shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags.ll
Florian Hahn f492eb9509
[VPlan] Make VPInstruction::AnyOf poison-safe. (#154156) 
AnyOf reduces multiple input vectors to a single boolean value. When
used for early-exit vectorization, we need to consider any lane after
the early exit being poison. Any poison lane would result in poison
after the AnyOf reduction. To prevent this, freeze all inputs to AnyOf.

Fixes https://github.com/llvm/llvm-project/issues/153946.
Fixes https://github.com/llvm/llvm-project/issues/155162.

https://alive2.llvm.org/ce/z/FD-XxA

PR: https://github.com/llvm/llvm-project/pull/154156
2025-08-25 18:55:23 +01:00

686 lines
34 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
; RUN: opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s
define float @fmax_ogt_with_select_1(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmax_ogt_with_select_1(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp ogt float [[L]], [[MAX]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[L]], float [[MAX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp ogt float %l, %max
%max.next = select i1 %cmp, float %l, float %max
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmax_ogt_with_select_2(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmax_ogt_with_select_2(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp ogt float [[MAX]], [[L]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[MAX]], float [[L]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp ogt float %max, %l
%max.next = select i1 %cmp, float %max, float %l
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmax_olt_with_select_1(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmax_olt_with_select_1(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp olt float [[L]], [[MAX]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[MAX]], float [[L]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp olt float %l, %max
%max.next = select i1 %cmp, float %max, float %l
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmax_olt_with_select_2(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmax_olt_with_select_2(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp olt float [[MAX]], [[L]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[L]], float [[MAX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp olt float %max, %l
%max.next = select i1 %cmp, float %l, float %max
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmax_ogt_with_select_store_result(ptr %src, ptr %dst, i64 %n) {
; CHECK-LABEL: define float @fmax_ogt_with_select_store_result(
; CHECK-SAME: ptr [[SRC:%.*]], ptr [[DST:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp ogt float [[L]], [[MAX]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[L]], float [[MAX]]
; CHECK-NEXT: store float [[MAX_NEXT]], ptr [[DST]], align 8
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp ogt float %l, %max
%max.next = select i1 %cmp, float %l, float %max
store float %max.next, ptr %dst, align 8
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmaxnum_1(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmaxnum_1(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[TMP4]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[VEC_PHI]])
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: [[TMP2:%.*]] = fcmp uno <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP10:%.*]] = freeze <4 x i1> [[TMP2]]
; CHECK-NEXT: [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP10]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[TMP3]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = or i1 [[TMP3]], [[TMP5]]
; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP7:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x float> [[VEC_PHI]], <4 x float> [[TMP4]]
; CHECK-NEXT: [[TMP9:%.*]] = select i1 [[TMP3]], i64 [[IV]], i64 [[N_VEC]]
; CHECK-NEXT: [[TMP8:%.*]] = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> [[TMP7]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: [[TMP11:%.*]] = xor i1 [[TMP3]], true
; CHECK-NEXT: [[TMP12:%.*]] = and i1 [[CMP_N]], [[TMP11]]
; CHECK-NEXT: br i1 [[TMP12]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP9]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP8]], %[[MIDDLE_BLOCK]] ], [ -1.000000e+07, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC1:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC1]], align 4
; CHECK-NEXT: [[MAX_NEXT]] = call float @llvm.maxnum.f32(float [[L]], float [[MAX]])
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ], [ [[TMP8]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%max.next = call float @llvm.maxnum.f32(float %l, float %max)
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmaxnum_2(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmaxnum_2(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[TMP4]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]])
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: [[TMP2:%.*]] = fcmp uno <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP10:%.*]] = freeze <4 x i1> [[TMP2]]
; CHECK-NEXT: [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP10]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[TMP3]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = or i1 [[TMP3]], [[TMP5]]
; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP7:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x float> [[VEC_PHI]], <4 x float> [[TMP4]]
; CHECK-NEXT: [[TMP9:%.*]] = select i1 [[TMP3]], i64 [[IV]], i64 [[N_VEC]]
; CHECK-NEXT: [[TMP8:%.*]] = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> [[TMP7]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: [[TMP11:%.*]] = xor i1 [[TMP3]], true
; CHECK-NEXT: [[TMP12:%.*]] = and i1 [[CMP_N]], [[TMP11]]
; CHECK-NEXT: br i1 [[TMP12]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP9]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP8]], %[[MIDDLE_BLOCK]] ], [ -1.000000e+07, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC1:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC1]], align 4
; CHECK-NEXT: [[MAX_NEXT]] = call float @llvm.maxnum.f32(float [[MAX]], float [[L]])
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ], [ [[TMP8]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%max.next = call float @llvm.maxnum.f32(float %max, float %l)
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmaxnum_induction_starts_at_10(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmaxnum_induction_starts_at_10(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[N]], -10
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[IV:%.*]] = add i64 10, [[INDEX]]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[TMP3]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[VEC_PHI]])
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: [[TMP5:%.*]] = fcmp uno <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP12:%.*]] = freeze <4 x i1> [[TMP5]]
; CHECK-NEXT: [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP12]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[TMP6]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP7:%.*]] = or i1 [[TMP6]], [[TMP4]]
; CHECK-NEXT: br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP8:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x float> [[VEC_PHI]], <4 x float> [[TMP3]]
; CHECK-NEXT: [[TMP9:%.*]] = select i1 [[TMP6]], i64 [[INDEX]], i64 [[N_VEC]]
; CHECK-NEXT: [[TMP10:%.*]] = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> [[TMP8]])
; CHECK-NEXT: [[TMP11:%.*]] = add i64 10, [[TMP9]]
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
; CHECK-NEXT: [[TMP13:%.*]] = xor i1 [[TMP6]], true
; CHECK-NEXT: [[TMP14:%.*]] = and i1 [[CMP_N]], [[TMP13]]
; CHECK-NEXT: br i1 [[TMP14]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP11]], %[[MIDDLE_BLOCK]] ], [ 10, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ -1.000000e+07, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC1:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC1]], align 4
; CHECK-NEXT: [[MAX_NEXT]] = call float @llvm.maxnum.f32(float [[L]], float [[MAX]])
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ], [ [[TMP10]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 10, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%max.next = call float @llvm.maxnum.f32(float %l, float %max)
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmaxnum_induction_starts_at_value(ptr %src, i64 %start, i64 %n) {
; CHECK-LABEL: define float @fmaxnum_induction_starts_at_value(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[START:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[TMP0:%.*]] = sub i64 [[N]], [[START]]
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[IV:%.*]] = add i64 [[START]], [[INDEX]]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[TMP3]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[VEC_PHI]])
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: [[TMP5:%.*]] = fcmp uno <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP12:%.*]] = freeze <4 x i1> [[TMP5]]
; CHECK-NEXT: [[TMP6:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP12]])
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[TMP6]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP7:%.*]] = or i1 [[TMP6]], [[TMP4]]
; CHECK-NEXT: br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP8:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x float> [[VEC_PHI]], <4 x float> [[TMP3]]
; CHECK-NEXT: [[TMP9:%.*]] = select i1 [[TMP6]], i64 [[INDEX]], i64 [[N_VEC]]
; CHECK-NEXT: [[TMP10:%.*]] = call float @llvm.vector.reduce.fmax.v4f32(<4 x float> [[TMP8]])
; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[START]], [[TMP9]]
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
; CHECK-NEXT: [[TMP13:%.*]] = xor i1 [[TMP6]], true
; CHECK-NEXT: [[TMP14:%.*]] = and i1 [[CMP_N]], [[TMP13]]
; CHECK-NEXT: br i1 [[TMP14]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP11]], %[[MIDDLE_BLOCK]] ], [ [[START]], %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ -1.000000e+07, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC1:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV1]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC1]], align 4
; CHECK-NEXT: [[MAX_NEXT]] = call float @llvm.maxnum.f32(float [[L]], float [[MAX]])
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ], [ [[TMP10]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ %start, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%max.next = call float @llvm.maxnum.f32(float %l, float %max)
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmaxnum_with_additional_add(ptr noalias %src, ptr noalias %src.2, i64 %n) {
; CHECK-LABEL: define float @fmaxnum_with_additional_add(
; CHECK-SAME: ptr noalias [[SRC:%.*]], ptr noalias [[SRC_2:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[SUM:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[SUM_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC_2:%.*]] = getelementptr inbounds nuw i32, ptr [[SRC_2]], i64 [[IV]]
; CHECK-NEXT: [[L_SRC_2:%.*]] = load i32, ptr [[GEP_SRC_2]], align 4
; CHECK-NEXT: [[SUM_NEXT]] = add i32 [[SUM]], [[L_SRC_2]]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[MAX_NEXT]] = call float @llvm.maxnum.f32(float [[L]], float [[MAX]])
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[SUM_NEXT_LCSSA:%.*]] = phi i32 [ [[SUM_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: store i32 [[SUM_NEXT_LCSSA]], ptr [[SRC_2]], align 4
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%sum = phi i32 [ 0, %entry ], [ %sum.next, %loop ]
%gep.src.2 = getelementptr inbounds nuw i32, ptr %src.2, i64 %iv
%l.src.2 = load i32, ptr %gep.src.2, align 4
%sum.next = add i32 %sum, %l.src.2
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%max.next = call float @llvm.maxnum.f32(float %l, float %max)
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
store i32 %sum.next, ptr %src.2
ret float %max.next
}
define float @fmax_select_with_blend(ptr %A, ptr %B) {
; CHECK-LABEL: define float @fmax_select_with_blend(
; CHECK-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ 0.000000e+00, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP_LATCH]] ]
; CHECK-NEXT: [[GEP_A:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
; CHECK-NEXT: [[L_A:%.*]] = load i32, ptr [[GEP_A]], align 4
; CHECK-NEXT: [[C_1:%.*]] = icmp eq i32 [[L_A]], 0
; CHECK-NEXT: br i1 [[C_1]], label %[[LOOP_THEN:.*]], label %[[LOOP_LATCH]]
; CHECK: [[LOOP_THEN]]:
; CHECK-NEXT: [[GEP_B:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_B]], align 4
; CHECK-NEXT: [[C_2:%.*]] = fcmp ogt float [[MAX]], [[L]]
; CHECK-NEXT: [[MAX_SEL:%.*]] = select i1 [[C_2]], float [[MAX]], float [[L]]
; CHECK-NEXT: br label %[[LOOP_LATCH]]
; CHECK: [[LOOP_LATCH]]:
; CHECK-NEXT: [[MAX_NEXT]] = phi float [ [[MAX_SEL]], %[[LOOP_THEN]] ], [ [[MAX]], %[[LOOP]] ]
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp ne i64 [[IV_NEXT]], 1000
; CHECK-NEXT: br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP_LATCH]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.latch ]
%max = phi float [ 0.000000e+00, %entry ], [ %max.next, %loop.latch ]
%gep.A = getelementptr inbounds i32, ptr %A, i64 %iv
%l.A = load i32, ptr %gep.A
%c.1 = icmp eq i32 %l.A, 0
br i1 %c.1, label %loop.then, label %loop.latch
loop.then:
%gep.B = getelementptr inbounds float, ptr %B, i64 %iv
%l = load float, ptr %gep.B
%c.2 = fcmp ogt float %max, %l
%max.sel = select i1 %c.2, float %max, float %l
br label %loop.latch
loop.latch:
%max.next = phi float [ %max.sel, %loop.then ], [ %max, %loop ]
%iv.next = add i64 %iv, 1
%ec = icmp ne i64 %iv.next, 1000
br i1 %ec, label %loop, label %exit
exit:
ret float %max.next
}
define float @fmax_with_select_and_load_store(ptr %src, ptr noalias %dst, i64 %n) {
; CHECK-LABEL: define float @fmax_with_select_and_load_store(
; CHECK-SAME: ptr [[SRC:%.*]], ptr noalias [[DST:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp ogt float [[L]], [[MAX]]
; CHECK-NEXT: [[IV_1:%.*]] = add i64 [[IV]], 1
; CHECK-NEXT: [[GEP_DST_1:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV_1]]
; CHECK-NEXT: [[L_2:%.*]] = load i32, ptr [[GEP_DST_1]], align 4
; CHECK-NEXT: [[GEP_DST_0:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[IV]]
; CHECK-NEXT: store i32 [[L_2]], ptr [[GEP_DST_0]], align 4
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[L]], float [[MAX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp ogt float %l, %max
%iv.1 = add i64 %iv, 1
%gep.dst.1 = getelementptr inbounds i32, ptr %dst, i64 %iv.1
%l.2 = load i32, ptr %gep.dst.1
%gep.dst.0 = getelementptr inbounds i32, ptr %dst, i64 %iv
store i32 %l.2, ptr %gep.dst.0
%max.next = select i1 %cmp, float %l, float %max
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmax_ugt_with_select_1(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmax_ugt_with_select_1(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp ugt float [[L]], [[MAX]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[L]], float [[MAX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp ugt float %l, %max
%max.next = select i1 %cmp, float %l, float %max
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
define float @fmax_oge_with_select_1(ptr %src, i64 %n) {
; CHECK-LABEL: define float @fmax_oge_with_select_1(
; CHECK-SAME: ptr [[SRC:%.*]], i64 [[N:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[MAX:%.*]] = phi float [ -1.000000e+07, %[[ENTRY]] ], [ [[MAX_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
; CHECK-NEXT: [[L:%.*]] = load float, ptr [[GEP_SRC]], align 4
; CHECK-NEXT: [[CMP:%.*]] = fcmp oge float [[L]], [[MAX]]
; CHECK-NEXT: [[MAX_NEXT]] = select i1 [[CMP]], float [[L]], float [[MAX]]
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: [[MAX_NEXT_LCSSA:%.*]] = phi float [ [[MAX_NEXT]], %[[LOOP]] ]
; CHECK-NEXT: ret float [[MAX_NEXT_LCSSA]]
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%max = phi float [ -1.000000e+07, %entry ], [ %max.next, %loop ]
%gep.src = getelementptr inbounds nuw float, ptr %src, i64 %iv
%l = load float, ptr %gep.src, align 4
%cmp = fcmp oge float %l, %max
%max.next = select i1 %cmp, float %l, float %max
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %n
br i1 %ec, label %exit, label %loop
exit:
ret float %max.next
}
Reference in New Issue View Git Blame Copy Permalink