RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 10:42:39 +01:00
Code Issues Projects Releases Wiki Activity

[Analysis] isDereferenceableAndAlignedPointer(): recurse into select's hands

Browse Source 
By doing this within the method itself,
we support traversing multiple levels of selects (TODO: PHI's),
fixing the SROA `std::clamp()` testcase.

Fixes https://bugs.llvm.org/show_bug.cgi?id=47271
Mostly fixes https://bugs.llvm.org/show_bug.cgi?id=49909
This commit is contained in:
Roman Lebedev 2021-04-10 00:31:56 +03:00
parent d3ca96c41b
commit f51bde334f
2 changed files with 15 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
lib/Analysis/Loads.cpp
View File

@ -59,6 +59,16 @@ static bool isDereferenceableAndAlignedPointer(
// Note that it is not safe to speculate into a malloc'd region because
// malloc may return null.
// Recurse into both hands of select.
if (const SelectInst *Sel = dyn_cast<SelectInst>(V)) {
return isDereferenceableAndAlignedPointer(Sel->getTrueValue(), Alignment,
Size, DL, CtxI, DT, TLI, Visited,
MaxDepth) &&
isDereferenceableAndAlignedPointer(Sel->getFalseValue(), Alignment,
Size, DL, CtxI, DT, TLI, Visited,
MaxDepth);
}
// bitcast instructions are no-ops as far as dereferenceability is concerned.
if (const BitCastOperator *BC = dyn_cast<BitCastOperator>(V)) {
if (BC->getSrcTy()->isPointerTy())

14
test/Transforms/SROA/std-clamp.ll
View File

@ -6,18 +6,14 @@ target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16
define float @_Z8stdclampfff(float %x, float %lo, float %hi) {
; CHECK-LABEL: @_Z8stdclampfff(
; CHECK-NEXT: bb:
; CHECK-NEXT: [[I:%.*]] = alloca float, align 4
; CHECK-NEXT: [[I3:%.*]] = alloca float, align 4
; CHECK-NEXT: [[I4:%.*]] = alloca float, align 4
; CHECK-NEXT: store float [[X:%.*]], float* [[I]], align 4
; CHECK-NEXT: store float [[LO:%.*]], float* [[I3]], align 4
; CHECK-NEXT: store float [[HI:%.*]], float* [[I4]], align 4
; CHECK-NEXT: [[I5:%.*]] = fcmp fast olt float [[X]], [[LO]]
; CHECK-NEXT: [[I5:%.*]] = fcmp fast olt float [[X:%.*]], [[LO:%.*]]
; CHECK-NEXT: [[I6:%.*]] = fcmp fast olt float [[HI]], [[X]]
; CHECK-NEXT: [[I7:%.*]] = select i1 [[I6]], float* [[I4]], float* [[I]]
; CHECK-NEXT: [[I8:%.*]] = select i1 [[I5]], float* [[I3]], float* [[I7]]
; CHECK-NEXT: [[I9:%.*]] = load float, float* [[I8]], align 4
; CHECK-NEXT: ret float [[I9]]
; CHECK-NEXT: [[I9_SROA_SPECULATE_LOAD_FALSE_SROA_SPECULATE_LOAD_TRUE:%.*]] = load float, float* [[I4]], align 4
; CHECK-NEXT: [[I9_SROA_SPECULATE_LOAD_FALSE_SROA_SPECULATED:%.*]] = select i1 [[I6]], float [[I9_SROA_SPECULATE_LOAD_FALSE_SROA_SPECULATE_LOAD_TRUE]], float [[X]]
; CHECK-NEXT: [[I9_SROA_SPECULATED:%.*]] = select i1 [[I5]], float [[LO]], float [[I9_SROA_SPECULATE_LOAD_FALSE_SROA_SPECULATED]]
; CHECK-NEXT: ret float [[I9_SROA_SPECULATED]]
;
bb:
%i = alloca float, align 4