Complete vst1 neon instructions (#1221)

Author: SparrowLii

Diff for: crates/core_arch/src/aarch64/neon/generated.rs

@@ -13066,7 +13066,7 @@ mod test {
         let a: [f64; 3] = [0., 1., 2.];
         let e: [f64; 2] = [1., 2.];
         let mut r: [f64; 2] = [0f64; 2];
-        vst1_f64_x2(r.as_mut_ptr(), vld1_f64_x2(a[1..].as_ptr()));
+        vst1_f64_x2(r.as_mut_ptr(), core::ptr::read_unaligned(a[1..].as_ptr().cast()));
         assert_eq!(r, e);
     }
 
@@ -13075,7 +13075,7 @@ mod test {
         let a: [f64; 5] = [0., 1., 2., 3., 4.];
         let e: [f64; 4] = [1., 2., 3., 4.];
         let mut r: [f64; 4] = [0f64; 4];
-        vst1q_f64_x2(r.as_mut_ptr(), vld1q_f64_x2(a[1..].as_ptr()));
+        vst1q_f64_x2(r.as_mut_ptr(), core::ptr::read_unaligned(a[1..].as_ptr().cast()));
         assert_eq!(r, e);
     }
 
@@ -13084,7 +13084,7 @@ mod test {
         let a: [f64; 4] = [0., 1., 2., 3.];
         let e: [f64; 3] = [1., 2., 3.];
         let mut r: [f64; 3] = [0f64; 3];
-        vst1_f64_x3(r.as_mut_ptr(), vld1_f64_x3(a[1..].as_ptr()));
+        vst1_f64_x3(r.as_mut_ptr(), core::ptr::read_unaligned(a[1..].as_ptr().cast()));
         assert_eq!(r, e);
     }
 
@@ -13093,7 +13093,7 @@ mod test {
         let a: [f64; 7] = [0., 1., 2., 3., 4., 5., 6.];
         let e: [f64; 6] = [1., 2., 3., 4., 5., 6.];
         let mut r: [f64; 6] = [0f64; 6];
-        vst1q_f64_x3(r.as_mut_ptr(), vld1q_f64_x3(a[1..].as_ptr()));
+        vst1q_f64_x3(r.as_mut_ptr(), core::ptr::read_unaligned(a[1..].as_ptr().cast()));
         assert_eq!(r, e);
     }
 
@@ -13102,7 +13102,7 @@ mod test {
         let a: [f64; 5] = [0., 1., 2., 3., 4.];
         let e: [f64; 4] = [1., 2., 3., 4.];
         let mut r: [f64; 4] = [0f64; 4];
-        vst1_f64_x4(r.as_mut_ptr(), vld1_f64_x4(a[1..].as_ptr()));
+        vst1_f64_x4(r.as_mut_ptr(), core::ptr::read_unaligned(a[1..].as_ptr().cast()));
         assert_eq!(r, e);
     }
 
@@ -13111,7 +13111,7 @@ mod test {
         let a: [f64; 9] = [0., 1., 2., 3., 4., 5., 6., 7., 8.];
         let e: [f64; 8] = [1., 2., 3., 4., 5., 6., 7., 8.];
         let mut r: [f64; 8] = [0f64; 8];
-        vst1q_f64_x4(r.as_mut_ptr(), vld1q_f64_x4(a[1..].as_ptr()));
+        vst1q_f64_x4(r.as_mut_ptr(), core::ptr::read_unaligned(a[1..].as_ptr().cast()));
         assert_eq!(r, e);
     }
 

Diff for: crates/core_arch/src/aarch64/neon/mod.rs

@@ -678,7 +678,7 @@ pub unsafe fn vld1_dup_f64(ptr: *const f64) -> float64x1_t {
 /// Load multiple single-element structures to one, two, three, or four registers
 #[inline]
 #[target_feature(enable = "neon")]
-#[cfg_attr(test, assert_instr(ldr))]
+#[cfg_attr(test, assert_instr(ld1r))]
 pub unsafe fn vld1q_dup_f64(ptr: *const f64) -> float64x2_t {
     let x = vld1q_lane_f64::<0>(ptr, transmute(f64x2::splat(0.)));
     simd_shuffle2!(x, x, [0, 0])
@@ -698,7 +698,7 @@ pub unsafe fn vld1_lane_f64<const LANE: i32>(ptr: *const f64, src: float64x1_t)
 #[inline]
 #[target_feature(enable = "neon")]
 #[rustc_legacy_const_generics(2)]
-#[cfg_attr(test, assert_instr(ldr, LANE = 1))]
+#[cfg_attr(test, assert_instr(ld1, LANE = 1))]
 pub unsafe fn vld1q_lane_f64<const LANE: i32>(ptr: *const f64, src: float64x2_t) -> float64x2_t {
     static_assert_imm1!(LANE);
     simd_insert(src, LANE as u32, *ptr)
@@ -886,7 +886,7 @@ pub unsafe fn vst1q_p16(ptr: *mut p16, a: poly16x8_t) {
 
 // Store multiple single-element structures from one, two, three, or four registers.
 #[inline]
-#[target_feature(enable = "neon")]
+#[target_feature(enable = "neon,aes")]
 #[cfg_attr(test, assert_instr(str))]
 #[allow(clippy::cast_ptr_alignment)]
 pub unsafe fn vst1_p64(ptr: *mut p64, a: poly64x1_t) {
@@ -895,7 +895,7 @@ pub unsafe fn vst1_p64(ptr: *mut p64, a: poly64x1_t) {
 
 // Store multiple single-element structures from one, two, three, or four registers.
 #[inline]
-#[target_feature(enable = "neon")]
+#[target_feature(enable = "neon,aes")]
 #[cfg_attr(test, assert_instr(str))]
 #[allow(clippy::cast_ptr_alignment)]
 pub unsafe fn vst1q_p64(ptr: *mut p64, a: poly64x2_t) {
@@ -4803,29 +4803,6 @@ mod tests {
         assert_eq!(r, e)
     }
 
-    #[simd_test(enable = "neon")]
-    unsafe fn test_vst1_p64() {
-        let mut vals = [0_u64; 2];
-        let a = u64x1::new(1);
-
-        vst1_p64(vals[1..].as_mut_ptr(), transmute(a));
-
-        assert_eq!(vals[0], 0);
-        assert_eq!(vals[1], 1);
-    }
-
-    #[simd_test(enable = "neon")]
-    unsafe fn test_vst1q_p64() {
-        let mut vals = [0_u64; 3];
-        let a = u64x2::new(1, 2);
-
-        vst1q_p64(vals[1..].as_mut_ptr(), transmute(a));
-
-        assert_eq!(vals[0], 0);
-        assert_eq!(vals[1], 1);
-        assert_eq!(vals[2], 2);
-    }
-
     #[simd_test(enable = "neon")]
     unsafe fn test_vst1_f64() {
         let mut vals = [0_f64; 2];

Diff for: crates/core_arch/src/arm/neon.rs

@@ -480,6 +480,22 @@ pub unsafe fn vst1q_p16(ptr: *mut p16, a: poly16x8_t) {
     vst1q_v8i16(ptr as *const i8, transmute(a), align_of::<p8>() as i32)
 }
 
+/// Store multiple single-element structures from one, two, three, or four registers.
+#[inline]
+#[target_feature(enable = "neon,aes,v8")]
+#[cfg_attr(test, assert_instr("vst1.64"))]
+pub unsafe fn vst1_p64(ptr: *mut p64, a: poly64x1_t) {
+    vst1_v1i64(ptr as *const i8, transmute(a), align_of::<p64>() as i32)
+}
+
+/// Store multiple single-element structures from one, two, three, or four registers.
+#[inline]
+#[target_feature(enable = "neon,aes,v8")]
+#[cfg_attr(test, assert_instr("vst1.64"))]
+pub unsafe fn vst1q_p64(ptr: *mut p64, a: poly64x2_t) {
+    vst1q_v2i64(ptr as *const i8, transmute(a), align_of::<p64>() as i32)
+}
+
 // Store multiple single-element structures from one, two, three, or four registers.
 #[inline]
 #[target_feature(enable = "neon,v7")]