Add GCC run to intrinsic-test

.github/workflows/main.yml

@@ -281,10 +281,19 @@ jobs:
           - armv7-unknown-linux-gnueabihf
           - x86_64-unknown-linux-gnu
         profile: [dev, release]
+        cc: [clang, gcc]
         include:
           - target: aarch64_be-unknown-linux-gnu
             build_std: true
-
+          - target: x86_64-unknown-linux-gnu
+            cc: icx
+            profile: dev
+          - target: x86_64-unknown-linux-gnu
+            cc: icx
+            profile: release
+        exclude:
+          - target: armv7-unknown-linux-gnueabihf
+            cc: gcc
     steps:
     - uses: actions/checkout@v6
     - name: Install Rust
@@ -301,7 +310,7 @@ jobs:
 
     # Configure some env vars based on matrix configuration
     - run: echo "PROFILE=${{ matrix.profile }}" >> $GITHUB_ENV
-    - run: ./ci/intrinsic-test-docker.sh ${{ matrix.target }}
+    - run: ./ci/intrinsic-test-docker.sh ${{ matrix.target }} ${{ matrix.cc }}
       if: ${{ !startsWith(matrix.target, 'thumb') }}
       env:
         TARGET: ${{ matrix.target }}

ci/docker/aarch64-unknown-linux-gnu/Dockerfile

@@ -15,7 +15,8 @@ RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.4-x86_64
 RUN mkdir llvm
 RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
 
-ENV PATH="/llvm/bin:$PATH"
+ENV CLANG_PATH="/llvm/bin/clang"
+ENV GCC_PATH=aarch64-linux-gnu-gcc
 
 ENV CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_LINKER=aarch64-linux-gnu-gcc \
     CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_RUNNER="qemu-aarch64 -cpu max -L /usr/aarch64-linux-gnu" \

ci/docker/aarch64_be-unknown-linux-gnu/Dockerfile

@@ -23,11 +23,12 @@ RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.4-x86_64
 RUN mkdir llvm
 RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
 
-ENV PATH="/llvm/bin:$PATH"
-
 ENV AARCH64_BE_TOOLCHAIN="/toolchains/${TOOLCHAIN}"
 ENV AARCH64_BE_LIBC="${AARCH64_BE_TOOLCHAIN}/aarch64_be-none-linux-gnu/libc"
 
+ENV CLANG_PATH="/llvm/bin/clang"
+ENV GCC_PATH="${AARCH64_BE_TOOLCHAIN}/bin/aarch64_be-none-linux-gnu-gcc"
+
 ENV CARGO_TARGET_AARCH64_BE_UNKNOWN_LINUX_GNU_LINKER="${AARCH64_BE_TOOLCHAIN}/bin/aarch64_be-none-linux-gnu-gcc"
 ENV CARGO_TARGET_AARCH64_BE_UNKNOWN_LINUX_GNU_RUNNER="qemu-aarch64_be -cpu max -L ${AARCH64_BE_LIBC}"
 ENV OBJDUMP="${AARCH64_BE_TOOLCHAIN}/bin/aarch64_be-none-linux-gnu-objdump"

ci/docker/armv7-unknown-linux-gnueabihf/Dockerfile

@@ -14,7 +14,8 @@ RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.4-x86_64
 RUN mkdir llvm
 RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
 
-ENV PATH="/llvm/bin:$PATH"
+ENV CLANG_PATH="/llvm/bin/clang"
+ENV GCC_PATH=arm-linux-gnueabihf-gcc
 
 ENV CARGO_TARGET_ARMV7_UNKNOWN_LINUX_GNUEABIHF_LINKER=arm-linux-gnueabihf-gcc \
     CARGO_TARGET_ARMV7_UNKNOWN_LINUX_GNUEABIHF_RUNNER="qemu-arm -cpu max -L /usr/arm-linux-gnueabihf" \

ci/docker/x86_64-unknown-linux-gnu/Dockerfile

@@ -6,7 +6,8 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
   make \
   ca-certificates \
   wget \
-  xz-utils
+  xz-utils \
+  gpg
 
 RUN wget http://ci-mirrors.rust-lang.org/sde-external-10.8.0-2026-03-15-lin.tar.xz -O sde.tar.xz
 RUN mkdir intel-sde
@@ -16,7 +17,18 @@ RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.4-x86_64
 RUN mkdir llvm
 RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
 
-ENV PATH="/llvm/bin:$PATH"
+RUN wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB |\
+    gpg --dearmor |\
+    tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null
+
+RUN echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" |\
+    tee /etc/apt/sources.list.d/oneAPI.list
+
+RUN apt-get update && apt-get install -y --no-install-recommends intel-oneapi-compiler-dpcpp-cpp
+
+ENV CLANG_PATH="/llvm/bin/clang"
+ENV GCC_PATH="gcc"
+ENV ICX_PATH="/opt/intel/oneapi/compiler/2026.0/bin/icx"
 
 ENV CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER="/intel-sde/sde64 \
             -cpuid-in /checkout/ci/docker/x86_64-unknown-linux-gnu/cpuid.def \

ci/intrinsic-test-docker.sh

@@ -5,8 +5,8 @@
 
 set -ex
 
-if [ $# -lt 1 ]; then
-    >&2 echo "Usage: $0 <TARGET>"
+if [ $# -lt 2 ]; then
+    >&2 echo "Usage: $0 <TARGET> <CC>"
     exit 1
 fi
 
@@ -29,7 +29,6 @@ run() {
       --user "$(id -u)":"$(id -g)" \
       --env CARGO_HOME=/cargo \
       --env CARGO_TARGET_DIR=/checkout/target \
-      --env TARGET="${1}" \
       --env PROFILE \
       --env "${HOST_LINKER}"="cc" \
       --env STDARCH_DISABLE_ASSERT_INSTR \
@@ -48,12 +47,12 @@ run() {
       --workdir /checkout \
       --privileged \
       stdarch \
-      sh -c "HOME=/tmp PATH=\$PATH:/rust/bin exec ci/intrinsic-test.sh"
+      sh -c "HOME=/tmp PATH=\$PATH:/rust/bin exec ci/intrinsic-test.sh ${1} ${2}"
 }
 
 if [ -z "$1" ]; then
   >&2 echo "No target specified!"
   exit 1
 else
-  run "${1}"
+  run "${1}" "${2}"
 fi

ci/intrinsic-test.sh

@@ -2,7 +2,29 @@
 
 set -ex
 
-: "${TARGET?The TARGET environment variable must be set.}"
+if [ $# -lt 2 ]; then
+    >&2 echo "Usage: $0 <TARGET> <CC>"
+    exit 1
+fi
+
+case ${2} in
+    clang)
+        export CC="${CLANG_PATH}"
+        CC_KIND=clang
+        ;;
+    gcc)
+        export CC="${GCC_PATH}"
+        CC_KIND=gcc
+        ;;
+    icx)
+        export CC="${ICX_PATH}"
+        CC_KIND=clang
+        ;;
+    *)
+        >&2 echo "Unknown compiler: ${2}"
+        exit 1
+        ;;
+esac
 
 export RUSTFLAGS="${RUSTFLAGS} -D warnings -Z merge-functions=disabled -Z verify-llvm-ir"
 export PROFILE="${PROFILE:="release"}"
@@ -12,49 +34,49 @@ echo "PROFILE=${PROFILE}"
 
 INTRINSIC_TEST="--manifest-path=crates/intrinsic-test/Cargo.toml"
 
-export CC="clang"
-
-case ${TARGET} in
+case ${1} in
     aarch64_be*)
         export CFLAGS="-I${AARCH64_BE_TOOLCHAIN}/aarch64_be-none-linux-gnu/libc/usr/include --sysroot={AARCH64_BE_TOOLCHAIN}/aarch64_be-none-linux-gnu/libc -Wno-nonportable-vector-initialization"
-        TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_aarch64_be.txt
+        ARCH=aarch64_be
         ;;
 
     aarch64*)
         export CFLAGS="-I/usr/aarch64-linux-gnu/include/"
-        TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_aarch64.txt
+        ARCH=aarch64
         ;;
 
     armv7*)
         export CFLAGS="-I/usr/arm-linux-gnueabihf/include/"
-        TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_arm.txt
+        ARCH=arm
         ;;
 
     x86_64*)
         export CFLAGS="-I/usr/include/x86_64-linux-gnu/"
-        TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_x86.txt
+        ARCH=x86
         ;;
     *)
         ;;
 
 esac
 
-case "${TARGET}" in
+case "${1}" in
     x86_64-unknown-linux-gnu*)
         env -u CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER \
             cargo run "${INTRINSIC_TEST}" --release \
             --bin intrinsic-test -- intrinsics_data/x86-intel.xml \
-            --skip "${TEST_SKIP_INTRINSICS}" \
-            --target "${TARGET}"
-
-        echo "${CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER}"
+            --skip "crates/intrinsic-test/missing_${ARCH}_common.txt" \
+            --skip "crates/intrinsic-test/missing_${ARCH}_${2}.txt" \
+            --target "${1}" \
+            --cc-kind "${CC_KIND}"
         ;;
     *)
         cargo run "${INTRINSIC_TEST}" --release \
             --bin intrinsic-test -- intrinsics_data/arm_intrinsics.json \
-            --skip "${TEST_SKIP_INTRINSICS}" \
-            --target "${TARGET}"
+            --skip "crates/intrinsic-test/missing_${ARCH}_common.txt" \
+            --skip "crates/intrinsic-test/missing_${ARCH}_${2}.txt" \
+            --target "${1}" \
+            --cc-kind "${CC_KIND}"
         ;;
 esac
 
-cargo test --manifest-path=rust_programs/Cargo.toml --target "${TARGET}" --profile "${PROFILE}"
+cargo test --manifest-path=rust_programs/Cargo.toml --target "${1}" --profile "${PROFILE}"

crates/intrinsic-test/missing_aarch64_be_clang.txt

@@ -0,0 +1,31 @@
+# Bad LLVM codegen for BE in O2 in clang, and release in rust (https://github.com/llvm/llvm-project/issues/166190)
+vcmla_laneq_f16
+vcmla_rot180_laneq_f16
+vcmla_rot270_laneq_f16
+vcmla_rot90_laneq_f16
+vcmlaq_lane_f16
+vcmlaq_laneq_f16
+vcmlaq_rot180_lane_f16
+vcmlaq_rot180_laneq_f16
+vcmlaq_rot270_lane_f16
+vcmlaq_rot270_laneq_f16
+vcmlaq_rot90_lane_f16
+vcmlaq_rot90_laneq_f16
+
+# Bad codegen for BE in O2 in clang, correct in rust. Same cause as above issue.
+vdot_lane_s32
+vdot_lane_u32
+vdot_laneq_s32
+vdot_laneq_u32
+vdotq_lane_s32
+vdotq_lane_u32
+vdotq_laneq_s32
+vdotq_laneq_u32
+vsudot_lane_s32
+vsudot_laneq_s32
+vsudotq_lane_s32
+vsudotq_laneq_s32
+vusdot_lane_s32
+vusdot_laneq_s32
+vusdotq_lane_s32
+vusdotq_laneq_s32

crates/intrinsic-test/missing_aarch64.txt → crates/intrinsic-test/missing_aarch64_be_common.txt

@@ -79,6 +79,3 @@ vcvtns_s64_f32
 vcvtns_u64_f32
 vcvtps_s64_f32
 vcvtps_u64_f32
-
-# Broken in Clang (fixed in https://github.com/llvm/llvm-project/pull/156029)
-vcvth_s16_f16

crates/intrinsic-test/missing_aarch64_be_gcc.txt

@@ -0,0 +1,20 @@
+# Broken in LLVM llvm/llvm-project#196999
+vmull_p64
+vmull_high_p64
+
+# Broken in LLVM llvm/llvm-project#197083
+vcvth_n_s32_f16
+vcvth_n_u32_f16
+vcvth_n_s64_f16
+vcvth_n_u64_f16
+vcvth_n_f16_s32
+vcvth_n_f16_u32
+vcvth_n_f16_s64
+vcvth_n_f16_u64
+
+# Broken in GCC https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125279
+vmaxh_f16
+vminh_f16
+
+# Rounding errors
+vfms_n_f64

crates/intrinsic-test/missing_aarch64_be.txt → crates/intrinsic-test/missing_aarch64_common.txt

@@ -1,43 +1,3 @@
-# Bad LLVM codegen for BE in O2 in clang, and release in rust (https://github.com/llvm/llvm-project/issues/166190)
-vcmla_lane_f16
-vcmla_laneq_f16
-vcmla_rot180_lane_f16
-vcmla_rot180_laneq_f16
-vcmla_rot270_lane_f16
-vcmla_rot270_laneq_f16
-vcmla_rot90_lane_f16
-vcmla_rot90_laneq_f16
-vcmlaq_lane_f16
-vcmlaq_laneq_f16
-vcmlaq_laneq_f32
-vcmlaq_rot180_lane_f16
-vcmlaq_rot180_laneq_f16
-vcmlaq_rot180_laneq_f32
-vcmlaq_rot270_lane_f16
-vcmlaq_rot270_laneq_f16
-vcmlaq_rot270_laneq_f32
-vcmlaq_rot90_lane_f16
-vcmlaq_rot90_laneq_f16
-vcmlaq_rot90_laneq_f32
-# Bad codegen for BE in O2 in clang, correct in rust. Same cause as above issue.
-vdot_lane_s32
-vdot_lane_u32
-vdot_laneq_s32
-vdot_laneq_u32
-vdotq_lane_s32
-vdotq_lane_u32
-vdotq_laneq_s32
-vdotq_laneq_u32
-vsudot_lane_s32
-vsudot_laneq_s32
-vsudotq_lane_s32
-vsudotq_laneq_s32
-vusdot_lane_s32
-vusdot_laneq_s32
-vusdotq_lane_s32
-vusdotq_laneq_s32
-
-# Below are in common to missing_aarch64.txt
 # Not supported by qemu (will throw illegal instruction)
 vamin_f16
 vaminq_f16
@@ -119,6 +79,3 @@ vcvtns_s64_f32
 vcvtns_u64_f32
 vcvtps_s64_f32
 vcvtps_u64_f32
-
-# Broken in Clang
-vcvth_s16_f16

crates/intrinsic-test/missing_aarch64_gcc.txt

@@ -0,0 +1,19 @@
+# Broken in GCC https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123584, fixed in GCC 16
+vxarq_u64
+
+# Broken in LLVM llvm/llvm-project#197083
+vcvth_n_s32_f16
+vcvth_n_u32_f16
+vcvth_n_s64_f16
+vcvth_n_u64_f16
+vcvth_n_f16_s32
+vcvth_n_f16_u32
+vcvth_n_f16_s64
+vcvth_n_f16_u64
+
+# Broken in GCC https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125279
+vmaxh_f16
+vminh_f16
+
+# Rounding errors
+vfms_n_f64

crates/intrinsic-test/missing_x86_clang.txt

@@ -0,0 +1,24 @@
+# not present in Clang
+_bswap
+_bswap64
+_mm_cvtsd_si64x
+_mm_cvtsi128_si64x
+_mm_cvtsi64x_sd
+_mm_cvtsi64x_si128
+_mm_cvttsd_si64x
+_popcnt32
+_popcnt64
+
+# Clang bug
+_mm512_mask_reduce_max_pd
+_mm512_mask_reduce_max_ps
+_mm512_mask_reduce_min_pd
+_mm512_mask_reduce_min_ps
+
+# Rounding errors in release mode
+_mm_maskz_fmadd_sd
+_mm_maskz_fmadd_ss
+_mm_maskz_fmsub_sd
+_mm_maskz_fmsub_ss
+_mm_maskz_fnmadd_sd
+_mm_maskz_fnmadd_ss