[RISCV] Handle zeroinitializer of vector tuple Type #113995
Merged
+79
−1
Conversation
This file contains hidden or 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
4vtomat
requested review from
asb,
preames,
kito-cheng,
topperc and
wangpc-pp
|
@llvm/pr-subscribers-backend-risc-v @llvm/pr-subscribers-llvm-selectiondag Author: Brandon Wu (4vtomat) ChangesIt doesn't make sense to add a new generic ISD to handle riscv tuple Note: If there's Full diff: https://github.com/llvm/llvm-project/pull/113995.diff 5 Files Affected:
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 1a86b3b51234d1..0ba0055f99c094 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -6288,6 +6288,9 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
return getNode(ISD::VECREDUCE_AND, DL, VT, N1);
break;
case ISD::SPLAT_VECTOR:
+ // RISC-V vector tuple type is not a vector type.
+ if (VT.isRISCVVectorTuple())
+ break;
assert(VT.isVector() && "Wrong return type!");
// FIXME: Hexagon uses i32 scalar for a floating point zero vector so allow
// that for now.
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index 8450553743074c..ca63b289ee20cb 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -1900,6 +1900,13 @@ SDValue SelectionDAGBuilder::getValueImpl(const Value *V) {
DAG.getConstant(0, getCurSDLoc(), MVT::nxv16i1));
}
+ if (VT.isRISCVVectorTuple()) {
+ assert(C->isNullValue() && "Can only zero this target type!");
+ return NodeMap[V] = DAG.getNode(
+ ISD::SPLAT_VECTOR, getCurSDLoc(), VT,
+ DAG.getConstant(0, getCurSDLoc(), MVT::getIntegerVT(8)));
+ }
+
VectorType *VecTy = cast<VectorType>(V->getType());
// Now that we know the number and type of the elements, get that number of
diff --git a/llvm/lib/IR/Type.cpp b/llvm/lib/IR/Type.cpp
index e311cde415174a..e5fe85dbd18e39 100644
--- a/llvm/lib/IR/Type.cpp
+++ b/llvm/lib/IR/Type.cpp
@@ -880,7 +880,8 @@ static TargetTypeInfo getTargetTypeInfo(const TargetExtType *Ty) {
RISCV::RVVBitsPerBlock / 8) *
Ty->getIntParameter(0);
return TargetTypeInfo(
- ScalableVectorType::get(Type::getInt8Ty(C), TotalNumElts));
+ ScalableVectorType::get(Type::getInt8Ty(C), TotalNumElts),
+ TargetExtType::HasZeroInit);
}
// DirectX resources
diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
index dc3f8254cb4e00..998f1e2188ff0b 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
@@ -66,6 +66,25 @@ void RISCVDAGToDAGISel::PreprocessISelDAG() {
VT.isInteger() ? RISCVISD::VMV_V_X_VL : RISCVISD::VFMV_V_F_VL;
SDLoc DL(N);
SDValue VL = CurDAG->getRegister(RISCV::X0, Subtarget->getXLenVT());
+
+ if (VT.isRISCVVectorTuple()) {
+ unsigned NF = VT.getRISCVVectorTupleNumFields();
+ unsigned NumScalElts = VT.getSizeInBits() / (NF * 8);
+ SDValue EltVal = CurDAG->getConstant(0, DL, Subtarget->getXLenVT());
+ MVT ScalTy =
+ MVT::getScalableVectorVT(MVT::getIntegerVT(8), NumScalElts);
+
+ SDValue Splat = CurDAG->getNode(RISCVISD::VMV_V_X_VL, DL, ScalTy,
+ CurDAG->getUNDEF(ScalTy), EltVal, VL);
+
+ Result = CurDAG->getUNDEF(VT);
+ for (unsigned i = 0; i < NF; ++i)
+ Result = CurDAG->getNode(RISCVISD::TUPLE_INSERT, DL, VT, Result,
+ Splat, CurDAG->getVectorIdxConstant(i, DL));
+
+ break;
+ }
+
SDValue Src = N->getOperand(0);
if (VT.isInteger())
Src = CurDAG->getNode(ISD::ANY_EXTEND, DL, Subtarget->getXLenVT(),
diff --git a/llvm/test/CodeGen/RISCV/vector-tuple-zeroinitializer.ll b/llvm/test/CodeGen/RISCV/vector-tuple-zeroinitializer.ll
new file mode 100644
index 00000000000000..88e1315d560b07
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/vector-tuple-zeroinitializer.ll
@@ -0,0 +1,40 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \
+; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK
+; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \
+; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK
+
+define target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @test_tuple_zero0() {
+; CHECK-LABEL: test_tuple_zero0:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vsetvli a0, zero, e8, m2, ta, ma
+; CHECK-NEXT: vmv.v.i v8, 0
+; CHECK-NEXT: vmv.v.i v10, 0
+; CHECK-NEXT: ret
+entry:
+ ret target("riscv.vector.tuple", <vscale x 16 x i8>, 2) zeroinitializer
+}
+
+define target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @test_tuple_zero1(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: test_tuple_zero1:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vsetvli a0, zero, e8, m2, ta, ma
+; CHECK-NEXT: vmv.v.i v10, 0
+; CHECK-NEXT: ret
+entry:
+ %1 = call target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv16i8_2t.nxv4i32(target("riscv.vector.tuple", <vscale x 16 x i8>, 2) zeroinitializer, <vscale x 4 x i32> %a, i32 0)
+ ret target("riscv.vector.tuple", <vscale x 16 x i8>, 2) %1
+}
+
+define target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @test_tuple_zero2(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: test_tuple_zero2:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vsetvli a0, zero, e8, m2, ta, ma
+; CHECK-NEXT: vmv.v.i v6, 0
+; CHECK-NEXT: vmv2r.v v10, v8
+; CHECK-NEXT: vmv2r.v v8, v6
+; CHECK-NEXT: ret
+entry:
+ %1 = call target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv16i8_2t.nxv4i32(target("riscv.vector.tuple", <vscale x 16 x i8>, 2) zeroinitializer, <vscale x 4 x i32> %a, i32 1)
+ ret target("riscv.vector.tuple", <vscale x 16 x i8>, 2) %1
+}
|
|
@llvm/pr-subscribers-llvm-ir Author: Brandon Wu (4vtomat) ChangesIt doesn't make sense to add a new generic ISD to handle riscv tuple Note: If there's Full diff: https://github.com/llvm/llvm-project/pull/113995.diff 5 Files Affected:
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 1a86b3b51234d1..0ba0055f99c094 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -6288,6 +6288,9 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
return getNode(ISD::VECREDUCE_AND, DL, VT, N1);
break;
case ISD::SPLAT_VECTOR:
+ // RISC-V vector tuple type is not a vector type.
+ if (VT.isRISCVVectorTuple())
+ break;
assert(VT.isVector() && "Wrong return type!");
// FIXME: Hexagon uses i32 scalar for a floating point zero vector so allow
// that for now.
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index 8450553743074c..ca63b289ee20cb 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -1900,6 +1900,13 @@ SDValue SelectionDAGBuilder::getValueImpl(const Value *V) {
DAG.getConstant(0, getCurSDLoc(), MVT::nxv16i1));
}
+ if (VT.isRISCVVectorTuple()) {
+ assert(C->isNullValue() && "Can only zero this target type!");
+ return NodeMap[V] = DAG.getNode(
+ ISD::SPLAT_VECTOR, getCurSDLoc(), VT,
+ DAG.getConstant(0, getCurSDLoc(), MVT::getIntegerVT(8)));
+ }
+
VectorType *VecTy = cast<VectorType>(V->getType());
// Now that we know the number and type of the elements, get that number of
diff --git a/llvm/lib/IR/Type.cpp b/llvm/lib/IR/Type.cpp
index e311cde415174a..e5fe85dbd18e39 100644
--- a/llvm/lib/IR/Type.cpp
+++ b/llvm/lib/IR/Type.cpp
@@ -880,7 +880,8 @@ static TargetTypeInfo getTargetTypeInfo(const TargetExtType *Ty) {
RISCV::RVVBitsPerBlock / 8) *
Ty->getIntParameter(0);
return TargetTypeInfo(
- ScalableVectorType::get(Type::getInt8Ty(C), TotalNumElts));
+ ScalableVectorType::get(Type::getInt8Ty(C), TotalNumElts),
+ TargetExtType::HasZeroInit);
}
// DirectX resources
diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
index dc3f8254cb4e00..998f1e2188ff0b 100644
--- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp
@@ -66,6 +66,25 @@ void RISCVDAGToDAGISel::PreprocessISelDAG() {
VT.isInteger() ? RISCVISD::VMV_V_X_VL : RISCVISD::VFMV_V_F_VL;
SDLoc DL(N);
SDValue VL = CurDAG->getRegister(RISCV::X0, Subtarget->getXLenVT());
+
+ if (VT.isRISCVVectorTuple()) {
+ unsigned NF = VT.getRISCVVectorTupleNumFields();
+ unsigned NumScalElts = VT.getSizeInBits() / (NF * 8);
+ SDValue EltVal = CurDAG->getConstant(0, DL, Subtarget->getXLenVT());
+ MVT ScalTy =
+ MVT::getScalableVectorVT(MVT::getIntegerVT(8), NumScalElts);
+
+ SDValue Splat = CurDAG->getNode(RISCVISD::VMV_V_X_VL, DL, ScalTy,
+ CurDAG->getUNDEF(ScalTy), EltVal, VL);
+
+ Result = CurDAG->getUNDEF(VT);
+ for (unsigned i = 0; i < NF; ++i)
+ Result = CurDAG->getNode(RISCVISD::TUPLE_INSERT, DL, VT, Result,
+ Splat, CurDAG->getVectorIdxConstant(i, DL));
+
+ break;
+ }
+
SDValue Src = N->getOperand(0);
if (VT.isInteger())
Src = CurDAG->getNode(ISD::ANY_EXTEND, DL, Subtarget->getXLenVT(),
diff --git a/llvm/test/CodeGen/RISCV/vector-tuple-zeroinitializer.ll b/llvm/test/CodeGen/RISCV/vector-tuple-zeroinitializer.ll
new file mode 100644
index 00000000000000..88e1315d560b07
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/vector-tuple-zeroinitializer.ll
@@ -0,0 +1,40 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \
+; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK
+; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \
+; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK
+
+define target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @test_tuple_zero0() {
+; CHECK-LABEL: test_tuple_zero0:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vsetvli a0, zero, e8, m2, ta, ma
+; CHECK-NEXT: vmv.v.i v8, 0
+; CHECK-NEXT: vmv.v.i v10, 0
+; CHECK-NEXT: ret
+entry:
+ ret target("riscv.vector.tuple", <vscale x 16 x i8>, 2) zeroinitializer
+}
+
+define target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @test_tuple_zero1(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: test_tuple_zero1:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vsetvli a0, zero, e8, m2, ta, ma
+; CHECK-NEXT: vmv.v.i v10, 0
+; CHECK-NEXT: ret
+entry:
+ %1 = call target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv16i8_2t.nxv4i32(target("riscv.vector.tuple", <vscale x 16 x i8>, 2) zeroinitializer, <vscale x 4 x i32> %a, i32 0)
+ ret target("riscv.vector.tuple", <vscale x 16 x i8>, 2) %1
+}
+
+define target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @test_tuple_zero2(<vscale x 4 x i32> %a) {
+; CHECK-LABEL: test_tuple_zero2:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vsetvli a0, zero, e8, m2, ta, ma
+; CHECK-NEXT: vmv.v.i v6, 0
+; CHECK-NEXT: vmv2r.v v10, v8
+; CHECK-NEXT: vmv2r.v v8, v6
+; CHECK-NEXT: ret
+entry:
+ %1 = call target("riscv.vector.tuple", <vscale x 16 x i8>, 2) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv16i8_2t.nxv4i32(target("riscv.vector.tuple", <vscale x 16 x i8>, 2) zeroinitializer, <vscale x 4 x i32> %a, i32 1)
+ ret target("riscv.vector.tuple", <vscale x 16 x i8>, 2) %1
+}
|
wangpc-pp
reviewed
Oct 29, 2024
Yeah, I think I added? Or do you mean we shouldn't add? |
Oh, never mind! Somehow the changes in |
wangpc-pp
reviewed
Oct 29, 2024
lukel97
reviewed
Oct 29, 2024
|
I'm clearly missing some context here, but... why do we need anything specific here to RISCVTupleType at all? At the IR level, why is this not simply a struct? I'd have assumed we already had the lowering for a first class aggregate, what am I missing here? |
It was a struct a couple months ago. Unfortunately, SelectionDAGBuilder decomposes structs into individual fields before call/return handling and before inline assembly handling. This made it very difficult to have an ABI for tuples or handle tuple operands in inline assembly. |
That was the context I was missing, thanks. |
4vtomat
force-pushed
the
riscv_tuple_zero_initializer
branch
from
87dd7c3 to
dbe42eb
Compare
|
I added a |
topperc
reviewed
Nov 19, 2024
4vtomat
force-pushed
the
riscv_tuple_zero_initializer
branch
from
dbe42eb to
663cd55
Compare
lukel97
reviewed
Nov 19, 2024
lukel97
reviewed
Nov 19, 2024
topperc
reviewed
Nov 26, 2024
It doesn't make sense to add a new generic ISD to handle riscv tuple type. Instead we use `SPLAT_VECTOR` for ISD and further lower to `VMV_V_X`. Note: If there's `visitSPLAT_VECTOR` in generic DAG combiner, it needs to skip riscv vector tuple type.
4vtomat
force-pushed
the
riscv_tuple_zero_initializer
branch
from
3fc09d4 to
5ffa3cb
Compare
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.
It doesn't make sense to add a new generic ISD to handle riscv tuple
type. Instead we use
SPLAT_VECTORfor ISD and further lower toVMV_V_X.Note: If there's
visitSPLAT_VECTORin generic DAG combiner, it needsto skip riscv vector tuple type.
Stack on #114329