maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/clang/test/CodeGenCXX/attr-likelihood-switch-bran...

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

359 lines
11 KiB
C++
Raw Normal View History

[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
// RUN: %clang_cc1 -O1 -disable-llvm-passes -emit-llvm %s -o - -triple=x86_64-linux-gnu | FileCheck %s
extern volatile int i;
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z8OneCaseLv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2:![0-9]+]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG]]
// CHECK-NEXT: ], !prof !6
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void OneCaseL() {
switch (i) {
[[likely]] case 1: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z8OneCaseUv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_BB:%.*]]
// CHECK-NEXT: ], !prof !7
// CHECK: sw.bb:
// CHECK-NEXT: [[TMP1:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: [[INC:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK-NEXT: store volatile i32 [[INC]], i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void OneCaseU() {
switch (i) {
[[unlikely]] case 1: ++i; break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z10TwoCasesLNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG]]
// CHECK-NEXT: i32 2, label [[SW_EPILOG]]
// CHECK-NEXT: ], !prof !8
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void TwoCasesLN() {
switch (i) {
[[likely]] case 1: break;
case 2: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z10TwoCasesUNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG]]
// CHECK-NEXT: i32 2, label [[SW_EPILOG]]
// CHECK-NEXT: ], !prof !9
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void TwoCasesUN() {
switch (i) {
[[unlikely]] case 1: break;
case 2: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z10TwoCasesLUv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG]]
// CHECK-NEXT: i32 2, label [[SW_EPILOG]]
// CHECK-NEXT: ], !prof !10
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void TwoCasesLU() {
switch (i) {
[[likely]] case 1: break;
[[unlikely]] case 2: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z20CasesFallthroughNNLNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_BB:%.*]]
// CHECK-NEXT: i32 2, label [[SW_BB]]
// CHECK-NEXT: i32 3, label [[SW_BB1:%.*]]
// CHECK-NEXT: i32 4, label [[SW_BB1]]
// CHECK-NEXT: ], !prof !11
// CHECK: sw.bb:
// CHECK-NEXT: br label [[SW_BB1]]
// CHECK: sw.bb1:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void CasesFallthroughNNLN() {
switch (i) {
case 1:
case 2:
[[likely]] case 3:
case 4: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z20CasesFallthroughNNUNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_BB:%.*]]
// CHECK-NEXT: i32 2, label [[SW_BB]]
// CHECK-NEXT: i32 3, label [[SW_BB1:%.*]]
// CHECK-NEXT: i32 4, label [[SW_BB1]]
// CHECK-NEXT: ], !prof !12
// CHECK: sw.bb:
// CHECK-NEXT: br label [[SW_BB1]]
// CHECK: sw.bb1:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void CasesFallthroughNNUN() {
switch (i) {
case 1:
case 2:
[[unlikely]] case 3:
case 4: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z28CasesFallthroughRangeSmallLNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_BB:%.*]]
// CHECK-NEXT: i32 2, label [[SW_BB]]
// CHECK-NEXT: i32 3, label [[SW_BB]]
// CHECK-NEXT: i32 4, label [[SW_BB]]
// CHECK-NEXT: i32 5, label [[SW_BB]]
// CHECK-NEXT: i32 102, label [[SW_BB1:%.*]]
// CHECK-NEXT: i32 103, label [[SW_BB2:%.*]]
// CHECK-NEXT: i32 104, label [[SW_BB2]]
// CHECK-NEXT: ], !prof !13
// CHECK: sw.bb:
// CHECK-NEXT: [[TMP1:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: [[INC:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK-NEXT: store volatile i32 [[INC]], i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: br label [[SW_BB1]]
// CHECK: sw.bb1:
// CHECK-NEXT: br label [[SW_BB2]]
// CHECK: sw.bb2:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void CasesFallthroughRangeSmallLN() {
switch (i) {
case 1 ... 5: ++i;
case 102:
[[likely]] case 103:
case 104: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z28CasesFallthroughRangeSmallUNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_EPILOG:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_BB:%.*]]
// CHECK-NEXT: i32 2, label [[SW_BB]]
// CHECK-NEXT: i32 3, label [[SW_BB]]
// CHECK-NEXT: i32 4, label [[SW_BB]]
// CHECK-NEXT: i32 5, label [[SW_BB]]
// CHECK-NEXT: i32 102, label [[SW_BB1:%.*]]
// CHECK-NEXT: i32 103, label [[SW_BB2:%.*]]
// CHECK-NEXT: i32 104, label [[SW_BB2]]
// CHECK-NEXT: ], !prof !14
// CHECK: sw.bb:
// CHECK-NEXT: [[TMP1:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: [[INC:%.*]] = add nsw i32 [[TMP1]], 1
// CHECK-NEXT: store volatile i32 [[INC]], i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: br label [[SW_BB1]]
// CHECK: sw.bb1:
// CHECK-NEXT: br label [[SW_BB2]]
// CHECK: sw.bb2:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void CasesFallthroughRangeSmallUN() {
switch (i) {
case 1 ... 5: ++i;
case 102:
[[unlikely]] case 103:
case 104: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z29CasesFallthroughRangeLargeLLNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_CASERANGE:%.*]] [
// CHECK-NEXT: i32 1003, label [[SW_BB1:%.*]]
// CHECK-NEXT: i32 104, label [[SW_BB1]]
// CHECK-NEXT: ], !prof !8
// CHECK: sw.bb:
// CHECK-NEXT: br label [[SW_BB1]]
// CHECK: sw.bb1:
// CHECK-NEXT: br label [[SW_EPILOG:%.*]]
// CHECK: sw.caserange:
// CHECK-NEXT: [[TMP1:%.*]] = sub i32 [[TMP0]], 0
// CHECK-NEXT: [[INBOUNDS:%.*]] = icmp ule i32 [[TMP1]], 64
// CHECK-NEXT: [[INBOUNDS_EXPVAL:%.*]] = call i1 @llvm.expect.i1(i1 [[INBOUNDS]], i1 true)
// CHECK-NEXT: br i1 [[INBOUNDS_EXPVAL]], label [[SW_BB:%.*]], label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void CasesFallthroughRangeLargeLLN() {
switch (i) {
[[likely]] case 0 ... 64:
[[likely]] case 1003:
case 104: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z29CasesFallthroughRangeLargeUUNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_CASERANGE:%.*]] [
// CHECK-NEXT: i32 1003, label [[SW_BB1:%.*]]
// CHECK-NEXT: i32 104, label [[SW_BB1]]
// CHECK-NEXT: ], !prof !9
// CHECK: sw.bb:
// CHECK-NEXT: br label [[SW_BB1]]
// CHECK: sw.bb1:
// CHECK-NEXT: br label [[SW_EPILOG:%.*]]
// CHECK: sw.caserange:
// CHECK-NEXT: [[TMP1:%.*]] = sub i32 [[TMP0]], 0
// CHECK-NEXT: [[INBOUNDS:%.*]] = icmp ule i32 [[TMP1]], 64
// CHECK-NEXT: [[INBOUNDS_EXPVAL:%.*]] = call i1 @llvm.expect.i1(i1 [[INBOUNDS]], i1 false)
// CHECK-NEXT: br i1 [[INBOUNDS_EXPVAL]], label [[SW_BB:%.*]], label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void CasesFallthroughRangeLargeUUN() {
switch (i) {
[[unlikely]] case 0 ... 64:
[[unlikely]] case 1003:
case 104: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z15OneCaseDefaultLv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_DEFAULT:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG:%.*]]
// CHECK-NEXT: ], !prof !15
// CHECK: sw.default:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void OneCaseDefaultL() {
switch (i) {
case 1: break;
[[likely]] default: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z15OneCaseDefaultUv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_DEFAULT:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG:%.*]]
// CHECK-NEXT: ], !prof !16
// CHECK: sw.default:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void OneCaseDefaultU() {
switch (i) {
case 1: break;
[[unlikely]] default: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z18TwoCasesDefaultLNLv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_DEFAULT:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG:%.*]]
// CHECK-NEXT: i32 2, label [[SW_EPILOG]]
// CHECK-NEXT: ], !prof !17
// CHECK: sw.default:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void TwoCasesDefaultLNL() {
switch (i) {
[[likely]] case 1: break;
case 2: break;
[[likely]] default: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z18TwoCasesDefaultLNNv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_DEFAULT:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG:%.*]]
// CHECK-NEXT: i32 2, label [[SW_EPILOG]]
// CHECK-NEXT: ], !prof !8
// CHECK: sw.default:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void TwoCasesDefaultLNN() {
switch (i) {
[[likely]] case 1: break;
case 2: break;
default: break;
}
}
[clang][CodeGen] Lower Likelihood attributes to @llvm.expect intrin instead of branch weights 08196e0b2e1f8aaa8a854585335c17ba479114df exposed LowerExpectIntrinsic's internal implementation detail in the form of LikelyBranchWeight/UnlikelyBranchWeight options to the outside. While this isn't incorrect from the results viewpoint, this is suboptimal from the layering viewpoint, and causes confusion - should transforms also use those weights, or should they use something else, D98898? So go back to status quo by making LikelyBranchWeight/UnlikelyBranchWeight internal again, and fixing all the code that used it directly, which currently is only clang codegen, thankfully, to emit proper @llvm.expect intrinsics instead.
2021-03-22 03:13:47 +08:00
// CHECK-LABEL: @_Z18TwoCasesDefaultLNUv(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[TMP0:%.*]] = load volatile i32, i32* @i, align 4, !tbaa [[TBAA2]]
// CHECK-NEXT: switch i32 [[TMP0]], label [[SW_DEFAULT:%.*]] [
// CHECK-NEXT: i32 1, label [[SW_EPILOG:%.*]]
// CHECK-NEXT: i32 2, label [[SW_EPILOG]]
// CHECK-NEXT: ], !prof !18
// CHECK: sw.default:
// CHECK-NEXT: br label [[SW_EPILOG]]
// CHECK: sw.epilog:
// CHECK-NEXT: ret void
//
[Sema, CodeGen] Implement [[likely]] and [[unlikely]] in SwitchStmt This implements the likelihood attribute for the switch statement. Based on the discussion in D85091 and D86559 it only handles the attribute when placed on the case labels or the default labels. It also marks the likelihood attribute as feature complete. There are more QoI patches in the pipeline. Differential Revision: https://reviews.llvm.org/D89210
2020-10-18 19:34:41 +08:00
void TwoCasesDefaultLNU() {
switch (i) {
[[likely]] case 1: break;
case 2: break;
[[unlikely]] default: break;
}
}