Merge pull request #2103 from heiher/vadda

loongarch: Use `intrinsics::simd` for vadda

Author: sayantn

crates/core_arch/src/loongarch64/lasx/generated.rs

@@ -91,14 +91,6 @@ unsafe extern "unadjusted" {
     fn __lasx_xvsat_wu(a: __v8u32, b: u32) -> __v8u32;
     #[link_name = "llvm.loongarch.lasx.xvsat.du"]
     fn __lasx_xvsat_du(a: __v4u64, b: u32) -> __v4u64;
-    #[link_name = "llvm.loongarch.lasx.xvadda.b"]
-    fn __lasx_xvadda_b(a: __v32i8, b: __v32i8) -> __v32i8;
-    #[link_name = "llvm.loongarch.lasx.xvadda.h"]
-    fn __lasx_xvadda_h(a: __v16i16, b: __v16i16) -> __v16i16;
-    #[link_name = "llvm.loongarch.lasx.xvadda.w"]
-    fn __lasx_xvadda_w(a: __v8i32, b: __v8i32) -> __v8i32;
-    #[link_name = "llvm.loongarch.lasx.xvadda.d"]
-    fn __lasx_xvadda_d(a: __v4i64, b: __v4i64) -> __v4i64;
     #[link_name = "llvm.loongarch.lasx.xvavg.b"]
     fn __lasx_xvavg_b(a: __v32i8, b: __v32i8) -> __v32i8;
     #[link_name = "llvm.loongarch.lasx.xvavg.h"]
@@ -1445,34 +1437,6 @@ pub fn lasx_xvsat_du<const IMM6: u32>(a: m256i) -> m256i {
     unsafe { transmute(__lasx_xvsat_du(transmute(a), IMM6)) }
 }
 
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvadda_b(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvadda_b(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvadda_h(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvadda_h(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvadda_w(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvadda_w(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvadda_d(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvadda_d(transmute(a), transmute(b))) }
-}
-
 #[inline]
 #[target_feature(enable = "lasx")]
 #[unstable(feature = "stdarch_loongarch", issue = "117427")]

crates/core_arch/src/loongarch64/lasx/portable.rs

@@ -148,6 +148,10 @@ impl_vvv!("lasx", lasx_xvssub_bu, is::simd_saturating_sub, m256i, u8x32);
 impl_vvv!("lasx", lasx_xvssub_hu, is::simd_saturating_sub, m256i, u16x16);
 impl_vvv!("lasx", lasx_xvssub_wu, is::simd_saturating_sub, m256i, u32x8);
 impl_vvv!("lasx", lasx_xvssub_du, is::simd_saturating_sub, m256i, u64x4);
+impl_vvv!("lasx", lasx_xvadda_b, ls::simd_adda, m256i, i8x32);
+impl_vvv!("lasx", lasx_xvadda_h, ls::simd_adda, m256i, i16x16);
+impl_vvv!("lasx", lasx_xvadda_w, ls::simd_adda, m256i, i32x8);
+impl_vvv!("lasx", lasx_xvadda_d, ls::simd_adda, m256i, i64x4);
 
 impl_vuv!("lasx", lasx_xvslli_b, is::simd_shl, m256i, i8x32);
 impl_vuv!("lasx", lasx_xvslli_h, is::simd_shl, m256i, i16x16);

crates/core_arch/src/loongarch64/lsx/generated.rs

@@ -91,14 +91,6 @@ unsafe extern "unadjusted" {
     fn __lsx_vsat_wu(a: __v4u32, b: u32) -> __v4u32;
     #[link_name = "llvm.loongarch.lsx.vsat.du"]
     fn __lsx_vsat_du(a: __v2u64, b: u32) -> __v2u64;
-    #[link_name = "llvm.loongarch.lsx.vadda.b"]
-    fn __lsx_vadda_b(a: __v16i8, b: __v16i8) -> __v16i8;
-    #[link_name = "llvm.loongarch.lsx.vadda.h"]
-    fn __lsx_vadda_h(a: __v8i16, b: __v8i16) -> __v8i16;
-    #[link_name = "llvm.loongarch.lsx.vadda.w"]
-    fn __lsx_vadda_w(a: __v4i32, b: __v4i32) -> __v4i32;
-    #[link_name = "llvm.loongarch.lsx.vadda.d"]
-    fn __lsx_vadda_d(a: __v2i64, b: __v2i64) -> __v2i64;
     #[link_name = "llvm.loongarch.lsx.vavg.b"]
     fn __lsx_vavg_b(a: __v16i8, b: __v16i8) -> __v16i8;
     #[link_name = "llvm.loongarch.lsx.vavg.h"]
@@ -1357,34 +1349,6 @@ pub fn lsx_vsat_du<const IMM6: u32>(a: m128i) -> m128i {
     unsafe { transmute(__lsx_vsat_du(transmute(a), IMM6)) }
 }
 
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vadda_b(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vadda_b(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vadda_h(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vadda_h(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vadda_w(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vadda_w(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vadda_d(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vadda_d(transmute(a), transmute(b))) }
-}
-
 #[inline]
 #[target_feature(enable = "lsx")]
 #[unstable(feature = "stdarch_loongarch", issue = "117427")]

crates/core_arch/src/loongarch64/lsx/portable.rs

@@ -148,6 +148,10 @@ impl_vvv!("lsx", lsx_vssub_bu, is::simd_saturating_sub, m128i, u8x16);
 impl_vvv!("lsx", lsx_vssub_hu, is::simd_saturating_sub, m128i, u16x8);
 impl_vvv!("lsx", lsx_vssub_wu, is::simd_saturating_sub, m128i, u32x4);
 impl_vvv!("lsx", lsx_vssub_du, is::simd_saturating_sub, m128i, u64x2);
+impl_vvv!("lsx", lsx_vadda_b, ls::simd_adda, m128i, i8x16);
+impl_vvv!("lsx", lsx_vadda_h, ls::simd_adda, m128i, i16x8);
+impl_vvv!("lsx", lsx_vadda_w, ls::simd_adda, m128i, i32x4);
+impl_vvv!("lsx", lsx_vadda_d, ls::simd_adda, m128i, i64x2);
 
 impl_vuv!("lsx", lsx_vslli_b, is::simd_shl, m128i, i8x16);
 impl_vuv!("lsx", lsx_vslli_h, is::simd_shl, m128i, i16x8);

crates/core_arch/src/loongarch64/simd.rs

@@ -45,6 +45,19 @@ impl_simd_ext!(i64x4, i64);
 impl_simd_ext!(u64x2, u64);
 impl_simd_ext!(u64x4, u64);
 
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_abs<T: Copy + const SimdExt>(a: T) -> T {
+    let m: T = is::simd_lt(a, ls::simd_splat(0));
+    is::simd_select(m, is::simd_neg(a), a)
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_adda<T: Copy + const SimdExt>(a: T, b: T) -> T {
+    is::simd_add(ls::simd_abs(a), ls::simd_abs(b))
+}
+
 #[inline(always)]
 #[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
 pub(super) const unsafe fn simd_andn<T: Copy + const SimdExt>(a: T, b: T) -> T {

crates/stdarch-gen-loongarch/lasx.spec

@@ -924,21 +924,25 @@ asm-fmts = xd, xj, ui6
 data-types = UV4DI, UV4DI, UQI
 
 /// lasx_xvadda_b
+impl = portable
 name = lasx_xvadda_b
 asm-fmts = xd, xj, xk
 data-types = V32QI, V32QI, V32QI
 
 /// lasx_xvadda_h
+impl = portable
 name = lasx_xvadda_h
 asm-fmts = xd, xj, xk
 data-types = V16HI, V16HI, V16HI
 
 /// lasx_xvadda_w
+impl = portable
 name = lasx_xvadda_w
 asm-fmts = xd, xj, xk
 data-types = V8SI, V8SI, V8SI
 
 /// lasx_xvadda_d
+impl = portable
 name = lasx_xvadda_d
 asm-fmts = xd, xj, xk
 data-types = V4DI, V4DI, V4DI

crates/stdarch-gen-loongarch/lsx.spec

@@ -924,21 +924,25 @@ asm-fmts = vd, vj, ui6
 data-types = UV2DI, UV2DI, UQI
 
 /// lsx_vadda_b
+impl = portable
 name = lsx_vadda_b
 asm-fmts = vd, vj, vk
 data-types = V16QI, V16QI, V16QI
 
 /// lsx_vadda_h
+impl = portable
 name = lsx_vadda_h
 asm-fmts = vd, vj, vk
 data-types = V8HI, V8HI, V8HI
 
 /// lsx_vadda_w
+impl = portable
 name = lsx_vadda_w
 asm-fmts = vd, vj, vk
 data-types = V4SI, V4SI, V4SI
 
 /// lsx_vadda_d
+impl = portable
 name = lsx_vadda_d
 asm-fmts = vd, vj, vk
 data-types = V2DI, V2DI, V2DI

crates/stdarch-gen-loongarch/src/portable-intrinsics.txt

@@ -219,6 +219,10 @@ lsx_vssub_bu
 lsx_vssub_hu
 lsx_vssub_wu
 lsx_vssub_du
+lsx_vadda_b
+lsx_vadda_h
+lsx_vadda_w
+lsx_vadda_d
 
 # LASX intrinsics
 lasx_xvsll_b
@@ -435,3 +439,7 @@ lasx_xvssub_bu
 lasx_xvssub_hu
 lasx_xvssub_wu
 lasx_xvssub_du
+lasx_xvadda_b
+lasx_xvadda_h
+lasx_xvadda_w
+lasx_xvadda_d