[llvm] [LV, VPlan] Check if plan is compatible to EVL transform (PR #92092)

Shih-Po Hung via llvm-commits llvm-commits at lists.llvm.org
Wed May 15 00:33:58 PDT 2024


https://github.com/arcbbb updated https://github.com/llvm/llvm-project/pull/92092

>From 3f3bf370a981e3215269b2b584ba3f75b3221749 Mon Sep 17 00:00:00 2001
From: ShihPo Hung <shihpo.hung at sifive.com>
Date: Tue, 14 May 2024 19:06:11 -0700
Subject: [PATCH 1/3] [LV][NFC] precommit test for EVL transform

A precommit test case to show vector loops generated from EVL transform
- This is a precommit test for https://github.com/llvm/llvm-project/pull/92092
---
 .../RISCV/evl-compatible-loops.ll             | 262 ++++++++++++++++++
 1 file changed, 262 insertions(+)
 create mode 100644 llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll

diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
new file mode 100644
index 0000000000000..e076ac2ca9f85
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
@@ -0,0 +1,262 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl \
+; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -force-target-instruction-cost=1 -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s
+
+; Check loops having VPWidenIntOrFpInductionRecipe
+define void @foo(ptr noalias %a, i64 %N) {
+; CHECK-LABEL: define void @foo(
+; CHECK-SAME: ptr noalias [[A:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
+; CHECK-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 2
+; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 2
+; CHECK-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP7]], 1
+; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP8]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 2
+; CHECK-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @llvm.experimental.stepvector.nxv2i64()
+; CHECK-NEXT:    [[TMP12:%.*]] = add <vscale x 2 x i64> [[TMP11]], zeroinitializer
+; CHECK-NEXT:    [[TMP13:%.*]] = mul <vscale x 2 x i64> [[TMP12]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP13]]
+; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 2
+; CHECK-NEXT:    [[TMP16:%.*]] = mul i64 1, [[TMP15]]
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP16]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDEX_EVL_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[INDUCTION]], [[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP17:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[TMP18:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP17]], i32 2, i1 true)
+; CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP19]]
+; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VEC_IND]], ptr align 8 [[TMP21]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), i32 [[TMP18]])
+; CHECK-NEXT:    [[TMP22:%.*]] = zext i32 [[TMP18]] to i64
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP22]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], [[TMP10]]
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[DOTSPLAT]]
+; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
+; CHECK-NEXT:    store i64 [[IV1]], ptr [[ARRAYIDX]], align 8
+; CHECK-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds i64, ptr %a, i64 %iv
+  store i64 %iv, ptr %arrayidx, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+  ret void
+}
+
+; Check loops having VPWidenPointerInductionRecipe
+define void @foo2(ptr noalias %a, ptr noalias %b, i64 %N) {
+; CHECK-LABEL: define void @foo2(
+; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
+; CHECK-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 2
+; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 2
+; CHECK-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP7]], 1
+; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP8]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP9:%.*]] = mul i64 [[N_VEC]], 8
+; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP9]]
+; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[B]], [[ENTRY]] ], [ [[PTR_IND:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDEX_EVL_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 2
+; CHECK-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], 1
+; CHECK-NEXT:    [[TMP15:%.*]] = mul i64 8, [[TMP14]]
+; CHECK-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP13]], 0
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP16]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x i64> @llvm.experimental.stepvector.nxv2i64()
+; CHECK-NEXT:    [[TMP18:%.*]] = add <vscale x 2 x i64> [[DOTSPLAT]], [[TMP17]]
+; CHECK-NEXT:    [[VECTOR_GEP:%.*]] = mul <vscale x 2 x i64> [[TMP18]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 8, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <vscale x 2 x i64> [[VECTOR_GEP]]
+; CHECK-NEXT:    [[TMP20:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[TMP21:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP20]], i32 2, i1 true)
+; CHECK-NEXT:    [[TMP22:%.*]] = add i64 [[EVL_BASED_IV]], 0
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP22]]
+; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds ptr, ptr [[TMP23]], i32 0
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2p0.p0(<vscale x 2 x ptr> [[TMP19]], ptr align 8 [[TMP24]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), i32 [[TMP21]])
+; CHECK-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP21]] to i64
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP25]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], [[TMP11]]
+; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 [[TMP15]]
+; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY1]] ]
+; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
+; CHECK-NEXT:    [[ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], [[FOR_BODY1]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds i8, ptr [[ADDR]], i64 8
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
+; CHECK-NEXT:    store ptr [[ADDR]], ptr [[ARRAYIDX]], align 8
+; CHECK-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %addr = phi ptr [ %incdec.ptr, %for.body ], [ %b, %entry ]
+  %incdec.ptr = getelementptr inbounds i8, ptr %addr, i64 8
+  %arrayidx = getelementptr inbounds i64, ptr %a, i64 %iv
+  store ptr %addr, ptr %arrayidx, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+  ret void
+}
+
+; Check loops having VPVectorPointerRecipe that access in reverse order
+define void @foo3(ptr noalias %a, i64 %N) {
+; CHECK-LABEL: define void @foo3(
+; CHECK-SAME: ptr noalias [[A:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
+; CHECK-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 2
+; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 2
+; CHECK-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP7]], 1
+; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP8]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP9:%.*]] = mul i64 [[N_VEC]], -8
+; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP9]]
+; CHECK-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
+; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT2]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[N]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT4]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDEX_EVL_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[EVL_BASED_IV]], -8
+; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP12]]
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[EVL_BASED_IV]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP13:%.*]] = call <vscale x 2 x i64> @llvm.experimental.stepvector.nxv2i64()
+; CHECK-NEXT:    [[TMP14:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP13]]
+; CHECK-NEXT:    [[VEC_IV:%.*]] = add <vscale x 2 x i64> [[BROADCAST_SPLAT]], [[TMP14]]
+; CHECK-NEXT:    [[TMP15:%.*]] = icmp ule <vscale x 2 x i64> [[VEC_IV]], [[BROADCAST_SPLAT3]]
+; CHECK-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP17:%.*]] = mul i64 [[TMP16]], 2
+; CHECK-NEXT:    [[TMP18:%.*]] = mul i64 0, [[TMP17]]
+; CHECK-NEXT:    [[TMP19:%.*]] = sub i64 1, [[TMP17]]
+; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i64 [[TMP18]]
+; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr i64, ptr [[TMP20]], i64 [[TMP19]]
+; CHECK-NEXT:    [[REVERSE:%.*]] = call <vscale x 2 x i1> @llvm.experimental.vector.reverse.nxv2i1(<vscale x 2 x i1> [[TMP15]])
+; CHECK-NEXT:    [[REVERSE6:%.*]] = call <vscale x 2 x i64> @llvm.experimental.vector.reverse.nxv2i64(<vscale x 2 x i64> [[BROADCAST_SPLAT5]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[REVERSE6]], ptr [[TMP21]], i32 8, <vscale x 2 x i1> [[REVERSE]])
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[EVL_BASED_IV]], [[TMP11]]
+; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_EVL_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[A]], [[ENTRY1]] ]
+; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY1]] ]
+; CHECK-NEXT:    [[ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], [[FOR_BODY1]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds i8, ptr [[ADDR]], i64 -8
+; CHECK-NEXT:    store i64 [[N]], ptr [[ADDR]], align 8
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %addr = phi ptr [ %incdec.ptr, %for.body ], [ %a, %entry ]
+  %incdec.ptr = getelementptr inbounds i8, ptr %addr, i64 -8
+  store i64 %N, ptr %addr, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
+
+for.cond.cleanup:
+  ret void
+}
+;.
+; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+; CHECK: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
+; CHECK: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
+; CHECK: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
+;.

>From a86656d16d5e0594863c7c31f39e966d7bd1765e Mon Sep 17 00:00:00 2001
From: ShihPo Hung <shihpo.hung at sifive.com>
Date: Tue, 14 May 2024 02:29:12 -0700
Subject: [PATCH 2/3] [LV, VPlan] Check if plan is compatible to EVL transform

Vector loops generated from the EVL transform may experience issues on
architectures where EVL differs from RuntimeVF at the second-to-last
iteration.

This patch introduces a check before applying EVL transform. If any recipes in
loop rely on RuntimeVF, the vectorizer stop to generate the plan.
---
 .../Transforms/Vectorize/LoopVectorize.cpp    |  29 ++++-
 llvm/lib/Transforms/Vectorize/VPlan.h         |   1 +
 .../RISCV/evl-compatible-loops.ll             | 121 ++----------------
 ...rize-force-tail-with-evl-gather-scatter.ll |  66 ++--------
 4 files changed, 48 insertions(+), 169 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index adae7caf5917c..22fc627dc9141 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -8539,6 +8539,29 @@ VPRecipeBuilder::tryToCreateWidenRecipe(Instruction *Instr,
   return tryToWiden(Instr, Operands, VPBB);
 }
 
+// EVL transform doesn't support backends where EVL diffs from RuntimeVF
+// in the second-to-last iteration.
+// Return false if the vector region has recipes relying on
+// RuntimeVF.
+static bool isCompatibleToEVLTransform(VPlan &Plan) {
+  auto HasAnyRuntimeVFUserInLoop = [](VPlan &Plan) -> bool {
+    for (auto &Phi : Plan.getVectorLoopRegion()->getEntryBasicBlock()->phis())
+      if (isa<VPWidenIntOrFpInductionRecipe>(&Phi) ||
+          isa<VPWidenPointerInductionRecipe>(&Phi))
+        return true;
+    for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(
+             vp_depth_first_deep(Plan.getVectorLoopRegion())))
+      for (VPRecipeBase &Recipe : *VPBB)
+        if (auto *VecPtrR = dyn_cast<VPVectorPointerRecipe>(&Recipe))
+          if (VecPtrR->isReverse())
+            return true;
+    return false;
+  };
+  if (HasAnyRuntimeVFUserInLoop(Plan))
+    return false;
+  return true;
+}
+
 void LoopVectorizationPlanner::buildVPlansWithVPRecipes(ElementCount MinVF,
                                                         ElementCount MaxVF) {
   assert(OrigLoop->isInnermost() && "Inner loop expected.");
@@ -8553,8 +8576,12 @@ void LoopVectorizationPlanner::buildVPlansWithVPRecipes(ElementCount MinVF,
             *Plan, CM.getMinimalBitwidths(), PSE.getSE()->getContext());
       VPlanTransforms::optimize(*Plan, *PSE.getSE());
       // TODO: try to put it close to addActiveLaneMask().
-      if (CM.foldTailWithEVL())
+      if (CM.foldTailWithEVL()) {
+        // Don't generate plan if the plan is not EVL-compatible
+        if (!isCompatibleToEVLTransform(*Plan))
+          break;
         VPlanTransforms::addExplicitVectorLength(*Plan);
+      }
       assert(verifyVPlanIsValid(*Plan) && "VPlan is invalid");
       VPlans.push_back(std::move(Plan));
     }
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index c74329a0bcc4a..94c342089fbdc 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1576,6 +1576,7 @@ class VPVectorPointerRecipe : public VPRecipeWithIRFlags {
 
   void execute(VPTransformState &State) override;
 
+  bool isReverse() { return IsReverse; }
   bool onlyFirstLaneUsed(const VPValue *Op) const override {
     assert(is_contained(operands(), Op) &&
            "Op must be an operand of the recipe");
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
index e076ac2ca9f85..4687e1dde1cc3 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
@@ -8,60 +8,14 @@ define void @foo(ptr noalias %a, i64 %N) {
 ; CHECK-LABEL: define void @foo(
 ; CHECK-SAME: ptr noalias [[A:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
-; CHECK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
-; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
-; CHECK-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
-; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 2
-; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 2
-; CHECK-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP7]], 1
-; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP8]]
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 2
-; CHECK-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @llvm.experimental.stepvector.nxv2i64()
-; CHECK-NEXT:    [[TMP12:%.*]] = add <vscale x 2 x i64> [[TMP11]], zeroinitializer
-; CHECK-NEXT:    [[TMP13:%.*]] = mul <vscale x 2 x i64> [[TMP12]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
-; CHECK-NEXT:    [[INDUCTION:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP13]]
-; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 2
-; CHECK-NEXT:    [[TMP16:%.*]] = mul i64 1, [[TMP15]]
-; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP16]], i64 0
-; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
-; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
-; CHECK:       vector.body:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDEX_EVL_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[INDUCTION]], [[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP17:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP18:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP17]], i32 2, i1 true)
-; CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[EVL_BASED_IV]], 0
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP19]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VEC_IND]], ptr align 8 [[TMP21]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), i32 [[TMP18]])
-; CHECK-NEXT:    [[TMP22:%.*]] = zext i32 [[TMP18]] to i64
-; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP22]], [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], [[TMP10]]
-; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[DOTSPLAT]]
-; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
-; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
-; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
 ; CHECK:       for.body:
-; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
 ; CHECK-NEXT:    store i64 [[IV1]], ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]]
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY1]]
 ; CHECK:       for.cond.cleanup:
 ; CHECK-NEXT:    ret void
 ;
@@ -85,67 +39,16 @@ define void @foo2(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; CHECK-LABEL: define void @foo2(
 ; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N]]
-; CHECK-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
-; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
-; CHECK-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
-; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 2
-; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 2
-; CHECK-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP7]], 1
-; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP8]]
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; CHECK-NEXT:    [[TMP9:%.*]] = mul i64 [[N_VEC]], 8
-; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP11:%.*]] = mul i64 [[TMP10]], 2
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
-; CHECK:       vector.body:
-; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[B]], [[ENTRY]] ], [ [[PTR_IND:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDEX_EVL_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP13:%.*]] = mul i64 [[TMP12]], 2
-; CHECK-NEXT:    [[TMP14:%.*]] = mul i64 [[TMP13]], 1
-; CHECK-NEXT:    [[TMP15:%.*]] = mul i64 8, [[TMP14]]
-; CHECK-NEXT:    [[TMP16:%.*]] = mul i64 [[TMP13]], 0
-; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP16]], i64 0
-; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[DOTSPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x i64> @llvm.experimental.stepvector.nxv2i64()
-; CHECK-NEXT:    [[TMP18:%.*]] = add <vscale x 2 x i64> [[DOTSPLAT]], [[TMP17]]
-; CHECK-NEXT:    [[VECTOR_GEP:%.*]] = mul <vscale x 2 x i64> [[TMP18]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 8, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <vscale x 2 x i64> [[VECTOR_GEP]]
-; CHECK-NEXT:    [[TMP20:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP21:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP20]], i32 2, i1 true)
-; CHECK-NEXT:    [[TMP22:%.*]] = add i64 [[EVL_BASED_IV]], 0
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP22]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds ptr, ptr [[TMP23]], i32 0
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2p0.p0(<vscale x 2 x ptr> [[TMP19]], ptr align 8 [[TMP24]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), i32 [[TMP21]])
-; CHECK-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP21]] to i64
-; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP25]], [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], [[TMP11]]
-; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 [[TMP15]]
-; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
-; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
-; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY1]] ]
-; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
 ; CHECK:       for.body:
-; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
-; CHECK-NEXT:    [[ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], [[FOR_BODY1]] ], [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_EVL_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[ADDR:%.*]] = phi ptr [ [[INCDEC_PTR:%.*]], [[FOR_BODY]] ], [ [[B]], [[ENTRY]] ]
 ; CHECK-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds i8, ptr [[ADDR]], i64 8
-; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[EVL_BASED_IV]]
 ; CHECK-NEXT:    store ptr [[ADDR]], ptr [[ARRAYIDX]], align 8
-; CHECK-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]]
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw nsw i64 [[EVL_BASED_IV]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDEX_EVL_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]]
 ; CHECK:       for.cond.cleanup:
 ; CHECK-NEXT:    ret void
 ;
@@ -217,7 +120,7 @@ define void @foo3(ptr noalias %a, i64 %N) {
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[REVERSE6]], ptr [[TMP21]], i32 8, <vscale x 2 x i1> [[REVERSE]])
 ; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[EVL_BASED_IV]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_EVL_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
@@ -231,7 +134,7 @@ define void @foo3(ptr noalias %a, i64 %N) {
 ; CHECK-NEXT:    store i64 [[N]], ptr [[ADDR]], align 8
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       for.cond.cleanup:
 ; CHECK-NEXT:    ret void
 ;
@@ -255,8 +158,4 @@ for.cond.cleanup:
 ; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
 ; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
 ; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
-; CHECK: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
-; CHECK: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
-; CHECK: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
-; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
 ;.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll
index ae01bdd371106..a52da79ee3963 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll
@@ -12,66 +12,18 @@
 define void @gather_scatter(ptr noalias %in, ptr noalias %out, ptr noalias %index, i64 %n) {
 ; IF-EVL-LABEL: @gather_scatter(
 ; IF-EVL-NEXT:  entry:
-; IF-EVL-NEXT:    [[TMP0:%.*]] = sub i64 -1, [[N:%.*]]
-; IF-EVL-NEXT:    [[TMP1:%.*]] = call i64 @llvm.vscale.i64()
-; IF-EVL-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2
-; IF-EVL-NEXT:    [[TMP3:%.*]] = icmp ult i64 [[TMP0]], [[TMP2]]
-; IF-EVL-NEXT:    br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
-; IF-EVL:       vector.ph:
-; IF-EVL-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
-; IF-EVL-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 2
-; IF-EVL-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
-; IF-EVL-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP6]], 2
-; IF-EVL-NEXT:    [[TMP8:%.*]] = sub i64 [[TMP7]], 1
-; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], [[TMP8]]
-; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
-; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; IF-EVL-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
-; IF-EVL-NEXT:    [[TMP10:%.*]] = mul i64 [[TMP9]], 2
-; IF-EVL-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @llvm.experimental.stepvector.nxv2i64()
-; IF-EVL-NEXT:    [[TMP12:%.*]] = add <vscale x 2 x i64> [[TMP11]], zeroinitializer
-; IF-EVL-NEXT:    [[TMP13:%.*]] = mul <vscale x 2 x i64> [[TMP12]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 1, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
-; IF-EVL-NEXT:    [[INDUCTION:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP13]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; IF-EVL-NEXT:    [[TMP15:%.*]] = mul i64 [[TMP14]], 2
-; IF-EVL-NEXT:    [[TMP16:%.*]] = mul i64 1, [[TMP15]]
-; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP16]], i64 0
-; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
-; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
-; IF-EVL:       vector.body:
-; IF-EVL-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
-; IF-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP17]], i32 2, i1 true)
-; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], <vscale x 2 x i64> [[VEC_IND]]
-; IF-EVL-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 2 x i64> @llvm.vp.gather.nxv2i64.nxv2p0(<vscale x 2 x ptr> align 8 [[TMP20]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), i32 [[TMP18]])
-; IF-EVL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_GATHER]]
-; IF-EVL-NEXT:    [[WIDE_MASKED_GATHER2:%.*]] = call <vscale x 2 x float> @llvm.vp.gather.nxv2f32.nxv2p0(<vscale x 2 x ptr> align 4 [[TMP21]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), i32 [[TMP18]])
-; IF-EVL-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_GATHER]]
-; IF-EVL-NEXT:    call void @llvm.vp.scatter.nxv2f32.nxv2p0(<vscale x 2 x float> [[WIDE_MASKED_GATHER2]], <vscale x 2 x ptr> align 4 [[TMP22]], <vscale x 2 x i1> shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer), i32 [[TMP18]])
-; IF-EVL-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP18]] to i64
-; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP23]], [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX1]], [[TMP10]]
-; IF-EVL-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
-; IF-EVL-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; IF-EVL-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
-; IF-EVL:       middle.block:
-; IF-EVL-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
-; IF-EVL:       scalar.ph:
-; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; IF-EVL-NEXT:    br label [[FOR_BODY:%.*]]
 ; IF-EVL:       for.body:
-; IF-EVL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; IF-EVL-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX]], i64 [[INDVARS_IV]]
-; IF-EVL-NEXT:    [[TMP25:%.*]] = load i64, ptr [[ARRAYIDX3]], align 8
-; IF-EVL-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN]], i64 [[TMP25]]
-; IF-EVL-NEXT:    [[TMP26:%.*]] = load float, ptr [[ARRAYIDX5]], align 4
-; IF-EVL-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT]], i64 [[TMP25]]
-; IF-EVL-NEXT:    store float [[TMP26]], ptr [[ARRAYIDX7]], align 4
+; IF-EVL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; IF-EVL-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], i64 [[INDVARS_IV]]
+; IF-EVL-NEXT:    [[TMP0:%.*]] = load i64, ptr [[ARRAYIDX3]], align 8
+; IF-EVL-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], i64 [[TMP0]]
+; IF-EVL-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX5]], align 4
+; IF-EVL-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], i64 [[TMP0]]
+; IF-EVL-NEXT:    store float [[TMP1]], ptr [[ARRAYIDX7]], align 4
 ; IF-EVL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N]]
-; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N:%.*]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]]
 ; IF-EVL:       for.end:
 ; IF-EVL-NEXT:    ret void
 ;

>From 07a62be11a036069421d6d799bf0f824fd9beec7 Mon Sep 17 00:00:00 2001
From: ShihPo Hung <shihpo.hung at sifive.com>
Date: Tue, 14 May 2024 20:32:56 -0700
Subject: [PATCH 3/3] Address comments

---
 .../Transforms/Vectorize/LoopVectorize.cpp    | 21 ++++++-------------
 1 file changed, 6 insertions(+), 15 deletions(-)

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 22fc627dc9141..3d5ed9085071f 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -8544,20 +8544,11 @@ VPRecipeBuilder::tryToCreateWidenRecipe(Instruction *Instr,
 // Return false if the vector region has recipes relying on
 // RuntimeVF.
 static bool isCompatibleToEVLTransform(VPlan &Plan) {
-  auto HasAnyRuntimeVFUserInLoop = [](VPlan &Plan) -> bool {
-    for (auto &Phi : Plan.getVectorLoopRegion()->getEntryBasicBlock()->phis())
-      if (isa<VPWidenIntOrFpInductionRecipe>(&Phi) ||
-          isa<VPWidenPointerInductionRecipe>(&Phi))
-        return true;
-    for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(
-             vp_depth_first_deep(Plan.getVectorLoopRegion())))
-      for (VPRecipeBase &Recipe : *VPBB)
-        if (auto *VecPtrR = dyn_cast<VPVectorPointerRecipe>(&Recipe))
-          if (VecPtrR->isReverse())
-            return true;
-    return false;
-  };
-  if (HasAnyRuntimeVFUserInLoop(Plan))
+  if (any_of(Plan.getVectorLoopRegion()->getEntryBasicBlock()->phis(),
+             [](VPRecipeBase &Phi) {
+               return (isa<VPWidenIntOrFpInductionRecipe>(&Phi) ||
+                       isa<VPWidenPointerInductionRecipe>(&Phi));
+             }))
     return false;
   return true;
 }
@@ -8577,7 +8568,7 @@ void LoopVectorizationPlanner::buildVPlansWithVPRecipes(ElementCount MinVF,
       VPlanTransforms::optimize(*Plan, *PSE.getSE());
       // TODO: try to put it close to addActiveLaneMask().
       if (CM.foldTailWithEVL()) {
-        // Don't generate plan if the plan is not EVL-compatible
+        // Discard the plan if it is not EVL-compatible
         if (!isCompatibleToEVLTransform(*Plan))
           break;
         VPlanTransforms::addExplicitVectorLength(*Plan);



More information about the llvm-commits mailing list