diff options
Diffstat (limited to 'llvm/test/Transforms/PGOProfile/chr.ll')
| -rw-r--r-- | llvm/test/Transforms/PGOProfile/chr.ll | 1912 |
1 files changed, 1912 insertions, 0 deletions
diff --git a/llvm/test/Transforms/PGOProfile/chr.ll b/llvm/test/Transforms/PGOProfile/chr.ll new file mode 100644 index 00000000000..399209181e8 --- /dev/null +++ b/llvm/test/Transforms/PGOProfile/chr.ll @@ -0,0 +1,1912 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; RUN: opt < %s -chr -instcombine -simplifycfg -S | FileCheck %s +; RUN: opt < %s -passes='require<profile-summary>,function(chr,instcombine,simplify-cfg)' -S | FileCheck %s + +declare void @foo() +declare void @bar() + +; Simple case. +; Roughly, +; t0 = *i +; if ((t0 & 1) != 0) // Likely true +; foo() +; if ((t0 & 2) != 0) // Likely true +; foo() +; -> +; t0 = *i +; if ((t0 & 3) != 0) { // Likely true +; foo() +; foo() +; } else { +; if ((t0 & 1) != 0) +; foo() +; if ((t0 & 2) != 0) +; foo() +; } +define void @test_chr_1(i32* %i) !prof !14 { +; CHECK-LABEL: @test_chr_1( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 0 +; CHECK-NEXT: br i1 [[TMP4]], label [[BB1_NONCHR:%.*]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: br i1 [[TMP6]], label [[BB3]], label [[BB2_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: ret void +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %3 = and i32 %0, 2 + %4 = icmp eq i32 %3, 0 + br i1 %4, label %bb3, label %bb2, !prof !15 + +bb2: + call void @foo() + br label %bb3 + +bb3: + ret void +} + +; Simple case with a cold block. +; Roughly, +; t0 = *i +; if ((t0 & 1) != 0) // Likely true +; foo() +; if ((t0 & 2) == 0) // Likely false +; bar() +; if ((t0 & 4) != 0) // Likely true +; foo() +; -> +; t0 = *i +; if ((t0 & 7) == 7) { // Likely true +; foo() +; foo() +; } else { +; if ((t0 & 1) != 0) +; foo() +; if ((t0 & 2) == 0) +; bar() +; if ((t0 & 4) != 0) +; foo() +; } +define void @test_chr_1_1(i32* %i) !prof !14 { +; CHECK-LABEL: @test_chr_1_1( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 7 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 7 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB5:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 0 +; CHECK-NEXT: br i1 [[TMP4]], label [[BB1_NONCHR:%.*]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: br i1 [[TMP6]], label [[BB2_NONCHR:%.*]], label [[BB3_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: call void @bar() +; CHECK-NEXT: br label [[BB3_NONCHR]] +; CHECK: bb3.nonchr: +; CHECK-NEXT: [[TMP7:%.*]] = and i32 [[TMP0]], 4 +; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0 +; CHECK-NEXT: br i1 [[TMP8]], label [[BB5]], label [[BB4_NONCHR:%.*]], !prof !16 +; CHECK: bb4.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB5]] +; CHECK: bb5: +; CHECK-NEXT: ret void +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %3 = and i32 %0, 2 + %4 = icmp eq i32 %3, 0 + br i1 %4, label %bb2, label %bb3, !prof !15 + +bb2: + call void @bar() + br label %bb3 + +bb3: + %5 = and i32 %0, 4 + %6 = icmp eq i32 %5, 0 + br i1 %6, label %bb5, label %bb4, !prof !15 + +bb4: + call void @foo() + br label %bb5 + +bb5: + ret void +} + +; With an aggregate bit check. +; Roughly, +; t0 = *i +; if ((t0 & 255) != 0) // Likely true +; if ((t0 & 1) != 0) // Likely true +; foo() +; if ((t0 & 2) != 0) // Likely true +; foo() +; -> +; t0 = *i +; if ((t0 & 3) != 0) { // Likely true +; foo() +; foo() +; } else if ((t0 & 255) != 0) +; if ((t0 & 1) != 0) +; foo() +; if ((t0 & 2) != 0) +; foo() +; } +define void @test_chr_2(i32* %i) !prof !14 { +; CHECK-LABEL: @test_chr_2( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB1:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb1: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB4:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 255 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 0 +; CHECK-NEXT: br i1 [[TMP4]], label [[BB4]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: br i1 [[TMP6]], label [[BB2_NONCHR:%.*]], label [[BB1_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: [[TMP7:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0 +; CHECK-NEXT: br i1 [[TMP8]], label [[BB4]], label [[BB3_NONCHR:%.*]], !prof !16 +; CHECK: bb3.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB4]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB2_NONCHR]] +; CHECK: bb4: +; CHECK-NEXT: ret void +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 255 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb4, label %bb0, !prof !15 + +bb0: + %3 = and i32 %0, 1 + %4 = icmp eq i32 %3, 0 + br i1 %4, label %bb2, label %bb1, !prof !15 + +bb1: + call void @foo() + br label %bb2 + +bb2: + %5 = and i32 %0, 2 + %6 = icmp eq i32 %5, 0 + br i1 %6, label %bb4, label %bb3, !prof !15 + +bb3: + call void @foo() + br label %bb4 + +bb4: + ret void +} + +; Split case. +; Roughly, +; t1 = *i +; if ((t1 & 1) != 0) // Likely true +; foo() +; if ((t1 & 2) != 0) // Likely true +; foo() +; t2 = *i +; if ((t2 & 4) != 0) // Likely true +; foo() +; if ((t2 & 8) != 0) // Likely true +; foo() +; -> +; t1 = *i +; if ((t1 & 3) != 0) { // Likely true +; foo() +; foo() +; } else { +; if ((t1 & 1) != 0) +; foo() +; if ((t1 & 2) != 0) +; foo() +; } +; t2 = *i +; if ((t2 & 12) != 0) { // Likely true +; foo() +; foo() +; } else { +; if ((t2 & 4) != 0) +; foo() +; if ((t2 & 8) != 0) +; foo() +; } +define void @test_chr_3(i32* %i) !prof !14 { +; CHECK-LABEL: @test_chr_3( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 0 +; CHECK-NEXT: br i1 [[TMP4]], label [[BB1_NONCHR:%.*]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: br i1 [[TMP6]], label [[BB3]], label [[BB2_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[TMP7:%.*]] = load i32, i32* [[I]], align 4 +; CHECK-NEXT: [[TMP8:%.*]] = and i32 [[TMP7]], 12 +; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP8]], 12 +; CHECK-NEXT: br i1 [[TMP9]], label [[BB4:%.*]], label [[BB3_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb4: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB7:%.*]] +; CHECK: bb3.split.nonchr: +; CHECK-NEXT: [[TMP10:%.*]] = and i32 [[TMP7]], 4 +; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 0 +; CHECK-NEXT: br i1 [[TMP11]], label [[BB5_NONCHR:%.*]], label [[BB4_NONCHR:%.*]], !prof !16 +; CHECK: bb4.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB5_NONCHR]] +; CHECK: bb5.nonchr: +; CHECK-NEXT: [[TMP12:%.*]] = and i32 [[TMP7]], 8 +; CHECK-NEXT: [[TMP13:%.*]] = icmp eq i32 [[TMP12]], 0 +; CHECK-NEXT: br i1 [[TMP13]], label [[BB7]], label [[BB6_NONCHR:%.*]], !prof !16 +; CHECK: bb6.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB7]] +; CHECK: bb7: +; CHECK-NEXT: ret void +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %3 = and i32 %0, 2 + %4 = icmp eq i32 %3, 0 + br i1 %4, label %bb3, label %bb2, !prof !15 + +bb2: + call void @foo() + br label %bb3 + +bb3: + %5 = load i32, i32* %i + %6 = and i32 %5, 4 + %7 = icmp eq i32 %6, 0 + br i1 %7, label %bb5, label %bb4, !prof !15 + +bb4: + call void @foo() + br label %bb5 + +bb5: + %8 = and i32 %5, 8 + %9 = icmp eq i32 %8, 0 + br i1 %9, label %bb7, label %bb6, !prof !15 + +bb6: + call void @foo() + br label %bb7 + +bb7: + ret void +} + +; Selects. +; Roughly, +; t0 = *i +; sum1 = (t0 & 1) ? sum0 : (sum0 + 42) // Likely false +; sum2 = (t0 & 2) ? sum1 : (sum1 + 43) // Likely false +; return sum2 +; -> +; t0 = *i +; if ((t0 & 3) == 3) +; return sum0 + 85 +; else { +; sum1 = (t0 & 1) ? sum0 : (sum0 + 42) +; sum2 = (t0 & 2) ? sum1 : (sum1 + 43) +; return sum2 +; } +define i32 @test_chr_4(i32* %i, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_4( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3 +; CHECK-NEXT: br i1 [[TMP2]], label [[ENTRY_SPLIT:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: entry.split: +; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[SUM0:%.*]], 85 +; CHECK-NEXT: ret i32 [[TMP3]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[SUM0]], 42 +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: [[SUM1_NONCHR:%.*]] = select i1 [[TMP6]], i32 [[SUM0]], i32 [[TMP4]], !prof !16 +; CHECK-NEXT: [[TMP7:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0 +; CHECK-NEXT: [[TMP9:%.*]] = add i32 [[SUM1_NONCHR]], 43 +; CHECK-NEXT: [[SUM2_NONCHR:%.*]] = select i1 [[TMP8]], i32 [[SUM1_NONCHR]], i32 [[TMP9]], !prof !16 +; CHECK-NEXT: ret i32 [[SUM2_NONCHR]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + %3 = add i32 %sum0, 42 + %sum1 = select i1 %2, i32 %sum0, i32 %3, !prof !15 + %4 = and i32 %0, 2 + %5 = icmp eq i32 %4, 0 + %6 = add i32 %sum1, 43 + %sum2 = select i1 %5, i32 %sum1, i32 %6, !prof !15 + ret i32 %sum2 +} + +; Selects + Brs +; Roughly, +; t0 = *i +; if ((t0 & 255) != 0) { // Likely true +; sum = (t0 & 1) ? sum0 : (sum0 + 42) // Likely false +; sum = (t0 & 2) ? sum : (sum + 43) // Likely false +; if ((t0 & 4) != 0) { // Likely true +; sum3 = sum + 44 +; sum = (t0 & 8) ? sum3 : (sum3 + 44) // Likely false +; } +; } +; return sum +; -> +; t0 = *i +; if ((t0 & 15) != 15) { // Likely true +; sum = sum0 + 173 +; } else if ((t0 & 255) != 0) { +; sum = (t0 & 1) ? sum0 : (sum0 + 42) +; sum = (t0 & 2) ? sum : (sum + 43) +; if ((t0 & 4) != 0) { +; sum3 = sum + 44 +; sum = (t0 & 8) ? sum3 : (sum3 + 44) +; } +; } +; return sum +define i32 @test_chr_5(i32* %i, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_5( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 15 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 15 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[SUM0:%.*]], 85 +; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[SUM0]], 173 +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 255 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: br i1 [[TMP6]], label [[BB3]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: [[TMP7:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0 +; CHECK-NEXT: [[TMP9:%.*]] = add i32 [[SUM0]], 42 +; CHECK-NEXT: [[SUM1_NONCHR:%.*]] = select i1 [[TMP8]], i32 [[SUM0]], i32 [[TMP9]], !prof !16 +; CHECK-NEXT: [[TMP10:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 0 +; CHECK-NEXT: [[TMP12:%.*]] = add i32 [[SUM1_NONCHR]], 43 +; CHECK-NEXT: [[SUM2_NONCHR:%.*]] = select i1 [[TMP11]], i32 [[SUM1_NONCHR]], i32 [[TMP12]], !prof !16 +; CHECK-NEXT: [[TMP13:%.*]] = and i32 [[TMP0]], 4 +; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i32 [[TMP13]], 0 +; CHECK-NEXT: br i1 [[TMP14]], label [[BB3]], label [[BB1_NONCHR:%.*]], !prof !16 +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP15:%.*]] = and i32 [[TMP0]], 8 +; CHECK-NEXT: [[TMP16:%.*]] = icmp eq i32 [[TMP15]], 0 +; CHECK-NEXT: [[SUM4_NONCHR_V:%.*]] = select i1 [[TMP16]], i32 44, i32 88, !prof !16 +; CHECK-NEXT: [[SUM4_NONCHR:%.*]] = add i32 [[SUM2_NONCHR]], [[SUM4_NONCHR_V]] +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[SUM6:%.*]] = phi i32 [ [[TMP4]], [[BB0]] ], [ [[SUM0]], [[ENTRY_SPLIT_NONCHR]] ], [ [[SUM2_NONCHR]], [[BB0_NONCHR]] ], [ [[SUM4_NONCHR]], [[BB1_NONCHR]] ] +; CHECK-NEXT: ret i32 [[SUM6]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 255 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb3, label %bb0, !prof !15 + +bb0: + %3 = and i32 %0, 1 + %4 = icmp eq i32 %3, 0 + %5 = add i32 %sum0, 42 + %sum1 = select i1 %4, i32 %sum0, i32 %5, !prof !15 + %6 = and i32 %0, 2 + %7 = icmp eq i32 %6, 0 + %8 = add i32 %sum1, 43 + %sum2 = select i1 %7, i32 %sum1, i32 %8, !prof !15 + %9 = and i32 %0, 4 + %10 = icmp eq i32 %9, 0 + br i1 %10, label %bb2, label %bb1, !prof !15 + +bb1: + %sum3 = add i32 %sum2, 44 + %11 = and i32 %0, 8 + %12 = icmp eq i32 %11, 0 + %13 = add i32 %sum3, 44 + %sum4 = select i1 %12, i32 %sum3, i32 %13, !prof !15 + br label %bb2 + +bb2: + %sum5 = phi i32 [ %sum2, %bb0 ], [ %sum4, %bb1 ] + br label %bb3 + +bb3: + %sum6 = phi i32 [ %sum0, %entry ], [ %sum5, %bb2 ] + ret i32 %sum6 +} + +; Selects + Brs with a scope split in the middle +; Roughly, +; t0 = *i +; if ((t0 & 255) != 0) { // Likely true +; sum = (t0 & 1) ? sum0 : (sum0 + 42) // Likely false +; sum = (t0 & 2) ? sum : (sum + 43) // Likely false +; if ((sum0 & 4) != 0) { // Likely true. The condition doesn't use v. +; sum3 = sum + 44 +; sum = (t0 & 8) ? sum3 : (sum3 + 44) // Likely false +; } +; } +; return sum +; -> +; t0 = *i +; if ((sum0 & 4) != 0 & (t0 & 11) != 11) { // Likely true +; sum = sum0 + 173 +; } else if ((t0 & 255) != 0) { +; sum = (t0 & 1) ? sum0 : (sum0 + 42) +; sum = (t0 & 2) ? sum : (sum + 43) +; if ((sum0 & 4) != 0) { +; sum3 = sum + 44 +; sum = (t0 & 8) ? sum3 : (sum3 + 44) +; } +; } +; return sum +define i32 @test_chr_5_1(i32* %i, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_5_1( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[SUM0:%.*]], 4 +; CHECK-NEXT: [[TMP2:%.*]] = icmp ne i32 [[TMP1]], 0 +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 11 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 11 +; CHECK-NEXT: [[TMP5:%.*]] = and i1 [[TMP4]], [[TMP2]] +; CHECK-NEXT: br i1 [[TMP5]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: [[TMP6:%.*]] = add i32 [[SUM0]], 85 +; CHECK-NEXT: [[TMP7:%.*]] = add i32 [[SUM0]], 173 +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP8:%.*]] = and i32 [[TMP0]], 255 +; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP8]], 0 +; CHECK-NEXT: br i1 [[TMP9]], label [[BB3]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: [[TMP10:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 0 +; CHECK-NEXT: [[TMP12:%.*]] = add i32 [[SUM0]], 42 +; CHECK-NEXT: [[SUM1_NONCHR:%.*]] = select i1 [[TMP11]], i32 [[SUM0]], i32 [[TMP12]], !prof !16 +; CHECK-NEXT: [[TMP13:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i32 [[TMP13]], 0 +; CHECK-NEXT: [[TMP15:%.*]] = add i32 [[SUM1_NONCHR]], 43 +; CHECK-NEXT: [[SUM2_NONCHR:%.*]] = select i1 [[TMP14]], i32 [[SUM1_NONCHR]], i32 [[TMP15]], !prof !16 +; CHECK-NEXT: [[TMP16:%.*]] = and i32 [[SUM0]], 4 +; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i32 [[TMP16]], 0 +; CHECK-NEXT: br i1 [[TMP17]], label [[BB3]], label [[BB1_NONCHR:%.*]], !prof !16 +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP18:%.*]] = and i32 [[TMP0]], 8 +; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i32 [[TMP18]], 0 +; CHECK-NEXT: [[SUM4_NONCHR_V:%.*]] = select i1 [[TMP19]], i32 44, i32 88, !prof !16 +; CHECK-NEXT: [[SUM4_NONCHR:%.*]] = add i32 [[SUM2_NONCHR]], [[SUM4_NONCHR_V]] +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[SUM6:%.*]] = phi i32 [ [[TMP7]], [[BB0]] ], [ [[SUM0]], [[ENTRY_SPLIT_NONCHR]] ], [ [[SUM2_NONCHR]], [[BB0_NONCHR]] ], [ [[SUM4_NONCHR]], [[BB1_NONCHR]] ] +; CHECK-NEXT: ret i32 [[SUM6]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 255 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb3, label %bb0, !prof !15 + +bb0: + %3 = and i32 %0, 1 + %4 = icmp eq i32 %3, 0 + %5 = add i32 %sum0, 42 + %sum1 = select i1 %4, i32 %sum0, i32 %5, !prof !15 + %6 = and i32 %0, 2 + %7 = icmp eq i32 %6, 0 + %8 = add i32 %sum1, 43 + %sum2 = select i1 %7, i32 %sum1, i32 %8, !prof !15 + %9 = and i32 %sum0, 4 ; Split + %10 = icmp eq i32 %9, 0 + br i1 %10, label %bb2, label %bb1, !prof !15 + +bb1: + %sum3 = add i32 %sum2, 44 + %11 = and i32 %0, 8 + %12 = icmp eq i32 %11, 0 + %13 = add i32 %sum3, 44 + %sum4 = select i1 %12, i32 %sum3, i32 %13, !prof !15 + br label %bb2 + +bb2: + %sum5 = phi i32 [ %sum2, %bb0 ], [ %sum4, %bb1 ] + br label %bb3 + +bb3: + %sum6 = phi i32 [ %sum0, %entry ], [ %sum5, %bb2 ] + ret i32 %sum6 +} + +; Selects + Brs, non-matching bases +; Roughly, +; i0 = *i +; j0 = *j +; if ((i0 & 255) != 0) { // Likely true +; sum = (i0 & 2) ? sum0 : (sum0 + 43) // Likely false +; if ((j0 & 4) != 0) { // Likely true. The condition uses j0, not i0. +; sum3 = sum + 44 +; sum = (i0 & 8) ? sum3 : (sum3 + 44) // Likely false +; } +; } +; return sum +; -> +; i0 = *i +; j0 = *j +; if ((j0 & 4) != 0 & (i0 & 10) != 10) { // Likely true +; sum = sum0 + 131 +; } else if ((i0 & 255) != 0) { +; sum = (i0 & 2) ? sum0 : (sum0 + 43) +; if ((j0 & 4) != 0) { +; sum3 = sum + 44 +; sum = (i0 & 8) ? sum3 : (sum3 + 44) +; } +; } +; return sum +define i32 @test_chr_6(i32* %i, i32* %j, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_6( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[I0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[J0:%.*]] = load i32, i32* [[J:%.*]], align 4 +; CHECK-NEXT: [[V9:%.*]] = and i32 [[J0]], 4 +; CHECK-NEXT: [[V10:%.*]] = icmp ne i32 [[V9]], 0 +; CHECK-NEXT: [[TMP0:%.*]] = and i32 [[I0]], 10 +; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[TMP0]], 10 +; CHECK-NEXT: [[TMP2:%.*]] = and i1 [[TMP1]], [[V10]] +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: [[V8:%.*]] = add i32 [[SUM0:%.*]], 43 +; CHECK-NEXT: [[V13:%.*]] = add i32 [[SUM0]], 131 +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[V1:%.*]] = and i32 [[I0]], 255 +; CHECK-NEXT: [[V2:%.*]] = icmp eq i32 [[V1]], 0 +; CHECK-NEXT: br i1 [[V2]], label [[BB3]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: [[V3_NONCHR:%.*]] = and i32 [[I0]], 2 +; CHECK-NEXT: [[V4_NONCHR:%.*]] = icmp eq i32 [[V3_NONCHR]], 0 +; CHECK-NEXT: [[V8_NONCHR:%.*]] = add i32 [[SUM0]], 43 +; CHECK-NEXT: [[SUM2_NONCHR:%.*]] = select i1 [[V4_NONCHR]], i32 [[SUM0]], i32 [[V8_NONCHR]], !prof !16 +; CHECK-NEXT: [[V9_NONCHR:%.*]] = and i32 [[J0]], 4 +; CHECK-NEXT: [[V10_NONCHR:%.*]] = icmp eq i32 [[V9_NONCHR]], 0 +; CHECK-NEXT: br i1 [[V10_NONCHR]], label [[BB3]], label [[BB1_NONCHR:%.*]], !prof !16 +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[V11_NONCHR:%.*]] = and i32 [[I0]], 8 +; CHECK-NEXT: [[V12_NONCHR:%.*]] = icmp eq i32 [[V11_NONCHR]], 0 +; CHECK-NEXT: [[SUM4_NONCHR_V:%.*]] = select i1 [[V12_NONCHR]], i32 44, i32 88, !prof !16 +; CHECK-NEXT: [[SUM4_NONCHR:%.*]] = add i32 [[SUM2_NONCHR]], [[SUM4_NONCHR_V]] +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[SUM6:%.*]] = phi i32 [ [[V13]], [[BB0]] ], [ [[SUM0]], [[ENTRY_SPLIT_NONCHR]] ], [ [[SUM2_NONCHR]], [[BB0_NONCHR]] ], [ [[SUM4_NONCHR]], [[BB1_NONCHR]] ] +; CHECK-NEXT: ret i32 [[SUM6]] +; +entry: + %i0 = load i32, i32* %i + %j0 = load i32, i32* %j + %v1 = and i32 %i0, 255 + %v2 = icmp eq i32 %v1, 0 + br i1 %v2, label %bb3, label %bb0, !prof !15 + +bb0: + %v3 = and i32 %i0, 2 + %v4 = icmp eq i32 %v3, 0 + %v8 = add i32 %sum0, 43 + %sum2 = select i1 %v4, i32 %sum0, i32 %v8, !prof !15 + %v9 = and i32 %j0, 4 + %v10 = icmp eq i32 %v9, 0 + br i1 %v10, label %bb2, label %bb1, !prof !15 + +bb1: + %sum3 = add i32 %sum2, 44 + %v11 = and i32 %i0, 8 + %v12 = icmp eq i32 %v11, 0 + %v13 = add i32 %sum3, 44 + %sum4 = select i1 %v12, i32 %sum3, i32 %v13, !prof !15 + br label %bb2 + +bb2: + %sum5 = phi i32 [ %sum2, %bb0 ], [ %sum4, %bb1 ] + br label %bb3 + +bb3: + %sum6 = phi i32 [ %sum0, %entry ], [ %sum5, %bb2 ] + ret i32 %sum6 +} + +; Selects + Brs, the branch condition can't be hoisted to be merged with a +; select. No CHR happens. +; Roughly, +; i0 = *i +; sum = ((i0 & 2) == 0) ? sum0 : (sum0 + 43) // Likely false +; foo(); +; j0 = *j +; if ((j0 & 4) != 0) { // Likely true +; foo(); +; sum = sum + 44 +; } +; return sum +; -> +; (no change) +define i32 @test_chr_7(i32* %i, i32* %j, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_7( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[I0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[V3:%.*]] = and i32 [[I0]], 2 +; CHECK-NEXT: [[V4:%.*]] = icmp eq i32 [[V3]], 0 +; CHECK-NEXT: [[V8:%.*]] = add i32 [[SUM0:%.*]], 43 +; CHECK-NEXT: [[SUM2:%.*]] = select i1 [[V4]], i32 [[SUM0]], i32 [[V8]], !prof !16 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[J0:%.*]] = load i32, i32* [[J:%.*]], align 4 +; CHECK-NEXT: [[V9:%.*]] = and i32 [[J0]], 4 +; CHECK-NEXT: [[V10:%.*]] = icmp eq i32 [[V9]], 0 +; CHECK-NEXT: br i1 [[V10]], label [[BB2:%.*]], label [[BB1:%.*]], !prof !16 +; CHECK: bb1: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[SUM4:%.*]] = add i32 [[SUM2]], 44 +; CHECK-NEXT: br label [[BB2]] +; CHECK: bb2: +; CHECK-NEXT: [[SUM5:%.*]] = phi i32 [ [[SUM2]], [[ENTRY:%.*]] ], [ [[SUM4]], [[BB1]] ] +; CHECK-NEXT: ret i32 [[SUM5]] +; +entry: + %i0 = load i32, i32* %i + %v3 = and i32 %i0, 2 + %v4 = icmp eq i32 %v3, 0 + %v8 = add i32 %sum0, 43 + %sum2 = select i1 %v4, i32 %sum0, i32 %v8, !prof !15 + call void @foo() + %j0 = load i32, i32* %j + %v9 = and i32 %j0, 4 + %v10 = icmp eq i32 %v9, 0 + br i1 %v10, label %bb2, label %bb1, !prof !15 ; %v10 can't be hoisted above the above select + +bb1: + call void @foo() + %sum4 = add i32 %sum2, 44 + br label %bb2 + +bb2: + %sum5 = phi i32 [ %sum2, %entry ], [ %sum4, %bb1 ] + ret i32 %sum5 +} + +; Selects + Brs, the branch condition can't be hoisted to be merged with the +; selects. Dropping the select. +; Roughly, +; i0 = *i +; sum = ((i0 & 2) == 0) ? sum0 : (sum0 + 43) // Likely false +; foo(); +; j0 = *j +; if ((j0 & 4) != 0) // Likely true +; foo() +; if ((j0 & 8) != 0) // Likely true +; foo() +; return sum +; -> +; i0 = *i +; sum = ((i0 & 2) == 0) ? sum0 : (sum0 + 43) // Likely false +; foo(); +; j0 = *j +; if ((j0 & 12) != 12) { // Likely true +; foo() +; foo() +; } else { +; if ((j0 & 4) != 0) +; foo() +; if ((j0 & 8) != 0) +; foo() +; } +; return sum +define i32 @test_chr_7_1(i32* %i, i32* %j, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_7_1( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[I0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[V3:%.*]] = and i32 [[I0]], 2 +; CHECK-NEXT: [[V4:%.*]] = icmp eq i32 [[V3]], 0 +; CHECK-NEXT: [[V8:%.*]] = add i32 [[SUM0:%.*]], 43 +; CHECK-NEXT: [[SUM2:%.*]] = select i1 [[V4]], i32 [[SUM0]], i32 [[V8]], !prof !16 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[J0:%.*]] = load i32, i32* [[J:%.*]], align 4 +; CHECK-NEXT: [[TMP0:%.*]] = and i32 [[J0]], 12 +; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[TMP0]], 12 +; CHECK-NEXT: br i1 [[TMP1]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[V9:%.*]] = and i32 [[J0]], 4 +; CHECK-NEXT: [[V10:%.*]] = icmp eq i32 [[V9]], 0 +; CHECK-NEXT: br i1 [[V10]], label [[BB1_NONCHR:%.*]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[V11_NONCHR:%.*]] = and i32 [[J0]], 8 +; CHECK-NEXT: [[V12_NONCHR:%.*]] = icmp eq i32 [[V11_NONCHR]], 0 +; CHECK-NEXT: br i1 [[V12_NONCHR]], label [[BB3]], label [[BB2_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: ret i32 [[SUM2]] +; +entry: + %i0 = load i32, i32* %i + %v3 = and i32 %i0, 2 + %v4 = icmp eq i32 %v3, 0 + %v8 = add i32 %sum0, 43 + %sum2 = select i1 %v4, i32 %sum0, i32 %v8, !prof !15 + call void @foo() + %j0 = load i32, i32* %j + %v9 = and i32 %j0, 4 + %v10 = icmp eq i32 %v9, 0 + br i1 %v10, label %bb1, label %bb0, !prof !15 ; %v10 can't be hoisted above the above select + +bb0: + call void @foo() + br label %bb1 + +bb1: + %v11 = and i32 %j0, 8 + %v12 = icmp eq i32 %v11, 0 + br i1 %v12, label %bb3, label %bb2, !prof !15 + +bb2: + call void @foo() + br label %bb3 + +bb3: + ret i32 %sum2 +} + +; Branches aren't biased enough. No CHR happens. +; Roughly, +; t0 = *i +; if ((t0 & 1) != 0) // Not biased +; foo() +; if ((t0 & 2) != 0) // Not biased +; foo() +; -> +; (no change) +define void @test_chr_8(i32* %i) !prof !14 { +; CHECK-LABEL: @test_chr_8( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 0 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB1:%.*]], label [[BB0:%.*]], !prof !17 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1]] +; CHECK: bb1: +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 0 +; CHECK-NEXT: br i1 [[TMP4]], label [[BB3:%.*]], label [[BB2:%.*]], !prof !17 +; CHECK: bb2: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: ret void +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !16 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %3 = and i32 %0, 2 + %4 = icmp eq i32 %3, 0 + br i1 %4, label %bb3, label %bb2, !prof !16 + +bb2: + call void @foo() + br label %bb3 + +bb3: + ret void +} + +; With an existing phi at the exit. +; Roughly, +; t = *i +; if ((t0 & 1) != 0) // Likely true +; foo() +; if ((t0 & 2) != 0) { // Likely true +; t = *j +; foo() +; } +; // There's a phi for t here. +; return t +; -> +; t = *i +; if ((t & 3) == 3) { // Likely true +; foo() +; t = *j +; foo() +; } else { +; if ((t & 1) != 0) +; foo() +; if ((t & 2) != 0) { +; t = *j +; foo() +; } +; } +; // There's a phi for t here. +; return t +define i32 @test_chr_9(i32* %i, i32* %j) !prof !14 { +; CHECK-LABEL: @test_chr_9( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[TMP3:%.*]] = load i32, i32* [[J:%.*]], align 4 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP4:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 0 +; CHECK-NEXT: br i1 [[TMP5]], label [[BB1_NONCHR:%.*]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP6:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i32 [[TMP6]], 0 +; CHECK-NEXT: br i1 [[TMP7]], label [[BB3]], label [[BB2_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: [[TMP8:%.*]] = load i32, i32* [[J]], align 4 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[TMP9:%.*]] = phi i32 [ [[TMP3]], [[BB0]] ], [ [[TMP0]], [[BB1_NONCHR]] ], [ [[TMP8]], [[BB2_NONCHR]] ] +; CHECK-NEXT: ret i32 [[TMP9]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %3 = and i32 %0, 2 + %4 = icmp eq i32 %3, 0 + br i1 %4, label %bb3, label %bb2, !prof !15 + +bb2: + %5 = load i32, i32* %j + call void @foo() + br label %bb3 + +bb3: + %6 = phi i32 [ %0, %bb1 ], [ %5, %bb2 ] + ret i32 %6 +} + +; With no phi at the exit, but the exit needs a phi inserted after CHR. +; Roughly, +; t0 = *i +; if ((t0 & 1) != 0) // Likely true +; foo() +; t1 = *j +; if ((t1 & 2) != 0) // Likely true +; foo() +; return (t1 * 42) - (t1 - 99) +; -> +; t0 = *i +; if ((t0 & 3) == 3) { // Likely true +; foo() +; t1 = *j +; foo() +; } else { +; if ((t0 & 1) != 0) +; foo() +; if ((t0 & 2) != 0) { +; t1 = *j +; foo() +; } +; } +; // A new phi for t1 is inserted here. +; return (t1 * 42) - (t1 - 99) +define i32 @test_chr_10(i32* %i, i32* %j) !prof !14 { +; CHECK-LABEL: @test_chr_10( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[TMP3:%.*]] = load i32, i32* [[J:%.*]], align 4 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP4:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 0 +; CHECK-NEXT: br i1 [[TMP5]], label [[BB1_NONCHR:%.*]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP6:%.*]] = load i32, i32* [[J]], align 4 +; CHECK-NEXT: [[TMP7:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0 +; CHECK-NEXT: br i1 [[TMP8]], label [[BB3]], label [[BB2_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[TMP9:%.*]] = phi i32 [ [[TMP3]], [[BB0]] ], [ [[TMP6]], [[BB2_NONCHR]] ], [ [[TMP6]], [[BB1_NONCHR]] ] +; CHECK-NEXT: [[TMP10:%.*]] = mul i32 [[TMP9]], 42 +; CHECK-NEXT: [[TMP11:%.*]] = add i32 [[TMP9]], -99 +; CHECK-NEXT: [[TMP12:%.*]] = add i32 [[TMP10]], [[TMP11]] +; CHECK-NEXT: ret i32 [[TMP12]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %3 = load i32, i32* %j + %4 = and i32 %0, 2 + %5 = icmp eq i32 %4, 0 + br i1 %5, label %bb3, label %bb2, !prof !15 + +bb2: + call void @foo() + br label %bb3 + +bb3: + %6 = mul i32 %3, 42 + %7 = sub i32 %3, 99 + %8 = add i32 %6, %7 + ret i32 %8 +} + +; Test a case where there are two use-def chain paths to the same value (t0) +; from the branch condition. This is a regression test for an old bug that +; caused a bad hoisting that moves (hoists) a value (%conv) twice to the end of +; the %entry block (once for %div and once for %mul16) and put a use ahead of +; its definition like: +; %entry: +; ... +; %div = fdiv double 1.000000e+00, %conv +; %conv = sitofp i32 %0 to double +; %mul16 = fmul double %div, %conv +; +; Roughly, +; t0 = *i +; if ((t0 & 1) != 0) // Likely true +; foo() +; // there are two use-def paths from the branch condition to t0. +; if ((1.0 / t0) * t0 < 1) // Likely true +; foo() +; -> +; t0 = *i +; if ((t0 & 1) != 0 & (1.0 / t0) * t0 > 0) { // Likely true +; foo() +; foo() +; } else { +; if ((t0 & 1) != 0) +; foo() +; if ((1.0 / t0) * t0 < 1) // Likely true +; foo() +; } +define void @test_chr_11(i32* %i, i32 %x) !prof !14 { +; CHECK-LABEL: @test_chr_11( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP2:%.*]] = icmp ne i32 [[TMP1]], 0 +; CHECK-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP0]] to double +; CHECK-NEXT: [[DIV:%.*]] = fdiv double 1.000000e+00, [[CONV]] +; CHECK-NEXT: [[MUL16:%.*]] = fmul double [[DIV]], [[CONV]] +; CHECK-NEXT: [[CONV717:%.*]] = fptosi double [[MUL16]] to i32 +; CHECK-NEXT: [[CMP18:%.*]] = icmp sgt i32 [[CONV717]], 0 +; CHECK-NEXT: [[TMP3:%.*]] = and i1 [[TMP2]], [[CMP18]] +; CHECK-NEXT: br i1 [[TMP3]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0_NONCHR:%.*]], label [[BB1_NONCHR:%.*]], !prof !18 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[CONV_NONCHR:%.*]] = sitofp i32 [[TMP0]] to double +; CHECK-NEXT: [[DIV_NONCHR:%.*]] = fdiv double 1.000000e+00, [[CONV_NONCHR]] +; CHECK-NEXT: [[MUL16_NONCHR:%.*]] = fmul double [[DIV_NONCHR]], [[CONV_NONCHR]] +; CHECK-NEXT: [[CONV717_NONCHR:%.*]] = fptosi double [[MUL16_NONCHR]] to i32 +; CHECK-NEXT: [[CMP18_NONCHR:%.*]] = icmp slt i32 [[CONV717_NONCHR]], 1 +; CHECK-NEXT: br i1 [[CMP18_NONCHR]], label [[BB3]], label [[BB2_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: ret void +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %conv = sitofp i32 %0 to double + %div = fdiv double 1.000000e+00, %conv + %mul16 = fmul double %div, %conv + %conv717 = fptosi double %mul16 to i32 + %cmp18 = icmp slt i32 %conv717, 1 + br i1 %cmp18, label %bb3, label %bb2, !prof !15 + +bb2: + call void @foo() + br label %bb3 + +bb3: + ret void +} + +; Selects + unrelated br only +define i32 @test_chr_12(i32* %i, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_12( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 255 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 0 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB3:%.*]], label [[BB0:%.*]], !prof !16 +; CHECK: bb0: +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 0 +; CHECK-NEXT: [[TMP5:%.*]] = add i32 [[SUM0:%.*]], 42 +; CHECK-NEXT: [[SUM1:%.*]] = select i1 [[TMP4]], i32 [[SUM0]], i32 [[TMP5]], !prof !16 +; CHECK-NEXT: [[TMP6:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i32 [[TMP6]], 0 +; CHECK-NEXT: [[TMP8:%.*]] = add i32 [[SUM1]], 43 +; CHECK-NEXT: [[SUM2:%.*]] = select i1 [[TMP7]], i32 [[SUM1]], i32 [[TMP8]], !prof !16 +; CHECK-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4 +; CHECK-NEXT: [[TMP10:%.*]] = icmp ne i32 [[TMP9]], 0 +; CHECK-NEXT: [[TMP11:%.*]] = and i32 [[TMP0]], 8 +; CHECK-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0 +; CHECK-NEXT: [[TMP13:%.*]] = and i1 [[TMP10]], [[TMP12]] +; CHECK-NEXT: br i1 [[TMP13]], label [[BB1:%.*]], label [[BB0_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb1: +; CHECK-NEXT: [[TMP14:%.*]] = add i32 [[SUM2]], 88 +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb0.split.nonchr: +; CHECK-NEXT: br i1 [[TMP10]], label [[BB1_NONCHR:%.*]], label [[BB3]], !prof !18 +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP15:%.*]] = and i32 [[TMP0]], 8 +; CHECK-NEXT: [[TMP16:%.*]] = icmp eq i32 [[TMP15]], 0 +; CHECK-NEXT: [[SUM4_NONCHR_V:%.*]] = select i1 [[TMP16]], i32 44, i32 88, !prof !16 +; CHECK-NEXT: [[SUM4_NONCHR:%.*]] = add i32 [[SUM2]], [[SUM4_NONCHR_V]] +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[SUM6:%.*]] = phi i32 [ [[SUM0]], [[ENTRY:%.*]] ], [ [[TMP14]], [[BB1]] ], [ [[SUM2]], [[BB0_SPLIT_NONCHR]] ], [ [[SUM4_NONCHR]], [[BB1_NONCHR]] ] +; CHECK-NEXT: ret i32 [[SUM6]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 255 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb3, label %bb0, !prof !15 + +bb0: + %3 = and i32 %0, 1 + %4 = icmp eq i32 %3, 0 + %5 = add i32 %sum0, 42 + %sum1 = select i1 %4, i32 %sum0, i32 %5, !prof !15 + %6 = and i32 %0, 2 + %7 = icmp eq i32 %6, 0 + %8 = add i32 %sum1, 43 + %sum2 = select i1 %7, i32 %sum1, i32 %8, !prof !15 + %9 = load i32, i32* %i + %10 = icmp eq i32 %9, 0 + br i1 %10, label %bb2, label %bb1, !prof !15 + +bb1: + %sum3 = add i32 %sum2, 44 + %11 = and i32 %0, 8 + %12 = icmp eq i32 %11, 0 + %13 = add i32 %sum3, 44 + %sum4 = select i1 %12, i32 %sum3, i32 %13, !prof !15 + br label %bb2 + +bb2: + %sum5 = phi i32 [ %sum2, %bb0 ], [ %sum4, %bb1 ] + br label %bb3 + +bb3: + %sum6 = phi i32 [ %sum0, %entry ], [ %sum5, %bb2 ] + ret i32 %sum6 +} + +; In the second CHR, a condition value depends on a trivial phi that's inserted +; by the first CHR. +; Roughly, +; i0 = *i +; v2 = (z != 1) ? pred : true // Likely false +; if (z == 0 & pred) // Likely false +; foo() +; j0 = *j +; sum2 = ((i0 & 2) == j0) ? sum0 : (sum0 + 43) // Likely false +; sum3 = ((i0 == j0) ? sum0 : (sum0 + 43) // Likely false +; foo() +; if ((i0 & 4) == 0) // Unbiased +; foo() +; return i0 + sum3 +; -> +; i0 = *i +; if (z != 1 & (z == 0 & pred)) // First CHR +; foo() +; // A trivial phi for i0 is inserted here by the first CHR (which gets removed +; // later) and the subsequent branch condition (for the second CHR) uses it. +; j0 = *j +; if ((i0 & 2) != j0 & i0 != j0) { // Second CHR +; sum3 = sum0 + 43 +; foo() +; if (i0 & 4) == 0) +; foo() +; } else { +; sum3 = (i0 == j0) ? sum0 : (sum0 + 43) +; foo() +; if (i0 & 4) == 0) +; foo() +; } +; return i0 + sum3 +define i32 @test_chr_14(i32* %i, i32* %j, i32 %sum0, i1 %pred, i32 %z) !prof !14 { +; CHECK-LABEL: @test_chr_14( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[I0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[V1:%.*]] = icmp ne i32 [[Z:%.*]], 1 +; CHECK-NEXT: [[V0:%.*]] = icmp eq i32 [[Z]], 0 +; CHECK-NEXT: [[V3_NONCHR:%.*]] = and i1 [[V0]], [[PRED:%.*]] +; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[V1]], [[V3_NONCHR]] +; CHECK-NEXT: br i1 [[OR_COND]], label [[BB0_NONCHR:%.*]], label [[BB1:%.*]], !prof !19 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1]] +; CHECK: bb1: +; CHECK-NEXT: [[J0:%.*]] = load i32, i32* [[J:%.*]], align 4 +; CHECK-NEXT: [[V6:%.*]] = and i32 [[I0]], 2 +; CHECK-NEXT: [[V4:%.*]] = icmp ne i32 [[V6]], [[J0]] +; CHECK-NEXT: [[V8:%.*]] = add i32 [[SUM0:%.*]], 43 +; CHECK-NEXT: [[V5:%.*]] = icmp ne i32 [[I0]], [[J0]] +; CHECK-NEXT: [[TMP0:%.*]] = and i1 [[V4]], [[V5]] +; CHECK-NEXT: br i1 [[TMP0]], label [[BB1_SPLIT:%.*]], label [[BB1_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb1.split: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[V9:%.*]] = and i32 [[I0]], 4 +; CHECK-NEXT: [[V10:%.*]] = icmp eq i32 [[V9]], 0 +; CHECK-NEXT: br i1 [[V10]], label [[BB3:%.*]], label [[BB2:%.*]] +; CHECK: bb2: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb1.split.nonchr: +; CHECK-NEXT: [[V5_NONCHR:%.*]] = icmp eq i32 [[I0]], [[J0]] +; CHECK-NEXT: [[SUM3_NONCHR:%.*]] = select i1 [[V5_NONCHR]], i32 [[SUM0]], i32 [[V8]], !prof !16 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[V9_NONCHR:%.*]] = and i32 [[I0]], 4 +; CHECK-NEXT: [[V10_NONCHR:%.*]] = icmp eq i32 [[V9_NONCHR]], 0 +; CHECK-NEXT: br i1 [[V10_NONCHR]], label [[BB3]], label [[BB2_NONCHR:%.*]] +; CHECK: bb2.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[TMP1:%.*]] = phi i32 [ [[V8]], [[BB2]] ], [ [[V8]], [[BB1_SPLIT]] ], [ [[SUM3_NONCHR]], [[BB2_NONCHR]] ], [ [[SUM3_NONCHR]], [[BB1_SPLIT_NONCHR]] ] +; CHECK-NEXT: [[V11:%.*]] = add i32 [[I0]], [[TMP1]] +; CHECK-NEXT: ret i32 [[V11]] +; +entry: + %i0 = load i32, i32* %i + %v0 = icmp eq i32 %z, 0 + %v1 = icmp ne i32 %z, 1 + %v2 = select i1 %v1, i1 %pred, i1 true, !prof !15 + %v3 = and i1 %v0, %pred + br i1 %v3, label %bb0, label %bb1, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %j0 = load i32, i32* %j + %v6 = and i32 %i0, 2 + %v4 = icmp eq i32 %v6, %j0 + %v8 = add i32 %sum0, 43 + %sum2 = select i1 %v4, i32 %sum0, i32 %v8, !prof !15 + %v5 = icmp eq i32 %i0, %j0 + %sum3 = select i1 %v5, i32 %sum0, i32 %v8, !prof !15 + call void @foo() + %v9 = and i32 %i0, 4 + %v10 = icmp eq i32 %v9, 0 + br i1 %v10, label %bb3, label %bb2 + +bb2: + call void @foo() + br label %bb3 + +bb3: + %v11 = add i32 %i0, %sum3 + ret i32 %v11 +} + +; Branch or selects depends on another select. No CHR happens. +; Roughly, +; i0 = *i +; if (z == 0 & ((z != 1) ? pred : true)) { // Likely false +; foo() +; j0 = *j +; sum2 = ((i0 & 2) == j0) ? sum0 : (sum0 + 43) // Likely false +; sum3 = (i0 == sum2) ? sum2 : (sum0 + 43) // Likely false. This depends on the +; // previous select. +; foo() +; if ((i0 & 4) == 0) // Unbiased +; foo() +; return i0 + sum3 +; -> +; (no change) +define i32 @test_chr_15(i32* %i, i32* %j, i32 %sum0, i1 %pred, i32 %z) !prof !14 { +; CHECK-LABEL: @test_chr_15( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[I0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[V0:%.*]] = icmp eq i32 [[Z:%.*]], 0 +; CHECK-NEXT: [[V3:%.*]] = and i1 [[V0]], [[PRED:%.*]] +; CHECK-NEXT: br i1 [[V3]], label [[BB0:%.*]], label [[BB1:%.*]], !prof !16 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1]] +; CHECK: bb1: +; CHECK-NEXT: [[J0:%.*]] = load i32, i32* [[J:%.*]], align 4 +; CHECK-NEXT: [[V6:%.*]] = and i32 [[I0]], 2 +; CHECK-NEXT: [[V4:%.*]] = icmp eq i32 [[V6]], [[J0]] +; CHECK-NEXT: [[V8:%.*]] = add i32 [[SUM0:%.*]], 43 +; CHECK-NEXT: [[SUM2:%.*]] = select i1 [[V4]], i32 [[SUM0]], i32 [[V8]], !prof !16 +; CHECK-NEXT: [[V5:%.*]] = icmp eq i32 [[I0]], [[SUM2]] +; CHECK-NEXT: [[SUM3:%.*]] = select i1 [[V5]], i32 [[SUM2]], i32 [[V8]], !prof !16 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[V9:%.*]] = and i32 [[I0]], 4 +; CHECK-NEXT: [[V10:%.*]] = icmp eq i32 [[V9]], 0 +; CHECK-NEXT: br i1 [[V10]], label [[BB3:%.*]], label [[BB2:%.*]] +; CHECK: bb2: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[V11:%.*]] = add i32 [[I0]], [[SUM3]] +; CHECK-NEXT: ret i32 [[V11]] +; +entry: + %i0 = load i32, i32* %i + %v0 = icmp eq i32 %z, 0 + %v1 = icmp ne i32 %z, 1 + %v2 = select i1 %v1, i1 %pred, i1 true, !prof !15 + %v3 = and i1 %v0, %v2 + br i1 %v3, label %bb0, label %bb1, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %j0 = load i32, i32* %j + %v6 = and i32 %i0, 2 + %v4 = icmp eq i32 %v6, %j0 + %v8 = add i32 %sum0, 43 + %sum2 = select i1 %v4, i32 %sum0, i32 %v8, !prof !15 + %v5 = icmp eq i32 %i0, %sum2 + %sum3 = select i1 %v5, i32 %sum2, i32 %v8, !prof !15 + call void @foo() + %v9 = and i32 %i0, 4 + %v10 = icmp eq i32 %v9, 0 + br i1 %v10, label %bb3, label %bb2 + +bb2: + call void @foo() + br label %bb3 + +bb3: + %v11 = add i32 %i0, %sum3 + ret i32 %v11 +} + +; With an existing phi at the exit but a value (%v40) is both alive and is an +; operand to a phi at the exit block. +; Roughly, +; t0 = *i +; if ((t0 & 1) != 0) // Likely true +; foo() +; v40 = t0 + 44 +; if ((t0 & 2) != 0) // Likely true +; v41 = t0 + 99 +; foo() +; } +; v42 = phi v40, v41 +; return v42 + v40 +; -> +; t0 = *i +; if ((t0 & 3) == 3) // Likely true +; foo() +; v40 = t0 + 44 +; v41 = t0 + 99 +; foo() +; } else { +; if ((t0 & 1) != 0) // Likely true +; foo() +; v40_nc = t0 + 44 +; if ((t0 & 2) != 0) // Likely true +; v41_nc = t0 + 99 +; foo() +; } +; } +; t7 = phi v40, v40_nc +; v42 = phi v41, v41_nc +; v43 = v42 + t7 +; return v43 +define i32 @test_chr_16(i32* %i) !prof !14 { +; CHECK-LABEL: @test_chr_16( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 3 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[V40:%.*]] = add i32 [[TMP0]], 44 +; CHECK-NEXT: [[V41:%.*]] = add i32 [[TMP0]], 99 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP3:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i32 [[TMP3]], 0 +; CHECK-NEXT: br i1 [[TMP4]], label [[BB1_NONCHR:%.*]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB1_NONCHR]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[V40_NONCHR:%.*]] = add i32 [[TMP0]], 44 +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 2 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: br i1 [[TMP6]], label [[BB3]], label [[BB2_NONCHR:%.*]], !prof !16 +; CHECK: bb2.nonchr: +; CHECK-NEXT: [[V41_NONCHR:%.*]] = add i32 [[TMP0]], 99 +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[TMP7:%.*]] = phi i32 [ [[V40]], [[BB0]] ], [ [[V40_NONCHR]], [[BB2_NONCHR]] ], [ [[V40_NONCHR]], [[BB1_NONCHR]] ] +; CHECK-NEXT: [[V42:%.*]] = phi i32 [ [[V41]], [[BB0]] ], [ [[V41_NONCHR]], [[BB2_NONCHR]] ], [ [[V40_NONCHR]], [[BB1_NONCHR]] ] +; CHECK-NEXT: [[V43:%.*]] = add i32 [[V42]], [[TMP7]] +; CHECK-NEXT: ret i32 [[V43]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 1 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + br label %bb1 + +bb1: + %v40 = add i32 %0, 44 + %3 = and i32 %0, 2 + %4 = icmp eq i32 %3, 0 + br i1 %4, label %bb3, label %bb2, !prof !15 + +bb2: + %v41 = add i32 %0, 99 + call void @foo() + br label %bb3 + +bb3: + %v42 = phi i32 [ %v41, %bb2 ], [ %v40, %bb1 ] + %v43 = add i32 %v42, %v40 + ret i32 %v43 +} + +; Two consecutive regions have an entry in the middle of them. No CHR happens. +; Roughly, +; if ((i & 4) == 0) { +; if (!j) +; goto bb1 +; } else { +; t0 = (i & 1) +; if (t0 != 0) // Likely true +; foo() +; s = (i & 1) + i +; } +; bb1: +; p = phi i, t0, s +; if ((i & 2) != 0) // Likely true +; foo() +; q = p + 2 +; } +; r = phi p, q, i +; return r +; -> +; (no change) +define i32 @test_chr_17(i32 %i, i1 %j) !prof !14 { +; CHECK-LABEL: @test_chr_17( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[V0:%.*]] = and i32 [[I:%.*]], 4 +; CHECK-NEXT: [[V1:%.*]] = icmp eq i32 [[V0]], 0 +; CHECK-NEXT: br i1 [[V1]], label [[BBE:%.*]], label [[BBQ:%.*]] +; CHECK: bbq: +; CHECK-NEXT: br i1 [[J:%.*]], label [[BB3:%.*]], label [[BB1:%.*]] +; CHECK: bbe: +; CHECK-NEXT: [[TMP0:%.*]] = and i32 [[I]], 1 +; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[TMP0]], 0 +; CHECK-NEXT: br i1 [[TMP1]], label [[BB1]], label [[BB0:%.*]], !prof !16 +; CHECK: bb0: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[S:%.*]] = add i32 [[TMP0]], [[I]] +; CHECK-NEXT: br label [[BB1]] +; CHECK: bb1: +; CHECK-NEXT: [[P:%.*]] = phi i32 [ [[I]], [[BBQ]] ], [ [[TMP0]], [[BBE]] ], [ [[S]], [[BB0]] ] +; CHECK-NEXT: [[TMP2:%.*]] = and i32 [[I]], 2 +; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i32 [[TMP2]], 0 +; CHECK-NEXT: br i1 [[TMP3]], label [[BB3]], label [[BB2:%.*]], !prof !16 +; CHECK: bb2: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[Q:%.*]] = add i32 [[P]], [[TMP2]] +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[R:%.*]] = phi i32 [ [[P]], [[BB1]] ], [ [[Q]], [[BB2]] ], [ [[I]], [[BBQ]] ] +; CHECK-NEXT: ret i32 [[R]] +; +entry: + %v0 = and i32 %i, 4 + %v1 = icmp eq i32 %v0, 0 + br i1 %v1, label %bbe, label %bbq + +bbq: + br i1 %j, label %bb3, label %bb1 + +bbe: + %0 = and i32 %i, 1 + %1 = icmp eq i32 %0, 0 + br i1 %1, label %bb1, label %bb0, !prof !15 + +bb0: + call void @foo() + %s = add i32 %0, %i + br label %bb1 + +bb1: + %p = phi i32 [ %i, %bbq ], [ %0, %bbe ], [ %s, %bb0 ] + %2 = and i32 %i, 2 + %3 = icmp eq i32 %2, 0 + br i1 %3, label %bb3, label %bb2, !prof !15 + +bb2: + call void @foo() + %q = add i32 %p, %2 + br label %bb3 + +bb3: + %r = phi i32 [ %p, %bb1 ], [ %q, %bb2 ], [ %i, %bbq ] + ret i32 %r +} + +; Select + br, there's a loop and we need to update the user of an inserted phi +; at the entry block. This is a regression test for a bug that's fixed. +; Roughly, +; do { +; inc1 = phi inc2, 0 +; li = *i +; sum1 = sum0 + 42 +; sum2 = ((li & 1) == 0) ? sum0 : sum1 // Likely false +; inc2 = inc1 + 1 +; if ((li & 4) != 0) // Likely true +; sum3 = sum2 + 44 +; sum4 = phi sum1, sum3 +; } while (inc2 != 100) // Likely true (loop back) +; return sum4 +; -> +; do { +; inc1 = phi tmp2, 0 // The first operand needed to be updated +; li = *i +; sum1 = sum0 + 42 +; if ((li & 5) == 5) { // Likely true +; inc2 = inc1 + 1 +; sum3 = sum0 + 86 +; } else { +; inc2_nc = inc1 + 1 +; if ((li & 4) == 0) +; sum2_nc = ((li & 1) == 0) ? sum0 : sum1 +; sum3_nc = sum2_nc + 44 +; } +; tmp2 = phi inc2, in2c_nc +; sum4 = phi sum3, sum3_nc, sum1 +; } while (tmp2 != 100) +; return sum4 +define i32 @test_chr_18(i32* %i, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_18( +; CHECK-NEXT: entry: +; CHECK-NEXT: br label [[BB0:%.*]] +; CHECK: bb0: +; CHECK-NEXT: [[INC1:%.*]] = phi i32 [ [[TMP2:%.*]], [[BB2:%.*]] ], [ 0, [[ENTRY:%.*]] ] +; CHECK-NEXT: [[LI:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[SUM1:%.*]] = add i32 [[SUM0:%.*]], 42 +; CHECK-NEXT: [[TMP0:%.*]] = and i32 [[LI]], 5 +; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[TMP0]], 5 +; CHECK-NEXT: br i1 [[TMP1]], label [[BB0_SPLIT:%.*]], label [[BB0_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0.split: +; CHECK-NEXT: [[INC2:%.*]] = add i32 [[INC1]], 1 +; CHECK-NEXT: [[SUM3:%.*]] = add i32 [[SUM0]], 86 +; CHECK-NEXT: br label [[BB2]] +; CHECK: bb0.split.nonchr: +; CHECK-NEXT: [[A4_NONCHR:%.*]] = and i32 [[LI]], 4 +; CHECK-NEXT: [[CMP4_NONCHR:%.*]] = icmp eq i32 [[A4_NONCHR]], 0 +; CHECK-NEXT: [[INC2_NONCHR:%.*]] = add i32 [[INC1]], 1 +; CHECK-NEXT: br i1 [[CMP4_NONCHR]], label [[BB2]], label [[BB1_NONCHR:%.*]], !prof !16 +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[A1:%.*]] = and i32 [[LI]], 1 +; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i32 [[A1]], 0 +; CHECK-NEXT: [[SUM2_NONCHR:%.*]] = select i1 [[CMP1]], i32 [[SUM0]], i32 [[SUM1]], !prof !16 +; CHECK-NEXT: [[SUM3_NONCHR:%.*]] = add i32 [[SUM2_NONCHR]], 44 +; CHECK-NEXT: br label [[BB2]] +; CHECK: bb2: +; CHECK-NEXT: [[TMP2]] = phi i32 [ [[INC2]], [[BB0_SPLIT]] ], [ [[INC2_NONCHR]], [[BB1_NONCHR]] ], [ [[INC2_NONCHR]], [[BB0_SPLIT_NONCHR]] ] +; CHECK-NEXT: [[SUM4:%.*]] = phi i32 [ [[SUM3]], [[BB0_SPLIT]] ], [ [[SUM3_NONCHR]], [[BB1_NONCHR]] ], [ [[SUM1]], [[BB0_SPLIT_NONCHR]] ] +; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[TMP2]], 100 +; CHECK-NEXT: br i1 [[CMP]], label [[BB3:%.*]], label [[BB0]], !prof !16 +; CHECK: bb3: +; CHECK-NEXT: ret i32 [[SUM4]] +; +entry: + br label %bb0 + +bb0: + %inc1 = phi i32 [ %inc2, %bb2 ], [ 0, %entry ] + %li = load i32, i32* %i + %a1 = and i32 %li, 1 + %cmp1 = icmp eq i32 %a1, 0 + %sum1 = add i32 %sum0, 42 + %sum2 = select i1 %cmp1, i32 %sum0, i32 %sum1, !prof !15 + %a4 = and i32 %li, 4 + %cmp4 = icmp eq i32 %a4, 0 + %inc2 = add i32 %inc1, 1 + br i1 %cmp4, label %bb2, label %bb1, !prof !15 + +bb1: + %sum3 = add i32 %sum2, 44 + br label %bb2 + +bb2: + %sum4 = phi i32 [ %sum1, %bb0 ], [ %sum3, %bb1 ] + %cmp = icmp eq i32 %inc2, 100 + br i1 %cmp, label %bb3, label %bb0, !prof !15 + +bb3: + ret i32 %sum4 +} + + +; Selects + Brs. Those share the condition value, which causes the +; targets/operands of the branch/select to be flipped. +; Roughly, +; t0 = *i +; if ((t0 & 255) != 0) { // Likely true +; sum1 = ((t0 & 1) == 0) ? sum0 : (sum0 + 42) // Likely false +; sum2 = ((t0 & 1) == 0) ? sum1 : (sum1 + 42) // Likely false +; if ((t0 & 1) != 0) { // Likely true +; sum3 = sum2 + 44 +; sum4 = ((t0 & 8) == 0) ? sum3 : (sum3 + 44) // Likely false +; } +; sum5 = phi sum2, sum4 +; } +; sum6 = phi sum0, sum5 +; return sum6 +; -> +; t0 = *i +; if ((t0 & 9) == 9) { // Likely true +; tmp3 = sum0 + 85 // Dead +; tmp4 = sum0 + 173 +; } else { +; if ((t0 & 255) != 0) { +; sum2_nc = ((t0 & 1) == 0) ? sum0 : (sum0 + 85) +; sum4_nc_v = ((t0 & 8) == 0) ? 44 : 88 +; sum4_nc = add sum2_nc + sum4_nc_v +; } +; } +; sum6 = phi tmp4, sum0, sum2_nc, sum4_nc +; return sum6 +define i32 @test_chr_19(i32* %i, i32 %sum0) !prof !14 { +; CHECK-LABEL: @test_chr_19( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP1:%.*]] = and i32 [[TMP0]], 9 +; CHECK-NEXT: [[TMP2:%.*]] = icmp eq i32 [[TMP1]], 9 +; CHECK-NEXT: br i1 [[TMP2]], label [[BB0:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: bb0: +; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[SUM0:%.*]], 85 +; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[SUM0]], 173 +; CHECK-NEXT: br label [[BB3:%.*]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[TMP5:%.*]] = and i32 [[TMP0]], 255 +; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i32 [[TMP5]], 0 +; CHECK-NEXT: br i1 [[TMP6]], label [[BB3]], label [[BB0_NONCHR:%.*]], !prof !16 +; CHECK: bb0.nonchr: +; CHECK-NEXT: [[TMP7:%.*]] = and i32 [[TMP0]], 1 +; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i32 [[TMP7]], 0 +; CHECK-NEXT: [[TMP9:%.*]] = add i32 [[SUM0]], 85 +; CHECK-NEXT: [[SUM2_NONCHR:%.*]] = select i1 [[TMP8]], i32 [[SUM0]], i32 [[TMP9]], !prof !16 +; CHECK-NEXT: br i1 [[TMP8]], label [[BB3]], label [[BB1_NONCHR:%.*]], !prof !16 +; CHECK: bb1.nonchr: +; CHECK-NEXT: [[TMP10:%.*]] = and i32 [[TMP0]], 8 +; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 0 +; CHECK-NEXT: [[SUM4_NONCHR_V:%.*]] = select i1 [[TMP11]], i32 44, i32 88, !prof !16 +; CHECK-NEXT: [[SUM4_NONCHR:%.*]] = add i32 [[SUM2_NONCHR]], [[SUM4_NONCHR_V]] +; CHECK-NEXT: br label [[BB3]] +; CHECK: bb3: +; CHECK-NEXT: [[SUM6:%.*]] = phi i32 [ [[TMP4]], [[BB0]] ], [ [[SUM0]], [[ENTRY_SPLIT_NONCHR]] ], [ [[SUM2_NONCHR]], [[BB0_NONCHR]] ], [ [[SUM4_NONCHR]], [[BB1_NONCHR]] ] +; CHECK-NEXT: ret i32 [[SUM6]] +; +entry: + %0 = load i32, i32* %i + %1 = and i32 %0, 255 + %2 = icmp eq i32 %1, 0 + br i1 %2, label %bb3, label %bb0, !prof !15 + +bb0: + %3 = and i32 %0, 1 + %4 = icmp eq i32 %3, 0 + %5 = add i32 %sum0, 42 + %sum1 = select i1 %4, i32 %sum0, i32 %5, !prof !15 + %6 = add i32 %sum1, 43 + %sum2 = select i1 %4, i32 %sum1, i32 %6, !prof !15 + br i1 %4, label %bb2, label %bb1, !prof !15 + +bb1: + %sum3 = add i32 %sum2, 44 + %7 = and i32 %0, 8 + %8 = icmp eq i32 %7, 0 + %9 = add i32 %sum3, 44 + %sum4 = select i1 %8, i32 %sum3, i32 %9, !prof !15 + br label %bb2 + +bb2: + %sum5 = phi i32 [ %sum2, %bb0 ], [ %sum4, %bb1 ] + br label %bb3 + +bb3: + %sum6 = phi i32 [ %sum0, %entry ], [ %sum5, %bb2 ] + ret i32 %sum6 +} + +; Selects. The exit block, which belongs to the top-level region, has a select +; and causes the top-level region to be the outermost CHR scope with the +; subscope that includes the entry block with two selects. The outermost CHR +; scope doesn't see the selects in the entry block as the entry block is in the +; subscope and incorrectly sets the CHR hoist point to the branch rather than +; the first select in the entry block and causes the CHR'ed selects ("select i1 +; false...") to incorrectly position above the CHR branch. This is testing +; against a quirk of how the region analysis handles the entry block. +; Roughly, +; i0 = *i +; sum2 = ((i0 & 2) == 0) ? sum0 : (sum0 + 43) // Likely false +; sum3 = ((i0 & 4) == 0) ? sum2 : (sum2 + 44) // Likely false +; if (j) +; foo() +; i5 = *i +; v13 = (i5 == 44) ? i5 : sum3 +; return v13 +; -> +; i0 = *i +; if ((i0 & 6) != 6) { // Likely true +; v9 = sum0 + 87 +; if (j) +; foo() +; } else { +; sum2.nc = ((i0 & 2) == 0) ? sum0 : (sum0 + 43) +; sum3.nc = ((i0 & 4) == 0) ? sum2.nc : (sum2.nc + 44) +; if (j) +; foo() +; } +; t2 = phi v9, sum3.nc +; i5 = *i +; v13 = (i5 == 44) ? 44 : t2 +; return v13 +define i32 @test_chr_20(i32* %i, i32 %sum0, i1 %j) !prof !14 { +; CHECK-LABEL: @test_chr_20( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[I0:%.*]] = load i32, i32* [[I:%.*]], align 4 +; CHECK-NEXT: [[TMP0:%.*]] = and i32 [[I0]], 6 +; CHECK-NEXT: [[TMP1:%.*]] = icmp eq i32 [[TMP0]], 6 +; CHECK-NEXT: br i1 [[TMP1]], label [[ENTRY_SPLIT:%.*]], label [[ENTRY_SPLIT_NONCHR:%.*]], !prof !15 +; CHECK: entry.split: +; CHECK-NEXT: [[V9:%.*]] = add i32 [[SUM0:%.*]], 87 +; CHECK-NEXT: br i1 [[J:%.*]], label [[BB1:%.*]], label [[BB4:%.*]] +; CHECK: bb1: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB4]] +; CHECK: entry.split.nonchr: +; CHECK-NEXT: [[V8:%.*]] = add i32 [[SUM0]], 43 +; CHECK-NEXT: [[V3:%.*]] = and i32 [[I0]], 2 +; CHECK-NEXT: [[V4:%.*]] = icmp eq i32 [[V3]], 0 +; CHECK-NEXT: [[SUM2_NONCHR:%.*]] = select i1 [[V4]], i32 [[SUM0]], i32 [[V8]], !prof !16 +; CHECK-NEXT: [[V6_NONCHR:%.*]] = and i32 [[I0]], 4 +; CHECK-NEXT: [[V5_NONCHR:%.*]] = icmp eq i32 [[V6_NONCHR]], 0 +; CHECK-NEXT: [[V9_NONCHR:%.*]] = add i32 [[SUM2_NONCHR]], 44 +; CHECK-NEXT: [[SUM3_NONCHR:%.*]] = select i1 [[V5_NONCHR]], i32 [[SUM2_NONCHR]], i32 [[V9_NONCHR]], !prof !16 +; CHECK-NEXT: br i1 [[J]], label [[BB1_NONCHR:%.*]], label [[BB4]] +; CHECK: bb1.nonchr: +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: br label [[BB4]] +; CHECK: bb4: +; CHECK-NEXT: [[TMP2:%.*]] = phi i32 [ [[V9]], [[BB1]] ], [ [[V9]], [[ENTRY_SPLIT]] ], [ [[SUM3_NONCHR]], [[BB1_NONCHR]] ], [ [[SUM3_NONCHR]], [[ENTRY_SPLIT_NONCHR]] ] +; CHECK-NEXT: [[I5:%.*]] = load i32, i32* [[I]], align 4 +; CHECK-NEXT: [[V12:%.*]] = icmp eq i32 [[I5]], 44 +; CHECK-NEXT: [[V13:%.*]] = select i1 [[V12]], i32 44, i32 [[TMP2]], !prof !16 +; CHECK-NEXT: ret i32 [[V13]] +; +entry: + %i0 = load i32, i32* %i + %v3 = and i32 %i0, 2 + %v4 = icmp eq i32 %v3, 0 + %v8 = add i32 %sum0, 43 + %sum2 = select i1 %v4, i32 %sum0, i32 %v8, !prof !15 + %v6 = and i32 %i0, 4 + %v5 = icmp eq i32 %v6, 0 + %v9 = add i32 %sum2, 44 + %sum3 = select i1 %v5, i32 %sum2, i32 %v9, !prof !15 + br i1 %j, label %bb1, label %bb4 + +bb1: + call void @foo() + br label %bb4 + +bb4: + %i5 = load i32, i32* %i + %v12 = icmp eq i32 %i5, 44 + %v13 = select i1 %v12, i32 %i5, i32 %sum3, !prof !15 + ret i32 %v13 +} + +!llvm.module.flags = !{!0} +!0 = !{i32 1, !"ProfileSummary", !1} +!1 = !{!2, !3, !4, !5, !6, !7, !8, !9} +!2 = !{!"ProfileFormat", !"InstrProf"} +!3 = !{!"TotalCount", i64 10000} +!4 = !{!"MaxCount", i64 10} +!5 = !{!"MaxInternalCount", i64 1} +!6 = !{!"MaxFunctionCount", i64 1000} +!7 = !{!"NumCounts", i64 3} +!8 = !{!"NumFunctions", i64 3} +!9 = !{!"DetailedSummary", !10} +!10 = !{!11, !12, !13} +!11 = !{i32 10000, i64 100, i32 1} +!12 = !{i32 999000, i64 100, i32 1} +!13 = !{i32 999999, i64 1, i32 2} + +!14 = !{!"function_entry_count", i64 100} +!15 = !{!"branch_weights", i32 0, i32 1} +!16 = !{!"branch_weights", i32 1, i32 1} +; CHECK: !15 = !{!"branch_weights", i32 1000, i32 0} +; CHECK: !16 = !{!"branch_weights", i32 0, i32 1} +; CHECK: !17 = !{!"branch_weights", i32 1, i32 1} +; CHECK: !18 = !{!"branch_weights", i32 1, i32 0} +; CHECK: !19 = !{!"branch_weights", i32 0, i32 1000} |

