[llvm] [X86] known-pow2.ll - adjust pow2_trunc_vec to show failure to pass through demanded elts mask to determine pow2orzero mask (PR #190170)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Thu Apr 2 06:17:37 PDT 2026
https://github.com/RKSimon created https://github.com/llvm/llvm-project/pull/190170
None
>From 6c79c1f4ff0d68de1d31457e2736d959de98a253 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Thu, 2 Apr 2026 14:16:35 +0100
Subject: [PATCH] [X86] known-pow2.ll - adjust pow2_trunc_vec to show failure
to pass through demanded elts mask to determine pow2orzero mask
---
llvm/test/CodeGen/X86/known-pow2.ll | 31 ++++++++++++++++-------------
1 file changed, 17 insertions(+), 14 deletions(-)
diff --git a/llvm/test/CodeGen/X86/known-pow2.ll b/llvm/test/CodeGen/X86/known-pow2.ll
index 629d7d9e4ae65..9e71dd890f81f 100644
--- a/llvm/test/CodeGen/X86/known-pow2.ll
+++ b/llvm/test/CodeGen/X86/known-pow2.ll
@@ -1367,26 +1367,29 @@ define i8 @pow2_trunc_fail(i32 %x, i32 %a){
ret i8 %r
}
-define i8 @pow2_trunc_vec(<4 x i32> %x, <4 x i32> %a) {
+define i8 @pow2_trunc_vec(i8 %x8, <4 x i32> %a, ptr %p) {
; CHECK-LABEL: pow2_trunc_vec:
; CHECK: # %bb.0:
-; CHECK-NEXT: pxor %xmm2, %xmm2
-; CHECK-NEXT: psubd %xmm1, %xmm2
-; CHECK-NEXT: pand %xmm1, %xmm2
-; CHECK-NEXT: pcmpeqd %xmm1, %xmm1
-; CHECK-NEXT: paddb %xmm2, %xmm1
+; CHECK-NEXT: movdqa {{.*#+}} xmm1 = [0,4294967295,4294967295,4294967295]
+; CHECK-NEXT: psubd %xmm0, %xmm1
; CHECK-NEXT: pand %xmm0, %xmm1
-; CHECK-NEXT: movd %xmm1, %eax
+; CHECK-NEXT: movd %xmm1, %ecx
+; CHECK-NEXT: pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
+; CHECK-NEXT: packuswb %xmm1, %xmm1
+; CHECK-NEXT: packuswb %xmm1, %xmm1
+; CHECK-NEXT: movd %xmm1, (%rsi)
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: divb %cl
+; CHECK-NEXT: movzbl %ah, %eax
; CHECK-NEXT: # kill: def $al killed $al killed $eax
; CHECK-NEXT: retq
- %a.splat = shufflevector <4 x i32> %a, <4 x i32> poison, <4 x i32> zeroinitializer
- %a.splat.neg = sub <4 x i32> zeroinitializer, %a.splat
- %y = and <4 x i32> %a.splat, %a.splat.neg
- %x8 = trunc <4 x i32> %x to <4 x i8>
+ %a.neg = sub <4 x i32> <i32 0, i32 -1, i32 -1, i32 -1>, %a
+ %y = and <4 x i32> %a, %a.neg
%y8 = trunc <4 x i32> %y to <4 x i8>
- %r = urem <4 x i8> %x8, %y8
- %ext = extractelement <4 x i8> %r, i8 0
- ret i8 %ext
+ store <4 x i8> %y8, ptr %p
+ %y8.elt = extractelement <4 x i8> %y8, i8 0
+ %r = urem i8 %x8, %y8.elt
+ ret i8 %r
}
define i8 @pow2_truncc_vec_fail(<4 x i32> %x, <4 x i32> %a) {
More information about the llvm-commits
mailing list