stdarch::x86: Fix intrinsics in x86

ci/intrinsic-test.sh

@@ -69,10 +69,10 @@ case ${TARGET} in
 
     x86_64-unknown-linux-gnu*)
         TEST_CPPFLAGS="-fuse-ld=lld -I/usr/include/x86_64-linux-gnu/"
-        TEST_CXX_COMPILER="clang++"
+        TEST_CXX_COMPILER="g++"
         TEST_RUNNER="${CARGO_TARGET_X86_64_UNKNOWN_LINUX_GNU_RUNNER}"
         TEST_SKIP_INTRINSICS=crates/intrinsic-test/missing_x86.txt
-        TEST_SAMPLE_INTRINSICS_PERCENTAGE=5
+        TEST_SAMPLE_INTRINSICS_PERCENTAGE=100
         ;;
     *)
         ;;

crates/core_arch/src/x86/avx512bw.rs

@@ -10681,7 +10681,7 @@ pub fn _mm_cvtepi16_epi8(a: __m128i) -> __m128i {
 pub fn _mm_mask_cvtepi16_epi8(src: __m128i, k: __mmask8, a: __m128i) -> __m128i {
     unsafe {
         let convert = _mm_cvtepi16_epi8(a).as_i8x16();
-        let k: __mmask16 = 0b11111111_11111111 & k as __mmask16;
+        let k: __mmask16 = 0b11111111_00000000 | k as __mmask16;
         transmute(simd_select_bitmask(k, convert, src.as_i8x16()))
     }
 }
@@ -20559,6 +20559,12 @@ mod tests {
         let r = _mm_mask_cvtepi16_epi8(src, 0b11111111, a);
         let e = _mm_set_epi8(0, 0, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 2, 2, 2, 2);
         assert_eq_m128i(r, e);
+
+        let src = _mm_set_epi8(1, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 1, 1);
+        let a = _mm_set1_epi16(3);
+        let r = _mm_mask_cvtepi16_epi8(src, 0b11011110, a);
+        let e = _mm_set_epi8(0, 0, 0, 0, 0, 0, 0, 0, 3, 3, 1, 3, 3, 3, 3, 1);
+        assert_eq_m128i(r, e);
     }
 
     #[simd_test(enable = "avx512bw,avx512vl")]

crates/intrinsic-test/missing_x86.txt

@@ -877,28 +877,28 @@ _mm512_mask_floor_ps
 _xgetbv
 
 # Miscellaneous issues that can be fixed first
-_kshiftli_mask16
-_kshiftli_mask32
-_kshiftli_mask64
-_kshiftli_mask8
-_kshiftri_mask16
-_kshiftri_mask32
-_kshiftri_mask64
-_kshiftri_mask8
-_mm256_castsi128_si256
-_mm256_extract_epi16
-_mm256_extract_epi8
-_mm512_castsi128_si512
-_mm512_castsi256_si512
+# _kshiftli_mask16
+# _kshiftli_mask32
+# _kshiftli_mask64
+# _kshiftli_mask8
+# _kshiftri_mask16
+# _kshiftri_mask32
+# _kshiftri_mask64
+# _kshiftri_mask8
+# _mm256_castsi128_si256
+# _mm256_extract_epi16
+# _mm256_extract_epi8
+# _mm512_castsi128_si512
+# _mm512_castsi256_si512
 # _mm512_conj_pch
-_mm512_mask_reduce_max_pd
-_mm512_mask_reduce_max_ps
-_mm512_mask_reduce_min_pd
-_mm512_mask_reduce_min_ps
-_mm_comineq_sh
-_mm_extract_epi16
-_mm_extract_epi8
-_mm_mask_cvtepi16_epi8
-_mm_mask_cvtpd_epi32
-_mm_mask_cvtpd_ps
-_mm_ucomineq_sh
+# _mm512_mask_reduce_max_pd
+# _mm512_mask_reduce_max_ps
+# _mm512_mask_reduce_min_pd
+# _mm512_mask_reduce_min_ps
+# _mm_comineq_sh
+# _mm_extract_epi16
+# _mm_extract_epi8
+# _mm_mask_cvtepi16_epi8
+# _mm_mask_cvtpd_epi32
+# _mm_mask_cvtpd_ps
+# _mm_ucomineq_sh

crates/intrinsic-test/src/arm/compile.rs

@@ -18,9 +18,11 @@ pub fn build_cpp_compilation(config: &ProcessedCli) -> Option<CppCompilation> {
         command = command.add_arch_flags(["faminmax", "lut", "sha3"]);
     }
 
-    if !cpp_compiler.contains("clang") {
-        command = command.add_extra_flag("-flax-vector-conversions");
-    }
+    command = if !cpp_compiler.contains("clang") {
+        command.add_extra_flag("-flax-vector-conversions")
+    } else {
+        command.add_extra_flag(format!("--target={}", config.target).as_str())
+    };
 
     let mut cpp_compiler = command.into_cpp_compilation();
 

crates/intrinsic-test/src/common/compile_c.rs

@@ -84,10 +84,6 @@ impl CompilationCommandBuilder {
 
         cpp_compiler.args(self.extra_flags);
 
-        if let Some(target) = &self.target {
-            cpp_compiler.arg(format!("--target={target}"));
-        }
-
         CppCompilation(cpp_compiler)
     }
 }

crates/intrinsic-test/src/x86/compile.rs

@@ -33,13 +33,17 @@ pub fn build_cpp_compilation(config: &ProcessedCli) -> Option<CppCompilation> {
             "-mavx512vbmi2",
             "-mavx512vnni",
             "-mavx512vpopcntdq",
-            "-ferror-limit=1000",
             "-std=c++23",
         ]);
 
-    if !cpp_compiler.contains("clang") {
-        command = command.add_extra_flag("-flax-vector-conversions");
-    }
+    command = if !cpp_compiler.contains("clang") {
+        command.add_extra_flags(vec!["-fmax-errors=1000", "-flax-vector-conversions"])
+    } else {
+        command.add_extra_flags(vec![
+            "-ferror-limit=1000",
+            format!("--target={}", config.target).as_str(),
+        ])
+    };
 
     let cpp_compiler = command.into_cpp_compilation();