| ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 |
| ; 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 [[CHAR_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 [[CHAR_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 [[CHAR_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 [[CHAR_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: [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i64 2 |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 4, !tbaa [[CHAR_TBAA0:![0-9]+]] |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP2]], align 4, !tbaa [[CHAR_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: [[TMP12:%.*]] = 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> [[TMP12]] to <2 x float>, !fpmath [[META3]] |
| ; INTERLEAVE-NEXT: [[TMP10:%.*]] = fptrunc <2 x double> [[TMP8]] to <2 x float>, !fpmath [[META3]] |
| ; INTERLEAVE-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 2 |
| ; INTERLEAVE-NEXT: store <2 x float> [[TMP9]], ptr [[TMP1]], align 4, !tbaa [[CHAR_TBAA0]] |
| ; INTERLEAVE-NEXT: store <2 x float> [[TMP10]], ptr [[TMP11]], align 4, !tbaa [[CHAR_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 [[CHAR_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 [[CHAR_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 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 [[CHAR_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: [[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 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: [[TMP1:%.*]] = getelementptr i64, ptr [[TMP0]], i64 2 |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4, !tbaa [[CHAR_TBAA0]] |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP1]], align 4, !tbaa [[CHAR_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]], i64 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: [[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 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 [[CHAR_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 [[LOOP9:![0-9]+]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: br label %[[EXIT:.*]] |
| ; 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 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: [[TMP1:%.*]] = getelementptr double, ptr [[TMP0]], i64 2 |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[CHAR_TBAA0]] |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP1]], align 8, !tbaa [[CHAR_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]], i64 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 [[LOOP9:![0-9]+]] |
| ; INTERLEAVE: [[MIDDLE_BLOCK]]: |
| ; INTERLEAVE-NEXT: br label %[[EXIT:.*]] |
| ; 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 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 [[LOOP10:![0-9]+]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: br label %[[EXIT:.*]] |
| ; 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 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]], i64 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]], i64 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 [[LOOP10:![0-9]+]] |
| ; INTERLEAVE: [[MIDDLE_BLOCK]]: |
| ; INTERLEAVE-NEXT: br label %[[EXIT:.*]] |
| ; 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 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 [[CHAR_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 [[LOOP11:![0-9]+]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: br label %[[EXIT:.*]] |
| ; 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 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: [[TMP1:%.*]] = getelementptr double, ptr [[TMP0]], i64 2 |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[CHAR_TBAA0]] |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP1]], align 8, !tbaa [[CHAR_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]], i64 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: [[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 [[LOOP12:![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 [[LOOP13:![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: [[TMP3:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 0 |
| ; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], <2 x i64> [[STEP_ADD]] |
| ; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 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]], i64 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 [[LOOP12:![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 [[LOOP13:![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 |
| } |
| |
| define void @noalias_metadata(ptr align 8 %dst, ptr align 8 %src) { |
| ; CHECK-LABEL: define void @noalias_metadata( |
| ; CHECK-SAME: ptr align 8 [[DST:%.*]], ptr align 8 [[SRC:%.*]]) { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: [[SRC4:%.*]] = ptrtoint ptr [[SRC]] to i64 |
| ; CHECK-NEXT: [[DST3:%.*]] = ptrtoint ptr [[DST]] to i64 |
| ; CHECK-NEXT: [[SRC2:%.*]] = ptrtoint ptr [[SRC]] to i64 |
| ; CHECK-NEXT: [[DST1:%.*]] = ptrtoint ptr [[DST]] to i64 |
| ; CHECK-NEXT: [[TMP2:%.*]] = sub i64 [[DST3]], [[SRC4]] |
| ; CHECK-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP2]], 3 |
| ; CHECK-NEXT: [[TMP4:%.*]] = add nuw nsw i64 [[TMP3]], 1 |
| ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP4]], 2 |
| ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] |
| ; CHECK: [[VECTOR_MEMCHECK]]: |
| ; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[DST]], i64 8 |
| ; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[DST1]], 8 |
| ; CHECK-NEXT: [[TMP22:%.*]] = sub i64 [[TMP5]], [[SRC2]] |
| ; CHECK-NEXT: [[SCEVGEP5:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP22]] |
| ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[DST]], [[SCEVGEP5]] |
| ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SRC]], [[SCEVGEP]] |
| ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP4]], 2 |
| ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP4]], [[N_MOD_VF]] |
| ; CHECK-NEXT: [[TMP23:%.*]] = mul i64 [[N_VEC]], 8 |
| ; CHECK-NEXT: [[TMP24:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP23]] |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[TMP26:%.*]] = mul i64 [[INDEX]], 8 |
| ; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP26]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x ptr>, ptr [[NEXT_GEP]], align 8, !alias.scope [[META14:![0-9]+]] |
| ; CHECK-NEXT: [[TMP7:%.*]] = extractelement <2 x ptr> [[WIDE_LOAD]], i32 1 |
| ; CHECK-NEXT: store ptr [[TMP7]], ptr [[DST]], align 8, !alias.scope [[META17:![0-9]+]], !noalias [[META19:![0-9]+]] |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP4]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] |
| ; CHECK: [[SCALAR_PH]]: |
| ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP24]], %[[MIDDLE_BLOCK]] ], [ [[SRC]], %[[ENTRY]] ], [ [[SRC]], %[[VECTOR_MEMCHECK]] ] |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] |
| ; CHECK-NEXT: [[VAL:%.*]] = load ptr, ptr [[PTR]], align 8 |
| ; CHECK-NEXT: store ptr [[VAL]], ptr [[DST]], align 8, !noalias [[META23:![0-9]+]] |
| ; CHECK-NEXT: [[PTR_NEXT]] = getelementptr inbounds i8, ptr [[PTR]], i64 8 |
| ; CHECK-NEXT: [[CMP:%.*]] = icmp eq ptr [[PTR]], [[DST]] |
| ; CHECK-NEXT: br i1 [[CMP]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP24:![0-9]+]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: ret void |
| ; |
| ; INTERLEAVE-LABEL: define void @noalias_metadata( |
| ; INTERLEAVE-SAME: ptr align 8 [[DST:%.*]], ptr align 8 [[SRC:%.*]]) { |
| ; INTERLEAVE-NEXT: [[ENTRY:.*]]: |
| ; INTERLEAVE-NEXT: [[SRC4:%.*]] = ptrtoint ptr [[SRC]] to i64 |
| ; INTERLEAVE-NEXT: [[DST3:%.*]] = ptrtoint ptr [[DST]] to i64 |
| ; INTERLEAVE-NEXT: [[SRC2:%.*]] = ptrtoint ptr [[SRC]] to i64 |
| ; INTERLEAVE-NEXT: [[DST1:%.*]] = ptrtoint ptr [[DST]] to i64 |
| ; INTERLEAVE-NEXT: [[TMP2:%.*]] = sub i64 [[DST3]], [[SRC4]] |
| ; INTERLEAVE-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP2]], 3 |
| ; INTERLEAVE-NEXT: [[TMP4:%.*]] = add nuw nsw i64 [[TMP3]], 1 |
| ; INTERLEAVE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP4]], 4 |
| ; INTERLEAVE-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] |
| ; INTERLEAVE: [[VECTOR_MEMCHECK]]: |
| ; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[DST]], i64 8 |
| ; INTERLEAVE-NEXT: [[TMP5:%.*]] = add i64 [[DST1]], 8 |
| ; INTERLEAVE-NEXT: [[TMP22:%.*]] = sub i64 [[TMP5]], [[SRC2]] |
| ; INTERLEAVE-NEXT: [[SCEVGEP5:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP22]] |
| ; INTERLEAVE-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[DST]], [[SCEVGEP5]] |
| ; INTERLEAVE-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SRC]], [[SCEVGEP]] |
| ; INTERLEAVE-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] |
| ; INTERLEAVE-NEXT: br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] |
| ; INTERLEAVE: [[VECTOR_PH]]: |
| ; INTERLEAVE-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP4]], 4 |
| ; INTERLEAVE-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP4]], [[N_MOD_VF]] |
| ; INTERLEAVE-NEXT: [[TMP23:%.*]] = mul i64 [[N_VEC]], 8 |
| ; INTERLEAVE-NEXT: [[TMP24:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP23]] |
| ; INTERLEAVE-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; INTERLEAVE: [[VECTOR_BODY]]: |
| ; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] |
| ; INTERLEAVE-NEXT: [[TMP26:%.*]] = mul i64 [[INDEX]], 8 |
| ; INTERLEAVE-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP26]] |
| ; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr ptr, ptr [[NEXT_GEP]], i64 2 |
| ; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <2 x ptr>, ptr [[TMP7]], align 8, !alias.scope [[META14:![0-9]+]] |
| ; INTERLEAVE-NEXT: [[TMP8:%.*]] = extractelement <2 x ptr> [[WIDE_LOAD]], i32 1 |
| ; INTERLEAVE-NEXT: store ptr [[TMP8]], ptr [[DST]], align 8, !alias.scope [[META17:![0-9]+]], !noalias [[META19:![0-9]+]] |
| ; INTERLEAVE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 |
| ; INTERLEAVE-NEXT: [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; INTERLEAVE-NEXT: br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; INTERLEAVE: [[MIDDLE_BLOCK]]: |
| ; INTERLEAVE-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP4]], [[N_VEC]] |
| ; INTERLEAVE-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] |
| ; INTERLEAVE: [[SCALAR_PH]]: |
| ; INTERLEAVE-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP24]], %[[MIDDLE_BLOCK]] ], [ [[SRC]], %[[ENTRY]] ], [ [[SRC]], %[[VECTOR_MEMCHECK]] ] |
| ; INTERLEAVE-NEXT: br label %[[LOOP:.*]] |
| ; INTERLEAVE: [[LOOP]]: |
| ; INTERLEAVE-NEXT: [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] |
| ; INTERLEAVE-NEXT: [[VAL:%.*]] = load ptr, ptr [[PTR]], align 8 |
| ; INTERLEAVE-NEXT: store ptr [[VAL]], ptr [[DST]], align 8, !noalias [[META23:![0-9]+]] |
| ; INTERLEAVE-NEXT: [[PTR_NEXT]] = getelementptr inbounds i8, ptr [[PTR]], i64 8 |
| ; INTERLEAVE-NEXT: [[CMP:%.*]] = icmp eq ptr [[PTR]], [[DST]] |
| ; INTERLEAVE-NEXT: br i1 [[CMP]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP24:![0-9]+]] |
| ; INTERLEAVE: [[EXIT]]: |
| ; INTERLEAVE-NEXT: ret void |
| ; |
| entry: |
| br label %loop |
| |
| loop: |
| %ptr = phi ptr [ %ptr.next, %loop ], [ %src, %entry ] |
| %val = load ptr, ptr %ptr, align 8 |
| store ptr %val, ptr %dst, align 8, !noalias !4 |
| %ptr.next = getelementptr inbounds i8, ptr %ptr, i64 8 |
| %cmp = icmp eq ptr %ptr, %dst |
| br i1 %cmp, label %exit, label %loop |
| |
| 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} |
| !4 = !{!5} |
| !5 = distinct !{!5, !6, !"g1"} |
| !6 = distinct !{!6, !"t2"} |
| |
| ;. |
| ; CHECK: [[CHAR_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: [[LOOP9]] = distinct !{[[LOOP9]], [[META5]], [[META6]]} |
| ; CHECK: [[LOOP10]] = distinct !{[[LOOP10]], [[META5]], [[META6]]} |
| ; CHECK: [[LOOP11]] = distinct !{[[LOOP11]], [[META5]], [[META6]]} |
| ; CHECK: [[LOOP12]] = distinct !{[[LOOP12]], [[META5]], [[META6]]} |
| ; CHECK: [[LOOP13]] = distinct !{[[LOOP13]], [[META6]], [[META5]]} |
| ; CHECK: [[META14]] = !{[[META15:![0-9]+]]} |
| ; CHECK: [[META15]] = distinct !{[[META15]], [[META16:![0-9]+]]} |
| ; CHECK: [[META16]] = distinct !{[[META16]], !"LVerDomain"} |
| ; CHECK: [[META17]] = !{[[META18:![0-9]+]]} |
| ; CHECK: [[META18]] = distinct !{[[META18]], [[META16]]} |
| ; CHECK: [[META19]] = !{[[META20:![0-9]+]], [[META15]]} |
| ; CHECK: [[META20]] = distinct !{[[META20]], [[META21:![0-9]+]], !"g1"} |
| ; CHECK: [[META21]] = distinct !{[[META21]], !"t2"} |
| ; CHECK: [[LOOP22]] = distinct !{[[LOOP22]], [[META5]], [[META6]]} |
| ; CHECK: [[META23]] = !{[[META20]]} |
| ; CHECK: [[LOOP24]] = distinct !{[[LOOP24]], [[META5]]} |
| ;. |
| ; INTERLEAVE: [[CHAR_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: [[LOOP9]] = distinct !{[[LOOP9]], [[META5]], [[META6]]} |
| ; INTERLEAVE: [[LOOP10]] = distinct !{[[LOOP10]], [[META5]], [[META6]]} |
| ; INTERLEAVE: [[LOOP11]] = distinct !{[[LOOP11]], [[META5]], [[META6]]} |
| ; INTERLEAVE: [[LOOP12]] = distinct !{[[LOOP12]], [[META5]], [[META6]]} |
| ; INTERLEAVE: [[LOOP13]] = distinct !{[[LOOP13]], [[META6]], [[META5]]} |
| ; INTERLEAVE: [[META14]] = !{[[META15:![0-9]+]]} |
| ; INTERLEAVE: [[META15]] = distinct !{[[META15]], [[META16:![0-9]+]]} |
| ; INTERLEAVE: [[META16]] = distinct !{[[META16]], !"LVerDomain"} |
| ; INTERLEAVE: [[META17]] = !{[[META18:![0-9]+]]} |
| ; INTERLEAVE: [[META18]] = distinct !{[[META18]], [[META16]]} |
| ; INTERLEAVE: [[META19]] = !{[[META20:![0-9]+]], [[META15]]} |
| ; INTERLEAVE: [[META20]] = distinct !{[[META20]], [[META21:![0-9]+]], !"g1"} |
| ; INTERLEAVE: [[META21]] = distinct !{[[META21]], !"t2"} |
| ; INTERLEAVE: [[LOOP22]] = distinct !{[[LOOP22]], [[META5]], [[META6]]} |
| ; INTERLEAVE: [[META23]] = !{[[META20]]} |
| ; INTERLEAVE: [[LOOP24]] = distinct !{[[LOOP24]], [[META5]]} |
| ;. |
