Redesign intrinsic-test to use simple comparison

.github/workflows/main.yml

@@ -272,12 +272,12 @@ jobs:
     name: Intrinsic Test
     runs-on: ubuntu-latest 
     strategy:
+      fail-fast: false
       matrix:
         target:
           - aarch64-unknown-linux-gnu
           - aarch64_be-unknown-linux-gnu
           - armv7-unknown-linux-gnueabihf
-          - arm-unknown-linux-gnueabihf
           - x86_64-unknown-linux-gnu
         profile: [dev, release]
         include:

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

@@ -1,17 +1,19 @@
 FROM ubuntu:25.10
 RUN apt-get update && apt-get install -y --no-install-recommends \
   gcc \
-  g++ \
   ca-certificates \
   libc6-dev \
   gcc-aarch64-linux-gnu \
-  g++-aarch64-linux-gnu \
   libc6-dev-arm64-cross \
   qemu-user \
   make \
   file \
-  clang \
-  lld
+  xz-utils \
+  wget
+
+RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.1-x86_64.tar.gz -O llvm.tar.xz
+RUN mkdir llvm
+RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
 
 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

@@ -2,17 +2,15 @@ FROM ubuntu:25.10
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
   gcc \
-  g++ \
   ca-certificates \
   libc6-dev \
   libc6-dev-arm64-cross \
   qemu-user \
   make \
   file \
-  clang \
   curl \
   xz-utils \
-  lld
+  wget
 
 ENV TOOLCHAIN="arm-gnu-toolchain-14.3.rel1-x86_64-aarch64_be-none-linux-gnu"
 
@@ -21,6 +19,10 @@ RUN curl -L "https://developer.arm.com/-/media/Files/downloads/gnu/14.3.rel1/bin
 RUN tar -xvf "${TOOLCHAIN}.tar.xz"
 RUN mkdir /toolchains && mv "./${TOOLCHAIN}" /toolchains
 
+RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.1-x86_64.tar.gz -O llvm.tar.xz
+RUN mkdir llvm
+RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
+
 ENV AARCH64_BE_TOOLCHAIN="/toolchains/${TOOLCHAIN}"
 ENV AARCH64_BE_LIBC="${AARCH64_BE_TOOLCHAIN}/aarch64_be-none-linux-gnu/libc"
 

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

@@ -7,7 +7,9 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
   libc6-dev-armhf-cross \
   qemu-user \
   make \
-  file
+  file \
+  clang \
+  lld
 ENV CARGO_TARGET_ARM_UNKNOWN_LINUX_GNUEABIHF_LINKER=arm-linux-gnueabihf-gcc \
     CARGO_TARGET_ARM_UNKNOWN_LINUX_GNUEABIHF_RUNNER="qemu-arm -cpu max -L /usr/arm-linux-gnueabihf" \
     OBJDUMP=arm-linux-gnueabihf-objdump

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

@@ -1,17 +1,19 @@
 FROM ubuntu:24.04
 RUN apt-get update && apt-get install -y --no-install-recommends \
   gcc \
-  g++ \
   ca-certificates \
   libc6-dev \
   gcc-arm-linux-gnueabihf \
-  g++-arm-linux-gnueabihf \
   libc6-dev-armhf-cross \
   qemu-user \
   make \
   file \
-  clang \
-  lld
+  wget
+
+RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.1-x86_64.tar.gz -O llvm.tar.xz
+RUN mkdir llvm
+RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
+
 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" \
     OBJDUMP=arm-linux-gnueabihf-objdump

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

@@ -6,15 +6,16 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
   make \
   ca-certificates \
   wget \
-  xz-utils \
-  clang \
-  libstdc++-14-dev \
-  build-essential \
-  lld
+  xz-utils
 
 RUN wget http://ci-mirrors.rust-lang.org/stdarch/sde-external-10.5.0-2026-01-13-lin.tar.xz -O sde.tar.xz
 RUN mkdir intel-sde
 RUN tar -xJf sde.tar.xz --strip-components=1 -C intel-sde
+
+RUN wget https://mirrors.edge.kernel.org/pub/tools/llvm/files/llvm-22.1.1-x86_64.tar.gz -O llvm.tar.xz
+RUN mkdir llvm
+RUN tar -xvf llvm.tar.xz --strip-components=1 -C llvm
+
 ENV CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER="/intel-sde/sde64 \
             -cpuid-in /checkout/ci/docker/x86_64-unknown-linux-gnu/cpuid.def \
             -rtm-mode full -tsx --"

ci/intrinsic-test.sh

@@ -5,127 +5,56 @@ set -ex
 : "${TARGET?The TARGET environment variable must be set.}"
 
 export RUSTFLAGS="${RUSTFLAGS} -D warnings -Z merge-functions=disabled -Z verify-llvm-ir"
-export HOST_RUSTFLAGS="${RUSTFLAGS}"
 export PROFILE="${PROFILE:="release"}"
 
-case ${TARGET} in
-    # On 32-bit use a static relocation model which avoids some extra
-    # instructions when dealing with static data, notably allowing some
-    # instruction assertion checks to pass below the 20 instruction limit. If
-    # this is the default, dynamic, then too many instructions are generated
-    # when we assert the instruction for a function and it causes tests to fail.
-    i686-* | i586-*)
-        export RUSTFLAGS="${RUSTFLAGS} -C relocation-model=static"
-        ;;
-    # Some x86_64 targets enable by default more features beyond SSE2,
-    # which cause some instruction assertion checks to fail.
-    x86_64-*)
-        export RUSTFLAGS="${RUSTFLAGS} -C target-feature=-sse3"
-        ;;
-    #Unoptimized build uses fast-isel which breaks with msa
-    mips-* | mipsel-*)
-	export RUSTFLAGS="${RUSTFLAGS} -C llvm-args=-fast-isel=false"
-	;;
-    armv7-*eabihf | thumbv7-*eabihf)
-        export RUSTFLAGS="${RUSTFLAGS} -Ctarget-feature=+neon"
-        ;;
-    # Some of our test dependencies use the deprecated `gcc` crates which
-    # doesn't detect RISC-V compilers automatically, so do it manually here.
-    riscv*)
-        export RUSTFLAGS="${RUSTFLAGS} -Ctarget-feature=+zk,+zks,+zbb,+zbc"
-        ;;
-esac
-
 echo "RUSTFLAGS=${RUSTFLAGS}"
-echo "OBJDUMP=${OBJDUMP}"
 echo "PROFILE=${PROFILE}"
 
 INTRINSIC_TEST="--manifest-path=crates/intrinsic-test/Cargo.toml"
 
-# Test targets compiled with extra features.
+export CC="/llvm/bin/clang"
+
 case ${TARGET} in
-    # Setup aarch64 & armv7 specific variables, the runner, along with some
-    # tests to skip
-    aarch64-unknown-linux-gnu*)
-        TEST_CPPFLAGS="-fuse-ld=lld -I/usr/aarch64-linux-gnu/include/ -I/usr/aarch64-linux-gnu/include/c++/9/aarch64-linux-gnu/"
-        TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_aarch64.txt
-        TEST_CXX_COMPILER="clang++"
-        TEST_RUNNER="${CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_RUNNER}"
-        : "${TEST_SAMPLE_INTRINSICS_PERCENTAGE:=100}"
+    aarch64_be*)
+        export CFLAGS="-I${AARCH64_BE_TOOLCHAIN}/aarch64_be-none-linux-gnu/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
         ;;
 
-    aarch64_be-unknown-linux-gnu*)
-        TEST_CPPFLAGS="-fuse-ld=lld"
-        TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_aarch64_be.txt
-        TEST_CXX_COMPILER="clang++"
-        TEST_RUNNER="${CARGO_TARGET_AARCH64_BE_UNKNOWN_LINUX_GNU_RUNNER}"
-        : "${TEST_SAMPLE_INTRINSICS_PERCENTAGE:=100}"
+    aarch64*)
+        export CFLAGS="-I/usr/aarch64-linux-gnu/include/"
+        TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_aarch64.txt
         ;;
 
-    armv7-unknown-linux-gnueabihf*)
-        TEST_CPPFLAGS="-fuse-ld=lld -I/usr/arm-linux-gnueabihf/include/ -I/usr/arm-linux-gnueabihf/include/c++/9/arm-linux-gnueabihf/"
+    armv7*)
+        export CFLAGS="-I/usr/arm-linux-gnueabihf/include/"
         TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_arm.txt
-        TEST_CXX_COMPILER="clang++"
-        TEST_RUNNER="${CARGO_TARGET_ARMV7_UNKNOWN_LINUX_GNUEABIHF_RUNNER}"
-        : "${TEST_SAMPLE_INTRINSICS_PERCENTAGE:=100}"
         ;;
 
-    x86_64-unknown-linux-gnu*)
-        TEST_CPPFLAGS="-fuse-ld=lld -I/usr/include/x86_64-linux-gnu/"
-        TEST_CXX_COMPILER="clang++"
-        TEST_RUNNER="${CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER}"
+    x86_64*)
+        export CFLAGS="-I/usr/include/x86_64-linux-gnu/"
         TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_x86.txt
-        : "${TEST_SAMPLE_INTRINSICS_PERCENTAGE:=20}"
         ;;
     *)
         ;;
 
 esac
 
-# Arm specific
 case "${TARGET}" in
-    aarch64-unknown-linux-gnu*|armv7-unknown-linux-gnueabihf*)
-        CPPFLAGS="${TEST_CPPFLAGS}" RUSTFLAGS="${HOST_RUSTFLAGS}" RUST_LOG=warn \
-            cargo run "${INTRINSIC_TEST}" --release  \
-            --bin intrinsic-test -- intrinsics_data/arm_intrinsics.json \
-            --runner "${TEST_RUNNER}" \
-            --cppcompiler "${TEST_CXX_COMPILER}" \
-            --skip "${TEST_SKIP_INTRINSICS}" \
-            --target "${TARGET}" \
-            --profile "${PROFILE}" \
-            --sample-percentage "${TEST_SAMPLE_INTRINSICS_PERCENTAGE}"
-        ;;
-
-    aarch64_be-unknown-linux-gnu*)
-        CPPFLAGS="${TEST_CPPFLAGS}" RUSTFLAGS="${HOST_RUSTFLAGS}" RUST_LOG=warn \
-            cargo run "${INTRINSIC_TEST}" --release  \
-            --bin intrinsic-test -- intrinsics_data/arm_intrinsics.json \
-            --runner "${TEST_RUNNER}" \
-            --cppcompiler "${TEST_CXX_COMPILER}" \
-            --skip "${TEST_SKIP_INTRINSICS}" \
-            --target "${TARGET}" \
-            --profile "${PROFILE}" \
-            --linker "${CARGO_TARGET_AARCH64_BE_UNKNOWN_LINUX_GNU_LINKER}" \
-            --cxx-toolchain-dir "${AARCH64_BE_TOOLCHAIN}" \
-            --sample-percentage "${TEST_SAMPLE_INTRINSICS_PERCENTAGE}"
-        ;;
-
     x86_64-unknown-linux-gnu*)
-        # `CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER` is not necessary for `intrinsic-test`
-        # because the binary needs to run directly on the host.
-        # Hence the use of `env -u`.
         env -u CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER \
-            CPPFLAGS="${TEST_CPPFLAGS}" RUSTFLAGS="${HOST_RUSTFLAGS}" \
-            RUST_LOG=warn RUST_BACKTRACE=1 \
             cargo run "${INTRINSIC_TEST}" --release \
             --bin intrinsic-test -- intrinsics_data/x86-intel.xml \
-            --runner "${TEST_RUNNER}" \
             --skip "${TEST_SKIP_INTRINSICS}" \
-            --cppcompiler "${TEST_CXX_COMPILER}" \
-            --target "${TARGET}" \
-            --profile "${PROFILE}" \
-            --sample-percentage "${TEST_SAMPLE_INTRINSICS_PERCENTAGE}"
+            --target "${TARGET}"
+
+        echo "${CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER}"
         ;;
-     *)
+    *)
+        cargo run "${INTRINSIC_TEST}" --release \
+            --bin intrinsic-test -- intrinsics_data/arm_intrinsics.json \
+            --skip "${TEST_SKIP_INTRINSICS}" \
+            --target "${TARGET}"
         ;;
 esac
+
+cargo test --manifest-path=rust_programs/Cargo.toml --target "${TARGET}" --profile "${PROFILE}" --no-fail-fast

crates/intrinsic-test/missing_x86.txt

@@ -52,8 +52,19 @@ _mm_sm3rnds2_epi32
 _xgetbv
 
 # top bits are undefined, unclear how to test these
+_mm256_castph128_ph256
+_mm256_castps128_ps256
+_mm256_castpd128_pd256
 _mm256_castsi128_si256
+
+_mm512_castph128_ph512
+_mm512_castps128_ps512
+_mm512_castpd128_pd512
 _mm512_castsi128_si512
+
+_mm512_castph256_ph512
+_mm512_castps256_ps512
+_mm512_castpd256_pd512
 _mm512_castsi256_si512
 
 # Clang bug
@@ -65,3 +76,7 @@ _mm512_mask_reduce_min_pd
 _mm512_mask_reduce_min_ps
 _mm_extract_epi16
 _mm_extract_epi8
+
+# TODO: fix
+_mm_movemask_epi8
+_mm_movemask_pd