[llvm] 23f0aa2 - [X86] movmsk-cmp.ll - update PR39665_c_ray test to match middle-end output (#213659)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Aug 3 06:26:18 PDT 2026
Author: Simon Pilgrim
Date: 2026-08-03T13:26:13Z
New Revision: 23f0aa2f2ea35a8e9b106f97d561482cbf6f17b5
URL: https://github.com/llvm/llvm-project/commit/23f0aa2f2ea35a8e9b106f97d561482cbf6f17b5
DIFF: https://github.com/llvm/llvm-project/commit/23f0aa2f2ea35a8e9b106f97d561482cbf6f17b5.diff
LOG: [X86] movmsk-cmp.ll - update PR39665_c_ray test to match middle-end output (#213659)
These now lower to vXi1 reduction (as bitcast) patterns
The PR39665_c_ray_opt test folds to the same IR, so I've merged the
tests
Added:
Modified:
llvm/test/CodeGen/X86/movmsk-cmp.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/X86/movmsk-cmp.ll b/llvm/test/CodeGen/X86/movmsk-cmp.ll
index 7f50cac5e4290..7669aa093458b 100644
--- a/llvm/test/CodeGen/X86/movmsk-cmp.ll
+++ b/llvm/test/CodeGen/X86/movmsk-cmp.ll
@@ -4370,94 +4370,31 @@ define i1 @movmsk_v2f64_var(<2 x double> %x, <2 x double> %y, i32 %z) {
ret i1 %val
}
-; TODO: We expect similar result as for PR39665_c_ray_opt,
-; but this is not the case in practice.
define i32 @PR39665_c_ray(<2 x double> %x, <2 x double> %y) {
; SSE-LABEL: PR39665_c_ray:
; SSE: # %bb.0:
-; SSE-NEXT: cmpltpd %xmm0, %xmm1
-; SSE-NEXT: movmskpd %xmm1, %ecx
-; SSE-NEXT: testb $2, %cl
-; SSE-NEXT: movl $42, %eax
-; SSE-NEXT: movl $99, %edx
-; SSE-NEXT: cmovel %edx, %eax
-; SSE-NEXT: testb $1, %cl
-; SSE-NEXT: cmovel %edx, %eax
-; SSE-NEXT: retq
-;
-; AVX1OR2-LABEL: PR39665_c_ray:
-; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX1OR2-NEXT: vmovmskpd %xmm0, %ecx
-; AVX1OR2-NEXT: testb $2, %cl
-; AVX1OR2-NEXT: movl $42, %eax
-; AVX1OR2-NEXT: movl $99, %edx
-; AVX1OR2-NEXT: cmovel %edx, %eax
-; AVX1OR2-NEXT: testb $1, %cl
-; AVX1OR2-NEXT: cmovel %edx, %eax
-; AVX1OR2-NEXT: retq
-;
-; KNL-LABEL: PR39665_c_ray:
-; KNL: # %bb.0:
-; KNL-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
-; KNL-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
-; KNL-NEXT: vcmpltpd %zmm0, %zmm1, %k0
-; KNL-NEXT: kmovw %k0, %ecx
-; KNL-NEXT: testb $2, %cl
-; KNL-NEXT: movl $42, %eax
-; KNL-NEXT: movl $99, %edx
-; KNL-NEXT: cmovel %edx, %eax
-; KNL-NEXT: testb $1, %cl
-; KNL-NEXT: cmovel %edx, %eax
-; KNL-NEXT: vzeroupper
-; KNL-NEXT: retq
-;
-; SKX-LABEL: PR39665_c_ray:
-; SKX: # %bb.0:
-; SKX-NEXT: vcmpltpd %xmm0, %xmm1, %k0
-; SKX-NEXT: kmovd %k0, %ecx
-; SKX-NEXT: testb $2, %cl
-; SKX-NEXT: movl $42, %eax
-; SKX-NEXT: movl $99, %edx
-; SKX-NEXT: cmovel %edx, %eax
-; SKX-NEXT: testb $1, %cl
-; SKX-NEXT: cmovel %edx, %eax
-; SKX-NEXT: retq
- %cmp = fcmp ogt <2 x double> %x, %y
- %e1 = extractelement <2 x i1> %cmp, i32 0
- %e2 = extractelement <2 x i1> %cmp, i32 1
- %u = and i1 %e1, %e2
- %r = select i1 %u, i32 42, i32 99
- ret i32 %r
-}
-
-define i32 @PR39665_c_ray_opt(<2 x double> %x, <2 x double> %y) {
-; SSE-LABEL: PR39665_c_ray_opt:
-; SSE: # %bb.0:
-; SSE-NEXT: cmpltpd %xmm0, %xmm1
+; SSE-NEXT: cmpnltpd %xmm0, %xmm1
; SSE-NEXT: movmskpd %xmm1, %eax
-; SSE-NEXT: cmpl $3, %eax
+; SSE-NEXT: testl %eax, %eax
; SSE-NEXT: movl $42, %ecx
; SSE-NEXT: movl $99, %eax
; SSE-NEXT: cmovel %ecx, %eax
; SSE-NEXT: retq
;
-; AVX1OR2-LABEL: PR39665_c_ray_opt:
+; AVX1OR2-LABEL: PR39665_c_ray:
; AVX1OR2: # %bb.0:
-; AVX1OR2-NEXT: vcmpltpd %xmm0, %xmm1, %xmm0
-; AVX1OR2-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX1OR2-NEXT: vtestpd %xmm1, %xmm0
+; AVX1OR2-NEXT: vcmpnltpd %xmm0, %xmm1, %xmm0
+; AVX1OR2-NEXT: vtestpd %xmm0, %xmm0
; AVX1OR2-NEXT: movl $42, %ecx
; AVX1OR2-NEXT: movl $99, %eax
-; AVX1OR2-NEXT: cmovbl %ecx, %eax
+; AVX1OR2-NEXT: cmovel %ecx, %eax
; AVX1OR2-NEXT: retq
;
-; KNL-LABEL: PR39665_c_ray_opt:
+; KNL-LABEL: PR39665_c_ray:
; KNL: # %bb.0:
; KNL-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; KNL-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
-; KNL-NEXT: vcmpltpd %zmm0, %zmm1, %k0
-; KNL-NEXT: knotw %k0, %k0
+; KNL-NEXT: vcmpnltpd %zmm0, %zmm1, %k0
; KNL-NEXT: kmovw %k0, %eax
; KNL-NEXT: testb $3, %al
; KNL-NEXT: movl $42, %ecx
@@ -4466,19 +4403,17 @@ define i32 @PR39665_c_ray_opt(<2 x double> %x, <2 x double> %y) {
; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
;
-; SKX-LABEL: PR39665_c_ray_opt:
+; SKX-LABEL: PR39665_c_ray:
; SKX: # %bb.0:
-; SKX-NEXT: vcmpltpd %xmm0, %xmm1, %k0
-; SKX-NEXT: kmovd %k0, %eax
-; SKX-NEXT: cmpb $3, %al
+; SKX-NEXT: vcmpnltpd %xmm0, %xmm1, %k0
+; SKX-NEXT: kortestb %k0, %k0
; SKX-NEXT: movl $42, %ecx
; SKX-NEXT: movl $99, %eax
; SKX-NEXT: cmovel %ecx, %eax
; SKX-NEXT: retq
- %cmp = fcmp ogt <2 x double> %x, %y
- %shift = shufflevector <2 x i1> %cmp, <2 x i1> poison, <2 x i32> <i32 1, i32 undef>
- %1 = and <2 x i1> %cmp, %shift
- %u = extractelement <2 x i1> %1, i64 0
+ %cmp = fcmp ule <2 x double> %x, %y
+ %bc = bitcast <2 x i1> %cmp to i2
+ %u = icmp eq i2 %bc, 0
%r = select i1 %u, i32 42, i32 99
ret i32 %r
}
@@ -4550,10 +4485,10 @@ define i32 @pr67287(<2 x i64> %broadcast.splatinsert25) {
; SSE2-NEXT: movd %xmm1, %ecx
; SSE2-NEXT: orb %al, %cl
; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB98_2
+; SSE2-NEXT: je .LBB97_2
; SSE2-NEXT: # %bb.1:
; SSE2-NEXT: movw $0, 0
-; SSE2-NEXT: .LBB98_2: # %middle.block
+; SSE2-NEXT: .LBB97_2: # %middle.block
; SSE2-NEXT: xorl %eax, %eax
; SSE2-NEXT: retq
;
@@ -4568,10 +4503,10 @@ define i32 @pr67287(<2 x i64> %broadcast.splatinsert25) {
; SSE41-NEXT: movd %xmm0, %ecx
; SSE41-NEXT: orb %al, %cl
; SSE41-NEXT: testb $1, %cl
-; SSE41-NEXT: je .LBB98_2
+; SSE41-NEXT: je .LBB97_2
; SSE41-NEXT: # %bb.1:
; SSE41-NEXT: movw $0, 0
-; SSE41-NEXT: .LBB98_2: # %middle.block
+; SSE41-NEXT: .LBB97_2: # %middle.block
; SSE41-NEXT: xorl %eax, %eax
; SSE41-NEXT: retq
;
@@ -4585,10 +4520,10 @@ define i32 @pr67287(<2 x i64> %broadcast.splatinsert25) {
; AVX1-NEXT: vmovd %xmm0, %ecx
; AVX1-NEXT: orb %al, %cl
; AVX1-NEXT: testb $1, %cl
-; AVX1-NEXT: je .LBB98_2
+; AVX1-NEXT: je .LBB97_2
; AVX1-NEXT: # %bb.1:
; AVX1-NEXT: movw $0, 0
-; AVX1-NEXT: .LBB98_2: # %middle.block
+; AVX1-NEXT: .LBB97_2: # %middle.block
; AVX1-NEXT: xorl %eax, %eax
; AVX1-NEXT: retq
;
@@ -4602,10 +4537,10 @@ define i32 @pr67287(<2 x i64> %broadcast.splatinsert25) {
; AVX2-NEXT: vmovd %xmm0, %ecx
; AVX2-NEXT: orb %al, %cl
; AVX2-NEXT: testb $1, %cl
-; AVX2-NEXT: je .LBB98_2
+; AVX2-NEXT: je .LBB97_2
; AVX2-NEXT: # %bb.1:
; AVX2-NEXT: movw $0, 0
-; AVX2-NEXT: .LBB98_2: # %middle.block
+; AVX2-NEXT: .LBB97_2: # %middle.block
; AVX2-NEXT: xorl %eax, %eax
; AVX2-NEXT: retq
;
@@ -4620,10 +4555,10 @@ define i32 @pr67287(<2 x i64> %broadcast.splatinsert25) {
; KNL-NEXT: setne %cl
; KNL-NEXT: orb %cl, %al
; KNL-NEXT: testb $1, %al
-; KNL-NEXT: je .LBB98_2
+; KNL-NEXT: je .LBB97_2
; KNL-NEXT: # %bb.1:
; KNL-NEXT: movw $0, 0
-; KNL-NEXT: .LBB98_2: # %middle.block
+; KNL-NEXT: .LBB97_2: # %middle.block
; KNL-NEXT: xorl %eax, %eax
; KNL-NEXT: vzeroupper
; KNL-NEXT: retq
@@ -4638,10 +4573,10 @@ define i32 @pr67287(<2 x i64> %broadcast.splatinsert25) {
; SKX-NEXT: kmovd %k0, %ecx
; SKX-NEXT: orb %al, %cl
; SKX-NEXT: testb $1, %cl
-; SKX-NEXT: je .LBB98_2
+; SKX-NEXT: je .LBB97_2
; SKX-NEXT: # %bb.1:
; SKX-NEXT: movw $0, 0
-; SKX-NEXT: .LBB98_2: # %middle.block
+; SKX-NEXT: .LBB97_2: # %middle.block
; SKX-NEXT: xorl %eax, %eax
; SKX-NEXT: retq
entry:
More information about the llvm-commits
mailing list