shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
llvm-project/llvm/test/Transforms/LoopVectorize/metadata.ll
Florian Hahn 89ae085859
[VPlan] Remove VPVectorPointer for part 0 after unrolling. (#149735) 
VPVectorPointer for part 0 is just the pointer operand. Simplify it
after unrolling. This removes a large number of redundant GEPs with
index 0.

PR: https://github.com/llvm/llvm-project/pull/149735
2025-07-27 13:53:26 +01:00

659 lines
39 KiB
LLVM
Raw Permalink Blame History

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
; RUN: opt -S < %s -passes=loop-vectorize -force-vector-interleave=1 -force-vector-width=2 | FileCheck %s
; RUN: opt -S < %s -passes=loop-vectorize -force-vector-interleave=2 -force-vector-width=2 | FileCheck --check-prefix=INTERLEAVE %s
; Make sure we are preserving existing metadata in the vectorized code.
define void @fp_math(ptr nocapture %a, ptr noalias %b, i64 %size) {
; CHECK-LABEL: define void @fp_math(
; CHECK-SAME: ptr captures(none) [[A:%.*]], ptr noalias [[B:%.*]], i64 [[SIZE:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SIZE]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SIZE]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SIZE]], [[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: [[TMP0:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 4, !tbaa [[TBAA0:![0-9]+]]
; CHECK-NEXT: [[TMP3:%.*]] = fadd <2 x double> [[WIDE_LOAD]], splat (double 9.900000e+01), !fpmath [[META3:![0-9]+]]
; CHECK-NEXT: [[TMP4:%.*]] = fcmp oge <2 x double> [[TMP3]], splat (double 1.000000e+01)
; CHECK-NEXT: [[TMP6:%.*]] = select <2 x i1> [[TMP4]], <2 x double> [[WIDE_LOAD]], <2 x double> zeroinitializer, !fpmath [[META3]]
; CHECK-NEXT: [[TMP5:%.*]] = fptrunc <2 x double> [[TMP6]] to <2 x float>, !fpmath [[META3]]
; CHECK-NEXT: store <2 x float> [[TMP5]], ptr [[TMP1]], align 4, !tbaa [[TBAA0]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SIZE]], [[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]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
; CHECK-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
; CHECK-NEXT: [[L_1:%.*]] = load double, ptr [[ARRAYIDX_1]], align 4, !tbaa [[TBAA0]]
; CHECK-NEXT: [[ADD:%.*]] = fadd double [[L_1]], 9.900000e+01, !fpmath [[META3]]
; CHECK-NEXT: [[C:%.*]] = fcmp oge double [[ADD]], 1.000000e+01
; CHECK-NEXT: [[SEL:%.*]] = select i1 [[C]], double [[L_1]], double 0.000000e+00, !fpmath [[META3]]
; CHECK-NEXT: [[T:%.*]] = fptrunc double [[SEL]] to float, !fpmath [[META3]]
; CHECK-NEXT: store float [[T]], ptr [[ARRAYIDX_2]], align 4, !tbaa [[TBAA0]]
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[SIZE]]
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
; INTERLEAVE-LABEL: define void @fp_math(
; INTERLEAVE-SAME: ptr captures(none) [[A:%.*]], ptr noalias [[B:%.*]], i64 [[SIZE:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SIZE]], 4
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SIZE]], 4
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub i64 [[SIZE]], [[N_MOD_VF]]
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i32 2
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 4, !tbaa [[TBAA0:![0-9]+]]
; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP3]], align 4, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = fadd <2 x double> [[WIDE_LOAD]], splat (double 9.900000e+01), !fpmath [[META3:![0-9]+]]
; INTERLEAVE-NEXT: [[TMP5:%.*]] = fadd <2 x double> [[WIDE_LOAD1]], splat (double 9.900000e+01), !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP6:%.*]] = fcmp oge <2 x double> [[TMP4]], splat (double 1.000000e+01)
; INTERLEAVE-NEXT: [[TMP7:%.*]] = fcmp oge <2 x double> [[TMP5]], splat (double 1.000000e+01)
; INTERLEAVE-NEXT: [[TMP11:%.*]] = select <2 x i1> [[TMP6]], <2 x double> [[WIDE_LOAD]], <2 x double> zeroinitializer, !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP8:%.*]] = select <2 x i1> [[TMP7]], <2 x double> [[WIDE_LOAD1]], <2 x double> zeroinitializer, !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP9:%.*]] = fptrunc <2 x double> [[TMP11]] to <2 x float>, !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP10:%.*]] = fptrunc <2 x double> [[TMP8]] to <2 x float>, !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 2
; INTERLEAVE-NEXT: store <2 x float> [[TMP9]], ptr [[TMP1]], align 4, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: store <2 x float> [[TMP10]], ptr [[TMP13]], align 4, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SIZE]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
; INTERLEAVE-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
; INTERLEAVE-NEXT: [[L_1:%.*]] = load double, ptr [[ARRAYIDX_1]], align 4, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[ADD:%.*]] = fadd double [[L_1]], 9.900000e+01, !fpmath [[META3]]
; INTERLEAVE-NEXT: [[C:%.*]] = fcmp oge double [[ADD]], 1.000000e+01
; INTERLEAVE-NEXT: [[SEL:%.*]] = select i1 [[C]], double [[L_1]], double 0.000000e+00, !fpmath [[META3]]
; INTERLEAVE-NEXT: [[T:%.*]] = fptrunc double [[SEL]] to float, !fpmath [[META3]]
; INTERLEAVE-NEXT: store float [[T]], ptr [[ARRAYIDX_2]], align 4, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[SIZE]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%arrayidx.1 = getelementptr inbounds double, ptr %a, i64 %iv
%arrayidx.2 = getelementptr inbounds float, ptr %b, i64 %iv
%l.1 = load double, ptr %arrayidx.1, align 4, !tbaa !3
%add = fadd double %l.1, 99.0, !fpmath !0
%c = fcmp oge double %add, 10.0
%sel = select i1 %c, double %l.1, double 0.0, !fpmath !0
%t = fptrunc double %sel to float, !fpmath !0
store float %t, ptr %arrayidx.2, !tbaa !3
%iv.next = add i64 %iv, 1
%exitcond = icmp ne i64 %iv.next, %size
br i1 %exitcond, label %loop, label %exit
exit:
ret void
}
; TODO: Range metadata not preserved
define void @widen_call_range(ptr noalias %a, ptr readonly %b) {
; CHECK-LABEL: define void @widen_call_range(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4, !tbaa [[TBAA0]]
; CHECK-NEXT: [[TMP2:%.*]] = call <2 x i64> @foo_vector_fixed2_nomask(<2 x i64> [[WIDE_LOAD]])
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: store <2 x i64> [[TMP2]], ptr [[TMP3]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: br label %[[EXIT:.*]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
; CHECK-NEXT: [[LOAD:%.*]] = load i64, ptr [[GEP]], align 4, !tbaa [[TBAA0]], !range [[RNG9:![0-9]+]]
; CHECK-NEXT: [[CALL:%.*]] = call i64 @foo(i64 [[LOAD]]) #[[ATTR1:[0-9]+]], !range [[RNG9]]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
; CHECK-NEXT: store i64 [[CALL]], ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP10:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
; INTERLEAVE-LABEL: define void @widen_call_range(
; INTERLEAVE-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[TMP0]], i32 2
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP2]], align 4, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = call <2 x i64> @foo_vector_fixed2_nomask(<2 x i64> [[WIDE_LOAD]])
; INTERLEAVE-NEXT: [[TMP4:%.*]] = call <2 x i64> @foo_vector_fixed2_nomask(<2 x i64> [[WIDE_LOAD1]])
; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 2
; INTERLEAVE-NEXT: store <2 x i64> [[TMP3]], ptr [[TMP5]], align 4
; INTERLEAVE-NEXT: store <2 x i64> [[TMP4]], ptr [[TMP7]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: br label %[[EXIT:.*]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
; INTERLEAVE-NEXT: [[LOAD:%.*]] = load i64, ptr [[GEP]], align 4, !tbaa [[TBAA0]], !range [[RNG9:![0-9]+]]
; INTERLEAVE-NEXT: [[CALL:%.*]] = call i64 @foo(i64 [[LOAD]]) #[[ATTR1:[0-9]+]], !range [[RNG9]]
; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
; INTERLEAVE-NEXT: store i64 [[CALL]], ptr [[ARRAYIDX]], align 4
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP10:![0-9]+]]
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%gep = getelementptr i64, ptr %b, i64 %iv
%load = load i64, ptr %gep, !range !1, !tbaa !3
%call = call i64 @foo(i64 %load) #0, !range !1
%arrayidx = getelementptr inbounds i64, ptr %a, i64 %iv
store i64 %call, ptr %arrayidx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond = icmp eq i64 %iv.next, 1024
br i1 %exitcond, label %exit, label %loop
exit:
ret void
}
define void @widen_call_fpmath(ptr noalias %a, ptr readonly %b) {
; CHECK-LABEL: define void @widen_call_fpmath(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
; CHECK-NEXT: [[TMP1:%.*]] = call <2 x double> @bar_vector_fixed2_nomask(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: store <2 x double> [[TMP1]], ptr [[TMP3]], align 8
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: br label %[[EXIT:.*]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr double, ptr [[B]], i64 [[IV]]
; CHECK-NEXT: [[LOAD:%.*]] = load double, ptr [[GEP]], align 8, !tbaa [[TBAA0]]
; CHECK-NEXT: [[CALL:%.*]] = call double @bar(double [[LOAD]]) #[[ATTR2:[0-9]+]], !fpmath [[META3]]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
; CHECK-NEXT: store double [[CALL]], ptr [[ARRAYIDX]], align 8
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP12:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
; INTERLEAVE-LABEL: define void @widen_call_fpmath(
; INTERLEAVE-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr double, ptr [[TMP0]], i32 2
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP2]], align 8, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = call <2 x double> @bar_vector_fixed2_nomask(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = call <2 x double> @bar_vector_fixed2_nomask(<2 x double> [[WIDE_LOAD1]]), !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 2
; INTERLEAVE-NEXT: store <2 x double> [[TMP3]], ptr [[TMP5]], align 8
; INTERLEAVE-NEXT: store <2 x double> [[TMP4]], ptr [[TMP7]], align 8
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: br label %[[EXIT:.*]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[GEP:%.*]] = getelementptr double, ptr [[B]], i64 [[IV]]
; INTERLEAVE-NEXT: [[LOAD:%.*]] = load double, ptr [[GEP]], align 8, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[CALL:%.*]] = call double @bar(double [[LOAD]]) #[[ATTR2:[0-9]+]], !fpmath [[META3]]
; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
; INTERLEAVE-NEXT: store double [[CALL]], ptr [[ARRAYIDX]], align 8
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP12:![0-9]+]]
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%gep = getelementptr double , ptr %b, i64 %iv
%load = load double, ptr %gep, !tbaa !3
%call = call double @bar(double %load) #1, !fpmath !0
%arrayidx = getelementptr inbounds double, ptr %a, i64 %iv
store double %call, ptr %arrayidx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond = icmp eq i64 %iv.next, 1024
br i1 %exitcond, label %exit, label %loop
exit:
ret void
}
define void @widen_intrinsic(ptr noalias %a, ptr readonly %b) {
; CHECK-LABEL: define void @widen_intrinsic(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4
; CHECK-NEXT: [[TMP2:%.*]] = call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[WIDE_LOAD]], i1 true)
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: store <2 x i64> [[TMP2]], ptr [[TMP3]], align 4
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: br label %[[EXIT:.*]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
; CHECK-NEXT: [[LOAD:%.*]] = load i64, ptr [[GEP]], align 4
; CHECK-NEXT: [[CALL:%.*]] = call i64 @llvm.abs.i64(i64 [[LOAD]], i1 true), !range [[RNG9]]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
; CHECK-NEXT: store i64 [[CALL]], ptr [[ARRAYIDX]], align 4
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP14:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
; INTERLEAVE-LABEL: define void @widen_intrinsic(
; INTERLEAVE-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[TMP0]], i32 2
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4
; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP2]], align 4
; INTERLEAVE-NEXT: [[TMP3:%.*]] = call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[WIDE_LOAD]], i1 true)
; INTERLEAVE-NEXT: [[TMP4:%.*]] = call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[WIDE_LOAD1]], i1 true)
; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 2
; INTERLEAVE-NEXT: store <2 x i64> [[TMP3]], ptr [[TMP5]], align 4
; INTERLEAVE-NEXT: store <2 x i64> [[TMP4]], ptr [[TMP7]], align 4
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: br label %[[EXIT:.*]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
; INTERLEAVE-NEXT: [[LOAD:%.*]] = load i64, ptr [[GEP]], align 4
; INTERLEAVE-NEXT: [[CALL:%.*]] = call i64 @llvm.abs.i64(i64 [[LOAD]], i1 true), !range [[RNG9]]
; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
; INTERLEAVE-NEXT: store i64 [[CALL]], ptr [[ARRAYIDX]], align 4
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP14:![0-9]+]]
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%gep = getelementptr i64, ptr %b, i64 %iv
%load = load i64, ptr %gep
%call = call i64 @llvm.abs.i64(i64 %load, i1 true), !range !1
%arrayidx = getelementptr inbounds i64, ptr %a, i64 %iv
store i64 %call, ptr %arrayidx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond = icmp eq i64 %iv.next, 1024
br i1 %exitcond, label %exit, label %loop
exit:
ret void
}
define void @widen_intrinsic_fpmath(ptr noalias %a, ptr readonly %b) {
; CHECK-LABEL: define void @widen_intrinsic_fpmath(
; CHECK-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
; CHECK-NEXT: [[TMP1:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
; CHECK-NEXT: store <2 x double> [[TMP1]], ptr [[TMP3]], align 8
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; CHECK-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: br label %[[EXIT:.*]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr double, ptr [[B]], i64 [[IV]]
; CHECK-NEXT: [[LOAD:%.*]] = load double, ptr [[GEP]], align 8, !tbaa [[TBAA0]]
; CHECK-NEXT: [[CALL:%.*]] = call double @llvm.sin.f64(double [[LOAD]]) #[[ATTR2]], !fpmath [[META3]]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
; CHECK-NEXT: store double [[CALL]], ptr [[ARRAYIDX]], align 8
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP16:![0-9]+]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
; INTERLEAVE-LABEL: define void @widen_intrinsic_fpmath(
; INTERLEAVE-SAME: ptr noalias [[A:%.*]], ptr readonly [[B:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr double, ptr [[TMP0]], i32 2
; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP2]], align 8, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[WIDE_LOAD1]]), !fpmath [[META3]]
; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 2
; INTERLEAVE-NEXT: store <2 x double> [[TMP3]], ptr [[TMP5]], align 8
; INTERLEAVE-NEXT: store <2 x double> [[TMP4]], ptr [[TMP7]], align 8
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: br label %[[EXIT:.*]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[GEP:%.*]] = getelementptr double, ptr [[B]], i64 [[IV]]
; INTERLEAVE-NEXT: [[LOAD:%.*]] = load double, ptr [[GEP]], align 8, !tbaa [[TBAA0]]
; INTERLEAVE-NEXT: [[CALL:%.*]] = call double @llvm.sin.f64(double [[LOAD]]) #[[ATTR2]], !fpmath [[META3]]
; INTERLEAVE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
; INTERLEAVE-NEXT: store double [[CALL]], ptr [[ARRAYIDX]], align 8
; INTERLEAVE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP16:![0-9]+]]
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%gep = getelementptr double , ptr %b, i64 %iv
%load = load double, ptr %gep, !tbaa !3
%call = call double @llvm.sin.f64(double %load) #1, !fpmath !0
%arrayidx = getelementptr inbounds double, ptr %a, i64 %iv
store double %call, ptr %arrayidx
%iv.next = add nuw nsw i64 %iv, 1
%exitcond = icmp eq i64 %iv.next, 1024
br i1 %exitcond, label %exit, label %loop
exit:
ret void
}
define void @unknown_metadata(ptr nocapture %a, ptr noalias %b, i64 %size) {
; CHECK-LABEL: define void @unknown_metadata(
; CHECK-SAME: ptr captures(none) [[A:%.*]], ptr noalias [[B:%.*]], i64 [[SIZE:%.*]]) {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SIZE]], 2
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SIZE]], 2
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[SIZE]], [[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_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP3:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[INDEX]], !custom_md [[META2:![0-9]+]]
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], <2 x i64> [[VEC_IND]]
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x ptr> [[TMP2]], i32 0
; CHECK-NEXT: store <2 x i32> [[TMP3]], ptr [[TMP1]], align 4
; CHECK-NEXT: store <2 x ptr> [[TMP2]], ptr [[TMP6]], align 8
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
; CHECK-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
; CHECK-NEXT: [[VEC_IND_NEXT2]] = add <2 x i32> [[TMP3]], splat (i32 2)
; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SIZE]], [[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]] ]
; CHECK-NEXT: br label %[[LOOP:.*]]
; CHECK: [[LOOP]]:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; CHECK-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]], !custom_md [[META2]]
; CHECK-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]], !custom_md [[META2]]
; CHECK-NEXT: [[T:%.*]] = trunc i64 [[IV]] to i32, !custom_md [[META3]]
; CHECK-NEXT: store i32 [[T]], ptr [[ARRAYIDX_2]], align 4, !custom_md [[META2]]
; CHECK-NEXT: store ptr [[ARRAYIDX_2]], ptr [[ARRAYIDX_1]], align 8, !custom_md [[META2]]
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1, !custom_md [[META2]]
; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[SIZE]], !custom_md [[META2]]
; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP18:![0-9]+]], !custom_md [[META2]]
; CHECK: [[EXIT]]:
; CHECK-NEXT: ret void
;
; INTERLEAVE-LABEL: define void @unknown_metadata(
; INTERLEAVE-SAME: ptr captures(none) [[A:%.*]], ptr noalias [[B:%.*]], i64 [[SIZE:%.*]]) {
; INTERLEAVE-NEXT: [[ENTRY:.*]]:
; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[SIZE]], 4
; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; INTERLEAVE: [[VECTOR_PH]]:
; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[SIZE]], 4
; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub i64 [[SIZE]], [[N_MOD_VF]]
; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]]
; INTERLEAVE: [[VECTOR_BODY]]:
; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[VEC_IND1:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], %[[VECTOR_BODY]] ]
; INTERLEAVE-NEXT: [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
; INTERLEAVE-NEXT: [[STEP_ADD3:%.*]] = add <2 x i32> [[VEC_IND1]], splat (i32 2)
; INTERLEAVE-NEXT: [[TMP0:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[INDEX]], !custom_md [[META2:![0-9]+]]
; INTERLEAVE-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], <2 x i64> [[VEC_IND]]
; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], <2 x i64> [[STEP_ADD]]
; INTERLEAVE-NEXT: [[TMP3:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 0
; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 2
; INTERLEAVE-NEXT: store <2 x i32> [[VEC_IND1]], ptr [[TMP3]], align 4
; INTERLEAVE-NEXT: store <2 x i32> [[STEP_ADD3]], ptr [[TMP5]], align 4
; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds ptr, ptr [[TMP0]], i32 2
; INTERLEAVE-NEXT: store <2 x ptr> [[TMP1]], ptr [[TMP0]], align 8
; INTERLEAVE-NEXT: store <2 x ptr> [[TMP2]], ptr [[TMP7]], align 8
; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; INTERLEAVE-NEXT: [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], splat (i64 2)
; INTERLEAVE-NEXT: [[VEC_IND_NEXT2]] = add <2 x i32> [[STEP_ADD3]], splat (i32 2)
; INTERLEAVE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
; INTERLEAVE: [[MIDDLE_BLOCK]]:
; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[SIZE]], [[N_VEC]]
; INTERLEAVE-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
; INTERLEAVE: [[SCALAR_PH]]:
; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; INTERLEAVE-NEXT: br label %[[LOOP:.*]]
; INTERLEAVE: [[LOOP]]:
; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
; INTERLEAVE-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]], !custom_md [[META2]]
; INTERLEAVE-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]], !custom_md [[META2]]
; INTERLEAVE-NEXT: [[T:%.*]] = trunc i64 [[IV]] to i32, !custom_md [[META3]]
; INTERLEAVE-NEXT: store i32 [[T]], ptr [[ARRAYIDX_2]], align 4, !custom_md [[META2]]
; INTERLEAVE-NEXT: store ptr [[ARRAYIDX_2]], ptr [[ARRAYIDX_1]], align 8, !custom_md [[META2]]
; INTERLEAVE-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1, !custom_md [[META2]]
; INTERLEAVE-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[IV_NEXT]], [[SIZE]], !custom_md [[META2]]
; INTERLEAVE-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP18:![0-9]+]], !custom_md [[META2]]
; INTERLEAVE: [[EXIT]]:
; INTERLEAVE-NEXT: ret void
;
entry:
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%arrayidx.1 = getelementptr inbounds ptr, ptr %a, i64 %iv, !custom_md !2
%arrayidx.2 = getelementptr inbounds i32, ptr %b, i64 %iv, !custom_md !2
%t = trunc i64 %iv to i32, !custom_md !0
store i32 %t, ptr %arrayidx.2, !custom_md !2
store ptr %arrayidx.2, ptr %arrayidx.1, !custom_md !2
%iv.next = add i64 %iv, 1, !custom_md !2
%exitcond = icmp ne i64 %iv.next, %size, !custom_md !2
br i1 %exitcond, label %loop, label %exit, !custom_md !2
exit:
ret void
}
declare i64 @foo(i64)
declare double @bar(double)
declare <2 x i64> @foo_vector_fixed2_nomask(<2 x i64>)
declare <2 x double> @bar_vector_fixed2_nomask(<2 x double>)
attributes #0 = { nounwind "vector-function-abi-variant"="_ZGV_LLVM_N2v_foo(foo_vector_fixed2_nomask)" }
attributes #1 = { nounwind "vector-function-abi-variant"="_ZGV_LLVM_N2v_bar(bar_vector_fixed2_nomask)" }
!0 = !{ float 2.5 }
!1 = !{ i64 0, i64 2 }
!2 = !{!"Simple C/C++ TBAA"}
!3 = !{!"omnipotent char", !2, i64 0}
;.
; CHECK: [[TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0, i64 0}
; CHECK: [[META1]] = !{!"omnipotent char", [[META2]]}
; CHECK: [[META2]] = !{!"Simple C/C++ TBAA"}
; CHECK: [[META3]] = !{float 2.500000e+00}
; CHECK: [[LOOP4]] = distinct !{[[LOOP4]], [[META5:![0-9]+]], [[META6:![0-9]+]]}
; CHECK: [[META5]] = !{!"llvm.loop.isvectorized", i32 1}
; CHECK: [[META6]] = !{!"llvm.loop.unroll.runtime.disable"}
; CHECK: [[LOOP7]] = distinct !{[[LOOP7]], [[META6]], [[META5]]}
; CHECK: [[LOOP8]] = distinct !{[[LOOP8]], [[META5]], [[META6]]}
; CHECK: [[RNG9]] = !{i64 0, i64 2}
; CHECK: [[LOOP10]] = distinct !{[[LOOP10]], [[META6]], [[META5]]}
; CHECK: [[LOOP11]] = distinct !{[[LOOP11]], [[META5]], [[META6]]}
; CHECK: [[LOOP12]] = distinct !{[[LOOP12]], [[META6]], [[META5]]}
; CHECK: [[LOOP13]] = distinct !{[[LOOP13]], [[META5]], [[META6]]}
; CHECK: [[LOOP14]] = distinct !{[[LOOP14]], [[META6]], [[META5]]}
; CHECK: [[LOOP15]] = distinct !{[[LOOP15]], [[META5]], [[META6]]}
; CHECK: [[LOOP16]] = distinct !{[[LOOP16]], [[META6]], [[META5]]}
; CHECK: [[LOOP17]] = distinct !{[[LOOP17]], [[META5]], [[META6]]}
; CHECK: [[LOOP18]] = distinct !{[[LOOP18]], [[META6]], [[META5]]}
;.
; INTERLEAVE: [[TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0, i64 0}
; INTERLEAVE: [[META1]] = !{!"omnipotent char", [[META2]]}
; INTERLEAVE: [[META2]] = !{!"Simple C/C++ TBAA"}
; INTERLEAVE: [[META3]] = !{float 2.500000e+00}
; INTERLEAVE: [[LOOP4]] = distinct !{[[LOOP4]], [[META5:![0-9]+]], [[META6:![0-9]+]]}
; INTERLEAVE: [[META5]] = !{!"llvm.loop.isvectorized", i32 1}
; INTERLEAVE: [[META6]] = !{!"llvm.loop.unroll.runtime.disable"}
; INTERLEAVE: [[LOOP7]] = distinct !{[[LOOP7]], [[META6]], [[META5]]}
; INTERLEAVE: [[LOOP8]] = distinct !{[[LOOP8]], [[META5]], [[META6]]}
; INTERLEAVE: [[RNG9]] = !{i64 0, i64 2}
; INTERLEAVE: [[LOOP10]] = distinct !{[[LOOP10]], [[META6]], [[META5]]}
; INTERLEAVE: [[LOOP11]] = distinct !{[[LOOP11]], [[META5]], [[META6]]}
; INTERLEAVE: [[LOOP12]] = distinct !{[[LOOP12]], [[META6]], [[META5]]}
; INTERLEAVE: [[LOOP13]] = distinct !{[[LOOP13]], [[META5]], [[META6]]}
; INTERLEAVE: [[LOOP14]] = distinct !{[[LOOP14]], [[META6]], [[META5]]}
; INTERLEAVE: [[LOOP15]] = distinct !{[[LOOP15]], [[META5]], [[META6]]}
; INTERLEAVE: [[LOOP16]] = distinct !{[[LOOP16]], [[META6]], [[META5]]}
; INTERLEAVE: [[LOOP17]] = distinct !{[[LOOP17]], [[META5]], [[META6]]}
; INTERLEAVE: [[LOOP18]] = distinct !{[[LOOP18]], [[META6]], [[META5]]}
;.
Reference in New Issue View Git Blame Copy Permalink