OpenMP: Don't include stdbool.h in builtin headers

[arsenm]

Summary

Pre-C99 didn't include bool, and C99 allows you to redefine true/false
apparently.

[ronlieb]

@mhalk to review

[ronlieb]

patch applied to local build resolves issues seen

[arsenm]

dee1f5b32c3a6a5694c3bb2fbf68d162447a5970

clang/lib/Headers/__clang_hip_libdevice_declares.h

 __device__ _Float16 __ocml_sin_f16(_Float16);
 __device__ __attribute__((const)) _Float16 __ocml_sqrt_f16(_Float16);
 __device__ __attribute__((const)) _Float16 __ocml_trunc_f16(_Float16);
 __device__ __attribute__((pure)) _Float16 __ocml_pown_f16(_Float16, int);
 
 typedef _Float16 __2f16 __attribute__((ext_vector_type(2)));
 typedef short __2i16 __attribute__((ext_vector_type(2)));
 
+// We need to match C99's bool and get an i1 in the IR.
+#ifdef __cplusplus
+typedef bool __ockl_bool;
+#else
+typedef _Bool __ockl_bool;
+#endif
+
 __device__ __attribute__((const)) float __ockl_fdot2(__2f16 a, __2f16 b,
-                                                     float c, bool s);
+                                                     float c, __ockl_bool s);
 __device__ __attribute__((const)) __2f16 __ocml_ceil_2f16(__2f16);
 __device__ __attribute__((const)) __2f16 __ocml_fabs_2f16(__2f16);
 __device__ __2f16 __ocml_cos_2f16(__2f16);
 __device__ __attribute__((pure)) __2f16 __ocml_exp_2f16(__2f16);
 __device__ __attribute__((pure)) __2f16 __ocml_exp10_2f16(__2f16);
 __device__ __attribute__((pure)) __2f16 __ocml_exp2_2f16(__2f16);
 __device__ __attribute__((const)) __2f16 __ocml_floor_2f16(__2f16);
 __device__ __attribute__((const))

clang/lib/Headers/openmp_wrappers/__clang_openmp_device_functions.h

 
 #pragma omp end declare variant
 
 #ifdef __AMDGCN__
 #pragma omp begin declare variant match(device = {arch(amdgcn)})
 
 // Import types which will be used by __clang_hip_libdevice_declares.h
 #ifndef __cplusplus
-#include <stdbool.h>
 #include <stdint.h>
 #endif
 
 #define __OPENMP_AMDGCN__
 #pragma push_macro("__device__")
 #define __device__
 
 /// Include declarations for libdevice functions.

clang/test/Headers/openmp-device-functions-bool.c

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 2
+// RUN: %clang_cc1 -x c -fopenmp -triple amdgcn-amd-amdhsa -aux-triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -internal-isystem %S/../../lib/Headers/openmp_wrappers -include __clang_openmp_device_functions.h -internal-isystem %S/../../lib/Headers/openmp_wrappers -internal-isystem %S/Inputs/include -emit-llvm %s -fopenmp-is-device  -o - | FileCheck %s --check-prefixes=CHECK,CHECK-C
+// RUN: %clang_cc1 -x c++ -fopenmp -triple amdgcn-amd-amdhsa -aux-triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -internal-isystem %S/../../lib/Headers/openmp_wrappers -include __clang_openmp_device_functions.h -internal-isystem %S/../../lib/Headers/openmp_wrappers -internal-isystem %S/Inputs/include -emit-llvm %s -fopenmp-is-device  -o - | FileCheck %s --check-prefixes=CHECK,CHECK-CPP
+
+// Test that we did not include <stdbool.h> in C, and OCKL functions using bool
+// produce an i1
+
+#ifdef __cplusplus
+typedef bool ockl_bool;
+#define EXTERN_C extern "C"
+#else
+typedef _Bool ockl_bool;
+#define EXTERN_C
+#endif
+
+#pragma omp begin declare target
+
+// CHECK-LABEL: define hidden float @test_fdot2
+// CHECK-SAME: (<2 x half> noundef [[A:%.*]], <2 x half> noundef [[B:%.*]], float noundef [[C:%.*]], i1 noundef zeroext [[S:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[RETVAL:%.*]] = alloca float, align 4, addrspace(5)
+// CHECK-NEXT:    [[A_ADDR:%.*]] = alloca <2 x half>, align 4, addrspace(5)
+// CHECK-NEXT:    [[B_ADDR:%.*]] = alloca <2 x half>, align 4, addrspace(5)
+// CHECK-NEXT:    [[C_ADDR:%.*]] = alloca float, align 4, addrspace(5)
+// CHECK-NEXT:    [[S_ADDR:%.*]] = alloca i8, align 1, addrspace(5)
+// CHECK-NEXT:    [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
+// CHECK-NEXT:    [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
+// CHECK-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// CHECK-NEXT:    [[C_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[C_ADDR]] to ptr
+// CHECK-NEXT:    [[S_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[S_ADDR]] to ptr
+// CHECK-NEXT:    store <2 x half> [[A]], ptr [[A_ADDR_ASCAST]], align 4
+// CHECK-NEXT:    store <2 x half> [[B]], ptr [[B_ADDR_ASCAST]], align 4
+// CHECK-NEXT:    store float [[C]], ptr [[C_ADDR_ASCAST]], align 4
+// CHECK-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[S]] to i8
+// CHECK-NEXT:    store i8 [[FROMBOOL]], ptr [[S_ADDR_ASCAST]], align 1
+// CHECK-NEXT:    [[TMP0:%.*]] = load <2 x half>, ptr [[A_ADDR_ASCAST]], align 4
+// CHECK-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr [[B_ADDR_ASCAST]], align 4
+// CHECK-NEXT:    [[TMP2:%.*]] = load float, ptr [[C_ADDR_ASCAST]], align 4
+// CHECK-NEXT:    [[TMP3:%.*]] = load i8, ptr [[S_ADDR_ASCAST]], align 1
+// CHECK-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP3]] to i1
+// CHECK-NEXT:    [[CALL:%.*]] = call float @__ockl_fdot2(<2 x half> noundef [[TMP0]], <2 x half> noundef [[TMP1]], float noundef [[TMP2]], i1 noundef zeroext [[TOBOOL]]) #[[ATTR2:[0-9]+]]
+// CHECK-NEXT:    ret float [[CALL]]
+//
+EXTERN_C float test_fdot2(__2f16 a, __2f16 b, float c, ockl_bool s) {
+  return __ockl_fdot2(a, b, c, s);
+}
+
+
+#ifndef __cplusplus
+
+enum my_bool {
+  false,
+  true
+};
+
+// CHECK-C-LABEL: define hidden i32 @use_my_bool
+// CHECK-C-SAME: (i32 noundef [[B:%.*]]) #[[ATTR0]] {
+// CHECK-C-NEXT:  entry:
+// CHECK-C-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4, addrspace(5)
+// CHECK-C-NEXT:    [[B_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
+// CHECK-C-NEXT:    [[T:%.*]] = alloca i32, align 4, addrspace(5)
+// CHECK-C-NEXT:    [[F:%.*]] = alloca i32, align 4, addrspace(5)
+// CHECK-C-NEXT:    [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
+// CHECK-C-NEXT:    [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
+// CHECK-C-NEXT:    [[T_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[T]] to ptr
+// CHECK-C-NEXT:    [[F_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[F]] to ptr
+// CHECK-C-NEXT:    store i32 [[B]], ptr [[B_ADDR_ASCAST]], align 4
+// CHECK-C-NEXT:    store volatile i32 1, ptr [[T_ASCAST]], align 4
+// CHECK-C-NEXT:    store volatile i32 0, ptr [[F_ASCAST]], align 4
+// CHECK-C-NEXT:    [[TMP0:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
+// CHECK-C-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP0]], 0
+// CHECK-C-NEXT:    [[LNOT:%.*]] = xor i1 [[TOBOOL]], true
+// CHECK-C-NEXT:    [[LNOT_EXT:%.*]] = zext i1 [[LNOT]] to i32
+// CHECK-C-NEXT:    ret i32 [[LNOT_EXT]]
+//
+enum my_bool use_my_bool(enum my_bool b) {
+  volatile enum my_bool t = true;
+  volatile enum my_bool f = false;
+
+  return !b;
+}
+
+#endif
+
+
+
+#pragma omp end declare target
+//// NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+// CHECK-CPP: {{.*}}