[LLVM][SelectionDAG] Ensure Constant[FP]SDnode only store references to scalar Constant{Int,FP}. #111005
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…to scalar Constant{Int,FP}.
This fixes a failure path when the use-constant-##-for-###-splat
IR options are enabled.
|
@llvm/pr-subscribers-backend-aarch64 @llvm/pr-subscribers-llvm-selectiondag Author: Paul Walker (paulwalker-arm) ChangesThis fixes a failure path when the use-constant-##-for-###-splat IR options are enabled. Full diff: https://github.com/llvm/llvm-project/pull/111005.diff 4 Files Affected:
diff --git a/llvm/include/llvm/CodeGen/SelectionDAGNodes.h b/llvm/include/llvm/CodeGen/SelectionDAGNodes.h
index 6067b3b29ea181..639e9311977502 100644
--- a/llvm/include/llvm/CodeGen/SelectionDAGNodes.h
+++ b/llvm/include/llvm/CodeGen/SelectionDAGNodes.h
@@ -1678,6 +1678,7 @@ class ConstantSDNode : public SDNode {
: SDNode(isTarget ? ISD::TargetConstant : ISD::Constant, 0, DebugLoc(),
VTs),
Value(val) {
+ assert(!isa<VectorType>(val->getType()) && "Unexpected vector type!");
ConstantSDNodeBits.IsOpaque = isOpaque;
}
@@ -1730,7 +1731,9 @@ class ConstantFPSDNode : public SDNode {
ConstantFPSDNode(bool isTarget, const ConstantFP *val, SDVTList VTs)
: SDNode(isTarget ? ISD::TargetConstantFP : ISD::ConstantFP, 0,
DebugLoc(), VTs),
- Value(val) {}
+ Value(val) {
+ assert(!isa<VectorType>(val->getType()) && "Unexpected vector type!");
+ }
public:
const APFloat& getValueAPF() const { return Value->getValueAPF(); }
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 498debf0955980..016d6990774549 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -1656,6 +1656,11 @@ SDValue SelectionDAG::getConstant(const ConstantInt &Val, const SDLoc &DL,
EVT EltVT = VT.getScalarType();
const ConstantInt *Elt = &Val;
+ // Vector splats are explicit within the DAG, with ConstantSDNode holding the
+ // to be splatted scalar ConstantInt.
+ if (isa<VectorType>(Elt->getType()))
+ Elt = ConstantInt::get(*getContext(), Elt->getValue());
+
// In some cases the vector type is legal but the element type is illegal and
// needs to be promoted, for example v8i8 on ARM. In this case, promote the
// inserted value (the type does not need to match the vector element type).
@@ -1809,6 +1814,12 @@ SDValue SelectionDAG::getConstantFP(const ConstantFP &V, const SDLoc &DL,
assert(VT.isFloatingPoint() && "Cannot create integer FP constant!");
EVT EltVT = VT.getScalarType();
+ const ConstantFP *Elt = &V;
+
+ // Vector splats are explicit within the DAG, with ConstantFPSDNode holding
+ // the to be splatted scalar ConstantFP.
+ if (isa<VectorType>(Elt->getType()))
+ Elt = ConstantFP::get(*getContext(), Elt->getValue());
// Do the map lookup using the actual bit pattern for the floating point
// value, so that we don't have problems with 0.0 comparing equal to -0.0, and
@@ -1817,7 +1828,7 @@ SDValue SelectionDAG::getConstantFP(const ConstantFP &V, const SDLoc &DL,
SDVTList VTs = getVTList(EltVT);
FoldingSetNodeID ID;
AddNodeIDNode(ID, Opc, VTs, {});
- ID.AddPointer(&V);
+ ID.AddPointer(Elt);
void *IP = nullptr;
SDNode *N = nullptr;
if ((N = FindNodeOrInsertPos(ID, DL, IP)))
@@ -1825,7 +1836,7 @@ SDValue SelectionDAG::getConstantFP(const ConstantFP &V, const SDLoc &DL,
return SDValue(N, 0);
if (!N) {
- N = newSDNode<ConstantFPSDNode>(isTarget, &V, VTs);
+ N = newSDNode<ConstantFPSDNode>(isTarget, Elt, VTs);
CSEMap.InsertNode(N, IP);
InsertNode(N);
}
diff --git a/llvm/test/CodeGen/AArch64/sve-fp-immediates-merging.ll b/llvm/test/CodeGen/AArch64/sve-fp-immediates-merging.ll
index e1d883b0e78997..905d110e001c85 100644
--- a/llvm/test/CodeGen/AArch64/sve-fp-immediates-merging.ll
+++ b/llvm/test/CodeGen/AArch64/sve-fp-immediates-merging.ll
@@ -1,5 +1,6 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s | FileCheck %s
+; RUN: llc -use-constant-fp-for-scalable-splat < %s | FileCheck %s
target triple = "aarch64-unknown-linux-gnu"
diff --git a/llvm/test/CodeGen/AArch64/sve-int-imm.ll b/llvm/test/CodeGen/AArch64/sve-int-imm.ll
index 4be1abe8420081..47f4f0181dfbdf 100644
--- a/llvm/test/CodeGen/AArch64/sve-int-imm.ll
+++ b/llvm/test/CodeGen/AArch64/sve-int-imm.ll
@@ -1,4 +1,5 @@
; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve < %s | FileCheck %s
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve -use-constant-int-for-scalable-splat < %s | FileCheck %s
;
; SVE Arith Vector Immediate Unpredicated CodeGen
|
arsenm
reviewed
Oct 3, 2024
david-arm
reviewed
Oct 10, 2024
| @@ -1678,6 +1678,7 @@ class ConstantSDNode : public SDNode { | |||
| : SDNode(isTarget ? ISD::TargetConstant : ISD::Constant, 0, DebugLoc(), | |||
| VTs), | |||
| Value(val) { | |||
| assert(!isa<VectorType>(val->getType()) && "Unexpected vector type!"); | |||
There was a problem hiding this comment.
This is just a thought, but is it worth tying these asserts down even further, i.e.
assert(val->getType()->isIntegerTy() && "Unexpected type!");
and similarly for the FP node? At the moment, these nodes still won't assert for any other unusual types.
There was a problem hiding this comment.
That's handled as part of ConstantInt construction and thus here we can safely assume val has a valid type for its class.
| @@ -1656,6 +1656,11 @@ SDValue SelectionDAG::getConstant(const ConstantInt &Val, const SDLoc &DL, | |||
| EVT EltVT = VT.getScalarType(); | |||
| const ConstantInt *Elt = &Val; | |||
|
|
|||
| // Vector splats are explicit within the DAG, with ConstantSDNode holding the | |||
| // to be splatted scalar ConstantInt. | |||
david-arm
approved these changes
Oct 11, 2024
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/73/builds/7074 Here is the relevant piece of the build log for the reference |
DanielCChen
pushed a commit
to DanielCChen/llvm-project
that referenced
this pull request
Oct 16, 2024
…to scalar Constant{Int,FP}. (llvm#111005)
This fixes a failure path when the use-constant-##-for-###-splat IR
options are enabled.
bricknerb
pushed a commit
to bricknerb/llvm-project
that referenced
this pull request
Oct 17, 2024
…to scalar Constant{Int,FP}. (llvm#111005)
This fixes a failure path when the use-constant-##-for-###-splat IR
options are enabled.
EricWF
pushed a commit
to efcs/llvm-project
that referenced
this pull request
Oct 22, 2024
…to scalar Constant{Int,FP}. (llvm#111005)
This fixes a failure path when the use-constant-##-for-###-splat IR
options are enabled.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This fixes a failure path when the use-constant-##-for-###-splat IR options are enabled.