[llvm] [X86] LowerABD - simplify i32/i64 to use sub+sub+cmov instead of repeating nodes via abs (PR #102174)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Fri Aug 9 04:08:53 PDT 2024
https://github.com/RKSimon updated https://github.com/llvm/llvm-project/pull/102174
>From 599bb21630c7f6671ca756d7f65d5210c90a2dd5 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Tue, 6 Aug 2024 17:06:23 +0100
Subject: [PATCH] [X86] LowerABD - simplify i32/i64 to use sub+sub+cmov instead
of repeating via abs
Using X86ISD::SUB nodes directly allows us to drive the X86ISD::CMOV node with exact flags instead of trying to cleanup the generic codegen via ICMP/SUBO nodes.
---
llvm/lib/Target/X86/X86ISelLowering.cpp | 20 +++-
llvm/test/CodeGen/X86/abds-neg.ll | 99 ++++++++-----------
llvm/test/CodeGen/X86/abds.ll | 124 ++++++++++--------------
llvm/test/CodeGen/X86/abdu-neg.ll | 93 ++++++++----------
llvm/test/CodeGen/X86/abdu.ll | 118 ++++++++++------------
5 files changed, 197 insertions(+), 257 deletions(-)
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 63690487632f88..28a47082b4af89 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -28435,13 +28435,27 @@ static SDValue LowerABD(SDValue Op, const X86Subtarget &Subtarget,
bool IsSigned = Op.getOpcode() == ISD::ABDS;
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
- // TODO: Move to TargetLowering expandABD() once we have ABD promotion.
if (VT.isScalarInteger()) {
+ // abds(lhs, rhs) -> select(slt(lhs,rhs),sub(rhs,lhs),sub(lhs,rhs))
+ // abdu(lhs, rhs) -> select(ult(lhs,rhs),sub(rhs,lhs),sub(lhs,rhs))
+ if (VT.bitsGE(MVT::i32)) {
+ SDVTList VTs = DAG.getVTList(VT, MVT::i32);
+ X86::CondCode CC = IsSigned ? X86::COND_S : X86::COND_B;
+ SDValue LHS = DAG.getFreeze(Op.getOperand(0));
+ SDValue RHS = DAG.getFreeze(Op.getOperand(1));
+ SDValue Diff0 = DAG.getNode(X86ISD::SUB, dl, VTs, LHS, RHS);
+ SDValue Diff1 = DAG.getNode(X86ISD::SUB, dl, VTs, RHS, LHS);
+ return DAG.getNode(X86ISD::CMOV, dl, VT, Diff1, Diff0,
+ DAG.getTargetConstant(CC, dl, MVT::i8),
+ Diff1.getValue(1));
+ }
+
+ // TODO: Move to TargetLowering expandABD() once we have ABD promotion.
+ // abds(lhs, rhs) -> trunc(abs(sub(sext(lhs), sext(rhs))))
+ // abdu(lhs, rhs) -> trunc(abs(sub(zext(lhs), zext(rhs))))
unsigned WideBits = std::max<unsigned>(2 * VT.getScalarSizeInBits(), 32u);
MVT WideVT = MVT::getIntegerVT(WideBits);
if (TLI.isTypeLegal(WideVT)) {
- // abds(lhs, rhs) -> trunc(abs(sub(sext(lhs), sext(rhs))))
- // abdu(lhs, rhs) -> trunc(abs(sub(zext(lhs), zext(rhs))))
unsigned ExtOpc = IsSigned ? ISD::SIGN_EXTEND : ISD::ZERO_EXTEND;
SDValue LHS = DAG.getNode(ExtOpc, dl, WideVT, Op.getOperand(0));
SDValue RHS = DAG.getNode(ExtOpc, dl, WideVT, Op.getOperand(1));
diff --git a/llvm/test/CodeGen/X86/abds-neg.ll b/llvm/test/CodeGen/X86/abds-neg.ll
index 541ed1cbf4fe89..72898d971107ff 100644
--- a/llvm/test/CodeGen/X86/abds-neg.ll
+++ b/llvm/test/CodeGen/X86/abds-neg.ll
@@ -139,28 +139,25 @@ define i16 @abd_ext_i16(i16 %a, i16 %b) nounwind {
define i16 @abd_ext_i16_i32(i16 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i16_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movswl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i16_i32:
; X64: # %bb.0:
-; X64-NEXT: # kill: def $edi killed $edi def $rdi
-; X64-NEXT: movswq %di, %rcx
-; X64-NEXT: movslq %esi, %rax
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movswl %di, %ecx
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %ecx, %esi
+; X64-NEXT: cmovnsl %esi, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $ax killed $ax killed $rax
+; X64-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NEXT: retq
%aext = sext i16 %a to i64
%bext = sext i32 %b to i64
@@ -205,26 +202,22 @@ define i16 @abd_ext_i16_undef(i16 %a, i16 %b) nounwind {
define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32:
; X64: # %bb.0:
-; X64-NEXT: movslq %esi, %rax
-; X64-NEXT: movslq %edi, %rcx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovnsl %esi, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
; X64-NEXT: retq
%aext = sext i32 %a to i64
%bext = sext i32 %b to i64
@@ -238,27 +231,23 @@ define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
; X86-LABEL: abd_ext_i32_i16:
; X86: # %bb.0:
-; X86-NEXT: movswl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_i16:
; X64: # %bb.0:
-; X64-NEXT: # kill: def $esi killed $esi def $rsi
-; X64-NEXT: movswq %si, %rax
-; X64-NEXT: movslq %edi, %rcx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movswl %si, %eax
+; X64-NEXT: movl %edi, %ecx
+; X64-NEXT: subl %eax, %ecx
+; X64-NEXT: subl %edi, %eax
+; X64-NEXT: cmovsl %ecx, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
; X64-NEXT: retq
%aext = sext i32 %a to i64
%bext = sext i16 %b to i64
@@ -272,26 +261,22 @@ define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
define i32 @abd_ext_i32_undef(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32_undef:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_undef:
; X64: # %bb.0:
-; X64-NEXT: movslq %esi, %rax
-; X64-NEXT: movslq %edi, %rcx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovnsl %esi, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
; X64-NEXT: retq
%aext = sext i32 %a to i64
%bext = sext i32 %b to i64
@@ -332,9 +317,8 @@ define i64 @abd_ext_i64(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovgq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovnsq %rsi, %rax
; X64-NEXT: negq %rax
; X64-NEXT: retq
%aext = sext i64 %a to i128
@@ -376,9 +360,8 @@ define i64 @abd_ext_i64_undef(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovgq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovnsq %rsi, %rax
; X64-NEXT: negq %rax
; X64-NEXT: retq
%aext = sext i64 %a to i128
diff --git a/llvm/test/CodeGen/X86/abds.ll b/llvm/test/CodeGen/X86/abds.ll
index fd47de5af05ab4..50f8bddb033b17 100644
--- a/llvm/test/CodeGen/X86/abds.ll
+++ b/llvm/test/CodeGen/X86/abds.ll
@@ -129,26 +129,23 @@ define i16 @abd_ext_i16(i16 %a, i16 %b) nounwind {
define i16 @abd_ext_i16_i32(i16 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i16_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movswl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i16_i32:
; X64: # %bb.0:
-; X64-NEXT: # kill: def $edi killed $edi def $rdi
-; X64-NEXT: movswq %di, %rcx
-; X64-NEXT: movslq %esi, %rax
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $ax killed $ax killed $rax
+; X64-NEXT: movswl %di, %ecx
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %ecx, %esi
+; X64-NEXT: cmovnsl %esi, %eax
+; X64-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NEXT: retq
%aext = sext i16 %a to i64
%bext = sext i32 %b to i64
@@ -191,24 +188,20 @@ define i16 @abd_ext_i16_undef(i16 %a, i16 %b) nounwind {
define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32:
; X64: # %bb.0:
-; X64-NEXT: movslq %esi, %rax
-; X64-NEXT: movslq %edi, %rcx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovnsl %esi, %eax
; X64-NEXT: retq
%aext = sext i32 %a to i64
%bext = sext i32 %b to i64
@@ -221,25 +214,21 @@ define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
; X86-LABEL: abd_ext_i32_i16:
; X86: # %bb.0:
-; X86-NEXT: movswl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movswl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_i16:
; X64: # %bb.0:
-; X64-NEXT: # kill: def $esi killed $esi def $rsi
-; X64-NEXT: movswq %si, %rax
-; X64-NEXT: movslq %edi, %rcx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movswl %si, %eax
+; X64-NEXT: movl %edi, %ecx
+; X64-NEXT: subl %eax, %ecx
+; X64-NEXT: subl %edi, %eax
+; X64-NEXT: cmovsl %ecx, %eax
; X64-NEXT: retq
%aext = sext i32 %a to i64
%bext = sext i16 %b to i64
@@ -252,24 +241,20 @@ define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
define i32 @abd_ext_i32_undef(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32_undef:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_undef:
; X64: # %bb.0:
-; X64-NEXT: movslq %esi, %rax
-; X64-NEXT: movslq %edi, %rcx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovnsl %esi, %eax
; X64-NEXT: retq
%aext = sext i32 %a to i64
%bext = sext i32 %b to i64
@@ -306,9 +291,8 @@ define i64 @abd_ext_i64(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovgq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovnsq %rsi, %rax
; X64-NEXT: retq
%aext = sext i64 %a to i128
%bext = sext i64 %b to i128
@@ -345,9 +329,8 @@ define i64 @abd_ext_i64_undef(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovgq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovnsq %rsi, %rax
; X64-NEXT: retq
%aext = sext i64 %a to i128
%bext = sext i64 %b to i128
@@ -542,24 +525,20 @@ define i16 @abd_minmax_i16(i16 %a, i16 %b) nounwind {
define i32 @abd_minmax_i32(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_minmax_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovlel %edx, %eax
+; X86-NEXT: cmovsl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_minmax_i32:
; X64: # %bb.0:
-; X64-NEXT: movslq %esi, %rax
-; X64-NEXT: movslq %edi, %rcx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovnsl %esi, %eax
; X64-NEXT: retq
%min = call i32 @llvm.smin.i32(i32 %a, i32 %b)
%max = call i32 @llvm.smax.i32(i32 %a, i32 %b)
@@ -594,9 +573,8 @@ define i64 @abd_minmax_i64(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovgq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovnsq %rsi, %rax
; X64-NEXT: retq
%min = call i64 @llvm.smin.i64(i64 %a, i64 %b)
%max = call i64 @llvm.smax.i64(i64 %a, i64 %b)
diff --git a/llvm/test/CodeGen/X86/abdu-neg.ll b/llvm/test/CodeGen/X86/abdu-neg.ll
index fe4525009ad5c8..9cb3e5e8bf0c26 100644
--- a/llvm/test/CodeGen/X86/abdu-neg.ll
+++ b/llvm/test/CodeGen/X86/abdu-neg.ll
@@ -139,27 +139,25 @@ define i16 @abd_ext_i16(i16 %a, i16 %b) nounwind {
define i16 @abd_ext_i16_i32(i16 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i16_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i16_i32:
; X64: # %bb.0:
-; X64-NEXT: movl %esi, %eax
; X64-NEXT: movzwl %di, %ecx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %ecx, %esi
+; X64-NEXT: cmovael %esi, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $ax killed $ax killed $rax
+; X64-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NEXT: retq
%aext = zext i16 %a to i64
%bext = zext i32 %b to i64
@@ -204,26 +202,22 @@ define i16 @abd_ext_i16_undef(i16 %a, i16 %b) nounwind {
define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32:
; X64: # %bb.0:
-; X64-NEXT: movl %esi, %eax
-; X64-NEXT: movl %edi, %ecx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovael %esi, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
; X64-NEXT: retq
%aext = zext i32 %a to i64
%bext = zext i32 %b to i64
@@ -237,26 +231,23 @@ define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
; X86-LABEL: abd_ext_i32_i16:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_i16:
; X64: # %bb.0:
-; X64-NEXT: movl %edi, %ecx
; X64-NEXT: movzwl %si, %eax
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movl %edi, %ecx
+; X64-NEXT: subl %eax, %ecx
+; X64-NEXT: subl %edi, %eax
+; X64-NEXT: cmovbl %ecx, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
; X64-NEXT: retq
%aext = zext i32 %a to i64
%bext = zext i16 %b to i64
@@ -270,26 +261,22 @@ define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
define i32 @abd_ext_i32_undef(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32_undef:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: negl %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_undef:
; X64: # %bb.0:
-; X64-NEXT: movl %esi, %eax
-; X64-NEXT: movl %edi, %ecx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovael %esi, %eax
; X64-NEXT: negl %eax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
; X64-NEXT: retq
%aext = zext i32 %a to i64
%bext = zext i32 %b to i64
@@ -324,9 +311,8 @@ define i64 @abd_ext_i64(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovaq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovaeq %rsi, %rax
; X64-NEXT: negq %rax
; X64-NEXT: retq
%aext = zext i64 %a to i128
@@ -362,9 +348,8 @@ define i64 @abd_ext_i64_undef(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovaq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovaeq %rsi, %rax
; X64-NEXT: negq %rax
; X64-NEXT: retq
%aext = zext i64 %a to i128
diff --git a/llvm/test/CodeGen/X86/abdu.ll b/llvm/test/CodeGen/X86/abdu.ll
index 2e2a1debc38cb1..e808e0f21babf2 100644
--- a/llvm/test/CodeGen/X86/abdu.ll
+++ b/llvm/test/CodeGen/X86/abdu.ll
@@ -129,25 +129,23 @@ define i16 @abd_ext_i16(i16 %a, i16 %b) nounwind {
define i16 @abd_ext_i16_i32(i16 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i16_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movzwl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: # kill: def $ax killed $ax killed $eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i16_i32:
; X64: # %bb.0:
-; X64-NEXT: movl %esi, %eax
; X64-NEXT: movzwl %di, %ecx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $ax killed $ax killed $rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %ecx, %esi
+; X64-NEXT: cmovael %esi, %eax
+; X64-NEXT: # kill: def $ax killed $ax killed $eax
; X64-NEXT: retq
%aext = zext i16 %a to i64
%bext = zext i32 %b to i64
@@ -190,24 +188,20 @@ define i16 @abd_ext_i16_undef(i16 %a, i16 %b) nounwind {
define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32:
; X64: # %bb.0:
-; X64-NEXT: movl %esi, %eax
-; X64-NEXT: movl %edi, %ecx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovael %esi, %eax
; X64-NEXT: retq
%aext = zext i32 %a to i64
%bext = zext i32 %b to i64
@@ -220,24 +214,21 @@ define i32 @abd_ext_i32(i32 %a, i32 %b) nounwind {
define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
; X86-LABEL: abd_ext_i32_i16:
; X86: # %bb.0:
-; X86-NEXT: movzwl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_i16:
; X64: # %bb.0:
-; X64-NEXT: movl %edi, %ecx
; X64-NEXT: movzwl %si, %eax
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movl %edi, %ecx
+; X64-NEXT: subl %eax, %ecx
+; X64-NEXT: subl %edi, %eax
+; X64-NEXT: cmovbl %ecx, %eax
; X64-NEXT: retq
%aext = zext i32 %a to i64
%bext = zext i16 %b to i64
@@ -250,24 +241,20 @@ define i32 @abd_ext_i32_i16(i32 %a, i16 %b) nounwind {
define i32 @abd_ext_i32_undef(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_ext_i32_undef:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_ext_i32_undef:
; X64: # %bb.0:
-; X64-NEXT: movl %esi, %eax
-; X64-NEXT: movl %edi, %ecx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovael %esi, %eax
; X64-NEXT: retq
%aext = zext i32 %a to i64
%bext = zext i32 %b to i64
@@ -296,9 +283,8 @@ define i64 @abd_ext_i64(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovaq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovaeq %rsi, %rax
; X64-NEXT: retq
%aext = zext i64 %a to i128
%bext = zext i64 %b to i128
@@ -327,9 +313,8 @@ define i64 @abd_ext_i64_undef(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovaq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovaeq %rsi, %rax
; X64-NEXT: retq
%aext = zext i64 %a to i128
%bext = zext i64 %b to i128
@@ -512,24 +497,20 @@ define i16 @abd_minmax_i16(i16 %a, i16 %b) nounwind {
define i32 @abd_minmax_i32(i32 %a, i32 %b) nounwind {
; X86-LABEL: abd_minmax_i32:
; X86: # %bb.0:
-; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: movl %eax, %edx
-; X86-NEXT: subl %ecx, %edx
-; X86-NEXT: negl %edx
+; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT: movl %ecx, %edx
+; X86-NEXT: subl %eax, %edx
; X86-NEXT: subl %ecx, %eax
-; X86-NEXT: cmovbel %edx, %eax
+; X86-NEXT: cmovbl %edx, %eax
; X86-NEXT: retl
;
; X64-LABEL: abd_minmax_i32:
; X64: # %bb.0:
-; X64-NEXT: movl %esi, %eax
-; X64-NEXT: movl %edi, %ecx
-; X64-NEXT: subq %rax, %rcx
-; X64-NEXT: movq %rcx, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: cmovsq %rcx, %rax
-; X64-NEXT: # kill: def $eax killed $eax killed $rax
+; X64-NEXT: movl %edi, %eax
+; X64-NEXT: subl %esi, %eax
+; X64-NEXT: subl %edi, %esi
+; X64-NEXT: cmovael %esi, %eax
; X64-NEXT: retq
%min = call i32 @llvm.umin.i32(i32 %a, i32 %b)
%max = call i32 @llvm.umax.i32(i32 %a, i32 %b)
@@ -556,9 +537,8 @@ define i64 @abd_minmax_i64(i64 %a, i64 %b) nounwind {
; X64: # %bb.0:
; X64-NEXT: movq %rdi, %rax
; X64-NEXT: subq %rsi, %rax
-; X64-NEXT: negq %rax
-; X64-NEXT: subq %rsi, %rdi
-; X64-NEXT: cmovaq %rdi, %rax
+; X64-NEXT: subq %rdi, %rsi
+; X64-NEXT: cmovaeq %rsi, %rax
; X64-NEXT: retq
%min = call i64 @llvm.umin.i64(i64 %a, i64 %b)
%max = call i64 @llvm.umax.i64(i64 %a, i64 %b)
More information about the llvm-commits
mailing list