llvm/test/Transforms/LoopVectorize/scalarize-masked-call.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 -S 2>&1 | FileCheck %s

target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"

define void @cond_call(ptr readonly %src, ptr noalias %dest, i64 %N) {
; CHECK-LABEL: @cond_call(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 2
; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
; CHECK:       vector.ph:
; CHECK-NEXT:    [[N_MOD_VF:%.*]] = and i64 [[N]], 1
; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_CALL_CONTINUE2:%.*]] ]
; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[SRC:%.*]], i64 [[INDEX]]
; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 8
; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <2 x i64> [[WIDE_LOAD]], splat (i64 5)
; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x i1> [[TMP1]], i64 0
; CHECK-NEXT:    br i1 [[TMP2]], label [[PRED_CALL_IF:%.*]], label [[PRED_CALL_CONTINUE:%.*]]
; CHECK:       pred.call.if:
; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x i64> [[WIDE_LOAD]], i64 0
; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @foo(i64 [[TMP3]]) #[[ATTR0:[0-9]+]]
; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <2 x i64> poison, i64 [[TMP4]], i64 0
; CHECK-NEXT:    br label [[PRED_CALL_CONTINUE]]
; CHECK:       pred.call.continue:
; CHECK-NEXT:    [[TMP6:%.*]] = phi <2 x i64> [ poison, [[VECTOR_BODY]] ], [ [[TMP5]], [[PRED_CALL_IF]] ]
; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i1> [[TMP1]], i64 1
; CHECK-NEXT:    br i1 [[TMP7]], label [[PRED_CALL_IF1:%.*]], label [[PRED_CALL_CONTINUE2]]
; CHECK:       pred.call.if1:
; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[WIDE_LOAD]], i64 1
; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @foo(i64 [[TMP8]]) #[[ATTR0]]
; CHECK-NEXT:    [[TMP10:%.*]] = insertelement <2 x i64> [[TMP6]], i64 [[TMP9]], i64 1
; CHECK-NEXT:    br label [[PRED_CALL_CONTINUE2]]
; CHECK:       pred.call.continue2:
; CHECK-NEXT:    [[TMP11:%.*]] = phi <2 x i64> [ [[TMP6]], [[PRED_CALL_CONTINUE]] ], [ [[TMP10]], [[PRED_CALL_IF1]] ]
; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP1]], <2 x i64> [[TMP11]], <2 x i64> [[WIDE_LOAD]]
; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[DEST:%.*]], i64 [[INDEX]]
; CHECK-NEXT:    store <2 x i64> [[PREDPHI]], ptr [[TMP12]], align 8
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       middle.block:
; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT:    br i1 [[CMP_N]], label [[END:%.*]], label [[SCALAR_PH]]
; CHECK:       scalar.ph:
; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
; CHECK:       for.body:
; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_LOOP:%.*]] ]
; CHECK-NEXT:    [[LD_ADDR:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT:    [[LD_VALUE:%.*]] = load i64, ptr [[LD_ADDR]], align 8
; CHECK-NEXT:    [[IFCOND:%.*]] = icmp ult i64 [[LD_VALUE]], 5
; CHECK-NEXT:    br i1 [[IFCOND]], label [[IF_THEN:%.*]], label [[FOR_LOOP]]
; CHECK:       if.then:
; CHECK-NEXT:    [[FOO_RET:%.*]] = call i64 @foo(i64 [[LD_VALUE]]) #[[ATTR0]]
; CHECK-NEXT:    br label [[FOR_LOOP]]
; CHECK:       for.loop:
; CHECK-NEXT:    [[ST_VALUE:%.*]] = phi i64 [ [[LD_VALUE]], [[FOR_BODY]] ], [ [[FOO_RET]], [[IF_THEN]] ]
; CHECK-NEXT:    [[ST_ADDR:%.*]] = getelementptr inbounds i64, ptr [[DEST]], i64 [[IV]]
; CHECK-NEXT:    store i64 [[ST_VALUE]], ptr [[ST_ADDR]], align 8
; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT:    br i1 [[LOOPCOND]], label [[END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       end:
; CHECK-NEXT:    ret void
;
entry:
  br label %for.body

for.body:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.loop ]
  %ld.addr = getelementptr inbounds i64, ptr %src, i64 %iv
  %ld.value = load i64, ptr %ld.addr, align 8
  %ifcond = icmp ult i64 %ld.value, 5
  br i1 %ifcond, label %if.then, label %for.loop

if.then:
  %foo.ret = call i64 @foo(i64 %ld.value) #0
  br label %for.loop

for.loop:
  %st.value = phi i64 [ %ld.value, %for.body ], [ %foo.ret, %if.then ]
  %st.addr = getelementptr inbounds i64, ptr %dest, i64 %iv
  store i64 %st.value, ptr %st.addr, align 8
  %iv.next = add nsw nuw i64 %iv, 1
  %loopcond = icmp eq i64 %iv.next, %N
  br i1 %loopcond, label %end, label %for.body

end:
  ret void
}

; Test for a masked scalar call to an intrinsic whose result is used as part of
; a load address.
define void @masked_umax_used_by_load_address(ptr noalias %src, ptr noalias %dst, i64 %n, i1 %c) {
; CHECK-LABEL: @masked_umax_used_by_load_address(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    br label [[VECTOR_PH:%.*]]
; CHECK:       vector.ph:
; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.umax.i64(i64 [[N:%.*]], i64 1)
; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i16, ptr [[SRC:%.*]], i64 [[TMP0]]
; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
; CHECK:       vector.body:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE2:%.*]] ]
; CHECK-NEXT:    br i1 [[C:%.*]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
; CHECK:       pred.load.if:
; CHECK-NEXT:    [[TMP2:%.*]] = load i16, ptr [[TMP1]], align 2
; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i16> poison, i16 [[TMP2]], i64 0
; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE]]
; CHECK:       pred.load.continue:
; CHECK-NEXT:    [[TMP4:%.*]] = phi <2 x i16> [ poison, [[VECTOR_BODY]] ], [ [[TMP3]], [[PRED_LOAD_IF]] ]
; CHECK-NEXT:    br i1 [[C]], label [[PRED_LOAD_IF1:%.*]], label [[PRED_LOAD_CONTINUE2]]
; CHECK:       pred.load.if1:
; CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[TMP1]], align 2
; CHECK-NEXT:    [[TMP6:%.*]] = insertelement <2 x i16> [[TMP4]], i16 [[TMP5]], i64 1
; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE2]]
; CHECK:       pred.load.continue2:
; CHECK-NEXT:    [[TMP7:%.*]] = phi <2 x i16> [ [[TMP4]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP6]], [[PRED_LOAD_IF1]] ]
; CHECK-NEXT:    [[PREDPHI:%.*]] = select i1 [[C]], <2 x i16> [[TMP7]], <2 x i16> zeroinitializer
; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i16, ptr [[DST:%.*]], i64 [[INDEX]]
; CHECK-NEXT:    store <2 x i16> [[PREDPHI]], ptr [[TMP8]], align 2
; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK:       middle.block:
; CHECK-NEXT:    br label [[EXIT:%.*]]
; CHECK:       exit:
; CHECK-NEXT:    ret void
;
entry:
  br label %loop.header

loop.header:
  %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
  br i1 %c, label %then, label %loop.latch

then:
  %umax = call i64 @llvm.umax.i64(i64 %n, i64 1)
  %gep.src = getelementptr i16, ptr %src, i64 %umax
  %l = load i16, ptr %gep.src, align 2
  br label %loop.latch

loop.latch:
  %m = phi i16 [ 0, %loop.header ], [ %l, %then ]
  %gep.dst = getelementptr i16, ptr %dst, i64 %iv
  store i16 %m, ptr %gep.dst, align 2
  %iv.next = add nuw nsw i64 %iv, 1
  %ec = icmp eq i64 %iv.next, 100
  br i1 %ec, label %exit, label %loop.header

exit:
  ret void
}

declare i64 @foo(i64) #0
declare <4 x i64> @vector_foo(<4 x i64>, <4 x i1>)

; We need a vector variant in order to allow for vectorization at present, but
; we want to test scalarization of conditional calls. If we provide a variant
; with a different number of lanes than the VF we force via
; "-force-vector-width=2", then it should pass the legality checks but
; scalarize. TODO: Remove the requirement to have a variant.
attributes #0 = { readonly nounwind "vector-function-abi-variant"="_ZGV_LLVM_M4v_foo(vector_foo)" }