replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[InstSimplify] fold 'fcmp nnan oge X, 0.0' when X is not negative 

This re-raises some of the open questions about how to apply and use fast-math-flags in IR from PR38086:
https://bugs.llvm.org/show_bug.cgi?id=38086
...but given the current implementation (no FMF on casts), this is likely the only way to predicate the 
transform.

This is part of solving PR39475:
https://bugs.llvm.org/show_bug.cgi?id=39475

Differential Revision: https://reviews.llvm.org/D53874

llvm-svn: 345725
This commit is contained in:
Sanjay Patel 2018-10-31 14:57:23 +00:00
parent 7d6d9eb688
commit 85cba3b6fb
3 changed files with 11 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Analysis/InstructionSimplify.cpp
View File

@ -3612,6 +3612,10 @@ static Value *SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
}
if (C->isZero()) {
switch (Pred) {
case FCmpInst::FCMP_OGE:
if (FMF.noNaNs() && CannotBeOrderedLessThanZero(LHS, Q.TLI))
return getTrue(RetTy);
break;
case FCmpInst::FCMP_UGE:
if (CannotBeOrderedLessThanZero(LHS, Q.TLI))
return getTrue(RetTy);

5
llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp
View File

@ -5433,8 +5433,8 @@ Instruction *InstCombiner::visitFCmpInst(FCmpInst &I) {
break;
// fabs(x) < 0 --> false
case FCmpInst::FCMP_OLT:
llvm_unreachable("handled by SimplifyFCmpInst");
// fabs(x) > 0 --> x != 0
llvm_unreachable("fcmp should have simplified");
// fabs(x) > 0 --> x != 0
case FCmpInst::FCMP_OGT:
return new FCmpInst(FCmpInst::FCMP_ONE, CI->getArgOperand(0), RHSC);
// fabs(x) <= 0 --> x == 0
@ -5442,6 +5442,7 @@ Instruction *InstCombiner::visitFCmpInst(FCmpInst &I) {
return new FCmpInst(FCmpInst::FCMP_OEQ, CI->getArgOperand(0), RHSC);
// fabs(x) >= 0 --> !isnan(x)
case FCmpInst::FCMP_OGE:
assert(!I.hasNoNaNs() && "fcmp should have simplified");
return new FCmpInst(FCmpInst::FCMP_ORD, CI->getArgOperand(0), RHSC);
// fabs(x) == 0 --> x == 0
// fabs(x) != 0 --> x != 0

16
llvm/test/Transforms/InstSimplify/floating-point-compare.ll
View File

@ -254,9 +254,7 @@ define <2 x i1> @UIToFP_is_nan_or_positive_or_zero_vec(<2 x i32> %x) {
define i1 @UIToFP_nnan_is_positive_or_zero(i32 %x) {
; CHECK-LABEL: @UIToFP_nnan_is_positive_or_zero(
; CHECK-NEXT: [[A:%.*]] = uitofp i32 [[X:%.*]] to float
; CHECK-NEXT: [[R:%.*]] = fcmp nnan oge float [[A]], 0.000000e+00
; CHECK-NEXT: ret i1 [[R]]
; CHECK-NEXT: ret i1 true
;
%a = uitofp i32 %x to float
%r = fcmp nnan oge float %a, 0.000000e+00
@ -265,9 +263,7 @@ define i1 @UIToFP_nnan_is_positive_or_zero(i32 %x) {
define <2 x i1> @UIToFP_nnan_is_positive_or_zero_vec(<2 x i32> %x) {
; CHECK-LABEL: @UIToFP_nnan_is_positive_or_zero_vec(
; CHECK-NEXT: [[A:%.*]] = uitofp <2 x i32> [[X:%.*]] to <2 x float>
; CHECK-NEXT: [[R:%.*]] = fcmp nnan oge <2 x float> [[A]], zeroinitializer
; CHECK-NEXT: ret <2 x i1> [[R]]
; CHECK-NEXT: ret <2 x i1> <i1 true, i1 true>
;
%a = uitofp <2 x i32> %x to <2 x float>
%r = fcmp nnan oge <2 x float> %a, zeroinitializer
@ -334,9 +330,7 @@ define <2 x i1> @fabs_is_nan_or_positive_or_zero_vec(<2 x double> %x) {
define i1 @fabs_nnan_is_positive_or_zero(double %x) {
; CHECK-LABEL: @fabs_nnan_is_positive_or_zero(
; CHECK-NEXT: [[FABS:%.*]] = tail call double @llvm.fabs.f64(double [[X:%.*]])
; CHECK-NEXT: [[CMP:%.*]] = fcmp nnan oge double [[FABS]], 0.000000e+00
; CHECK-NEXT: ret i1 [[CMP]]
; CHECK-NEXT: ret i1 true
;
%fabs = tail call double @llvm.fabs.f64(double %x)
%cmp = fcmp nnan oge double %fabs, 0.0
@ -345,9 +339,7 @@ define i1 @fabs_nnan_is_positive_or_zero(double %x) {
define <2 x i1> @fabs_nnan_is_positive_or_zero_vec(<2 x double> %x) {
; CHECK-LABEL: @fabs_nnan_is_positive_or_zero_vec(
; CHECK-NEXT: [[FABS:%.*]] = tail call <2 x double> @llvm.fabs.v2f64(<2 x double> [[X:%.*]])
; CHECK-NEXT: [[CMP:%.*]] = fcmp nnan oge <2 x double> [[FABS]], zeroinitializer
; CHECK-NEXT: ret <2 x i1> [[CMP]]
; CHECK-NEXT: ret <2 x i1> <i1 true, i1 true>
;
%fabs = tail call <2 x double> @llvm.fabs.v2f64(<2 x double> %x)
%cmp = fcmp nnan oge <2 x double> %fabs, zeroinitializer