[llvm] b10721e - [AArch64] A few extra rshrn intrinsic tests. NFC

[David Green via llvm-commits]

Author: David Green
Date: 2023-09-26T17:13:27+01:00
New Revision: b10721e94127bb62acdd8be35fab6249da26d975

URL: https://github.com/llvm/llvm-project/commit/b10721e94127bb62acdd8be35fab6249da26d975
DIFF: https://github.com/llvm/llvm-project/commit/b10721e94127bb62acdd8be35fab6249da26d975.diff

LOG: [AArch64] A few extra rshrn intrinsic tests. NFC

Added: 
    

Modified: 
    llvm/test/CodeGen/AArch64/arm64-vshift.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/AArch64/arm64-vshift.ll b/llvm/test/CodeGen/AArch64/arm64-vshift.ll
index 4fe1fe8a0ebc98f..ef54f6d2bb1828f 100644
--- a/llvm/test/CodeGen/AArch64/arm64-vshift.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-vshift.ll
@@ -3528,4 +3528,47 @@ entry:
   ret void
 }
 
+define <4 x i32> @sext_rshrn(<4 x i32> noundef %a) {
+; CHECK-LABEL: sext_rshrn:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    movi.4s v1, #16, lsl #8
+; CHECK-NEXT:    add.4s v0, v0, v1
+; CHECK-NEXT:    ushr.4s v0, v0, #13
+; CHECK-NEXT:    shl.4s v0, v0, #16
+; CHECK-NEXT:    sshr.4s v0, v0, #16
+; CHECK-NEXT:    ret
+entry:
+  %vrshrn_n1 = tail call <4 x i16> @llvm.aarch64.neon.rshrn.v4i16(<4 x i32> %a, i32 13)
+  %vmovl.i = sext <4 x i16> %vrshrn_n1 to <4 x i32>
+  ret <4 x i32> %vmovl.i
+}
+
+define <4 x i32> @zext_rshrn(<4 x i32> noundef %a) {
+; CHECK-LABEL: zext_rshrn:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    movi.4s v1, #16, lsl #8
+; CHECK-NEXT:    add.4s v0, v0, v1
+; CHECK-NEXT:    ushr.4s v0, v0, #13
+; CHECK-NEXT:    bic.4s v0, #7, lsl #16
+; CHECK-NEXT:    ret
+entry:
+  %vrshrn_n1 = tail call <4 x i16> @llvm.aarch64.neon.rshrn.v4i16(<4 x i32> %a, i32 13)
+  %vmovl.i = zext <4 x i16> %vrshrn_n1 to <4 x i32>
+  ret <4 x i32> %vmovl.i
+}
+
+define <4 x i16> @mul_rshrn(<4 x i32> noundef %a) {
+; CHECK-LABEL: mul_rshrn:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    mov w8, #4099 // =0x1003
+; CHECK-NEXT:    dup.4s v1, w8
+; CHECK-NEXT:    add.4s v0, v0, v1
+; CHECK-NEXT:    shrn.4h v0, v0, #13
+; CHECK-NEXT:    ret
+entry:
+  %b = add <4 x i32> %a, <i32 3, i32 3, i32 3, i32 3>
+  %vrshrn_n1 = tail call <4 x i16> @llvm.aarch64.neon.rshrn.v4i16(<4 x i32> %b, i32 13)
+  ret <4 x i16> %vrshrn_n1
+}
+
 declare <2 x i64> @llvm.aarch64.neon.addp.v2i64(<2 x i64>, <2 x i64>)