Update fma.rs

Author: TDecking

crates/core_arch/src/x86/fma.rs

@@ -83,7 +83,11 @@ pub unsafe fn _mm256_fmadd_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
 #[cfg_attr(test, assert_instr(vfmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fmadd_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
-    vfmaddsd(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtsd_f64(a).mul_add(_mm_cvtsd_f64(b), _mm_cvtsd_f64(c))
+    )
 }
 
 /// Multiplies the lower single-precision (32-bit) floating-point elements in
@@ -97,7 +101,11 @@ pub unsafe fn _mm_fmadd_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
 #[cfg_attr(test, assert_instr(vfmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fmadd_ss(a: __m128, b: __m128, c: __m128) -> __m128 {
-    vfmaddss(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtss_f32(a).mul_add(_mm_cvtss_f32(b), _mm_cvtss_f32(c))
+    )
 }
 
 /// Multiplies packed double-precision (64-bit) floating-point elements in `a`
@@ -161,7 +169,7 @@ pub unsafe fn _mm256_fmaddsub_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
 #[cfg_attr(test, assert_instr(vfmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fmsub_pd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
-    vfmsubpd(a, b, c)
+    simd_fma(a, b, simd_neg(c))
 }
 
 /// Multiplies packed double-precision (64-bit) floating-point elements in `a`
@@ -173,7 +181,7 @@ pub unsafe fn _mm_fmsub_pd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
 #[cfg_attr(test, assert_instr(vfmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_fmsub_pd(a: __m256d, b: __m256d, c: __m256d) -> __m256d {
-    vfmsubpd256(a, b, c)
+    simd_fma(a, b, simd_neg(c))
 }
 
 /// Multiplies packed single-precision (32-bit) floating-point elements in `a`
@@ -185,7 +193,7 @@ pub unsafe fn _mm256_fmsub_pd(a: __m256d, b: __m256d, c: __m256d) -> __m256d {
 #[cfg_attr(test, assert_instr(vfmsub213ps))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fmsub_ps(a: __m128, b: __m128, c: __m128) -> __m128 {
-    vfmsubps(a, b, c)
+    simd_fma(a, b, simd_neg(c))
 }
 
 /// Multiplies packed single-precision (32-bit) floating-point elements in `a`
@@ -197,7 +205,7 @@ pub unsafe fn _mm_fmsub_ps(a: __m128, b: __m128, c: __m128) -> __m128 {
 #[cfg_attr(test, assert_instr(vfmsub213ps))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_fmsub_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
-    vfmsubps256(a, b, c)
+    simd_fma(a, b, simd_neg(c))
 }
 
 /// Multiplies the lower double-precision (64-bit) floating-point elements in
@@ -211,7 +219,11 @@ pub unsafe fn _mm256_fmsub_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
 #[cfg_attr(test, assert_instr(vfmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fmsub_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
-    vfmsubsd(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtsd_f64(a).mul_add(_mm_cvtsd_f64(b), -_mm_cvtsd_f64(c))
+    )
 }
 
 /// Multiplies the lower single-precision (32-bit) floating-point elements in
@@ -225,7 +237,11 @@ pub unsafe fn _mm_fmsub_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
 #[cfg_attr(test, assert_instr(vfmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fmsub_ss(a: __m128, b: __m128, c: __m128) -> __m128 {
-    vfmsubss(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtss_f32(a).mul_add(_mm_cvtss_f32(b), -_mm_cvtss_f32(c))
+    )
 }
 
 /// Multiplies packed double-precision (64-bit) floating-point elements in `a`
@@ -289,7 +305,7 @@ pub unsafe fn _mm256_fmsubadd_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
 #[cfg_attr(test, assert_instr(vfnmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmadd_pd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
-    vfnmaddpd(a, b, c)
+    simd_fma(simd_neg(a), b, c)
 }
 
 /// Multiplies packed double-precision (64-bit) floating-point elements in `a`
@@ -301,7 +317,7 @@ pub unsafe fn _mm_fnmadd_pd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
 #[cfg_attr(test, assert_instr(vfnmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_fnmadd_pd(a: __m256d, b: __m256d, c: __m256d) -> __m256d {
-    vfnmaddpd256(a, b, c)
+    simd_fma(simd_neg(a), b, c)
 }
 
 /// Multiplies packed single-precision (32-bit) floating-point elements in `a`
@@ -313,7 +329,7 @@ pub unsafe fn _mm256_fnmadd_pd(a: __m256d, b: __m256d, c: __m256d) -> __m256d {
 #[cfg_attr(test, assert_instr(vfnmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmadd_ps(a: __m128, b: __m128, c: __m128) -> __m128 {
-    vfnmaddps(a, b, c)
+    simd_fma(simd_neg(a), b, c)
 }
 
 /// Multiplies packed single-precision (32-bit) floating-point elements in `a`
@@ -325,7 +341,7 @@ pub unsafe fn _mm_fnmadd_ps(a: __m128, b: __m128, c: __m128) -> __m128 {
 #[cfg_attr(test, assert_instr(vfnmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_fnmadd_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
-    vfnmaddps256(a, b, c)
+    simd_fma(simd_neg(a), b, c)
 }
 
 /// Multiplies the lower double-precision (64-bit) floating-point elements in
@@ -339,7 +355,11 @@ pub unsafe fn _mm256_fnmadd_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
 #[cfg_attr(test, assert_instr(vfnmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmadd_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
-    vfnmaddsd(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtsd_f64(a).mul_add(-_mm_cvtsd_f64(b), _mm_cvtsd_f64(c))
+    )
 }
 
 /// Multiplies the lower single-precision (32-bit) floating-point elements in
@@ -353,7 +373,11 @@ pub unsafe fn _mm_fnmadd_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
 #[cfg_attr(test, assert_instr(vfnmadd))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmadd_ss(a: __m128, b: __m128, c: __m128) -> __m128 {
-    vfnmaddss(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtss_f32(a).mul_add(-_mm_cvtss_f32(b), _mm_cvtss_f32(c))
+    )
 }
 
 /// Multiplies packed double-precision (64-bit) floating-point elements in `a`
@@ -366,7 +390,7 @@ pub unsafe fn _mm_fnmadd_ss(a: __m128, b: __m128, c: __m128) -> __m128 {
 #[cfg_attr(test, assert_instr(vfnmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmsub_pd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
-    vfnmsubpd(a, b, c)
+    simd_fma(simd_neg(a), b, simd_neg(c))
 }
 
 /// Multiplies packed double-precision (64-bit) floating-point elements in `a`
@@ -379,7 +403,7 @@ pub unsafe fn _mm_fnmsub_pd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
 #[cfg_attr(test, assert_instr(vfnmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_fnmsub_pd(a: __m256d, b: __m256d, c: __m256d) -> __m256d {
-    vfnmsubpd256(a, b, c)
+    simd_fma(simd_neg(a), b, simd_neg(c))
 }
 
 /// Multiplies packed single-precision (32-bit) floating-point elements in `a`
@@ -392,7 +416,7 @@ pub unsafe fn _mm256_fnmsub_pd(a: __m256d, b: __m256d, c: __m256d) -> __m256d {
 #[cfg_attr(test, assert_instr(vfnmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmsub_ps(a: __m128, b: __m128, c: __m128) -> __m128 {
-    vfnmsubps(a, b, c)
+    simd_fma(simd_neg(a), b, simd_neg(c))
 }
 
 /// Multiplies packed single-precision (32-bit) floating-point elements in `a`
@@ -405,7 +429,7 @@ pub unsafe fn _mm_fnmsub_ps(a: __m128, b: __m128, c: __m128) -> __m128 {
 #[cfg_attr(test, assert_instr(vfnmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm256_fnmsub_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
-    vfnmsubps256(a, b, c)
+    simd_fma(simd_neg(a), b, simd_neg(c))
 }
 
 /// Multiplies the lower double-precision (64-bit) floating-point elements in
@@ -420,7 +444,11 @@ pub unsafe fn _mm256_fnmsub_ps(a: __m256, b: __m256, c: __m256) -> __m256 {
 #[cfg_attr(test, assert_instr(vfnmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmsub_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
-    vfnmsubsd(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtsd_f64(a).mul_add(-_mm_cvtsd_f64(b), -_mm_cvtsd_f64(c))
+    )
 }
 
 /// Multiplies the lower single-precision (32-bit) floating-point elements in
@@ -435,15 +463,15 @@ pub unsafe fn _mm_fnmsub_sd(a: __m128d, b: __m128d, c: __m128d) -> __m128d {
 #[cfg_attr(test, assert_instr(vfnmsub))]
 #[stable(feature = "simd_x86", since = "1.27.0")]
 pub unsafe fn _mm_fnmsub_ss(a: __m128, b: __m128, c: __m128) -> __m128 {
-    vfnmsubss(a, b, c)
+    simd_insert!(
+        a,
+        0,
+        _mm_cvtss_f32(a).mul_add(-_mm_cvtss_f32(b), -_mm_cvtss_f32(c))
+    )
 }
 
 #[allow(improper_ctypes)]
 extern "C" {
-    #[link_name = "llvm.x86.fma.vfmadd.sd"]
-    fn vfmaddsd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
-    #[link_name = "llvm.x86.fma.vfmadd.ss"]
-    fn vfmaddss(a: __m128, b: __m128, c: __m128) -> __m128;
     #[link_name = "llvm.x86.fma.vfmaddsub.pd"]
     fn vfmaddsubpd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
     #[link_name = "llvm.x86.fma.vfmaddsub.pd.256"]
@@ -452,18 +480,6 @@ extern "C" {
     fn vfmaddsubps(a: __m128, b: __m128, c: __m128) -> __m128;
     #[link_name = "llvm.x86.fma.vfmaddsub.ps.256"]
     fn vfmaddsubps256(a: __m256, b: __m256, c: __m256) -> __m256;
-    #[link_name = "llvm.x86.fma.vfmsub.pd"]
-    fn vfmsubpd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
-    #[link_name = "llvm.x86.fma.vfmsub.pd.256"]
-    fn vfmsubpd256(a: __m256d, b: __m256d, c: __m256d) -> __m256d;
-    #[link_name = "llvm.x86.fma.vfmsub.ps"]
-    fn vfmsubps(a: __m128, b: __m128, c: __m128) -> __m128;
-    #[link_name = "llvm.x86.fma.vfmsub.ps.256"]
-    fn vfmsubps256(a: __m256, b: __m256, c: __m256) -> __m256;
-    #[link_name = "llvm.x86.fma.vfmsub.sd"]
-    fn vfmsubsd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
-    #[link_name = "llvm.x86.fma.vfmsub.ss"]
-    fn vfmsubss(a: __m128, b: __m128, c: __m128) -> __m128;
     #[link_name = "llvm.x86.fma.vfmsubadd.pd"]
     fn vfmsubaddpd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
     #[link_name = "llvm.x86.fma.vfmsubadd.pd.256"]
@@ -472,30 +488,6 @@ extern "C" {
     fn vfmsubaddps(a: __m128, b: __m128, c: __m128) -> __m128;
     #[link_name = "llvm.x86.fma.vfmsubadd.ps.256"]
     fn vfmsubaddps256(a: __m256, b: __m256, c: __m256) -> __m256;
-    #[link_name = "llvm.x86.fma.vfnmadd.pd"]
-    fn vfnmaddpd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
-    #[link_name = "llvm.x86.fma.vfnmadd.pd.256"]
-    fn vfnmaddpd256(a: __m256d, b: __m256d, c: __m256d) -> __m256d;
-    #[link_name = "llvm.x86.fma.vfnmadd.ps"]
-    fn vfnmaddps(a: __m128, b: __m128, c: __m128) -> __m128;
-    #[link_name = "llvm.x86.fma.vfnmadd.ps.256"]
-    fn vfnmaddps256(a: __m256, b: __m256, c: __m256) -> __m256;
-    #[link_name = "llvm.x86.fma.vfnmadd.sd"]
-    fn vfnmaddsd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
-    #[link_name = "llvm.x86.fma.vfnmadd.ss"]
-    fn vfnmaddss(a: __m128, b: __m128, c: __m128) -> __m128;
-    #[link_name = "llvm.x86.fma.vfnmsub.pd"]
-    fn vfnmsubpd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
-    #[link_name = "llvm.x86.fma.vfnmsub.pd.256"]
-    fn vfnmsubpd256(a: __m256d, b: __m256d, c: __m256d) -> __m256d;
-    #[link_name = "llvm.x86.fma.vfnmsub.ps"]
-    fn vfnmsubps(a: __m128, b: __m128, c: __m128) -> __m128;
-    #[link_name = "llvm.x86.fma.vfnmsub.ps.256"]
-    fn vfnmsubps256(a: __m256, b: __m256, c: __m256) -> __m256;
-    #[link_name = "llvm.x86.fma.vfnmsub.sd"]
-    fn vfnmsubsd(a: __m128d, b: __m128d, c: __m128d) -> __m128d;
-    #[link_name = "llvm.x86.fma.vfnmsub.ss"]
-    fn vfnmsubss(a: __m128, b: __m128, c: __m128) -> __m128;
 }
 
 #[cfg(test)]