Auto merge of #154094 - folkertdev:aarch64-arm-load-store, r=<try>

add neon load/store assembly test


try-job: aarch64-apple

Author: bors

tests/assembly-llvm/aarch64-arm-load-store.rs

@@ -0,0 +1,78 @@
+//@ assembly-output: emit-asm
+//
+//@ revisions: AARCH64
+//@[AARCH64] compile-flags: -Copt-level=3
+//@[AARCH64] only-aarch64-unknown-linux-gnu
+//
+//@ revisions: ARMV7
+//@[ARMV7] compile-flags: -Copt-level=3
+//@[ARMV7] only-arm
+//@[ARMV7] ignore-thumb
+//@[ARMV7] ignore-android
+#![crate_type = "lib"]
+#![cfg_attr(target_arch = "arm", feature(arm_target_feature, stdarch_arm_neon_intrinsics))]
+
+#[cfg(target_arch = "aarch64")]
+use std::arch::aarch64::*;
+#[cfg(target_arch = "arm")]
+use std::arch::arm::*;
+
+// Loads of 3 are error-prone because a `repr(simd)` type's size is always rounded up to the next
+// power of 2. Hence, using `read_unaligned` and `write_unaligned` on such types is invalid, it
+// would go out of bounds.
+#[unsafe(no_mangle)]
+#[cfg_attr(target_arch = "arm", target_feature(enable = "neon,v7"))]
+fn test_vld3q_f32(ptr: *const f32) -> float32x4x3_t {
+    // AARCH64-LABEL: test_vld3q_f32
+    // AARCH64: ld3 { v0.4s, v1.4s, v2.4s }, [x0]
+    // AARCH64: stp q0, q1, [x8]
+    // AARCH64: str q2, [x8, #32]
+    // AARCH64: ret
+    //
+    // ARMV7-LABEL: test_vld3q_f32
+    // ARMV7: vld3.32 {d16, d18, d20}, [r1]!
+    // ARMV7: vld3.32 {d17, d19, d21}, [r1]
+    // ARMV7: vst1.32 {d16, d17}, [r0]!
+    // ARMV7: vst1.32 {d18, d19}, [r0]!
+    // ARMV7: vst1.64 {d20, d21}, [r0]
+    // ARMV7: bx lr
+    unsafe { vld3q_f32(ptr) }
+}
+
+#[unsafe(no_mangle)]
+#[cfg_attr(target_arch = "arm", target_feature(enable = "neon,v7"))]
+fn test_vld3q_s32(ptr: *const i32) -> int32x4x3_t {
+    // AARCH64-LABEL: test_vld3q_s32
+    // AARCH64: ld3 { v0.4s, v1.4s, v2.4s }, [x0]
+    // AARCH64: stp q0, q1, [x8]
+    // AARCH64: str q2, [x8, #32]
+    // AARCH64: ret
+    //
+    // ARMV7-LABEL: test_vld3q_s32
+    // ARMV7: vld3.32 {d16, d18, d20}, [r1]!
+    // ARMV7: vld3.32 {d17, d19, d21}, [r1]
+    // ARMV7: vst1.32 {d16, d17}, [r0]!
+    // ARMV7: vst1.32 {d18, d19}, [r0]!
+    // ARMV7: vst1.64 {d20, d21}, [r0]
+    // ARMV7: bx lr
+    unsafe { vld3q_s32(ptr) }
+}
+
+#[unsafe(no_mangle)]
+#[cfg_attr(target_arch = "arm", target_feature(enable = "neon,v7"))]
+fn test_vld3q_u32(ptr: *const u32) -> uint32x4x3_t {
+    // AARCH64-LABEL: test_vld3q_u32
+    // AARCH64: ld3 { v0.4s, v1.4s, v2.4s }, [x0]
+    // AARCH64: stp q0, q1, [x8]
+    // AARCH64: str q2, [x8, #32]
+    // AARCH64: ret
+    //
+    // ARMV7-LABEL: test_vld3q_u32
+    // ARMV7: vld3.32 {d16, d18, d20}, [r1]!
+    // ARMV7: vld3.32 {d17, d19, d21}, [r1]
+    // ARMV7: vst1.32 {d16, d17}, [r0]!
+    // ARMV7: vst1.32 {d18, d19}, [r0]!
+    // ARMV7: vst1.64 {d20, d21}, [r0]
+    // ARMV7: bx lr
+    unsafe { vld3q_u32(ptr) }
+}