[llvm] [X86] known-never-zero.ll - add sdiv/udiv vector test coverage for #183047 (PR #184350)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Tue Mar 3 06:38:57 PST 2026
https://github.com/RKSimon created https://github.com/llvm/llvm-project/pull/184350
None
>From c5ab3fb7151cbf1c858d43534124d25eb8c83d87 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Tue, 3 Mar 2026 14:38:06 +0000
Subject: [PATCH] [X86] known-never-zero.ll - add sdiv/udiv vector coverage for
#183047
---
llvm/test/CodeGen/X86/known-never-zero.ll | 146 ++++++++++++++++++++++
1 file changed, 146 insertions(+)
diff --git a/llvm/test/CodeGen/X86/known-never-zero.ll b/llvm/test/CodeGen/X86/known-never-zero.ll
index 5cebf17a43fc2..25c0d87110aba 100644
--- a/llvm/test/CodeGen/X86/known-never-zero.ll
+++ b/llvm/test/CodeGen/X86/known-never-zero.ll
@@ -1478,6 +1478,79 @@ define i32 @udiv_known_nonzero(i32 %xx, i32 %y) {
ret i32 %r
}
+define i32 @udiv_known_nonzero_vec(<4 x i32> %xx, <4 x i32> %y, ptr %p) nounwind {
+; X86-LABEL: udiv_known_nonzero_vec:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: por {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-NEXT: pshufd {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; X86-NEXT: movd %xmm2, %esi
+; X86-NEXT: movl $-1, %eax
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: divl %esi
+; X86-NEXT: movd %eax, %xmm3
+; X86-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; X86-NEXT: movd %xmm2, %esi
+; X86-NEXT: movl $-1, %eax
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: divl %esi
+; X86-NEXT: movd %eax, %xmm2
+; X86-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; X86-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
+; X86-NEXT: movd %xmm3, %esi
+; X86-NEXT: movl $-1, %eax
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: divl %esi
+; X86-NEXT: movd %eax, %xmm3
+; X86-NEXT: movd %xmm1, %esi
+; X86-NEXT: movd %xmm0, %eax
+; X86-NEXT: xorl %edx, %edx
+; X86-NEXT: divl %esi
+; X86-NEXT: movd %eax, %xmm0
+; X86-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; X86-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; X86-NEXT: movdqa %xmm0, (%ecx)
+; X86-NEXT: rep bsfl %eax, %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+;
+; X64-LABEL: udiv_known_nonzero_vec:
+; X64: # %bb.0:
+; X64-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-NEXT: vpextrd $1, %xmm1, %ecx
+; X64-NEXT: vpextrd $1, %xmm0, %eax
+; X64-NEXT: xorl %edx, %edx
+; X64-NEXT: divl %ecx
+; X64-NEXT: movl %eax, %ecx
+; X64-NEXT: vmovd %xmm1, %esi
+; X64-NEXT: vmovd %xmm0, %eax
+; X64-NEXT: xorl %edx, %edx
+; X64-NEXT: divl %esi
+; X64-NEXT: vmovd %eax, %xmm2
+; X64-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; X64-NEXT: vpextrd $2, %xmm1, %ecx
+; X64-NEXT: vpextrd $2, %xmm0, %eax
+; X64-NEXT: xorl %edx, %edx
+; X64-NEXT: divl %ecx
+; X64-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; X64-NEXT: vpextrd $3, %xmm1, %ecx
+; X64-NEXT: vpextrd $3, %xmm0, %eax
+; X64-NEXT: xorl %edx, %edx
+; X64-NEXT: divl %ecx
+; X64-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
+; X64-NEXT: vmovdqa %xmm0, (%rdi)
+; X64-NEXT: vmovd %xmm0, %eax
+; X64-NEXT: rep bsfl %eax, %eax
+; X64-NEXT: retq
+ %x = or <4 x i32> %xx, <i32 64, i32 -1, i32 -1, i32 -1>
+ %z = udiv exact <4 x i32> %x, %y
+ store <4 x i32> %z, ptr %p
+ %e = extractelement <4 x i32> %z, i32 0
+ %r = call i32 @llvm.cttz.i32(i32 %e, i1 false)
+ ret i32 %r
+}
+
define i32 @udiv_maybe_zero(i32 %x, i32 %y) {
; X86-LABEL: udiv_maybe_zero:
; X86: # %bb.0:
@@ -1527,6 +1600,79 @@ define i32 @sdiv_known_nonzero(i32 %xx, i32 %y) {
ret i32 %r
}
+define i32 @sdiv_known_nonzero_vec(<4 x i32> %xx, <4 x i32> %y, ptr %p) nounwind {
+; X86-LABEL: sdiv_known_nonzero_vec:
+; X86: # %bb.0:
+; X86-NEXT: pushl %esi
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: por {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-NEXT: pshufd {{.*#+}} xmm2 = xmm1[3,3,3,3]
+; X86-NEXT: movd %xmm2, %esi
+; X86-NEXT: movl $-1, %eax
+; X86-NEXT: cltd
+; X86-NEXT: idivl %esi
+; X86-NEXT: movd %eax, %xmm3
+; X86-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; X86-NEXT: movd %xmm2, %esi
+; X86-NEXT: movl $-1, %eax
+; X86-NEXT: cltd
+; X86-NEXT: idivl %esi
+; X86-NEXT: movd %eax, %xmm2
+; X86-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1]
+; X86-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,1,1]
+; X86-NEXT: movd %xmm3, %esi
+; X86-NEXT: movl $-1, %eax
+; X86-NEXT: cltd
+; X86-NEXT: idivl %esi
+; X86-NEXT: movd %eax, %xmm3
+; X86-NEXT: movd %xmm1, %esi
+; X86-NEXT: movd %xmm0, %eax
+; X86-NEXT: cltd
+; X86-NEXT: idivl %esi
+; X86-NEXT: movd %eax, %xmm0
+; X86-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1]
+; X86-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; X86-NEXT: movdqa %xmm0, (%ecx)
+; X86-NEXT: rep bsfl %eax, %eax
+; X86-NEXT: popl %esi
+; X86-NEXT: retl
+;
+; X64-LABEL: sdiv_known_nonzero_vec:
+; X64: # %bb.0:
+; X64-NEXT: vpor {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-NEXT: vpextrd $1, %xmm1, %ecx
+; X64-NEXT: vpextrd $1, %xmm0, %eax
+; X64-NEXT: cltd
+; X64-NEXT: idivl %ecx
+; X64-NEXT: movl %eax, %ecx
+; X64-NEXT: vmovd %xmm1, %esi
+; X64-NEXT: vmovd %xmm0, %eax
+; X64-NEXT: cltd
+; X64-NEXT: idivl %esi
+; X64-NEXT: vmovd %eax, %xmm2
+; X64-NEXT: vpinsrd $1, %ecx, %xmm2, %xmm2
+; X64-NEXT: vpextrd $2, %xmm1, %ecx
+; X64-NEXT: vpextrd $2, %xmm0, %eax
+; X64-NEXT: cltd
+; X64-NEXT: idivl %ecx
+; X64-NEXT: vpinsrd $2, %eax, %xmm2, %xmm2
+; X64-NEXT: vpextrd $3, %xmm1, %ecx
+; X64-NEXT: vpextrd $3, %xmm0, %eax
+; X64-NEXT: cltd
+; X64-NEXT: idivl %ecx
+; X64-NEXT: vpinsrd $3, %eax, %xmm2, %xmm0
+; X64-NEXT: vmovdqa %xmm0, (%rdi)
+; X64-NEXT: vmovd %xmm0, %eax
+; X64-NEXT: rep bsfl %eax, %eax
+; X64-NEXT: retq
+ %x = or <4 x i32> %xx, <i32 64, i32 -1, i32 -1, i32 -1>
+ %z = sdiv exact <4 x i32> %x, %y
+ store <4 x i32> %z, ptr %p
+ %e = extractelement <4 x i32> %z, i32 0
+ %r = call i32 @llvm.cttz.i32(i32 %e, i1 false)
+ ret i32 %r
+}
+
define i32 @sdiv_maybe_zero(i32 %x, i32 %y) {
; X86-LABEL: sdiv_maybe_zero:
; X86: # %bb.0:
More information about the llvm-commits
mailing list