[PATCH] D158613: [AArch64] Mark known zero for high 16-bits of uaddlv intrinsic output with v8i8

[JinGu Kang via Phabricator via llvm-commits]

jaykang10 updated this revision to Diff 554251.
jaykang10 added a comment.

Fixed wrong bound for v16i8.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D158613/new/

https://reviews.llvm.org/D158613

Files:
  llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
  llvm/test/CodeGen/AArch64/aarch64-neon-vector-insert-uaddlv.ll
  llvm/test/CodeGen/AArch64/neon-addlv.ll


Index: llvm/test/CodeGen/AArch64/neon-addlv.ll
===================================================================
--- llvm/test/CodeGen/AArch64/neon-addlv.ll
+++ llvm/test/CodeGen/AArch64/neon-addlv.ll
@@ -153,8 +153,8 @@
 
 declare i32 @llvm.aarch64.neon.uaddlv.i32.v8i8(<8 x i8>) nounwind readnone
 
-define i32 @uaddlv_known_bits(<8 x i8> %a) {
-; CHECK-LABEL: uaddlv_known_bits:
+define i32 @uaddlv_known_bits_v8i8(<8 x i8> %a) {
+; CHECK-LABEL: uaddlv_known_bits_v8i8:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    uaddlv h0, v0.8b
 ; CHECK-NEXT:    fmov w0, s0
@@ -163,3 +163,17 @@
   %tmp2 = and i32 %tmp1, 65535
   ret i32 %tmp2
 }
+
+declare i32 @llvm.aarch64.neon.uaddlv.i32.v16i8(<16 x i8>) nounwind readnone
+
+define i32 @uaddlv_known_bits_v16i8(<16 x i8> %a) {
+; CHECK-LABEL: uaddlv_known_bits_v16i8:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    uaddlv h0, v0.16b
+; CHECK-NEXT:    fmov w0, s0
+; CHECK-NEXT:    ret
+entry:
+  %vaddlv.i = tail call i32 @llvm.aarch64.neon.uaddlv.i32.v16i8(<16 x i8> %a)
+  %0 = and i32 %vaddlv.i, 65535
+  ret i32 %0
+}
Index: llvm/test/CodeGen/AArch64/aarch64-neon-vector-insert-uaddlv.ll
===================================================================
--- llvm/test/CodeGen/AArch64/aarch64-neon-vector-insert-uaddlv.ll
+++ llvm/test/CodeGen/AArch64/aarch64-neon-vector-insert-uaddlv.ll
@@ -300,7 +300,7 @@
 ; CHECK-NEXT:    stp xzr, xzr, [x0, #16]
 ; CHECK-NEXT:    uaddlv.8b h1, v0
 ; CHECK-NEXT:    mov.h v0[0], v1[0]
-; CHECK-NEXT:    bic.4h v0, #255, lsl #8
+; CHECK-NEXT:    bic.4h v0, #7, lsl #8
 ; CHECK-NEXT:    ushll.4s v0, v0, #0
 ; CHECK-NEXT:    ucvtf.4s v0, v0
 ; CHECK-NEXT:    str q0, [x0]
Index: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
===================================================================
--- llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -2165,9 +2165,10 @@
     case Intrinsic::aarch64_neon_uaddlv: {
       MVT VT = Op.getOperand(1).getValueType().getSimpleVT();
       unsigned BitWidth = Known.getBitWidth();
-      if (VT == MVT::v8i8) {
-        assert(BitWidth >= 16 && "Unexpected width!");
-        APInt Mask = APInt::getHighBitsSet(BitWidth, BitWidth - 16);
+      if (VT == MVT::v8i8 || VT == MVT::v16i8) {
+        unsigned Bound = (VT == MVT::v8i8) ?  11 : 12;
+        assert(BitWidth >= Bound && "Unexpected width!");
+        APInt Mask = APInt::getHighBitsSet(BitWidth, BitWidth - Bound);
         Known.Zero |= Mask;
       }
       break;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D158613.554251.patch
Type: text/x-patch
Size: 2518 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20230829/8e9c948a/attachment.bin>