[llvm] [LV]Support dropping of nneg flag for zext widencast recipes. (PR #74112)

[Florian Hahn via llvm-commits]

================
@@ -0,0 +1,104 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -passes=loop-vectorize -S -mtriple=riscv64-unknown-linux-gnu -mattr=+v < %s | FileCheck %s
+
+define void @test() {
+; CHECK-LABEL: define void @test(
+; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[TMP0]], 2
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 0, [[TMP1]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; CHECK:       vector.scevcheck:
+; CHECK-NEXT:    br i1 true, label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = mul i64 [[TMP2]], 2
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 0, [[TMP3]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 0, [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP4:%.*]] = call <vscale x 2 x i32> @llvm.experimental.stepvector.nxv2i32()
+; CHECK-NEXT:    [[TMP5:%.*]] = add <vscale x 2 x i32> [[TMP4]], zeroinitializer
+; CHECK-NEXT:    [[TMP6:%.*]] = mul <vscale x 2 x i32> [[TMP5]], shufflevector (<vscale x 2 x i32> insertelement (<vscale x 2 x i32> poison, i32 1, i64 0), <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer)
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <vscale x 2 x i32> zeroinitializer, [[TMP6]]
+; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vscale.i32()
+; CHECK-NEXT:    [[TMP8:%.*]] = mul i32 [[TMP7]], 2
+; CHECK-NEXT:    [[TMP9:%.*]] = mul i32 1, [[TMP8]]
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 2 x i32> poison, i32 [[TMP9]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <vscale x 2 x i32> [[DOTSPLATINSERT]], <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP10:%.*]] = zext <vscale x 2 x i32> [[VEC_IND]] to <vscale x 2 x i64>
+; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <vscale x 2 x i64> [[TMP10]], i32 0
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr double, ptr null, i64 [[TMP11]]
+; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr double, ptr [[TMP12]], i32 0
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x double> @llvm.masked.load.nxv2f64.p0(ptr [[TMP13]], i32 8, <vscale x 2 x i1> zeroinitializer, <vscale x 2 x double> poison)
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <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), <vscale x 2 x double> zeroinitializer, <vscale x 2 x double> [[WIDE_MASKED_LOAD]]
+; CHECK-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
+; CHECK-NEXT:    [[TMP15:%.*]] = mul i32 [[TMP14]], 2
+; CHECK-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP15]], 1
+; CHECK-NEXT:    [[TMP17:%.*]] = extractelement <vscale x 2 x double> [[PREDPHI]], i32 [[TMP16]]
+; CHECK-NEXT:    store double [[TMP17]], ptr null, align 8
+; CHECK-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP19:%.*]] = mul i64 [[TMP18]], 2
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP19]]
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i32> [[VEC_IND]], [[DOTSPLAT]]
+; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 0, [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    br label [[BODY:%.*]]
+; CHECK:       body:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[NEXT:%.*]], [[ELSE:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
+; CHECK-NEXT:    [[TMP21:%.*]] = trunc i64 [[IV]] to i32
+; CHECK-NEXT:    br i1 false, label [[THEN:%.*]], label [[ELSE]]
+; CHECK:       then:
+; CHECK-NEXT:    [[ZEXT:%.*]] = zext nneg i32 [[TMP21]] to i64
+; CHECK-NEXT:    [[IDX1:%.*]] = getelementptr double, ptr null, i64 [[ZEXT]]
+; CHECK-NEXT:    [[IDX2:%.*]] = getelementptr double, ptr null, i64 [[ZEXT]]
+; CHECK-NEXT:    [[TMP22:%.*]] = load double, ptr [[IDX2]], align 8
+; CHECK-NEXT:    br label [[ELSE]]
+; CHECK:       else:
+; CHECK-NEXT:    [[PHI:%.*]] = phi double [ [[TMP22]], [[THEN]] ], [ 0.000000e+00, [[BODY]] ]
+; CHECK-NEXT:    store double [[PHI]], ptr null, align 8
+; CHECK-NEXT:    [[NEXT]] = add i64 [[IV]], 1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i64 [[NEXT]], 0
+; CHECK-NEXT:    br i1 [[CMP]], label [[EXIT]], label [[BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %body
+
+body:
+  %iv = phi i64 [ %next, %else ], [ 0, %entry ]
+  %0 = trunc i64 %iv to i32
+  br i1 false, label %then, label %else
----------------
fhahn wrote:

please update the condition here to use a loop dependent value; if the condition is false, it will `zext nneg` will always execute and wouldn't need to be dropped.

https://github.com/llvm/llvm-project/pull/74112