From: Craig Topper <craig.topper@intel.com>
Date: Mon, 14 Oct 2019 06:47:56 +0000 (+0000)
Subject: [X86] Teach EmitTest to handle ISD::SSUBO/USUBO in order to use the Z flag from the... 
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=eca2f6ad30e9137009551cbf81bf53c416af2dd7;p=llvm

[X86] Teach EmitTest to handle ISD::SSUBO/USUBO in order to use the Z flag from the subtract directly during isel.

This prevents isel from emitting a TEST instruction that
optimizeCompareInstr will need to remove later.

In some of the modified tests, the SUB gets duplicated due to
the flags being needed in two places and being clobbered in
between. optimizeCompareInstr was able to optimize away the TEST
that was using the result of one of them, but optimizeCompareInstr
doesn't know to turn SUB into CMP after removing the TEST. It
only knows how to turn SUB into CMP if the result was already
dead.

With this change the TEST never exists, so optimizeCompareInstr
doesn't have to remove it. Then it can just turn the SUB into
CMP immediately.

Fixes PR43649.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@374755 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index a340eff9789..c649a90ddff 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -20079,6 +20079,13 @@ static SDValue EmitTest(SDValue Op, unsigned X86CC, const SDLoc &dl,
   case X86ISD::XOR:
   case X86ISD::AND:
     return SDValue(Op.getNode(), 1);
+  case ISD::SSUBO:
+  case ISD::USUBO: {
+    // /USUBO/SSUBO will become a X86ISD::SUB and we can use its Z flag.
+    SDVTList VTs = DAG.getVTList(Op.getValueType(), MVT::i32);
+    return DAG.getNode(X86ISD::SUB, dl, VTs, Op->getOperand(0),
+                       Op->getOperand(1)).getValue(1);
+  }
   default:
   default_case:
     break;
diff --git a/test/CodeGen/X86/known-bits.ll b/test/CodeGen/X86/known-bits.ll
index d9a7921131f..8e2e9c05fdc 100644
--- a/test/CodeGen/X86/known-bits.ll
+++ b/test/CodeGen/X86/known-bits.ll
@@ -190,26 +190,22 @@ define {i32, i1} @knownbits_uaddo_saddo(i64 %a0, i64 %a1) nounwind {
 define {i32, i1} @knownbits_usubo_ssubo(i64 %a0, i64 %a1) nounwind {
 ; X32-LABEL: knownbits_usubo_ssubo:
 ; X32:       # %bb.0:
-; X32-NEXT:    pushl %ebx
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X32-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X32-NEXT:    movl %ecx, %edx
-; X32-NEXT:    subl %eax, %edx
-; X32-NEXT:    setb %bl
-; X32-NEXT:    testl %eax, %eax
-; X32-NEXT:    setns %al
+; X32-NEXT:    cmpl %eax, %ecx
+; X32-NEXT:    setb %dh
+; X32-NEXT:    setns %dl
 ; X32-NEXT:    testl %ecx, %ecx
 ; X32-NEXT:    setns %cl
-; X32-NEXT:    cmpb %al, %cl
-; X32-NEXT:    setne %al
-; X32-NEXT:    testl %edx, %edx
-; X32-NEXT:    setns %dl
 ; X32-NEXT:    cmpb %dl, %cl
+; X32-NEXT:    setne %ch
+; X32-NEXT:    testl %eax, %eax
+; X32-NEXT:    setns %al
+; X32-NEXT:    cmpb %al, %cl
 ; X32-NEXT:    setne %dl
-; X32-NEXT:    andb %al, %dl
-; X32-NEXT:    orb %bl, %dl
+; X32-NEXT:    andb %ch, %dl
+; X32-NEXT:    orb %dh, %dl
 ; X32-NEXT:    xorl %eax, %eax
-; X32-NEXT:    popl %ebx
 ; X32-NEXT:    retl
 ;
 ; X64-LABEL: knownbits_usubo_ssubo:
diff --git a/test/CodeGen/X86/ssub_sat.ll b/test/CodeGen/X86/ssub_sat.ll
index 62724e981f7..51ca6d8a794 100644
--- a/test/CodeGen/X86/ssub_sat.ll
+++ b/test/CodeGen/X86/ssub_sat.ll
@@ -12,24 +12,20 @@ declare <4 x i32> @llvm.ssub.sat.v4i32(<4 x i32>, <4 x i32>)
 define i32 @func(i32 %x, i32 %y) nounwind {
 ; X86-LABEL: func:
 ; X86:       # %bb.0:
-; X86-NEXT:    pushl %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    xorl %ecx, %ecx
-; X86-NEXT:    movl %eax, %esi
-; X86-NEXT:    subl %edx, %esi
+; X86-NEXT:    cmpl %edx, %eax
 ; X86-NEXT:    setns %cl
 ; X86-NEXT:    addl $2147483647, %ecx # imm = 0x7FFFFFFF
 ; X86-NEXT:    subl %edx, %eax
 ; X86-NEXT:    cmovol %ecx, %eax
-; X86-NEXT:    popl %esi
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func:
 ; X64:       # %bb.0:
 ; X64-NEXT:    xorl %eax, %eax
-; X64-NEXT:    movl %edi, %ecx
-; X64-NEXT:    subl %esi, %ecx
+; X64-NEXT:    cmpl %esi, %edi
 ; X64-NEXT:    setns %al
 ; X64-NEXT:    addl $2147483647, %eax # imm = 0x7FFFFFFF
 ; X64-NEXT:    subl %esi, %edi
@@ -79,8 +75,7 @@ define i64 @func2(i64 %x, i64 %y) nounwind {
 ; X64-LABEL: func2:
 ; X64:       # %bb.0:
 ; X64-NEXT:    xorl %ecx, %ecx
-; X64-NEXT:    movq %rdi, %rax
-; X64-NEXT:    subq %rsi, %rax
+; X64-NEXT:    cmpq %rsi, %rdi
 ; X64-NEXT:    setns %cl
 ; X64-NEXT:    movabsq $9223372036854775807, %rax # imm = 0x7FFFFFFFFFFFFFFF
 ; X64-NEXT:    addq %rcx, %rax
@@ -94,25 +89,21 @@ define i64 @func2(i64 %x, i64 %y) nounwind {
 define i16 @func16(i16 %x, i16 %y) nounwind {
 ; X86-LABEL: func16:
 ; X86:       # %bb.0:
-; X86-NEXT:    pushl %esi
 ; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    xorl %ecx, %ecx
-; X86-NEXT:    movl %eax, %esi
-; X86-NEXT:    subw %dx, %si
+; X86-NEXT:    cmpw %dx, %ax
 ; X86-NEXT:    setns %cl
 ; X86-NEXT:    addl $32767, %ecx # imm = 0x7FFF
 ; X86-NEXT:    subw %dx, %ax
 ; X86-NEXT:    cmovol %ecx, %eax
 ; X86-NEXT:    # kill: def $ax killed $ax killed $eax
-; X86-NEXT:    popl %esi
 ; X86-NEXT:    retl
 ;
 ; X64-LABEL: func16:
 ; X64:       # %bb.0:
 ; X64-NEXT:    xorl %eax, %eax
-; X64-NEXT:    movl %edi, %ecx
-; X64-NEXT:    subw %si, %cx
+; X64-NEXT:    cmpw %si, %di
 ; X64-NEXT:    setns %al
 ; X64-NEXT:    addl $32767, %eax # imm = 0x7FFF
 ; X64-NEXT:    subw %si, %di
@@ -129,8 +120,7 @@ define i8 @func8(i8 %x, i8 %y) nounwind {
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %al
 ; X86-NEXT:    movb {{[0-9]+}}(%esp), %dl
 ; X86-NEXT:    xorl %ecx, %ecx
-; X86-NEXT:    movb %al, %ah
-; X86-NEXT:    subb %dl, %ah
+; X86-NEXT:    cmpb %dl, %al
 ; X86-NEXT:    setns %cl
 ; X86-NEXT:    addl $127, %ecx
 ; X86-NEXT:    subb %dl, %al
@@ -142,8 +132,7 @@ define i8 @func8(i8 %x, i8 %y) nounwind {
 ; X64-LABEL: func8:
 ; X64:       # %bb.0:
 ; X64-NEXT:    xorl %ecx, %ecx
-; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    subb %sil, %al
+; X64-NEXT:    cmpb %sil, %dil
 ; X64-NEXT:    setns %cl
 ; X64-NEXT:    addl $127, %ecx
 ; X64-NEXT:    subb %sil, %dil
@@ -163,8 +152,7 @@ define i4 @func3(i4 %x, i4 %y) nounwind {
 ; X86-NEXT:    shlb $4, %dl
 ; X86-NEXT:    shlb $4, %al
 ; X86-NEXT:    xorl %ecx, %ecx
-; X86-NEXT:    movb %al, %ah
-; X86-NEXT:    subb %dl, %ah
+; X86-NEXT:    cmpb %dl, %al
 ; X86-NEXT:    setns %cl
 ; X86-NEXT:    addl $127, %ecx
 ; X86-NEXT:    subb %dl, %al
@@ -179,8 +167,7 @@ define i4 @func3(i4 %x, i4 %y) nounwind {
 ; X64-NEXT:    shlb $4, %sil
 ; X64-NEXT:    shlb $4, %dil
 ; X64-NEXT:    xorl %ecx, %ecx
-; X64-NEXT:    movl %edi, %eax
-; X64-NEXT:    subb %sil, %al
+; X64-NEXT:    cmpb %sil, %dil
 ; X64-NEXT:    setns %cl
 ; X64-NEXT:    addl $127, %ecx
 ; X64-NEXT:    subb %sil, %dil
@@ -196,15 +183,13 @@ define i4 @func3(i4 %x, i4 %y) nounwind {
 define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; X86-LABEL: vec:
 ; X86:       # %bb.0:
-; X86-NEXT:    pushl %ebp
 ; X86-NEXT:    pushl %ebx
 ; X86-NEXT:    pushl %edi
 ; X86-NEXT:    pushl %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edx
 ; X86-NEXT:    xorl %eax, %eax
-; X86-NEXT:    movl %ecx, %esi
-; X86-NEXT:    subl %edx, %esi
+; X86-NEXT:    cmpl %edx, %ecx
 ; X86-NEXT:    setns %al
 ; X86-NEXT:    addl $2147483647, %eax # imm = 0x7FFFFFFF
 ; X86-NEXT:    subl %edx, %ecx
@@ -212,8 +197,7 @@ define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; X86-NEXT:    cmovol %eax, %ecx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
 ; X86-NEXT:    xorl %eax, %eax
-; X86-NEXT:    movl %edx, %edi
-; X86-NEXT:    subl %esi, %edi
+; X86-NEXT:    cmpl %esi, %edx
 ; X86-NEXT:    setns %al
 ; X86-NEXT:    addl $2147483647, %eax # imm = 0x7FFFFFFF
 ; X86-NEXT:    subl %esi, %edx
@@ -221,8 +205,7 @@ define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; X86-NEXT:    cmovol %eax, %edx
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
 ; X86-NEXT:    xorl %eax, %eax
-; X86-NEXT:    movl %esi, %ebx
-; X86-NEXT:    subl %edi, %ebx
+; X86-NEXT:    cmpl %edi, %esi
 ; X86-NEXT:    setns %al
 ; X86-NEXT:    addl $2147483647, %eax # imm = 0x7FFFFFFF
 ; X86-NEXT:    subl %edi, %esi
@@ -230,8 +213,7 @@ define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; X86-NEXT:    cmovol %eax, %esi
 ; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-NEXT:    xorl %ebx, %ebx
-; X86-NEXT:    movl %edi, %ebp
-; X86-NEXT:    subl %eax, %ebp
+; X86-NEXT:    cmpl %eax, %edi
 ; X86-NEXT:    setns %bl
 ; X86-NEXT:    addl $2147483647, %ebx # imm = 0x7FFFFFFF
 ; X86-NEXT:    subl %eax, %edi
@@ -244,7 +226,6 @@ define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
 ; X86-NEXT:    popl %esi
 ; X86-NEXT:    popl %edi
 ; X86-NEXT:    popl %ebx
-; X86-NEXT:    popl %ebp
 ; X86-NEXT:    retl $4
 ;
 ; X64-LABEL: vec:
diff --git a/test/CodeGen/X86/ssub_sat_vec.ll b/test/CodeGen/X86/ssub_sat_vec.ll
index 43e8fca8304..73fbb7c0029 100644
--- a/test/CodeGen/X86/ssub_sat_vec.ll
+++ b/test/CodeGen/X86/ssub_sat_vec.ll
@@ -408,30 +408,28 @@ define void @v12i16(<12 x i16>* %px, <12 x i16>* %py, <12 x i16>* %pz) nounwind
 define void @v1i8(<1 x i8>* %px, <1 x i8>* %py, <1 x i8>* %pz) nounwind {
 ; SSE-LABEL: v1i8:
 ; SSE:       # %bb.0:
-; SSE-NEXT:    movb (%rdi), %cl
-; SSE-NEXT:    movb (%rsi), %dil
+; SSE-NEXT:    movb (%rdi), %al
+; SSE-NEXT:    movb (%rsi), %cl
 ; SSE-NEXT:    xorl %esi, %esi
-; SSE-NEXT:    movl %ecx, %eax
-; SSE-NEXT:    subb %dil, %al
+; SSE-NEXT:    cmpb %cl, %al
 ; SSE-NEXT:    setns %sil
 ; SSE-NEXT:    addl $127, %esi
-; SSE-NEXT:    subb %dil, %cl
-; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    subb %cl, %al
+; SSE-NEXT:    movzbl %al, %eax
 ; SSE-NEXT:    cmovol %esi, %eax
 ; SSE-NEXT:    movb %al, (%rdx)
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: v1i8:
 ; AVX:       # %bb.0:
-; AVX-NEXT:    movb (%rdi), %cl
-; AVX-NEXT:    movb (%rsi), %dil
+; AVX-NEXT:    movb (%rdi), %al
+; AVX-NEXT:    movb (%rsi), %cl
 ; AVX-NEXT:    xorl %esi, %esi
-; AVX-NEXT:    movl %ecx, %eax
-; AVX-NEXT:    subb %dil, %al
+; AVX-NEXT:    cmpb %cl, %al
 ; AVX-NEXT:    setns %sil
 ; AVX-NEXT:    addl $127, %esi
-; AVX-NEXT:    subb %dil, %cl
-; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    subb %cl, %al
+; AVX-NEXT:    movzbl %al, %eax
 ; AVX-NEXT:    cmovol %esi, %eax
 ; AVX-NEXT:    movb %al, (%rdx)
 ; AVX-NEXT:    retq
@@ -448,8 +446,7 @@ define void @v1i16(<1 x i16>* %px, <1 x i16>* %py, <1 x i16>* %pz) nounwind {
 ; SSE-NEXT:    movzwl (%rdi), %eax
 ; SSE-NEXT:    movzwl (%rsi), %ecx
 ; SSE-NEXT:    xorl %esi, %esi
-; SSE-NEXT:    movl %eax, %edi
-; SSE-NEXT:    subw %cx, %di
+; SSE-NEXT:    cmpw %cx, %ax
 ; SSE-NEXT:    setns %sil
 ; SSE-NEXT:    addl $32767, %esi # imm = 0x7FFF
 ; SSE-NEXT:    subw %cx, %ax
@@ -462,8 +459,7 @@ define void @v1i16(<1 x i16>* %px, <1 x i16>* %py, <1 x i16>* %pz) nounwind {
 ; AVX-NEXT:    movzwl (%rdi), %eax
 ; AVX-NEXT:    movzwl (%rsi), %ecx
 ; AVX-NEXT:    xorl %esi, %esi
-; AVX-NEXT:    movl %eax, %edi
-; AVX-NEXT:    subw %cx, %di
+; AVX-NEXT:    cmpw %cx, %ax
 ; AVX-NEXT:    setns %sil
 ; AVX-NEXT:    addl $32767, %esi # imm = 0x7FFF
 ; AVX-NEXT:    subw %cx, %ax