[llvm] d561b6f - [InstCombine] Fold (select C, (gep Ptr, Idx), Ptr) -> (gep Ptr, (select C, Idx, 0)) (PR50183) (REAPPLIED)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Wed Jul 14 04:25:54 PDT 2021
Author: Simon Pilgrim
Date: 2021-07-14T12:21:01+01:00
New Revision: d561b6fbdbe6d1da05fd92003a4ac1e37bf4b8bc
URL: https://github.com/llvm/llvm-project/commit/d561b6fbdbe6d1da05fd92003a4ac1e37bf4b8bc
DIFF: https://github.com/llvm/llvm-project/commit/d561b6fbdbe6d1da05fd92003a4ac1e37bf4b8bc.diff
LOG: [InstCombine] Fold (select C, (gep Ptr, Idx), Ptr) -> (gep Ptr, (select C, Idx, 0)) (PR50183) (REAPPLIED)
As discussed on PR50183, we already fold to prefer 'select-of-idx' vs 'select-of-gep':
define <4 x i32>* @select0a(<4 x i32>* %a0, i64 %a1, i1 %a2, i64 %a3) {
%gep0 = getelementptr inbounds <4 x i32>, <4 x i32>* %a0, i64 %a1
%gep1 = getelementptr inbounds <4 x i32>, <4 x i32>* %a0, i64 %a3
%sel = select i1 %a2, <4 x i32>* %gep0, <4 x i32>* %gep1
ret <4 x i32>* %sel
}
-->
define <4 x i32>* @select1a(<4 x i32>* %a0, i64 %a1, i1 %a2, i64 %a3) {
%sel = select i1 %a2, i64 %a1, i64 %a3
%gep = getelementptr inbounds <4 x i32>, <4 x i32>* %a0, i64 %sel
ret <4 x i32>* %gep
}
This patch adds basic handling for the 'fallthrough' cases where the gep idx == 0 has been folded away to the base address:
define <4 x i32>* @select0(<4 x i32>* %a0, i64 %a1, i1 %a2) {
%gep = getelementptr inbounds <4 x i32>, <4 x i32>* %a0, i64 %a1
%sel = select i1 %a2, <4 x i32>* %a0, <4 x i32>* %gep
ret <4 x i32>* %sel
}
-->
define <4 x i32>* @select1(<4 x i32>* %a0, i64 %a1, i1 %a2) {
%sel = select i1 %a2, i64 0, i64 %a1
%gep = getelementptr inbounds <4 x i32>, <4 x i32>* %a0, i64 %sel
ret <4 x i32>* %gep
}
Reapplied with a fix for the bpf "-bpf-disable-avoid-speculation" tests
Differential Revision: https://reviews.llvm.org/D105901
Added:
Modified:
llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
llvm/test/CodeGen/BPF/adjust-opt-speculative1.ll
llvm/test/CodeGen/BPF/adjust-opt-speculative2.ll
llvm/test/Transforms/InstCombine/select-gep.ll
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
index eadc826c7190d..fb9750d59e656 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSelect.cpp
@@ -2933,6 +2933,33 @@ Instruction *InstCombinerImpl::visitSelectInst(SelectInst &SI) {
if (Instruction *I = foldSelectExtConst(SI))
return I;
+ // Fold (select C, (gep Ptr, Idx), Ptr) -> (gep Ptr, (select C, Idx, 0))
+ // Fold (select C, Ptr, (gep Ptr, Idx)) -> (gep Ptr, (select C, 0, Idx))
+ auto SelectGepWithBase = [&](GetElementPtrInst *Gep, Value *Base,
+ bool Swap) -> GetElementPtrInst * {
+ Value *Ptr = Gep->getPointerOperand();
+ if (Gep->getNumOperands() != 2 || Gep->getPointerOperand() != Base ||
+ !Gep->hasOneUse())
+ return nullptr;
+ Type *ElementType = Gep->getResultElementType();
+ Value *Idx = Gep->getOperand(1);
+ Value *NewT = Idx;
+ Value *NewF = Constant::getNullValue(Idx->getType());
+ if (Swap)
+ std::swap(NewT, NewF);
+ Value *NewSI =
+ Builder.CreateSelect(CondVal, NewT, NewF, SI.getName() + ".idx", &SI);
+ return Gep->isInBounds()
+ ? GetElementPtrInst::CreateInBounds(ElementType, Ptr, {NewSI})
+ : GetElementPtrInst::Create(ElementType, Ptr, {NewSI});
+ };
+ if (auto *TrueGep = dyn_cast<GetElementPtrInst>(TrueVal))
+ if (auto *NewGep = SelectGepWithBase(TrueGep, FalseVal, false))
+ return NewGep;
+ if (auto *FalseGep = dyn_cast<GetElementPtrInst>(FalseVal))
+ if (auto *NewGep = SelectGepWithBase(FalseGep, TrueVal, true))
+ return NewGep;
+
// See if we can fold the select into one of our operands.
if (SelType->isIntOrIntVectorTy() || SelType->isFPOrFPVectorTy()) {
if (Instruction *FoldI = foldSelectIntoOp(SI, TrueVal, FalseVal))
diff --git a/llvm/test/CodeGen/BPF/adjust-opt-speculative1.ll b/llvm/test/CodeGen/BPF/adjust-opt-speculative1.ll
index 0d29fff08eed7..afb820c2d6366 100644
--- a/llvm/test/CodeGen/BPF/adjust-opt-speculative1.ll
+++ b/llvm/test/CodeGen/BPF/adjust-opt-speculative1.ll
@@ -49,12 +49,12 @@ if.end: ; preds = %if.then, %entry
; CHECK: [[LABEL]]:
; CHECK: r0 = [[REG6]]
-; CHECK-DISABLE: r0 = [[REG6]]
-; CHECK-DISABLE: r0 += [[REG1:r[0-9]+]]
-; CHECK-DISABLE: [[REG2:r[0-9]+]] = 8
-; CHECK-DISABLE: if [[REG2]] > [[REG1]] goto [[LABEL:.*]]
-; CHECK-DISABLE: r0 = [[REG6]]
+; CHECK-DISABLE: [[REG1:r[0-9]+]] = 8
+; CHECK-DISABLE: if [[REG1]] > r0 goto [[LABEL:.*]]
+; CHECK-DISABLE: r0 = 0
; CHECK-DISABLE: [[LABEL]]:
+; CHECK-DISABLE: [[REG6]] += r0
+; CHECK-DISABLE: r0 = [[REG6]]
; CHECK-COMMON: exit
diff --git a/llvm/test/CodeGen/BPF/adjust-opt-speculative2.ll b/llvm/test/CodeGen/BPF/adjust-opt-speculative2.ll
index 3b5cbea9f84b6..b1a53d18271e0 100644
--- a/llvm/test/CodeGen/BPF/adjust-opt-speculative2.ll
+++ b/llvm/test/CodeGen/BPF/adjust-opt-speculative2.ll
@@ -56,12 +56,14 @@ if.end: ; preds = %if.then, %entry
; CHECK-DISABLE: [[REG1:r[0-9]+]] = r0
; CHECK-DISABLE: [[REG1]] <<= 32
; CHECK-DISABLE: [[REG1]] >>= 32
-; CHECK-DISABLE: r0 = [[REG6]]
-; CHECK-DISABLE: r0 += [[REG1]]
; CHECK-DISABLE: [[REG2:r[0-9]+]] = 8
; CHECK-DISABLE: if [[REG2]] > [[REG1]] goto [[LABEL:.*]]
-; CHECK-DISABLE: r0 = [[REG6]]
+; CHECK-DISABLE: r0 = 0
; CHECK-DISABLE: [[LABEL]]:
+; CHECK-DISABLE: r0 <<= 32
+; CHECK-DISABLE: r0 >>= 32
+; CHECK-DISABLE: [[REG6]] += r0
+; CHECK-DISABLE: r0 = [[REG6]]
; CHECK-COMMON: exit
diff --git a/llvm/test/Transforms/InstCombine/select-gep.ll b/llvm/test/Transforms/InstCombine/select-gep.ll
index 77710cee416d2..bba67091b3fa2 100644
--- a/llvm/test/Transforms/InstCombine/select-gep.ll
+++ b/llvm/test/Transforms/InstCombine/select-gep.ll
@@ -74,9 +74,9 @@ define i32* @test2(i32* %p, i64 %x, i64 %y) {
; PR50183
define i32* @test2a(i32* %p, i64 %x, i64 %y) {
; CHECK-LABEL: @test2a(
-; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, i32* [[P:%.*]], i64 [[X:%.*]]
-; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[X]], [[Y:%.*]]
-; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[CMP]], i32* [[GEP]], i32* [[P]]
+; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT: [[SELECT_IDX:%.*]] = select i1 [[CMP]], i64 [[X]], i64 0
+; CHECK-NEXT: [[SELECT:%.*]] = getelementptr inbounds i32, i32* [[P:%.*]], i64 [[SELECT_IDX]]
; CHECK-NEXT: ret i32* [[SELECT]]
;
%gep = getelementptr inbounds i32, i32* %p, i64 %x
@@ -88,9 +88,9 @@ define i32* @test2a(i32* %p, i64 %x, i64 %y) {
; PR50183
define i32* @test2b(i32* %p, i64 %x, i64 %y) {
; CHECK-LABEL: @test2b(
-; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, i32* [[P:%.*]], i64 [[X:%.*]]
-; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[X]], [[Y:%.*]]
-; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[CMP]], i32* [[P]], i32* [[GEP]]
+; CHECK-NEXT: [[CMP:%.*]] = icmp ugt i64 [[X:%.*]], [[Y:%.*]]
+; CHECK-NEXT: [[SELECT_IDX:%.*]] = select i1 [[CMP]], i64 0, i64 [[X]]
+; CHECK-NEXT: [[SELECT:%.*]] = getelementptr inbounds i32, i32* [[P:%.*]], i64 [[SELECT_IDX]]
; CHECK-NEXT: ret i32* [[SELECT]]
;
%gep = getelementptr inbounds i32, i32* %p, i64 %x
More information about the llvm-commits
mailing list