loongarch: Use intrinsics::simd for vpack{ev,od}

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

@@ -147,22 +147,6 @@ unsafe extern "unadjusted" {
     fn __lasx_xvhsubw_wu_hu(a: __v16u16, b: __v16u16) -> __v8i32;
     #[link_name = "llvm.loongarch.lasx.xvhsubw.du.wu"]
     fn __lasx_xvhsubw_du_wu(a: __v8u32, b: __v8u32) -> __v4i64;
-    #[link_name = "llvm.loongarch.lasx.xvpackev.b"]
-    fn __lasx_xvpackev_b(a: __v32i8, b: __v32i8) -> __v32i8;
-    #[link_name = "llvm.loongarch.lasx.xvpackev.h"]
-    fn __lasx_xvpackev_h(a: __v16i16, b: __v16i16) -> __v16i16;
-    #[link_name = "llvm.loongarch.lasx.xvpackev.w"]
-    fn __lasx_xvpackev_w(a: __v8i32, b: __v8i32) -> __v8i32;
-    #[link_name = "llvm.loongarch.lasx.xvpackev.d"]
-    fn __lasx_xvpackev_d(a: __v4i64, b: __v4i64) -> __v4i64;
-    #[link_name = "llvm.loongarch.lasx.xvpackod.b"]
-    fn __lasx_xvpackod_b(a: __v32i8, b: __v32i8) -> __v32i8;
-    #[link_name = "llvm.loongarch.lasx.xvpackod.h"]
-    fn __lasx_xvpackod_h(a: __v16i16, b: __v16i16) -> __v16i16;
-    #[link_name = "llvm.loongarch.lasx.xvpackod.w"]
-    fn __lasx_xvpackod_w(a: __v8i32, b: __v8i32) -> __v8i32;
-    #[link_name = "llvm.loongarch.lasx.xvpackod.d"]
-    fn __lasx_xvpackod_d(a: __v4i64, b: __v4i64) -> __v4i64;
     #[link_name = "llvm.loongarch.lasx.xvshuf.b"]
     fn __lasx_xvshuf_b(a: __v32i8, b: __v32i8, c: __v32i8) -> __v32i8;
     #[link_name = "llvm.loongarch.lasx.xvshuf.h"]
@@ -1567,62 +1551,6 @@ pub fn lasx_xvhsubw_du_wu(a: m256i, b: m256i) -> m256i {
     unsafe { transmute(__lasx_xvhsubw_du_wu(transmute(a), transmute(b))) }
 }
 
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackev_b(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackev_b(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackev_h(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackev_h(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackev_w(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackev_w(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackev_d(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackev_d(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackod_b(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackod_b(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackod_h(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackod_h(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackod_w(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackod_w(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lasx_xvpackod_d(a: m256i, b: m256i) -> m256i {
-    unsafe { transmute(__lasx_xvpackod_d(transmute(a), transmute(b))) }
-}
-
 #[inline]
 #[target_feature(enable = "lasx")]
 #[unstable(feature = "stdarch_loongarch", issue = "117427")]

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

@@ -205,6 +205,68 @@ pub(super) const unsafe fn simd_replve0_q<T: Copy>(a: T) -> T {
     simd_shuffle!(a, a, [0, 1, 0, 1])
 }
 
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_b<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(
+        b,
+        a,
+        [
+            0, 32, 2, 34, 4, 36, 6, 38, 8, 40, 10, 42, 12, 44, 14, 46,
+            16, 48, 18, 50, 20, 52, 22, 54, 24, 56, 26, 58, 28, 60, 30, 62
+        ]
+    )
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_h<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [0, 16, 2, 18, 4, 20, 6, 22, 8, 24, 10, 26, 12, 28, 14, 30])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_w<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [0, 8, 2, 10, 4, 12, 6, 14])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_d<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [0, 4, 2, 6])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_b<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(
+        b,
+        a,
+        [
+            1, 33, 3, 35, 5, 37, 7, 39, 9, 41, 11, 43, 13, 45, 15, 47,
+            17, 49, 19, 51, 21, 53, 23, 55, 25, 57, 27, 59, 29, 61, 31, 63
+        ]
+    )
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_h<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [1, 17, 3, 19, 5, 21, 7, 23, 9, 25, 11, 27, 13, 29, 15, 31])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_w<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [1, 9, 3, 11, 5, 13, 7, 15])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_d<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [1, 5, 3, 7])
+}
+
 impl_vv!("lasx", lasx_xvpcnt_b, is::simd_ctpop, m256i, i8x32);
 impl_vv!("lasx", lasx_xvpcnt_h, is::simd_ctpop, m256i, i16x16);
 impl_vv!("lasx", lasx_xvpcnt_w, is::simd_ctpop, m256i, i32x8);
@@ -381,6 +443,14 @@ impl_vvv!("lasx", lasx_xvilvl_b, simd_ilvl_b, m256i, i8x32);
 impl_vvv!("lasx", lasx_xvilvl_h, simd_ilvl_h, m256i, i16x16);
 impl_vvv!("lasx", lasx_xvilvl_w, simd_ilvl_w, m256i, i32x8);
 impl_vvv!("lasx", lasx_xvilvl_d, simd_ilvl_d, m256i, i64x4);
+impl_vvv!("lasx", lasx_xvpackev_b, simd_packev_b, m256i, i8x32);
+impl_vvv!("lasx", lasx_xvpackev_h, simd_packev_h, m256i, i16x16);
+impl_vvv!("lasx", lasx_xvpackev_w, simd_packev_w, m256i, i32x8);
+impl_vvv!("lasx", lasx_xvpackev_d, simd_packev_d, m256i, i64x4);
+impl_vvv!("lasx", lasx_xvpackod_b, simd_packod_b, m256i, i8x32);
+impl_vvv!("lasx", lasx_xvpackod_h, simd_packod_h, m256i, i16x16);
+impl_vvv!("lasx", lasx_xvpackod_w, simd_packod_w, m256i, i32x8);
+impl_vvv!("lasx", lasx_xvpackod_d, simd_packod_d, 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

@@ -155,22 +155,6 @@ unsafe extern "unadjusted" {
     fn __lsx_vreplve_w(a: __v4i32, b: i32) -> __v4i32;
     #[link_name = "llvm.loongarch.lsx.vreplve.d"]
     fn __lsx_vreplve_d(a: __v2i64, b: i32) -> __v2i64;
-    #[link_name = "llvm.loongarch.lsx.vpackev.b"]
-    fn __lsx_vpackev_b(a: __v16i8, b: __v16i8) -> __v16i8;
-    #[link_name = "llvm.loongarch.lsx.vpackev.h"]
-    fn __lsx_vpackev_h(a: __v8i16, b: __v8i16) -> __v8i16;
-    #[link_name = "llvm.loongarch.lsx.vpackev.w"]
-    fn __lsx_vpackev_w(a: __v4i32, b: __v4i32) -> __v4i32;
-    #[link_name = "llvm.loongarch.lsx.vpackev.d"]
-    fn __lsx_vpackev_d(a: __v2i64, b: __v2i64) -> __v2i64;
-    #[link_name = "llvm.loongarch.lsx.vpackod.b"]
-    fn __lsx_vpackod_b(a: __v16i8, b: __v16i8) -> __v16i8;
-    #[link_name = "llvm.loongarch.lsx.vpackod.h"]
-    fn __lsx_vpackod_h(a: __v8i16, b: __v8i16) -> __v8i16;
-    #[link_name = "llvm.loongarch.lsx.vpackod.w"]
-    fn __lsx_vpackod_w(a: __v4i32, b: __v4i32) -> __v4i32;
-    #[link_name = "llvm.loongarch.lsx.vpackod.d"]
-    fn __lsx_vpackod_d(a: __v2i64, b: __v2i64) -> __v2i64;
     #[link_name = "llvm.loongarch.lsx.vshuf.h"]
     fn __lsx_vshuf_h(a: __v8i16, b: __v8i16, c: __v8i16) -> __v8i16;
     #[link_name = "llvm.loongarch.lsx.vshuf.w"]
@@ -1517,62 +1501,6 @@ pub fn lsx_vreplve_d(a: m128i, b: i32) -> m128i {
     unsafe { transmute(__lsx_vreplve_d(transmute(a), transmute(b))) }
 }
 
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackev_b(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackev_b(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackev_h(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackev_h(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackev_w(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackev_w(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackev_d(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackev_d(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackod_b(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackod_b(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackod_h(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackod_h(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackod_w(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackod_w(transmute(a), transmute(b))) }
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub fn lsx_vpackod_d(a: m128i, b: m128i) -> m128i {
-    unsafe { transmute(__lsx_vpackod_d(transmute(a), transmute(b))) }
-}
-
 #[inline]
 #[target_feature(enable = "lsx")]
 #[unstable(feature = "stdarch_loongarch", issue = "117427")]

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

@@ -125,6 +125,54 @@ pub(crate) const unsafe fn simd_replvei_d<const I: u32, T: Copy>(a: T) -> T {
     simd_shuffle!(a, a, [I, I])
 }
 
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_b<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [0, 16, 2, 18, 4, 20, 6, 22, 8, 24, 10, 26, 12, 28, 14, 30])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_h<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [0, 8, 2, 10, 4, 12, 6, 14])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_w<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [0, 4, 2, 6])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packev_d<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [0, 2])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_b<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [1, 17, 3, 19, 5, 21, 7, 23, 9, 25, 11, 27, 13, 29, 15, 31])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_h<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [1, 9, 3, 11, 5, 13, 7, 15])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_w<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [1, 5, 3, 7])
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(crate) const unsafe fn simd_packod_d<T: Copy>(a: T, b: T) -> T {
+    simd_shuffle!(b, a, [1, 3])
+}
+
 impl_vv!("lsx", lsx_vpcnt_b, is::simd_ctpop, m128i, i8x16);
 impl_vv!("lsx", lsx_vpcnt_h, is::simd_ctpop, m128i, i16x8);
 impl_vv!("lsx", lsx_vpcnt_w, is::simd_ctpop, m128i, i32x4);
@@ -296,6 +344,14 @@ impl_vvv!("lsx", lsx_vilvl_b, simd_ilvl_b, m128i, i8x16);
 impl_vvv!("lsx", lsx_vilvl_h, simd_ilvl_h, m128i, i16x8);
 impl_vvv!("lsx", lsx_vilvl_w, simd_ilvl_w, m128i, i32x4);
 impl_vvv!("lsx", lsx_vilvl_d, simd_ilvl_d, m128i, i64x2);
+impl_vvv!("lsx", lsx_vpackev_b, simd_packev_b, m128i, i8x16);
+impl_vvv!("lsx", lsx_vpackev_h, simd_packev_h, m128i, i16x8);
+impl_vvv!("lsx", lsx_vpackev_w, simd_packev_w, m128i, i32x4);
+impl_vvv!("lsx", lsx_vpackev_d, simd_packev_d, m128i, i64x2);
+impl_vvv!("lsx", lsx_vpackod_b, simd_packod_b, m128i, i8x16);
+impl_vvv!("lsx", lsx_vpackod_h, simd_packod_h, m128i, i16x8);
+impl_vvv!("lsx", lsx_vpackod_w, simd_packod_w, m128i, i32x4);
+impl_vvv!("lsx", lsx_vpackod_d, simd_packod_d, 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/stdarch-gen-loongarch/lasx.spec

@@ -1520,41 +1520,49 @@ asm-fmts = xd, xj, xk
 data-types = V4DI, V4DI, V4DI
 
 /// lasx_xvpackev_b
+impl = portable
 name = lasx_xvpackev_b
 asm-fmts = xd, xj, xk
 data-types = V32QI, V32QI, V32QI
 
 /// lasx_xvpackev_h
+impl = portable
 name = lasx_xvpackev_h
 asm-fmts = xd, xj, xk
 data-types = V16HI, V16HI, V16HI
 
 /// lasx_xvpackev_w
+impl = portable
 name = lasx_xvpackev_w
 asm-fmts = xd, xj, xk
 data-types = V8SI, V8SI, V8SI
 
 /// lasx_xvpackev_d
+impl = portable
 name = lasx_xvpackev_d
 asm-fmts = xd, xj, xk
 data-types = V4DI, V4DI, V4DI
 
 /// lasx_xvpackod_b
+impl = portable
 name = lasx_xvpackod_b
 asm-fmts = xd, xj, xk
 data-types = V32QI, V32QI, V32QI
 
 /// lasx_xvpackod_h
+impl = portable
 name = lasx_xvpackod_h
 asm-fmts = xd, xj, xk
 data-types = V16HI, V16HI, V16HI
 
 /// lasx_xvpackod_w
+impl = portable
 name = lasx_xvpackod_w
 asm-fmts = xd, xj, xk
 data-types = V8SI, V8SI, V8SI
 
 /// lasx_xvpackod_d
+impl = portable
 name = lasx_xvpackod_d
 asm-fmts = xd, xj, xk
 data-types = V4DI, V4DI, V4DI