loongarch: Use intrinsics::simd for v{ld,st}[x]

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

@@ -455,10 +455,6 @@ unsafe extern "unadjusted" {
     fn __lasx_xvfrintrm_s(a: __v8f32) -> __v8f32;
     #[link_name = "llvm.loongarch.lasx.xvfrintrm.d"]
     fn __lasx_xvfrintrm_d(a: __v4f64) -> __v4f64;
-    #[link_name = "llvm.loongarch.lasx.xvld"]
-    fn __lasx_xvld(a: *const i8, b: i32) -> __v32i8;
-    #[link_name = "llvm.loongarch.lasx.xvst"]
-    fn __lasx_xvst(a: __v32i8, b: *mut i8, c: i32);
     #[link_name = "llvm.loongarch.lasx.xvstelm.b"]
     fn __lasx_xvstelm_b(a: __v32i8, b: *mut i8, c: i32, d: u32);
     #[link_name = "llvm.loongarch.lasx.xvstelm.h"]
@@ -489,10 +485,6 @@ unsafe extern "unadjusted" {
     fn __lasx_xvssrln_w_d(a: __v4i64, b: __v4i64) -> __v8i32;
     #[link_name = "llvm.loongarch.lasx.xvldi"]
     fn __lasx_xvldi(a: i32) -> __v4i64;
-    #[link_name = "llvm.loongarch.lasx.xvldx"]
-    fn __lasx_xvldx(a: *const i8, b: i64) -> __v32i8;
-    #[link_name = "llvm.loongarch.lasx.xvstx"]
-    fn __lasx_xvstx(a: __v32i8, b: *mut i8, c: i64);
     #[link_name = "llvm.loongarch.lasx.xvextl.qu.du"]
     fn __lasx_xvextl_qu_du(a: __v4u64) -> __v4u64;
     #[link_name = "llvm.loongarch.lasx.vext2xv.h.b"]
@@ -2665,24 +2657,6 @@ pub fn lasx_xvfrintrm_d(a: m256d) -> m256d {
     unsafe { transmute(__lasx_xvfrintrm_d(transmute(a))) }
 }
 
-#[inline]
-#[target_feature(enable = "lasx")]
-#[rustc_legacy_const_generics(1)]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lasx_xvld<const IMM_S12: i32>(mem_addr: *const i8) -> m256i {
-    static_assert_simm_bits!(IMM_S12, 12);
-    transmute(__lasx_xvld(mem_addr, IMM_S12))
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[rustc_legacy_const_generics(2)]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lasx_xvst<const IMM_S12: i32>(a: m256i, mem_addr: *mut i8) {
-    static_assert_simm_bits!(IMM_S12, 12);
-    __lasx_xvst(transmute(a), mem_addr, IMM_S12)
-}
-
 #[inline]
 #[target_feature(enable = "lasx")]
 #[rustc_legacy_const_generics(2, 3)]
@@ -2810,20 +2784,6 @@ pub fn lasx_xvldi<const IMM_S13: i32>() -> m256i {
     unsafe { transmute(__lasx_xvldi(IMM_S13)) }
 }
 
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lasx_xvldx(mem_addr: *const i8, b: i64) -> m256i {
-    transmute(__lasx_xvldx(mem_addr, transmute(b)))
-}
-
-#[inline]
-#[target_feature(enable = "lasx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lasx_xvstx(a: m256i, mem_addr: *mut i8, b: i64) {
-    __lasx_xvstx(transmute(a), mem_addr, transmute(b))
-}
-
 #[inline]
 #[target_feature(enable = "lasx")]
 #[unstable(feature = "stdarch_loongarch", issue = "117427")]

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

@@ -339,6 +339,10 @@ impl_gv!("lasx", lasx_xvreplgr2vr_h, ls::simd_splat, m256i, i16x16, i32);
 impl_gv!("lasx", lasx_xvreplgr2vr_w, ls::simd_splat, m256i, i32x8, i32);
 impl_gv!("lasx", lasx_xvreplgr2vr_d, ls::simd_splat, m256i, i64x4, i64);
 
+impl_ggv!("lasx", lasx_xvldx, simd_ldx, m256i, i8x32, *const i8, i64, unsafe);
+
+impl_gsv!("lasx", lasx_xvld, simd_ld, m256i, i8x32, *const i8, 12, const, unsafe);
+
 impl_sv!("lasx", lasx_xvrepli_b, ls::simd_splat, m256i, i8x32, 10);
 impl_sv!("lasx", lasx_xvrepli_h, ls::simd_splat, m256i, i16x16, 10);
 impl_sv!("lasx", lasx_xvrepli_w, ls::simd_splat, m256i, i32x8, 10);
@@ -499,6 +503,10 @@ 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_vgg!("lasx", lasx_xvstx, simd_stx, m256i, i8x32, *mut i8, i64, unsafe);
+
+impl_vgs!("lasx", lasx_xvst, simd_st, m256i, i8x32, *mut i8, 12, const, unsafe);
+
 impl_vuv!("lasx", lasx_xvslli_b, is::simd_shl, m256i, i8x32);
 impl_vuv!("lasx", lasx_xvslli_h, is::simd_shl, m256i, i16x16);
 impl_vuv!("lasx", lasx_xvslli_w, is::simd_shl, m256i, i32x8);

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

@@ -789,10 +789,6 @@ unsafe extern "unadjusted" {
     fn __lsx_vssrarni_du_q(a: __v2u64, b: __v2i64, c: u32) -> __v2u64;
     #[link_name = "llvm.loongarch.lsx.vpermi.w"]
     fn __lsx_vpermi_w(a: __v4i32, b: __v4i32, c: u32) -> __v4i32;
-    #[link_name = "llvm.loongarch.lsx.vld"]
-    fn __lsx_vld(a: *const i8, b: i32) -> __v16i8;
-    #[link_name = "llvm.loongarch.lsx.vst"]
-    fn __lsx_vst(a: __v16i8, b: *mut i8, c: i32);
     #[link_name = "llvm.loongarch.lsx.vssrlrn.b.h"]
     fn __lsx_vssrlrn_b_h(a: __v8i16, b: __v8i16) -> __v16i8;
     #[link_name = "llvm.loongarch.lsx.vssrlrn.h.w"]
@@ -809,10 +805,6 @@ unsafe extern "unadjusted" {
     fn __lsx_vldi(a: i32) -> __v2i64;
     #[link_name = "llvm.loongarch.lsx.vshuf.b"]
     fn __lsx_vshuf_b(a: __v16i8, b: __v16i8, c: __v16i8) -> __v16i8;
-    #[link_name = "llvm.loongarch.lsx.vldx"]
-    fn __lsx_vldx(a: *const i8, b: i64) -> __v16i8;
-    #[link_name = "llvm.loongarch.lsx.vstx"]
-    fn __lsx_vstx(a: __v16i8, b: *mut i8, c: i64);
     #[link_name = "llvm.loongarch.lsx.vextl.qu.du"]
     fn __lsx_vextl_qu_du(a: __v2u64) -> __v2u64;
     #[link_name = "llvm.loongarch.lsx.bnz.b"]
@@ -3876,24 +3868,6 @@ pub fn lsx_vpermi_w<const IMM8: u32>(a: m128i, b: m128i) -> m128i {
     unsafe { transmute(__lsx_vpermi_w(transmute(a), transmute(b), IMM8)) }
 }
 
-#[inline]
-#[target_feature(enable = "lsx")]
-#[rustc_legacy_const_generics(1)]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lsx_vld<const IMM_S12: i32>(mem_addr: *const i8) -> m128i {
-    static_assert_simm_bits!(IMM_S12, 12);
-    transmute(__lsx_vld(mem_addr, IMM_S12))
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[rustc_legacy_const_generics(2)]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lsx_vst<const IMM_S12: i32>(a: m128i, mem_addr: *mut i8) {
-    static_assert_simm_bits!(IMM_S12, 12);
-    __lsx_vst(transmute(a), mem_addr, IMM_S12)
-}
-
 #[inline]
 #[target_feature(enable = "lsx")]
 #[unstable(feature = "stdarch_loongarch", issue = "117427")]
@@ -3952,20 +3926,6 @@ pub fn lsx_vshuf_b(a: m128i, b: m128i, c: m128i) -> m128i {
     unsafe { transmute(__lsx_vshuf_b(transmute(a), transmute(b), transmute(c))) }
 }
 
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lsx_vldx(mem_addr: *const i8, b: i64) -> m128i {
-    transmute(__lsx_vldx(mem_addr, transmute(b)))
-}
-
-#[inline]
-#[target_feature(enable = "lsx")]
-#[unstable(feature = "stdarch_loongarch", issue = "117427")]
-pub unsafe fn lsx_vstx(a: m128i, mem_addr: *mut i8, b: i64) {
-    __lsx_vstx(transmute(a), mem_addr, transmute(b))
-}
-
 #[inline]
 #[target_feature(enable = "lsx")]
 #[unstable(feature = "stdarch_loongarch", issue = "117427")]

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

@@ -233,6 +233,10 @@ impl_gv!("lsx", lsx_vreplgr2vr_h, ls::simd_splat, m128i, i16x8, i32);
 impl_gv!("lsx", lsx_vreplgr2vr_w, ls::simd_splat, m128i, i32x4, i32);
 impl_gv!("lsx", lsx_vreplgr2vr_d, ls::simd_splat, m128i, i64x2, i64);
 
+impl_ggv!("lsx", lsx_vldx, simd_ldx, m128i, i8x16, *const i8, i64, unsafe);
+
+impl_gsv!("lsx", lsx_vld, simd_ld, m128i, i8x16, *const i8, 12, const, unsafe);
+
 impl_sv!("lsx", lsx_vrepli_b, ls::simd_splat, m128i, i8x16, 10);
 impl_sv!("lsx", lsx_vrepli_h, ls::simd_splat, m128i, i16x8, 10);
 impl_sv!("lsx", lsx_vrepli_w, ls::simd_splat, m128i, i32x4, 10);
@@ -393,6 +397,10 @@ 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_vgg!("lsx", lsx_vstx, simd_stx, m128i, i8x16, *mut i8, i64, unsafe);
+
+impl_vgs!("lsx", lsx_vst, simd_st, m128i, i8x16, *mut i8, 12, const, unsafe);
+
 impl_vuv!("lsx", lsx_vslli_b, is::simd_shl, m128i, i8x16);
 impl_vuv!("lsx", lsx_vslli_h, is::simd_shl, m128i, i16x8);
 impl_vuv!("lsx", lsx_vslli_w, is::simd_shl, m128i, i32x4);

crates/core_arch/src/loongarch64/simd.rs

@@ -108,6 +108,20 @@ pub(super) const unsafe fn simd_fnmsub<T: Copy>(a: T, b: T, c: T) -> T {
     is::simd_neg(ls::simd_fmsub(a, b, c))
 }
 
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(super) const unsafe fn simd_ld<const I: i32, T: Copy>(a: *const i8) -> T {
+    let a = a.offset(I as isize) as *const T;
+    core::ptr::read_unaligned(a)
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(super) const unsafe fn simd_ldx<T: Copy>(a: *const i8, b: i64) -> T {
+    let a = a.offset(b as isize) as *const T;
+    core::ptr::read_unaligned(a)
+}
+
 #[inline(always)]
 #[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
 pub(super) const unsafe fn simd_madd<T: Copy>(a: T, b: T, c: T) -> T {
@@ -158,6 +172,20 @@ pub(super) const unsafe fn simd_splat<T: Copy + const SimdExt>(a: i64) -> T {
     T::splat(a)
 }
 
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(super) const unsafe fn simd_st<const I: i32, T: Copy>(a: T, b: *mut i8) {
+    let b = b.offset(I as isize) as *mut T;
+    core::ptr::write_unaligned(b, a);
+}
+
+#[inline(always)]
+#[rustc_const_unstable(feature = "stdarch_const_helpers", issue = "none")]
+pub(super) const unsafe fn simd_stx<T: Copy>(a: T, b: *mut i8, c: i64) {
+    let b = b.offset(c as isize) as *mut T;
+    core::ptr::write_unaligned(b, a);
+}
+
 macro_rules! impl_vv {
     ($ft:literal, $name:ident, $op:path, $oty:ty, $ity:ty) => {
         #[inline]
@@ -191,6 +219,36 @@ macro_rules! impl_gv {
 
 pub(super) use impl_gv;
 
+macro_rules! impl_ggv {
+    ($ft:literal, $name:ident, $op:path, $oty:ty, $ity:ident, $gty:ty, $xty:ty, unsafe) => {
+        #[inline]
+        #[target_feature(enable = $ft)]
+        #[unstable(feature = "stdarch_loongarch", issue = "117427")]
+        pub unsafe fn $name(a: $gty, b: $xty) -> $oty {
+            let r: $ity = $op(a, b);
+            transmute(r)
+        }
+    };
+}
+
+pub(super) use impl_ggv;
+
+macro_rules! impl_gsv {
+    ($ft:literal, $name:ident, $op:ident, $oty:ty, $ity:ident, $gty:ty, $ibs:expr, const, unsafe) => {
+        #[inline]
+        #[target_feature(enable = $ft)]
+        #[rustc_legacy_const_generics(1)]
+        #[unstable(feature = "stdarch_loongarch", issue = "117427")]
+        pub unsafe fn $name<const IMM: i32>(a: $gty) -> $oty {
+            static_assert_simm_bits!(IMM, $ibs);
+            let r: $ity = $op::<IMM, _>(a);
+            transmute(r)
+        }
+    };
+}
+
+pub(super) use impl_gsv;
+
 macro_rules! impl_sv {
     ($ft:literal, $name:ident, $op:path, $oty:ty, $ity:ident, $ibs:expr) => {
         #[inline]
@@ -227,6 +285,34 @@ macro_rules! impl_vvv {
 
 pub(super) use impl_vvv;
 
+macro_rules! impl_vgg {
+    ($ft:literal, $name:ident, $op:path, $oty:ty, $ity:ident, $gty:ty, $xty:ty, unsafe) => {
+        #[inline]
+        #[target_feature(enable = $ft)]
+        #[unstable(feature = "stdarch_loongarch", issue = "117427")]
+        pub unsafe fn $name(a: $oty, b: $gty, c: $xty) {
+            $op(a, b, c);
+        }
+    };
+}
+
+pub(super) use impl_vgg;
+
+macro_rules! impl_vgs {
+    ($ft:literal, $name:ident, $op:ident, $oty:ty, $ity:ident, $gty:ty, $ibs:expr, const, unsafe) => {
+        #[inline]
+        #[target_feature(enable = $ft)]
+        #[rustc_legacy_const_generics(2)]
+        #[unstable(feature = "stdarch_loongarch", issue = "117427")]
+        pub unsafe fn $name<const IMM: i32>(a: $oty, b: $gty) {
+            static_assert_simm_bits!(IMM, $ibs);
+            $op::<IMM, _>(a, b);
+        }
+    };
+}
+
+pub(super) use impl_vgs;
+
 macro_rules! impl_vuv {
     ($ft:literal, $name:ident, $op:path, $oty:ty, $ity:ident) => {
         #[inline]

crates/stdarch-gen-loongarch/lasx.spec

@@ -2590,11 +2590,13 @@ asm-fmts = xd, xj
 data-types = V4DF, V4DF
 
 /// lasx_xvld
+impl = portable
 name = lasx_xvld
 asm-fmts = xd, rj, si12
 data-types = V32QI, CVPOINTER, SI
 
 /// lasx_xvst
+impl = portable
 name = lasx_xvst
 asm-fmts = xd, rj, si12
 data-types = VOID, V32QI, CVPOINTER, SI
@@ -2681,11 +2683,13 @@ asm-fmts = xd, i13
 data-types = V4DI, HI
 
 /// lasx_xvldx
+impl = portable
 name = lasx_xvldx
 asm-fmts = xd, rj, rk
 data-types = V32QI, CVPOINTER, DI
 
 /// lasx_xvstx
+impl = portable
 name = lasx_xvstx
 asm-fmts = xd, rj, rk
 data-types = VOID, V32QI, CVPOINTER, DI

crates/stdarch-gen-loongarch/lsx.spec

@@ -3503,11 +3503,13 @@ asm-fmts = vd, vj, ui8
 data-types = V4SI, V4SI, V4SI, USI
 
 /// lsx_vld
+impl = portable
 name = lsx_vld
 asm-fmts = vd, rj, si12
 data-types = V16QI, CVPOINTER, SI
 
 /// lsx_vst
+impl = portable
 name = lsx_vst
 asm-fmts = vd, rj, si12
 data-types = VOID, V16QI, CVPOINTER, SI
@@ -3559,11 +3561,13 @@ asm-fmts = vd, vj, vk, va
 data-types = V16QI, V16QI, V16QI, V16QI
 
 /// lsx_vldx
+impl = portable
 name = lsx_vldx
 asm-fmts = vd, rj, rk
 data-types = V16QI, CVPOINTER, DI
 
 /// lsx_vstx
+impl = portable
 name = lsx_vstx
 asm-fmts = vd, rj, rk
 data-types = VOID, V16QI, CVPOINTER, DI

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

@@ -263,6 +263,10 @@ lsx_vshuf4i_b
 lsx_vshuf4i_h
 lsx_vshuf4i_w
 lsx_vshuf4i_d
+lsx_vld
+lsx_vst
+lsx_vldx
+lsx_vstx
 
 # LASX intrinsics
 lasx_xvsll_b
@@ -527,3 +531,7 @@ lasx_xvpackod_d
 lasx_xvshuf4i_b
 lasx_xvshuf4i_h
 lasx_xvshuf4i_w
+lasx_xvld
+lasx_xvst
+lasx_xvldx
+lasx_xvstx