamp
@@ -99,24 +107,11 @@ elif [ "$CONTAINER_RUNTIME" = "docker" ]; then
     fi
 
     # Build run command with proper GPU access
-    # Get video and render group IDs from host
-    VIDEO_GID=$(getent group video | cut -d: -f3)
-    RENDER_GID=$(getent group render | cut -d: -f3)
-
     RUN_CMD="docker run --rm --network=host --device=/dev/kfd --device=/dev/dri"
     RUN_CMD="$RUN_CMD --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
     RUN_CMD="$RUN_CMD -v ${PWD}:/iris_workspace -w /iris_workspace"
     RUN_CMD="$RUN_CMD --shm-size=16G --ulimit memlock=-1 --ulimit stack=67108864"
-    RUN_CMD="$RUN_CMD --user $(id -u):$(id -g)"
-
-    # Add video and render groups for GPU access
-    if [ -n "$VIDEO_GID" ]; then
-        RUN_CMD="$RUN_CMD --group-add $VIDEO_GID"
-    fi
-    if [ -n "$RENDER_GID" ]; then
-        RUN_CMD="$RUN_CMD --group-add $RENDER_GID"
-    fi
-
+
     RUN_CMD="$RUN_CMD -e HOME=/iris_workspace"
     RUN_CMD="$RUN_CMD --entrypoint bash"
 

.github/scripts/examples_config.json

@@ -0,0 +1,5 @@
+{
+  "31_message_passing": {
+    "required_ranks": 2
+  }
+}