[X86] Add test showing failure to remove freeze from all_of pattern

llvm · Sep 3, 2024 · 6c59dfb · 6c59dfb
1 parent a70d999
commit 6c59dfb
Showing 1 changed file with 67 additions and 12 deletions.
diff --git a/llvm/test/CodeGen/X86/movmsk-cmp.ll b/llvm/test/CodeGen/X86/movmsk-cmp.ll
@@ -4425,6 +4425,61 @@ define i32 @PR39665_c_ray_opt(<2 x double> %x, <2 x double> %y) {
   ret i32 %r
 }
 
+define i32 @PR39665_c_ray_select(<2 x double> %x, <2 x double> %y) {
+; SSE-LABEL: PR39665_c_ray_select:
+; SSE:       # %bb.0:
+; SSE-NEXT:    cmpltpd %xmm0, %xmm1
+; SSE-NEXT:    psllq $63, %xmm1
+; SSE-NEXT:    movmskpd %xmm1, %eax
+; SSE-NEXT:    cmpl $3, %eax
+; SSE-NEXT:    movl $42, %ecx
+; SSE-NEXT:    movl $99, %eax
+; SSE-NEXT:    cmovel %ecx, %eax
+; SSE-NEXT:    retq
+;
+; AVX1OR2-LABEL: PR39665_c_ray_select:
+; AVX1OR2:       # %bb.0:
+; AVX1OR2-NEXT:    vcmpltpd %xmm0, %xmm1, %xmm0
+; AVX1OR2-NEXT:    vpsllq $63, %xmm0, %xmm0
+; AVX1OR2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1OR2-NEXT:    vtestpd %xmm1, %xmm0
+; AVX1OR2-NEXT:    movl $42, %ecx
+; AVX1OR2-NEXT:    movl $99, %eax
+; AVX1OR2-NEXT:    cmovbl %ecx, %eax
+; AVX1OR2-NEXT:    retq
+;
+; KNL-LABEL: PR39665_c_ray_select:
+; 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:    kmovw %k0, %eax
+; KNL-NEXT:    testb $3, %al
+; KNL-NEXT:    movl $42, %ecx
+; KNL-NEXT:    movl $99, %eax
+; KNL-NEXT:    cmovel %ecx, %eax
+; KNL-NEXT:    vzeroupper
+; KNL-NEXT:    retq
+;
+; SKX-LABEL: PR39665_c_ray_select:
+; SKX:       # %bb.0:
+; SKX-NEXT:    vcmpltpd %xmm0, %xmm1, %k0
+; SKX-NEXT:    knotw %k0, %k0
+; SKX-NEXT:    kmovd %k0, %eax
+; SKX-NEXT:    testb $3, %al
+; SKX-NEXT:    movl $42, %ecx
+; SKX-NEXT:    movl $99, %eax
+; SKX-NEXT:    cmovel %ecx, %eax
+; SKX-NEXT:    retq
+  %cmp = fcmp fast ogt <2 x double> %x, %y
+  %b0 = extractelement <2 x i1> %cmp, i64 0
+  %b1 = extractelement <2 x i1> %cmp, i64 1
+  %b = select i1 %b0, i1 %b1, i1 false
+  %r = select i1 %b, i32 42, i32 99
+  ret i32 %r
+}
+
 define i32 @pr67287(<2 x i64> %broadcast.splatinsert25) {
 ; SSE2-LABEL: pr67287:
 ; SSE2:       # %bb.0: # %entry
@@ -4438,10 +4493,10 @@ 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 .LBB97_2
+; SSE2-NEXT:    je .LBB98_2
 ; SSE2-NEXT:  # %bb.1:
 ; SSE2-NEXT:    movw $0, 0
-; SSE2-NEXT:  .LBB97_2: # %middle.block
+; SSE2-NEXT:  .LBB98_2: # %middle.block
 ; SSE2-NEXT:    xorl %eax, %eax
 ; SSE2-NEXT:    retq
 ;
@@ -4456,10 +4511,10 @@ 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 .LBB97_2
+; SSE41-NEXT:    je .LBB98_2
 ; SSE41-NEXT:  # %bb.1:
 ; SSE41-NEXT:    movw $0, 0
-; SSE41-NEXT:  .LBB97_2: # %middle.block
+; SSE41-NEXT:  .LBB98_2: # %middle.block
 ; SSE41-NEXT:    xorl %eax, %eax
 ; SSE41-NEXT:    retq
 ;
@@ -4473,10 +4528,10 @@ 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 .LBB97_2
+; AVX1-NEXT:    je .LBB98_2
 ; AVX1-NEXT:  # %bb.1:
 ; AVX1-NEXT:    movw $0, 0
-; AVX1-NEXT:  .LBB97_2: # %middle.block
+; AVX1-NEXT:  .LBB98_2: # %middle.block
 ; AVX1-NEXT:    xorl %eax, %eax
 ; AVX1-NEXT:    retq
 ;
@@ -4490,10 +4545,10 @@ 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 .LBB97_2
+; AVX2-NEXT:    je .LBB98_2
 ; AVX2-NEXT:  # %bb.1:
 ; AVX2-NEXT:    movw $0, 0
-; AVX2-NEXT:  .LBB97_2: # %middle.block
+; AVX2-NEXT:  .LBB98_2: # %middle.block
 ; AVX2-NEXT:    xorl %eax, %eax
 ; AVX2-NEXT:    retq
 ;
@@ -4508,10 +4563,10 @@ 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 .LBB97_2
+; KNL-NEXT:    je .LBB98_2
 ; KNL-NEXT:  # %bb.1:
 ; KNL-NEXT:    movw $0, 0
-; KNL-NEXT:  .LBB97_2: # %middle.block
+; KNL-NEXT:  .LBB98_2: # %middle.block
 ; KNL-NEXT:    xorl %eax, %eax
 ; KNL-NEXT:    vzeroupper
 ; KNL-NEXT:    retq
@@ -4526,10 +4581,10 @@ 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 .LBB97_2
+; SKX-NEXT:    je .LBB98_2
 ; SKX-NEXT:  # %bb.1:
 ; SKX-NEXT:    movw $0, 0
-; SKX-NEXT:  .LBB97_2: # %middle.block
+; SKX-NEXT:  .LBB98_2: # %middle.block
 ; SKX-NEXT:    xorl %eax, %eax
 ; SKX-NEXT:    retq
 entry: