[llvm] [LV][EVL]Explicitly model AVL as sub, original TC, EVL_PHI. (PR #108869)

[Alexey Bataev via llvm-commits]

https://github.com/alexey-bataev updated https://github.com/llvm/llvm-project/pull/108869

>From 40887db66d4c2519808ab99510914cea04e82d90 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev at outlook.com>
Date: Mon, 16 Sep 2024 18:41:18 +0000
Subject: [PATCH] =?UTF-8?q?[=F0=9D=98=80=F0=9D=97=BD=F0=9D=97=BF]=20initia?=
 =?UTF-8?q?l=20version?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Created using spr 1.3.5
---
 llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp      |  5 +----
 llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp   | 13 +++++++++----
 .../RISCV/vplan-vp-intrinsics-reduction.ll          |  3 ++-
 .../LoopVectorize/RISCV/vplan-vp-intrinsics.ll      |  3 ++-
 4 files changed, 14 insertions(+), 10 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 9068ccf519c55c..24468643f8d47c 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -484,10 +484,7 @@ Value *VPInstruction::generatePerPart(VPTransformState &State, unsigned Part) {
     // TODO: Restructure this code with an explicit remainder loop, vsetvli can
     // be outside of the main loop.
     assert(Part == 0 && "No unrolling expected for predicated vectorization.");
-    // Compute VTC - IV as the AVL (requested vector length).
-    Value *Index = State.get(getOperand(0), VPIteration(0, 0));
-    Value *TripCount = State.get(getOperand(1), VPIteration(0, 0));
-    Value *AVL = State.Builder.CreateSub(TripCount, Index);
+    Value *AVL = State.get(getOperand(0), VPIteration(0, 0));
     Value *EVL = GetEVL(State, AVL);
     return EVL;
   }
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 1d84550010017f..fe18ce858594f7 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -1387,7 +1387,8 @@ static void transformRecipestoEVLRecipes(VPlan &Plan, VPValue &EVL) {
 /// ...
 /// %EVLPhi = EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI [ %StartV, %vector.ph ],
 ///                                               [ %NextEVLIV, %vector.body ]
-/// %VPEVL = EXPLICIT-VECTOR-LENGTH %EVLPhi, original TC
+/// %AVL = sub original TC, %EVLPhi
+/// %VPEVL = EXPLICIT-VECTOR-LENGTH %AVL
 /// ...
 /// %NextEVLIV = add IVSize (cast i32 %VPEVVL to IVSize), %EVLPhi
 /// ...
@@ -1417,9 +1418,13 @@ bool VPlanTransforms::tryAddExplicitVectorLength(VPlan &Plan) {
   // Create the ExplicitVectorLengthPhi recipe in the main loop.
   auto *EVLPhi = new VPEVLBasedIVPHIRecipe(StartV, DebugLoc());
   EVLPhi->insertAfter(CanonicalIVPHI);
-  auto *VPEVL = new VPInstruction(VPInstruction::ExplicitVectorLength,
-                                  {EVLPhi, Plan.getTripCount()});
-  VPEVL->insertBefore(*Header, Header->getFirstNonPhi());
+  // Compute original TC - IV as the AVL (requested vector length).
+  auto *AVL = new VPInstruction(Instruction::Sub, {Plan.getTripCount(), EVLPhi},
+                                DebugLoc(), "avl");
+  AVL->insertBefore(*Header, Header->getFirstNonPhi());
+  auto *VPEVL =
+      new VPInstruction(VPInstruction::ExplicitVectorLength, AVL, DebugLoc());
+  VPEVL->insertAfter(AVL);
 
   auto *CanonicalIVIncrement =
       cast<VPInstruction>(CanonicalIVPHI->getBackedgeValue());
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll
index 11405a1c911584..90c209cf3f5186 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll
@@ -39,7 +39,8 @@ define i32 @reduction(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-INLOOP-NEXT:    EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
 ; IF-EVL-INLOOP-NEXT:    EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]> = phi ir<0>, vp<[[IV_NEXT:%[0-9]+]]>
 ; IF-EVL-INLOOP-NEXT:    WIDEN-REDUCTION-PHI ir<[[RDX_PHI:%.+]]> = phi ir<%start>, ir<[[RDX_NEXT:%.+]]>
-; IF-EVL-INLOOP-NEXT:    EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[EVL_PHI]]>, ir<%n>
+; IF-EVL-INLOOP-NEXT:    EMIT vp<[[AVL:%.+]]> = sub ir<%n>, vp<[[EVL_PHI]]>
+; IF-EVL-INLOOP-NEXT:    EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
 ; IF-EVL-INLOOP-NEXT:    vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
 ; IF-EVL-INLOOP-NEXT:    CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%a>, vp<[[ST]]>
 ; IF-EVL-INLOOP-NEXT:    vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll
index 04b3ba52cbefc6..26ce02e44ea437 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll
@@ -23,7 +23,8 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; IF-EVL-NEXT:  vector.body:
 ; IF-EVL-NEXT:    EMIT vp<[[IV:%[0-9]+]]> = CANONICAL-INDUCTION
 ; IF-EVL-NEXT:    EXPLICIT-VECTOR-LENGTH-BASED-IV-PHI vp<[[EVL_PHI:%[0-9]+]]> = phi ir<0>, vp<[[IV_NEXT:%[0-9]+]]>
-; IF-EVL-NEXT:    EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[EVL_PHI]]>, ir<%N>
+; IF-EVL-NEXT:    EMIT vp<[[AVL:%.+]]> = sub ir<%N>, vp<[[EVL_PHI]]>
+; IF-EVL-NEXT:    EMIT vp<[[EVL:%.+]]> = EXPLICIT-VECTOR-LENGTH vp<[[AVL]]>
 ; IF-EVL-NEXT:    vp<[[ST:%[0-9]+]]> = SCALAR-STEPS vp<[[EVL_PHI]]>, ir<1>
 ; IF-EVL-NEXT:    CLONE ir<[[GEP1:%.+]]> = getelementptr inbounds ir<%b>, vp<[[ST]]>
 ; IF-EVL-NEXT:    vp<[[PTR1:%[0-9]+]]> = vector-pointer ir<[[GEP1]]>