apple-fcloutier
d28a8b0b3f
The main follow-up item to https://github.com/llvm/llvm-project/pull/160790 was changing -O0 codegen to convert in-memory i8 bool values to i1 with the `nonzero` rule (`icmp ne i8 %val, 0`) rather than the `truncate` rule (`trunc i8 %val to i1`). Bool values can only be `true` or `false`. While they are notionally a single bit, the smallest addressable unit is CHAR_BIT bits large, and CHAR_BIT is typically 8. Programming errors (such as memcpying a random byte to a `bool`) can cause the 8-bit storage for a `bool` value to have a bit pattern that is different from `true` or `false`, which then leads to undefined behavior. Clang has historically taken advantage of this in optimized builds (everything other than -O0) by attaching range metadata to `bool` loads to assume that the value loaded can only be 0 or 1. This leads to exploitable security issues, and the correct behavior is not always easy to explain to C developers. To remedy this situation, Clang accepted a [-fstrict-bool](https://discourse.llvm.org/t/defining-what-happens-when-a-bool-isn-t-0-or-1/86778) switch to control whether it can assume that loaded bool values are always necessarily 0 or 1. By default, it does (maintaining the status quo), and users must specify `-fno-strict-bool` to opt out of that behavior. When opting out, users can optionally request that bool i8 values are converted to i1 either by truncation or by comparing to 0. The default is comparing to 0. However, since `-O0` alone _technically_ uses -fstrict-bool, unoptimized builds convert i8 bool values to i1 with a `trunc` operation, whereas `-O1 -fno-strict-bool` converts i8 bool values to i1 with `icmp ne 0`. This is a surprising inconsistency. This PR changes -O0 codegen to align with -fno-strict-bool. This is achieved with a single-line change: ``` bool isConvertingBoolWithCmp0() const { switch (getLoadBoolFromMem()) { case BoolFromMem::Strict: + return !isOptimizedBuild(); case BoolFromMem::Truncate: ``` However, it impacts a _very large_ number of tests, so we agreed to move it out of the -fstrict-bool PR to reduce the chances we would have to back out the whole thing for this secondary item. This PR does the change and modifies the tests accordingly. I expect that it will go stale rather quickly. If this needs more discussion, I'll only update it once we reach consensus.
133 lines
6.9 KiB
C++
133 lines
6.9 KiB
C++
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --check-globals --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _ --global-value-regex "\.offload_.*" --global-hex-value-regex ".offload_maptypes.*" --version 3
|
|
|
|
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-llvm %s -o - \
|
|
// RUN: | FileCheck %s
|
|
|
|
// Check same results after serialization round-trip
|
|
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -emit-pch -o %t %s
|
|
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -include-pch %t -emit-llvm %s -o - \
|
|
// RUN: | FileCheck %s
|
|
|
|
// expected-no-diagnostics
|
|
#ifndef HEADER
|
|
#define HEADER
|
|
|
|
extern bool foo(bool);
|
|
|
|
int if_logical() {
|
|
bool a = foo(true);
|
|
bool b = foo(true);
|
|
int pp = 42;
|
|
int *p = &pp;
|
|
#pragma omp target data if(a && b) map(to: p[0])
|
|
{
|
|
p[0]++;
|
|
}
|
|
if (p[0])
|
|
return 1;
|
|
return 0;
|
|
}
|
|
|
|
int main() {
|
|
return if_logical();
|
|
}
|
|
|
|
#endif
|
|
// &p[0], &p[0], sizeof(p0]), TO
|
|
// &p, &p[0], sizeof(void*), ATTACH
|
|
|
|
//.
|
|
// CHECK: @.offload_sizes = private unnamed_addr constant [2 x i64] [i64 4, i64 8]
|
|
// CHECK: @.offload_maptypes = private unnamed_addr constant [2 x i64] [i64 [[#0x1]], i64 [[#0x4000]]]
|
|
//.
|
|
// CHECK-LABEL: define dso_local noundef i32 @_Z10if_logicalv(
|
|
// CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
|
|
// CHECK-NEXT: [[A:%.*]] = alloca i8, align 1
|
|
// CHECK-NEXT: [[B:%.*]] = alloca i8, align 1
|
|
// CHECK-NEXT: [[PP:%.*]] = alloca i32, align 4
|
|
// CHECK-NEXT: [[P:%.*]] = alloca ptr, align 8
|
|
// CHECK-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x ptr], align 8
|
|
// CHECK-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x ptr], align 8
|
|
// CHECK-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x ptr], align 8
|
|
// CHECK-NEXT: [[CALL:%.*]] = call noundef zeroext i1 @_Z3foob(i1 noundef zeroext true)
|
|
// CHECK-NEXT: [[STOREDV:%.*]] = zext i1 [[CALL]] to i8
|
|
// CHECK-NEXT: store i8 [[STOREDV]], ptr [[A]], align 1
|
|
// CHECK-NEXT: [[CALL1:%.*]] = call noundef zeroext i1 @_Z3foob(i1 noundef zeroext true)
|
|
// CHECK-NEXT: [[STOREDV2:%.*]] = zext i1 [[CALL1]] to i8
|
|
// CHECK-NEXT: store i8 [[STOREDV2]], ptr [[B]], align 1
|
|
// CHECK-NEXT: store i32 42, ptr [[PP]], align 4
|
|
// CHECK-NEXT: store ptr [[PP]], ptr [[P]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A]], align 1
|
|
// CHECK-NEXT: [[LOADEDV:%.*]] = icmp ne i8 [[TMP0]], 0
|
|
// CHECK-NEXT: br i1 [[LOADEDV]], label [[LAND_RHS:%.*]], label [[LAND_END:%.*]]
|
|
// CHECK: land.rhs:
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i8, ptr [[B]], align 1
|
|
// CHECK-NEXT: [[LOADEDV3:%.*]] = icmp ne i8 [[TMP1]], 0
|
|
// CHECK-NEXT: br label [[LAND_END]]
|
|
// CHECK: land.end:
|
|
// CHECK-NEXT: [[TMP2:%.*]] = phi i1 [ false, [[ENTRY:%.*]] ], [ [[LOADEDV3]], [[LAND_RHS]] ]
|
|
// CHECK-NEXT: br i1 [[TMP2]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
|
|
// CHECK: omp_if.then:
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[P]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[P]], align 8
|
|
// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i64 0
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP3]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[ARRAYIDX]], ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr null, ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
|
|
// CHECK-NEXT: store ptr [[P]], ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
|
|
// CHECK-NEXT: store ptr [[ARRAYIDX]], ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
|
|
// CHECK-NEXT: store ptr null, ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
|
|
// CHECK-NEXT: call void @__tgt_target_data_begin_mapper(ptr @[[GLOB1:[0-9]+]], i64 -1, i32 2, ptr [[TMP11]], ptr [[TMP12]], ptr @.offload_sizes, ptr @.offload_maptypes, ptr null, ptr null)
|
|
// CHECK-NEXT: br label [[OMP_IF_END:%.*]]
|
|
// CHECK: omp_if.else:
|
|
// CHECK-NEXT: br label [[OMP_IF_END]]
|
|
// CHECK: omp_if.end:
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load ptr, ptr [[P]], align 8
|
|
// CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i64 0
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[ARRAYIDX4]], align 4
|
|
// CHECK-NEXT: [[INC:%.*]] = add nsw i32 [[TMP14]], 1
|
|
// CHECK-NEXT: store i32 [[INC]], ptr [[ARRAYIDX4]], align 4
|
|
// CHECK-NEXT: br i1 [[TMP2]], label [[OMP_IF_THEN5:%.*]], label [[OMP_IF_ELSE6:%.*]]
|
|
// CHECK: omp_if.then5:
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds [2 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
|
|
// CHECK-NEXT: call void @__tgt_target_data_end_mapper(ptr @[[GLOB1]], i64 -1, i32 2, ptr [[TMP15]], ptr [[TMP16]], ptr @.offload_sizes, ptr @.offload_maptypes, ptr null, ptr null)
|
|
// CHECK-NEXT: br label [[OMP_IF_END7:%.*]]
|
|
// CHECK: omp_if.else6:
|
|
// CHECK-NEXT: br label [[OMP_IF_END7]]
|
|
// CHECK: omp_if.end7:
|
|
// CHECK-NEXT: [[TMP17:%.*]] = load ptr, ptr [[P]], align 8
|
|
// CHECK-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i64 0
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX8]], align 4
|
|
// CHECK-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP18]], 0
|
|
// CHECK-NEXT: br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
|
|
// CHECK: if.then:
|
|
// CHECK-NEXT: store i32 1, ptr [[RETVAL]], align 4
|
|
// CHECK-NEXT: br label [[RETURN:%.*]]
|
|
// CHECK: if.end:
|
|
// CHECK-NEXT: store i32 0, ptr [[RETVAL]], align 4
|
|
// CHECK-NEXT: br label [[RETURN]]
|
|
// CHECK: return:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = load i32, ptr [[RETVAL]], align 4
|
|
// CHECK-NEXT: ret i32 [[TMP19]]
|
|
//
|
|
//
|
|
// CHECK-LABEL: define dso_local noundef i32 @main(
|
|
// CHECK-SAME: ) #[[ATTR3:[0-9]+]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[RETVAL]], align 4
|
|
// CHECK-NEXT: [[CALL:%.*]] = call noundef i32 @_Z10if_logicalv()
|
|
// CHECK-NEXT: ret i32 [[CALL]]
|
|
//
|