[llvm] [LV][VPlan] Extract the implementation of transform Recipe to EVLRecipe into a small function. NFC (PR #119510)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Dec 20 19:05:25 PST 2024
https://github.com/LiqinWeng updated https://github.com/llvm/llvm-project/pull/119510
>From 7f27bc1618674f5240757e392917efc02869d31b Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Sat, 21 Dec 2024 10:28:41 +0800
Subject: [PATCH 1/2] [LV][VPlan] Use opcode to retrieve the VPID of the
CallRecipe, rather than underlying instruction.
This patch may cause the flags in the CallRecipe to be lost after EVL transformation, and it has been addressed in the patch: #119847
---
llvm/lib/Transforms/Vectorize/VPlan.h | 3 ++
.../Transforms/Vectorize/VPlanTransforms.cpp | 21 ++++++-----
...vectorize-force-tail-with-evl-reduction.ll | 2 +-
.../RISCV/vplan-vp-call-intrinsics.ll | 36 +++++++++----------
4 files changed, 32 insertions(+), 30 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index d26995883cb905..bfc5b90b804306 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1688,6 +1688,9 @@ class VPWidenIntrinsicRecipe : public VPRecipeWithIRFlags {
InstructionCost computeCost(ElementCount VF,
VPCostContext &Ctx) const override;
+ /// Return the ID of the intrinsic.
+ Intrinsic::ID getVectorIntrinsicID() const { return VectorIntrinsicID; }
+
/// Return the scalar return type of the intrinsic.
Type *getResultType() const { return ResultTy; }
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index aacb27f9325d07..76618e2c8f150a 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1487,24 +1487,23 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
return new VPReductionEVLRecipe(*Red, EVL, NewMask);
})
.Case<VPWidenIntrinsicRecipe>(
- [&](VPWidenIntrinsicRecipe *CInst) -> VPRecipeBase * {
- auto *CI = cast<CallInst>(CInst->getUnderlyingInstr());
+ [&](VPWidenIntrinsicRecipe *CallR) -> VPRecipeBase * {
Intrinsic::ID VPID = VPIntrinsic::getForIntrinsic(
- CI->getCalledFunction()->getIntrinsicID());
- if (VPID == Intrinsic::not_intrinsic)
- return nullptr;
-
- SmallVector<VPValue *> Ops(CInst->operands());
+ CallR->getVectorIntrinsicID());
+ assert(VPID != Intrinsic::not_intrinsic &&
+ "Expected vp.casts Instrinsic");
assert(VPIntrinsic::getMaskParamPos(VPID) &&
VPIntrinsic::getVectorLengthParamPos(VPID) &&
"Expected VP intrinsic");
- VPValue *Mask = Plan.getOrAddLiveIn(ConstantInt::getTrue(
- IntegerType::getInt1Ty(CI->getContext())));
+
+ SmallVector<VPValue *> Ops(CallR->operands());
+ VPValue *Mask =
+ Plan.getOrAddLiveIn(ConstantInt::getTrue(Ctx));
Ops.push_back(Mask);
Ops.push_back(&EVL);
return new VPWidenIntrinsicRecipe(
- *CI, VPID, Ops, TypeInfo.inferScalarType(CInst),
- CInst->getDebugLoc());
+ VPID, Ops, TypeInfo.inferScalarType(CallR),
+ CallR->getDebugLoc());
})
.Case<VPWidenCastRecipe>(
[&](VPWidenCastRecipe *CastR) -> VPRecipeBase * {
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll
index af5a62e5f480d6..7557c10892d6d4 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll
@@ -1727,7 +1727,7 @@ define float @fmuladd(ptr %a, ptr %b, i64 %n, float %start) {
; IF-EVL-NEXT: [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[TMP11]]
; IF-EVL-NEXT: [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
; IF-EVL-NEXT: [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
-; IF-EVL-NEXT: [[TMP16:%.*]] = call reassoc <vscale x 4 x float> @llvm.vp.fmuladd.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VP_OP_LOAD1]], <vscale x 4 x float> [[VEC_PHI]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT: [[TMP16:%.*]] = call <vscale x 4 x float> @llvm.vp.fmuladd.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VP_OP_LOAD1]], <vscale x 4 x float> [[VEC_PHI]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
; IF-EVL-NEXT: [[TMP17]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x float> [[TMP16]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP10]])
; IF-EVL-NEXT: [[TMP18:%.*]] = zext i32 [[TMP10]] to i64
; IF-EVL-NEXT: [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
index 5c47f07f4436c3..05c687a2053533 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
@@ -27,10 +27,10 @@ define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) {
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%c>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
; IF-EVL-NEXT: WIDEN ir<[[LD2:%.+]]> = vp.load vp<[[PTR2]]>, vp<[[EVL]]>
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[SMAX:%.+]]> = call llvm.vp.smax(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[SMAX:%.+]]> = call llvm.vp.smax(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP3:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR3:%[0-9]+]]> = vector-pointer ir<[[GEP3]]>
-; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, ir<[[SMAX]]>, vp<[[EVL]]>
+; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, vp<[[SMAX]]>, vp<[[EVL]]>
; IF-EVL-NEXT: SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -80,10 +80,10 @@ define void @vp_smin(ptr %a, ptr %b, ptr %c, i64 %N) {
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%c>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
; IF-EVL-NEXT: WIDEN ir<[[LD2:%.+]]> = vp.load vp<[[PTR2]]>, vp<[[EVL]]>
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[SMIN:%.+]]> = call llvm.vp.smin(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[SMIN:%.+]]> = call llvm.vp.smin(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP3:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR3:%[0-9]+]]> = vector-pointer ir<[[GEP3]]>
-; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, ir<[[SMIN]]>, vp<[[EVL]]>
+; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, vp<[[SMIN]]>, vp<[[EVL]]>
; IF-EVL-NEXT: SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -133,10 +133,10 @@ define void @vp_umax(ptr %a, ptr %b, ptr %c, i64 %N) {
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%c>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
; IF-EVL-NEXT: WIDEN ir<[[LD2:%.+]]> = vp.load vp<[[PTR2]]>, vp<[[EVL]]>
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[UMAX:%.+]]> = call llvm.vp.umax(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[UMAX:%.+]]> = call llvm.vp.umax(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP3:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR3:%[0-9]+]]> = vector-pointer ir<[[GEP3]]>
-; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, ir<[[UMAX]]>, vp<[[EVL]]>
+; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, vp<[[UMAX]]>, vp<[[EVL]]>
; IF-EVL-NEXT: SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -186,10 +186,10 @@ define void @vp_umin(ptr %a, ptr %b, ptr %c, i64 %N) {
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%c>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
; IF-EVL-NEXT: WIDEN ir<[[LD2:%.+]]> = vp.load vp<[[PTR2]]>, vp<[[EVL]]>
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[UMIN:%.+]]> = call llvm.vp.umin(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[UMIN:%.+]]> = call llvm.vp.umin(ir<[[LD1]]>, ir<[[LD2]]>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP3:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR3:%[0-9]+]]> = vector-pointer ir<[[GEP3]]>
-; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, ir<[[UMIN]]>, vp<[[EVL]]>
+; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR3]]>, vp<[[UMIN]]>, vp<[[EVL]]>
; IF-EVL-NEXT: SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -236,10 +236,10 @@ define void @vp_ctlz(ptr %a, ptr %b, i64 %N) {
; IF-EVL-NEXT: CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
; IF-EVL-NEXT: WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[CTLZ:%.+]]> = call llvm.vp.ctlz(ir<[[LD1]]>, ir<true>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[CTLZ:%.+]]> = call llvm.vp.ctlz(ir<[[LD1]]>, ir<true>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, ir<[[CTLZ]]>, vp<[[EVL]]>
+; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, vp<[[CTLZ]]>, vp<[[EVL]]>
; IF-EVL-NEXT: SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -284,10 +284,10 @@ define void @vp_cttz(ptr %a, ptr %b, i64 %N) {
; IF-EVL-NEXT: CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
; IF-EVL-NEXT: WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[CTTZ:%.+]]> = call llvm.vp.cttz(ir<[[LD1]]>, ir<true>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[CTTZ:%.+]]> = call llvm.vp.cttz(ir<[[LD1]]>, ir<true>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, ir<[[CTTZ]]>, vp<[[EVL]]>
+; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, vp<[[CTTZ]]>, vp<[[EVL]]>
; IF-EVL-NEXT: SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
@@ -333,8 +333,8 @@ define void @vp_lrint(ptr %a, ptr %b, i64 %N) {
; IF-EVL-NEXT: vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
; IF-EVL-NEXT: WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[LRINT:%.+]]> = call llvm.vp.lrint(vp<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
-; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LRINT]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[LRINT:%.+]]> = call llvm.vp.lrint(vp<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(vp<[[LRINT]]>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, vp<[[TRUNC]]>, vp<[[EVL]]>
@@ -385,8 +385,8 @@ define void @vp_llrint(ptr %a, ptr %b, i64 %N) {
; IF-EVL-NEXT: vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
; IF-EVL-NEXT: WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[FPEXT:%.+]]> = call llvm.vp.fpext(ir<[[LD1]]>, ir<true>, vp<[[EVL]]>)
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[LLRINT:%.+]]> = call llvm.vp.llrint(vp<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
-; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(ir<[[LLRINT]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[LLRINT:%.+]]> = call llvm.vp.llrint(vp<[[FPEXT]]>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[TRUNC:%.+]]> = call llvm.vp.trunc(vp<[[LLRINT]]>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, vp<[[TRUNC]]>, vp<[[EVL]]>
@@ -436,10 +436,10 @@ define void @vp_abs(ptr %a, ptr %b, i64 %N) {
; IF-EVL-NEXT: CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
; IF-EVL-NEXT: WIDEN ir<[[LD1:%.+]]> = vp.load vp<[[PTR1]]>, vp<[[EVL]]>
-; IF-EVL-NEXT: WIDEN-INTRINSIC ir<[[ABS:%.+]]> = call llvm.vp.abs(ir<[[LD1]]>, ir<true>, ir<true>, vp<[[EVL]]>)
+; IF-EVL-NEXT: WIDEN-INTRINSIC vp<[[ABS:%.+]]> = call llvm.vp.abs(ir<[[LD1]]>, ir<true>, ir<true>, vp<[[EVL]]>)
; IF-EVL-NEXT: CLONE ir<[[GEP2:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
; IF-EVL-NEXT: vp<[[PTR2:%[0-9]+]]> = vector-pointer ir<[[GEP2]]>
-; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, ir<[[ABS]]>, vp<[[EVL]]>
+; IF-EVL-NEXT: WIDEN vp.store vp<[[PTR2]]>, vp<[[ABS]]>, vp<[[EVL]]>
; IF-EVL-NEXT: SCALAR-CAST vp<[[CAST:%[0-9]+]]> = zext vp<[[EVL]]> to i64
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT]]> = add vp<[[CAST]]>, vp<[[EVL_PHI]]>
; IF-EVL-NEXT: EMIT vp<[[IV_NEXT_EXIT:%.+]]> = add vp<[[IV]]>, vp<[[VFUF]]>
>From d1279c227f8651ebfc8b6a48e6fe2e40959a4f7e Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Wed, 11 Dec 2024 13:58:41 +0800
Subject: [PATCH 2/2] [LV][VPlan] Extract the implementation of transform
Recipe to EVLRecipe into a small function
---
.../Transforms/Vectorize/VPlanTransforms.cpp | 186 +++++++++---------
1 file changed, 89 insertions(+), 97 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 76618e2c8f150a..4e4be48749d930 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1442,13 +1442,93 @@ void VPlanTransforms::addActiveLaneMask(
HeaderMask->replaceAllUsesWith(LaneMask);
}
+// Try to convert \p CurRecipe to a corresponding EVL-based recipe. Returns
+// nullptr if no EVL-based recipe could be created.
+// \p HeaderMask Header Mask.
+// \p CurRecipe Recipe to be transform.
+// \p TypeInfo VPlan-based type analysis.
+// \p AllOneMask The vector mask parameter of vector-predication intrinsics.
+// \p EVL The explicit vector length parameter of vector-predication
+// intrinsics.
+static VPRecipeBase *createEVLRecipe(VPValue *HeaderMask,
+ VPRecipeBase &CurRecipe,
+ VPTypeAnalysis &TypeInfo,
+ VPValue &AllOneMask, VPValue &EVL) {
+ using namespace llvm::VPlanPatternMatch;
+ auto GetNewMask = [&](VPValue *OrigMask) -> VPValue * {
+ assert(OrigMask && "Unmasked recipe when folding tail");
+ return HeaderMask == OrigMask ? nullptr : OrigMask;
+ };
+
+ return TypeSwitch<VPRecipeBase *, VPRecipeBase *>(&CurRecipe)
+ .Case<VPWidenLoadRecipe>([&](VPWidenLoadRecipe *L) {
+ VPValue *NewMask = GetNewMask(L->getMask());
+ return new VPWidenLoadEVLRecipe(*L, EVL, NewMask);
+ })
+ .Case<VPWidenStoreRecipe>([&](VPWidenStoreRecipe *S) {
+ VPValue *NewMask = GetNewMask(S->getMask());
+ return new VPWidenStoreEVLRecipe(*S, EVL, NewMask);
+ })
+ .Case<VPWidenRecipe>([&](VPWidenRecipe *W) -> VPRecipeBase * {
+ unsigned Opcode = W->getOpcode();
+ if (!Instruction::isBinaryOp(Opcode) && !Instruction::isUnaryOp(Opcode))
+ return nullptr;
+ return new VPWidenEVLRecipe(*W, EVL);
+ })
+ .Case<VPReductionRecipe>([&](VPReductionRecipe *Red) {
+ VPValue *NewMask = GetNewMask(Red->getCondOp());
+ return new VPReductionEVLRecipe(*Red, EVL, NewMask);
+ })
+ .Case<VPWidenIntrinsicRecipe, VPWidenCastRecipe>(
+ [&](auto *CR) -> VPRecipeBase * {
+ Intrinsic::ID VPID;
+ if (auto *CallR = dyn_cast<VPWidenIntrinsicRecipe>(CR))
+ VPID =
+ VPIntrinsic::getForIntrinsic(CallR->getVectorIntrinsicID());
+ else if (auto *CastR = dyn_cast<VPWidenCastRecipe>(CR))
+ VPID = VPIntrinsic::getForOpcode(CastR->getOpcode());
+ assert(VPID != Intrinsic::not_intrinsic && "Expected VP intrinsic");
+ assert(VPIntrinsic::getMaskParamPos(VPID) &&
+ VPIntrinsic::getVectorLengthParamPos(VPID) &&
+ "Expected VP intrinsic");
+
+ SmallVector<VPValue *> Ops(CR->operands());
+ Ops.push_back(&AllOneMask);
+ Ops.push_back(&EVL);
+ return new VPWidenIntrinsicRecipe(
+ VPID, Ops, TypeInfo.inferScalarType(CR), CR->getDebugLoc());
+ })
+ .Case<VPWidenSelectRecipe>([&](VPWidenSelectRecipe *Sel) {
+ SmallVector<VPValue *> Ops(Sel->operands());
+ Ops.push_back(&EVL);
+ return new VPWidenIntrinsicRecipe(Intrinsic::vp_select, Ops,
+ TypeInfo.inferScalarType(Sel),
+ Sel->getDebugLoc());
+ })
+ .Case<VPInstruction>([&](VPInstruction *VPI) -> VPRecipeBase * {
+ VPValue *LHS, *RHS;
+ // Transform select with a header mask condition
+ // select(header_mask, LHS, RHS)
+ // into vector predication merge.
+ // vp.merge(all-true, LHS, RHS, EVL)
+ if (!match(VPI, m_Select(m_Specific(HeaderMask), m_VPValue(LHS),
+ m_VPValue(RHS))))
+ return nullptr;
+ // Use all true as the condition because this transformation is
+ // limited to selects whose condition is a header mask.
+ return new VPWidenIntrinsicRecipe(
+ Intrinsic::vp_merge, {&AllOneMask, LHS, RHS, &EVL},
+ TypeInfo.inferScalarType(LHS), VPI->getDebugLoc());
+ })
+ .Default([&](VPRecipeBase *R) { return nullptr; });
+}
+
/// Replace recipes with their EVL variants.
static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
- using namespace llvm::VPlanPatternMatch;
Type *CanonicalIVType = Plan.getCanonicalIV()->getScalarType();
VPTypeAnalysis TypeInfo(CanonicalIVType);
LLVMContext &Ctx = CanonicalIVType->getContext();
- SmallVector<VPValue *> HeaderMasks = collectAllHeaderMasks(Plan);
+ VPValue *AllOneMask = Plan.getOrAddLiveIn(ConstantInt::getTrue(Ctx));
for (VPUser *U : Plan.getVF().users()) {
if (auto *R = dyn_cast<VPReverseVectorPointerRecipe>(U))
@@ -1460,110 +1540,22 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
for (VPValue *HeaderMask : collectAllHeaderMasks(Plan)) {
for (VPUser *U : collectUsersRecursively(HeaderMask)) {
auto *CurRecipe = cast<VPRecipeBase>(U);
- auto GetNewMask = [&](VPValue *OrigMask) -> VPValue * {
- assert(OrigMask && "Unmasked recipe when folding tail");
- return HeaderMask == OrigMask ? nullptr : OrigMask;
- };
-
- VPRecipeBase *NewRecipe =
- TypeSwitch<VPRecipeBase *, VPRecipeBase *>(CurRecipe)
- .Case<VPWidenLoadRecipe>([&](VPWidenLoadRecipe *L) {
- VPValue *NewMask = GetNewMask(L->getMask());
- return new VPWidenLoadEVLRecipe(*L, EVL, NewMask);
- })
- .Case<VPWidenStoreRecipe>([&](VPWidenStoreRecipe *S) {
- VPValue *NewMask = GetNewMask(S->getMask());
- return new VPWidenStoreEVLRecipe(*S, EVL, NewMask);
- })
- .Case<VPWidenRecipe>([&](VPWidenRecipe *W) -> VPRecipeBase * {
- unsigned Opcode = W->getOpcode();
- if (!Instruction::isBinaryOp(Opcode) &&
- !Instruction::isUnaryOp(Opcode))
- return nullptr;
- return new VPWidenEVLRecipe(*W, EVL);
- })
- .Case<VPReductionRecipe>([&](VPReductionRecipe *Red) {
- VPValue *NewMask = GetNewMask(Red->getCondOp());
- return new VPReductionEVLRecipe(*Red, EVL, NewMask);
- })
- .Case<VPWidenIntrinsicRecipe>(
- [&](VPWidenIntrinsicRecipe *CallR) -> VPRecipeBase * {
- Intrinsic::ID VPID = VPIntrinsic::getForIntrinsic(
- CallR->getVectorIntrinsicID());
- assert(VPID != Intrinsic::not_intrinsic &&
- "Expected vp.casts Instrinsic");
- assert(VPIntrinsic::getMaskParamPos(VPID) &&
- VPIntrinsic::getVectorLengthParamPos(VPID) &&
- "Expected VP intrinsic");
-
- SmallVector<VPValue *> Ops(CallR->operands());
- VPValue *Mask =
- Plan.getOrAddLiveIn(ConstantInt::getTrue(Ctx));
- Ops.push_back(Mask);
- Ops.push_back(&EVL);
- return new VPWidenIntrinsicRecipe(
- VPID, Ops, TypeInfo.inferScalarType(CallR),
- CallR->getDebugLoc());
- })
- .Case<VPWidenCastRecipe>(
- [&](VPWidenCastRecipe *CastR) -> VPRecipeBase * {
- Intrinsic::ID VPID =
- VPIntrinsic::getForOpcode(CastR->getOpcode());
- assert(VPID != Intrinsic::not_intrinsic &&
- "Expected vp.casts Instrinsic");
-
- SmallVector<VPValue *> Ops(CastR->operands());
- assert(VPIntrinsic::getMaskParamPos(VPID) &&
- VPIntrinsic::getVectorLengthParamPos(VPID) &&
- "Expected VP intrinsic");
- VPValue *Mask =
- Plan.getOrAddLiveIn(ConstantInt::getTrue(Ctx));
- Ops.push_back(Mask);
- Ops.push_back(&EVL);
- return new VPWidenIntrinsicRecipe(
- VPID, Ops, TypeInfo.inferScalarType(CastR),
- CastR->getDebugLoc());
- })
- .Case<VPWidenSelectRecipe>([&](VPWidenSelectRecipe *Sel) {
- SmallVector<VPValue *> Ops(Sel->operands());
- Ops.push_back(&EVL);
- return new VPWidenIntrinsicRecipe(Intrinsic::vp_select, Ops,
- TypeInfo.inferScalarType(Sel),
- Sel->getDebugLoc());
- })
- .Case<VPInstruction>([&](VPInstruction *VPI) -> VPRecipeBase * {
- VPValue *LHS, *RHS;
- // Transform select with a header mask condition
- // select(header_mask, LHS, RHS)
- // into vector predication merge.
- // vp.merge(all-true, LHS, RHS, EVL)
- if (!match(VPI, m_Select(m_Specific(HeaderMask), m_VPValue(LHS),
- m_VPValue(RHS))))
- return nullptr;
- // Use all true as the condition because this transformation is
- // limited to selects whose condition is a header mask.
- VPValue *AllTrue =
- Plan.getOrAddLiveIn(ConstantInt::getTrue(Ctx));
- return new VPWidenIntrinsicRecipe(
- Intrinsic::vp_merge, {AllTrue, LHS, RHS, &EVL},
- TypeInfo.inferScalarType(LHS), VPI->getDebugLoc());
- })
- .Default([&](VPRecipeBase *R) { return nullptr; });
-
- if (!NewRecipe)
+ VPRecipeBase *EVLRecipe =
+ createEVLRecipe(HeaderMask, *CurRecipe, TypeInfo, *AllOneMask, EVL);
+ if (!EVLRecipe)
continue;
- [[maybe_unused]] unsigned NumDefVal = NewRecipe->getNumDefinedValues();
+ [[maybe_unused]] unsigned NumDefVal = EVLRecipe->getNumDefinedValues();
assert(NumDefVal == CurRecipe->getNumDefinedValues() &&
"New recipe must define the same number of values as the "
"original.");
assert(
NumDefVal <= 1 &&
"Only supports recipes with a single definition or without users.");
- NewRecipe->insertBefore(CurRecipe);
- if (isa<VPSingleDefRecipe, VPWidenLoadEVLRecipe>(NewRecipe)) {
+ EVLRecipe->insertBefore(CurRecipe);
+ if (isa<VPSingleDefRecipe, VPWidenLoadEVLRecipe>(EVLRecipe)) {
VPValue *CurVPV = CurRecipe->getVPSingleValue();
- CurVPV->replaceAllUsesWith(NewRecipe->getVPSingleValue());
+ CurVPV->replaceAllUsesWith(EVLRecipe->getVPSingleValue());
}
// Defer erasing recipes till the end so that we don't invalidate the
// VPTypeAnalysis cache.
More information about the llvm-commits
mailing list