add bmi1

Author: folkertdev

.github/workflows/checks.yaml

@@ -479,7 +479,7 @@ jobs:
       - name: Test public C api with NULL arguments
         run: "cargo +nightly miri nextest run -j4 -p test-libz-rs-sys --target ${{ matrix.target }} null::"
         env:
-          RUSTFLAGS: "-Ctarget-feature=+avx2,+bmi2"
+          RUSTFLAGS: "-Ctarget-feature=+avx2,+bmi2,+bmi1"
       - name: Test allocator with miri
         run: "cargo +nightly miri nextest run -j4 -p zlib-rs --target ${{ matrix.target }} allocate::"
       - name: Test gz logic with miri

zlib-rs/src/adler32/avx2.rs

@@ -70,6 +70,7 @@ pub fn adler32_avx2(adler: u32, src: &[u8]) -> u32 {
 
 #[target_feature(enable = "avx2")]
 #[target_feature(enable = "bmi2")]
+#[target_feature(enable = "bmi1")]
 unsafe fn adler32_avx2_help(adler: u32, src: &[u8]) -> u32 {
     if src.is_empty() {
         return adler;

zlib-rs/src/cpu_features.rs

@@ -39,8 +39,9 @@ pub fn is_enabled_avx2_and_bmi2() -> bool {
             0 => false,
             1 => true,
             _ => {
-                let detected =
-                    std::is_x86_feature_detected!("avx2") && std::is_x86_feature_detected!("bmi2");
+                let detected = std::is_x86_feature_detected!("avx2")
+                    && std::is_x86_feature_detected!("bmi1")
+                    && std::is_x86_feature_detected!("bmi2");
                 CACHE.store(u32::from(detected), Ordering::Relaxed);
                 detected
             }

zlib-rs/src/deflate/compare256.rs

@@ -181,6 +181,7 @@ mod avx2 {
     /// Behavior is undefined if the `avx` target feature is not enabled
     #[target_feature(enable = "avx2")]
     #[target_feature(enable = "bmi2")]
+    #[target_feature(enable = "bmi1")]
     pub unsafe fn compare256(src0: &[u8; 256], src1: &[u8; 256]) -> usize {
         let src0 = src0.chunks_exact(32);
         let src1 = src1.chunks_exact(32);

zlib-rs/src/deflate/slide_hash.rs

@@ -55,6 +55,7 @@ mod avx2 {
     /// Behavior is undefined if the `avx2` target feature is not enabled
     #[target_feature(enable = "avx2")]
     #[target_feature(enable = "bmi2")]
+    #[target_feature(enable = "bmi1")]
     pub unsafe fn slide_hash_chain(table: &mut [u16], wsize: u16) {
         // 64 means that 4 256-bit values can be processed per iteration.
         // That appear to be the optimal amount for avx2.

zlib-rs/src/inflate.rs

@@ -1827,6 +1827,7 @@ fn inflate_fast_help(state: &mut State, start: usize) {
 #[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
 #[target_feature(enable = "avx2")]
 #[target_feature(enable = "bmi2")]
+#[target_feature(enable = "bmi1")]
 unsafe fn inflate_fast_help_avx2(state: &mut State, start: usize) {
     inflate_fast_help_impl::<{ CpuFeatures::AVX2 }>(state, start);
 }