6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-07-20 10:30:03 +00:00
Code Issues Projects Releases Wiki Activity

Allow ScalarEvolution to catch more min/max cases

Browse Source 
For the attached test case different types are used in the ICmpInst
  and SelectInst that represent the min/max expressions. However, if the
  ICmpInst type is smaller a comparison with the sign/zero extended
  operands would have yielded the same result. This situation might
  arise after the instruction combination pass was applied.

  Differential Revision: http://reviews.llvm.org/D7338


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@228572 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Johannes Doerfert 2015-02-09 12:34:23 +00:00
parent 7e44cee3e6
commit f8c4fd6005
2 changed files with 78 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

34
lib/Analysis/ScalarEvolution.cpp
View File

@ -4297,9 +4297,10 @@ const SCEV *ScalarEvolution::createSCEV(Value *V) {
case ICmpInst::ICMP_SGE:
// a >s b ? a+x : b+x -> smax(a, b)+x
// a >s b ? b+x : a+x -> smin(a, b)+x
if (LHS->getType() == U->getType()) {
const SCEV *LS = getSCEV(LHS);
const SCEV *RS = getSCEV(RHS);
if (getTypeSizeInBits(LHS->getType()) <=
getTypeSizeInBits(U->getType())) {
const SCEV *LS = getNoopOrSignExtend(getSCEV(LHS), U->getType());
const SCEV *RS = getNoopOrSignExtend(getSCEV(RHS), U->getType());
const SCEV *LA = getSCEV(U->getOperand(1));
const SCEV *RA = getSCEV(U->getOperand(2));
const SCEV *LDiff = getMinusSCEV(LA, LS);
@ -4320,9 +4321,10 @@ const SCEV *ScalarEvolution::createSCEV(Value *V) {
case ICmpInst::ICMP_UGE:
// a >u b ? a+x : b+x -> umax(a, b)+x
// a >u b ? b+x : a+x -> umin(a, b)+x
if (LHS->getType() == U->getType()) {
const SCEV *LS = getSCEV(LHS);
const SCEV *RS = getSCEV(RHS);
if (getTypeSizeInBits(LHS->getType()) <=
getTypeSizeInBits(U->getType())) {
const SCEV *LS = getNoopOrZeroExtend(getSCEV(LHS), U->getType());
const SCEV *RS = getNoopOrZeroExtend(getSCEV(RHS), U->getType());
const SCEV *LA = getSCEV(U->getOperand(1));
const SCEV *RA = getSCEV(U->getOperand(2));
const SCEV *LDiff = getMinusSCEV(LA, LS);
@ -4337,11 +4339,11 @@ const SCEV *ScalarEvolution::createSCEV(Value *V) {
break;
case ICmpInst::ICMP_NE:
// n != 0 ? n+x : 1+x -> umax(n, 1)+x
if (LHS->getType() == U->getType() &&
isa<ConstantInt>(RHS) &&
cast<ConstantInt>(RHS)->isZero()) {
const SCEV *One = getConstant(LHS->getType(), 1);
const SCEV *LS = getSCEV(LHS);
if (getTypeSizeInBits(LHS->getType()) <=
getTypeSizeInBits(U->getType()) &&
isa<ConstantInt>(RHS) && cast<ConstantInt>(RHS)->isZero()) {
const SCEV *One = getConstant(U->getType(), 1);
const SCEV *LS = getNoopOrZeroExtend(getSCEV(LHS), U->getType());
const SCEV *LA = getSCEV(U->getOperand(1));
const SCEV *RA = getSCEV(U->getOperand(2));
const SCEV *LDiff = getMinusSCEV(LA, LS);
@ -4352,11 +4354,11 @@ const SCEV *ScalarEvolution::createSCEV(Value *V) {
break;
case ICmpInst::ICMP_EQ:
// n == 0 ? 1+x : n+x -> umax(n, 1)+x
if (LHS->getType() == U->getType() &&
isa<ConstantInt>(RHS) &&
cast<ConstantInt>(RHS)->isZero()) {
const SCEV *One = getConstant(LHS->getType(), 1);
const SCEV *LS = getSCEV(LHS);
if (getTypeSizeInBits(LHS->getType()) <=
getTypeSizeInBits(U->getType()) &&
isa<ConstantInt>(RHS) && cast<ConstantInt>(RHS)->isZero()) {
const SCEV *One = getConstant(U->getType(), 1);
const SCEV *LS = getNoopOrZeroExtend(getSCEV(LHS), U->getType());
const SCEV *LA = getSCEV(U->getOperand(1));
const SCEV *RA = getSCEV(U->getOperand(2));
const SCEV *LDiff = getMinusSCEV(LA, One);

53
test/Analysis/ScalarEvolution/min-max-exprs.ll Normal file
View File

@ -0,0 +1,53 @@
; RUN: opt -scalar-evolution -analyze < %s | FileCheck %s
;
; This checks if the min and max expressions are properly recognized by
; ScalarEvolution even though they the ICmpInst and SelectInst have different
; types.
;
; #define max(a, b) (a > b ? a : b)
; #define min(a, b) (a < b ? a : b)
;
; void f(int *A, int N) {
; for (int i = 0; i < N; i++) {
; A[max(0, i - 3)] = A[min(N, i + 3)] * 2;
; }
; }
;
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
define void @f(i32* %A, i32 %N) {
bb:
br label %bb1
bb1: ; preds = %bb2, %bb
%i.0 = phi i32 [ 0, %bb ], [ %tmp23, %bb2 ]
%i.0.1 = sext i32 %i.0 to i64
%tmp = icmp slt i32 %i.0, %N
br i1 %tmp, label %bb2, label %bb24
bb2: ; preds = %bb1
%tmp3 = add nuw nsw i32 %i.0, 3
%tmp4 = icmp slt i32 %tmp3, %N
%tmp5 = sext i32 %tmp3 to i64
%tmp6 = sext i32 %N to i64
%tmp9 = select i1 %tmp4, i64 %tmp5, i64 %tmp6
; min(N, i+3)
; CHECK: select i1 %tmp4, i64 %tmp5, i64 %tmp6
; CHECK-NEXT: --> (-1 + (-1 * ((-1 + (-1 * (sext i32 {3,+,1}<nw><%bb1> to i64))) smax (-1 + (-1 * (sext i32 %N to i64))))))
%tmp11 = getelementptr inbounds i32* %A, i64 %tmp9
%tmp12 = load i32* %tmp11, align 4
%tmp13 = shl nsw i32 %tmp12, 1
%tmp14 = icmp sge i32 3, %i.0
%tmp17 = add nsw i64 %i.0.1, -3
%tmp19 = select i1 %tmp14, i64 0, i64 %tmp17
; max(0, i - 3)
; CHECK: select i1 %tmp14, i64 0, i64 %tmp17
; CHECK-NEXT: --> (-3 + (3 smax {0,+,1}<nuw><nsw><%bb1>))
%tmp21 = getelementptr inbounds i32* %A, i64 %tmp19
store i32 %tmp13, i32* %tmp21, align 4
%tmp23 = add nuw nsw i32 %i.0, 1
br label %bb1
bb24: ; preds = %bb1
ret void
}