Match clang for _mm512_abs_epi32 intrinsics

This commit updates stdsimd's codegen to match Clang's for the
`_mm512_abs_epi32` intrinsic (and masked versions) which doesn't use any
LLVM intrinsic calls, but rather raw SIMD operations.

These are built on top of the new `simd_select_bitmask` intrinsic
introduced recently to the compiler!

Author: alexcrichton

.appveyor.yml

@@ -13,7 +13,7 @@ environment:
 install:
   # Install rust, x86_64-pc-windows-msvc host
   - appveyor-retry appveyor DownloadFile https://win.rustup.rs/ -FileName rustup-init.exe
-  - rustup-init.exe -y --default-host x86_64-pc-windows-msvc --default-toolchain nightly-2018-10-20
+  - rustup-init.exe -y --default-host x86_64-pc-windows-msvc --default-toolchain nightly
   - set PATH=%PATH%;C:\Users\appveyor\.cargo\bin
   - if NOT "%TARGET%" == "x86_64-pc-windows-msvc" rustup target add %TARGET%
   - rustc -vV

coresimd/simd_llvm.rs

@@ -49,6 +49,7 @@ extern "platform-intrinsic" {
     pub fn simd_reduce_any<T>(x: T) -> bool;
 
     pub fn simd_select<M, T>(m: M, a: T, b: T) -> T;
+    pub fn simd_select_bitmask<M, T>(m: M, a: T, b: T) -> T;
 
     pub fn simd_fmin<T>(a: T, b: T) -> T;
     pub fn simd_fmax<T>(a: T, b: T) -> T;

coresimd/x86/avx512f.rs

@@ -1,4 +1,5 @@
 use coresimd::simd::*;
+use coresimd::simd_llvm::*;
 use coresimd::x86::*;
 use mem;
 
@@ -12,7 +13,11 @@ use stdsimd_test::assert_instr;
 #[target_feature(enable = "avx512f")]
 #[cfg_attr(test, assert_instr(vpabsd))]
 pub unsafe fn _mm512_abs_epi32(a: __m512i) -> __m512i {
-    mem::transmute(pabsd(a.as_i32x16(), _mm512_setzero_si512().as_i32x16(), -1))
+    let a = a.as_i32x16();
+    let zero: i32x16 = mem::zeroed();
+    let sub = simd_sub(zero, a);
+    let cmp: i32x16 = simd_gt(a, zero);
+    mem::transmute(simd_select(cmp, a, sub))
 }
 
 /// Compute the absolute value of packed 32-bit integers in `a`, and store the
@@ -24,7 +29,8 @@ pub unsafe fn _mm512_abs_epi32(a: __m512i) -> __m512i {
 #[target_feature(enable = "avx512f")]
 #[cfg_attr(test, assert_instr(vpabsd))]
 pub unsafe fn _mm512_mask_abs_epi32(src: __m512i, k: __mmask16, a: __m512i) -> __m512i {
-    mem::transmute(pabsd(a.as_i32x16(), src.as_i32x16(), k))
+    let abs = _mm512_abs_epi32(a).as_i32x16();
+    mem::transmute(simd_select_bitmask(k, abs, src.as_i32x16()))
 }
 
 /// Compute the absolute value of packed 32-bit integers in `a`, and store the
@@ -36,7 +42,9 @@ pub unsafe fn _mm512_mask_abs_epi32(src: __m512i, k: __mmask16, a: __m512i) -> _
 #[target_feature(enable = "avx512f")]
 #[cfg_attr(test, assert_instr(vpabsd))]
 pub unsafe fn _mm512_maskz_abs_epi32(k: __mmask16, a: __m512i) -> __m512i {
-    mem::transmute(pabsd(a.as_i32x16(), _mm512_setzero_si512().as_i32x16(), k))
+    let abs = _mm512_abs_epi32(a).as_i32x16();
+    let zero = _mm512_setzero_si512().as_i32x16();
+    mem::transmute(simd_select_bitmask(k, abs, zero))
 }
 
 /// Return vector of type `__m512i` with all elements set to zero.
@@ -77,12 +85,6 @@ pub unsafe fn _mm512_setr_epi32(
     mem::transmute(r)
 }
 
-#[allow(improper_ctypes)]
-extern "C" {
-    #[link_name = "llvm.x86.avx512.mask.pabs.d.512"]
-    fn pabsd(a: i32x16, b: i32x16, c: i16) -> i32x16;
-}
-
 #[cfg(test)]
 mod tests {
     use std;