1; __arm_streaming function, no special flags. Should by default not vectorize. 2; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_enabled/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S - | FileCheck %s --check-prefix=CHECK 3; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_enabled/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S - | FileCheck %s --check-prefix=CHECK 4; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_body/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S - | FileCheck %s --check-prefix=CHECK 5; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_body/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S - | FileCheck %s --check-prefix=CHECK 6 7; __arm_streaming function, force use of scalable autovec. Should vectorize. 8; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_enabled/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S -enable-scalable-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-SCALABLE 9; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_enabled/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S -enable-scalable-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-SCALABLE 10; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_body/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S -enable-scalable-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-SCALABLE 11; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_body/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S -enable-scalable-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-SCALABLE 12 13; __arm_streaming function, force use of fixed-width autovec. Should vectorize. 14; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_enabled/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S -enable-fixedwidth-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-FIXEDWIDTH 15; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_enabled/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S -enable-fixedwidth-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-FIXEDWIDTH 16; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_body/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S -enable-fixedwidth-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-FIXEDWIDTH 17; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_body/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S -enable-fixedwidth-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-FIXEDWIDTH 18 19; __arm_streaming_compatible function, no special flags. Should by default not vectorize. 20; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_compatible/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S - | FileCheck %s --check-prefix=CHECK 21; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_compatible/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S - | FileCheck %s --check-prefix=CHECK 22 23; __arm_streaming_compatible function, force use of scalable autovec. Can only vectorize if +sve is available. 24; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_compatible/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S -enable-scalable-autovec-in-streaming-mode - | FileCheck %s 25; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_compatible/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S -enable-scalable-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-SCALABLE 26 27; __arm_streaming_compatible function, force use of fixed-width autovec. Can only vectorize if +sve is available. 28; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_compatible/ %s | opt -mattr="+sme" -passes=loop-vectorize,slp-vectorizer -S -enable-fixedwidth-autovec-in-streaming-mode - | FileCheck %s 29; RUN: sed -e s/REPLACE_PSTATE_MACRO/aarch64_pstate_sm_compatible/ %s | opt -mattr="+sme,+sve" -passes=loop-vectorize,slp-vectorizer -S -enable-fixedwidth-autovec-in-streaming-mode - | FileCheck %s --check-prefix=CHECK-FORCE-FIXEDWIDTH 30 31target triple = "aarch64-unknown-linux-gnu" 32 33attributes #0 = { vscale_range(1,16) "REPLACE_PSTATE_MACRO" } 34 35define void @test_fixedwidth_loopvec(ptr noalias %dst, ptr readonly %src, i32 %N) #0 { 36; CHECK-LABEL: @test_fixedwidth_loopvec 37; CHECK-NOT: <{{[1-9]+}} x i32> 38; CHECK-FORCE-FIXEDWIDTH-LABEL: @test_fixedwidth_loopvec 39; CHECK-FORCE-FIXEDWIDTH: <{{[1-9]+}} x i32> 40entry: 41 %cmp6 = icmp sgt i32 %N, 0 42 br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup 43 44for.body.preheader: ; preds = %entry 45 %wide.trip.count = zext i32 %N to i64 46 br label %for.body 47 48for.cond.cleanup: ; preds = %for.body, %entry 49 ret void 50 51for.body: ; preds = %for.body.preheader, %for.body 52 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 53 %arrayidx = getelementptr inbounds i32, ptr %src, i64 %indvars.iv 54 %0 = load i32, ptr %arrayidx, align 4 55 %add = add nsw i32 %0, 42 56 %arrayidx2 = getelementptr inbounds i32, ptr %dst, i64 %indvars.iv 57 store i32 %add, ptr %arrayidx2, align 4 58 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 59 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count 60 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body, !llvm.loop !0 61} 62 63!0 = distinct !{!0, !1, !2, !3} 64!1 = !{!"llvm.loop.mustprogress"} 65!2 = !{!"llvm.loop.interleave.count", i32 1} 66!3 = !{!"llvm.loop.vectorize.scalable.enable", i1 false} 67 68define void @test_scalable_loopvec(ptr noalias %dst, ptr readonly %src, i32 %N) #0 { 69; CHECK-LABEL: @test_scalable_loopvec 70; CHECK-NOT: <vscale x {{[1-9]+}} x i32> 71; CHECK-FORCE-SCALABLE-LABEL: @test_fixedwidth_loopvec 72; CHECK-FORCE-SCALABLE-LABEL: <vscale x {{[1-9]+}} x i32> 73entry: 74 %cmp6 = icmp sgt i32 %N, 0 75 br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup 76 77for.body.preheader: ; preds = %entry 78 %wide.trip.count = zext i32 %N to i64 79 br label %for.body 80 81for.cond.cleanup: ; preds = %for.body, %entry 82 ret void 83 84for.body: ; preds = %for.body.preheader, %for.body 85 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 86 %arrayidx = getelementptr inbounds i32, ptr %src, i64 %indvars.iv 87 %0 = load i32, ptr %arrayidx, align 4 88 %add = add nsw i32 %0, 42 89 %arrayidx2 = getelementptr inbounds i32, ptr %dst, i64 %indvars.iv 90 store i32 %add, ptr %arrayidx2, align 4 91 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 92 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count 93 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body, !llvm.loop !4 94} 95 96!4 = distinct !{!4, !5, !6, !7} 97!5 = !{!"llvm.loop.mustprogress"} 98!6 = !{!"llvm.loop.interleave.count", i32 1} 99!7 = !{!"llvm.loop.vectorize.scalable.enable", i1 true} 100 101define void @test_slp(ptr noalias %dst, ptr readonly %src, i32 %N) #0 { 102; CHECK-LABEL: @test_slp 103; CHECK-NOT: <{{[1-9]+}} x i32> 104; CHECK-FORCE-FIXEDWIDTH-LABEL: @test_slp 105; CHECK-FORCE-FIXEDWIDTH: <{{[1-9]+}} x i32> 106entry: 107 %cmp6 = icmp sgt i32 %N, 0 108 br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup 109 110for.body.preheader: ; preds = %entry 111 %wide.trip.count = zext i32 %N to i64 112 br label %for.body 113 114for.cond.cleanup: ; preds = %for.body, %entry 115 ret void 116 117for.body: ; preds = %for.body.preheader, %for.body 118 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 119 %arrayidx = getelementptr inbounds i32, ptr %src, i64 %indvars.iv 120 %0 = load i32, ptr %arrayidx, align 4 121 %add = add nsw i32 %0, 42 122 %arrayidx2 = getelementptr inbounds i32, ptr %dst, i64 %indvars.iv 123 store i32 %add, ptr %arrayidx2, align 4 124 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 125 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count 126 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body, !llvm.loop !8 127} 128 129!8 = distinct !{!8, !9, !10, !11} 130!9 = !{!"llvm.loop.mustprogress"} 131!10 = !{!"llvm.loop.interleave.count", i32 4} 132!11 = !{!"llvm.loop.vectorize.width", i32 1} 133