[llvm-branch-commits] [libclc] cda3c27 - Merge branch 'main' into revert-152096-revert-151622-libc-fix-integeration-alloc
via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Tue Aug 5 19:05:10 PDT 2025
Author: Schrodinger ZHU Yifan
Date: 2025-08-05T22:05:02-04:00
New Revision: cda3c279733bc534550f757162e65794c288b981
URL: https://github.com/llvm/llvm-project/commit/cda3c279733bc534550f757162e65794c288b981
DIFF: https://github.com/llvm/llvm-project/commit/cda3c279733bc534550f757162e65794c288b981.diff
LOG: Merge branch 'main' into revert-152096-revert-151622-libc-fix-integeration-alloc
Added:
libclc/clc/include/clc/mem_fence/clc_mem_fence.h
libclc/clc/include/clc/synchronization/clc_work_group_barrier.h
libclc/clc/lib/amdgcn/mem_fence/clc_mem_fence.cl
libclc/clc/lib/amdgcn/synchronization/clc_work_group_barrier.cl
libclc/clc/lib/ptx-nvidiacl/mem_fence/clc_mem_fence.cl
libclc/clc/lib/ptx-nvidiacl/synchronization/clc_work_group_barrier.cl
libclc/opencl/include/clc/opencl/synchronization/utils.h
Modified:
clang/lib/Sema/SemaModule.cpp
libclc/clc/lib/amdgcn/SOURCES
libclc/clc/lib/ptx-nvidiacl/SOURCES
libclc/opencl/include/clc/opencl/synchronization/cl_mem_fence_flags.h
libclc/opencl/lib/amdgcn/mem_fence/fence.cl
libclc/opencl/lib/amdgcn/synchronization/barrier.cl
libclc/opencl/lib/ptx-nvidiacl/mem_fence/fence.cl
libclc/opencl/lib/ptx-nvidiacl/synchronization/barrier.cl
llvm/cmake/modules/LLVMProcessSources.cmake
llvm/include/llvm/CodeGen/TargetLowering.h
llvm/include/llvm/IR/RuntimeLibcalls.td
llvm/include/llvm/Support/DebugLog.h
llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
llvm/lib/Target/AArch64/AArch64Subtarget.h
llvm/lib/Target/ARM/ARMISelLowering.cpp
llvm/test/CodeGen/AArch64/stacksmash-arm64ec.ll
Removed:
################################################################################
diff --git a/clang/lib/Sema/SemaModule.cpp b/clang/lib/Sema/SemaModule.cpp
index b137549b8f859..ff9f85f960d93 100644
--- a/clang/lib/Sema/SemaModule.cpp
+++ b/clang/lib/Sema/SemaModule.cpp
@@ -1222,7 +1222,8 @@ bool ExposureChecker::isTULocal(const NamedDecl *D) {
// [basic.link]p15.5
// - a specialization of a template whose (possibly instantiated) declaration
// is an exposure.
- if (checkExposure(PrimaryTemplate, /*Diag=*/false))
+ if (ExposureSet.count(PrimaryTemplate) ||
+ checkExposure(PrimaryTemplate, /*Diag=*/false))
return true;
// Avoid calling checkExposure again since it is expensive.
diff --git a/libclc/clc/include/clc/mem_fence/clc_mem_fence.h b/libclc/clc/include/clc/mem_fence/clc_mem_fence.h
new file mode 100644
index 0000000000000..2321634c76842
--- /dev/null
+++ b/libclc/clc/include/clc/mem_fence/clc_mem_fence.h
@@ -0,0 +1,17 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CLC_MEM_FENCE_CLC_MEM_FENCE_H__
+#define __CLC_MEM_FENCE_CLC_MEM_FENCE_H__
+
+#include <clc/internal/clc.h>
+
+_CLC_OVERLOAD _CLC_DECL void __clc_mem_fence(int memory_scope,
+ int memory_order);
+
+#endif // __CLC_MEM_FENCE_CLC_MEM_FENCE_H__
diff --git a/libclc/clc/include/clc/synchronization/clc_work_group_barrier.h b/libclc/clc/include/clc/synchronization/clc_work_group_barrier.h
new file mode 100644
index 0000000000000..5f864e1057b8b
--- /dev/null
+++ b/libclc/clc/include/clc/synchronization/clc_work_group_barrier.h
@@ -0,0 +1,17 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CLC_SYNCHRONIZATION_CLC_WORK_GROUP_BARRIER_H__
+#define __CLC_SYNCHRONIZATION_CLC_WORK_GROUP_BARRIER_H__
+
+#include <clc/internal/clc.h>
+
+_CLC_OVERLOAD _CLC_DECL void __clc_work_group_barrier(int memory_scope,
+ int memory_order);
+
+#endif // __CLC_SYNCHRONIZATION_CLC_WORK_GROUP_BARRIER_H__
diff --git a/libclc/clc/lib/amdgcn/SOURCES b/libclc/clc/lib/amdgcn/SOURCES
index d91f08533e149..76c3266e3af7b 100644
--- a/libclc/clc/lib/amdgcn/SOURCES
+++ b/libclc/clc/lib/amdgcn/SOURCES
@@ -1,4 +1,6 @@
math/clc_ldexp_override.cl
+mem_fence/clc_mem_fence.cl
+synchronization/clc_work_group_barrier.cl
workitem/clc_get_global_offset.cl
workitem/clc_get_global_size.cl
workitem/clc_get_group_id.cl
diff --git a/libclc/clc/lib/amdgcn/mem_fence/clc_mem_fence.cl b/libclc/clc/lib/amdgcn/mem_fence/clc_mem_fence.cl
new file mode 100644
index 0000000000000..9e6460313718e
--- /dev/null
+++ b/libclc/clc/lib/amdgcn/mem_fence/clc_mem_fence.cl
@@ -0,0 +1,37 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <clc/mem_fence/clc_mem_fence.h>
+
+void __clc_amdgcn_s_waitcnt(unsigned flags);
+
+// s_waitcnt takes 16bit argument with a combined number of maximum allowed
+// pending operations:
+// [12:8] LGKM -- LDS, GDS, Konstant (SMRD), Messages
+// [7] -- undefined
+// [6:4] -- exports, GDS, and mem write
+// [3:0] -- vector memory operations
+
+// Newer clang supports __builtin_amdgcn_s_waitcnt
+#if __clang_major__ >= 5
+#define __waitcnt(x) __builtin_amdgcn_s_waitcnt(x)
+#else
+#define __waitcnt(x) __clc_amdgcn_s_waitcnt(x)
+_CLC_DEF void __clc_amdgcn_s_waitcnt(unsigned) __asm("llvm.amdgcn.s.waitcnt");
+#endif
+
+_CLC_OVERLOAD _CLC_DEF void __clc_mem_fence(int memory_scope,
+ int memory_order) {
+ if (memory_scope & __MEMORY_SCOPE_DEVICE) {
+ // scalar loads are counted with LGKM but we don't know whether
+ // the compiler turned any loads to scalar
+ __waitcnt(0);
+ } else if (memory_scope & __MEMORY_SCOPE_WRKGRP)
+ __waitcnt(0xff); // LGKM is [12:8]
+}
+#undef __waitcnt
diff --git a/libclc/clc/lib/amdgcn/synchronization/clc_work_group_barrier.cl b/libclc/clc/lib/amdgcn/synchronization/clc_work_group_barrier.cl
new file mode 100644
index 0000000000000..ff3628fa7c339
--- /dev/null
+++ b/libclc/clc/lib/amdgcn/synchronization/clc_work_group_barrier.cl
@@ -0,0 +1,16 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <clc/mem_fence/clc_mem_fence.h>
+#include <clc/synchronization/clc_work_group_barrier.h>
+
+_CLC_OVERLOAD _CLC_DEF void __clc_work_group_barrier(int memory_scope,
+ int memory_order) {
+ __clc_mem_fence(memory_scope, memory_order);
+ __builtin_amdgcn_s_barrier();
+}
diff --git a/libclc/clc/lib/ptx-nvidiacl/SOURCES b/libclc/clc/lib/ptx-nvidiacl/SOURCES
index 05368c5e4d4e3..b6f50654f89c5 100644
--- a/libclc/clc/lib/ptx-nvidiacl/SOURCES
+++ b/libclc/clc/lib/ptx-nvidiacl/SOURCES
@@ -1,3 +1,5 @@
+mem_fence/clc_mem_fence.cl
+synchronization/clc_work_group_barrier.cl
workitem/clc_get_global_id.cl
workitem/clc_get_group_id.cl
workitem/clc_get_local_id.cl
diff --git a/libclc/clc/lib/ptx-nvidiacl/mem_fence/clc_mem_fence.cl b/libclc/clc/lib/ptx-nvidiacl/mem_fence/clc_mem_fence.cl
new file mode 100644
index 0000000000000..b3e2375e755a2
--- /dev/null
+++ b/libclc/clc/lib/ptx-nvidiacl/mem_fence/clc_mem_fence.cl
@@ -0,0 +1,15 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <clc/mem_fence/clc_mem_fence.h>
+
+_CLC_OVERLOAD _CLC_DEF void __clc_mem_fence(int memory_scope,
+ int memory_order) {
+ if (memory_scope & (__MEMORY_SCOPE_DEVICE | __MEMORY_SCOPE_WRKGRP))
+ __nvvm_membar_cta();
+}
diff --git a/libclc/clc/lib/ptx-nvidiacl/synchronization/clc_work_group_barrier.cl b/libclc/clc/lib/ptx-nvidiacl/synchronization/clc_work_group_barrier.cl
new file mode 100644
index 0000000000000..6cb37a38f06ac
--- /dev/null
+++ b/libclc/clc/lib/ptx-nvidiacl/synchronization/clc_work_group_barrier.cl
@@ -0,0 +1,14 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <clc/synchronization/clc_work_group_barrier.h>
+
+_CLC_OVERLOAD _CLC_DEF void __clc_work_group_barrier(int memory_scope,
+ int memory_order) {
+ __syncthreads();
+}
diff --git a/libclc/opencl/include/clc/opencl/synchronization/cl_mem_fence_flags.h b/libclc/opencl/include/clc/opencl/synchronization/cl_mem_fence_flags.h
index 6636515fca47d..7b2f701c1ff99 100644
--- a/libclc/opencl/include/clc/opencl/synchronization/cl_mem_fence_flags.h
+++ b/libclc/opencl/include/clc/opencl/synchronization/cl_mem_fence_flags.h
@@ -13,5 +13,6 @@ typedef uint cl_mem_fence_flags;
#define CLK_LOCAL_MEM_FENCE 1
#define CLK_GLOBAL_MEM_FENCE 2
+#define CLK_IMAGE_MEM_FENCE 4
#endif // __CLC_OPENCL_SYNCHRONIZATION_CL_MEM_FENCE_FLAGS_H__
diff --git a/libclc/opencl/include/clc/opencl/synchronization/utils.h b/libclc/opencl/include/clc/opencl/synchronization/utils.h
new file mode 100644
index 0000000000000..cf3baf28cb5f1
--- /dev/null
+++ b/libclc/opencl/include/clc/opencl/synchronization/utils.h
@@ -0,0 +1,24 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CLC_OPENCL_SYNCHRONIZATION_UTILS_H__
+#define __CLC_OPENCL_SYNCHRONIZATION_UTILS_H__
+
+#include <clc/internal/clc.h>
+#include <clc/opencl/synchronization/cl_mem_fence_flags.h>
+
+_CLC_INLINE int getCLCMemoryScope(cl_mem_fence_flags flag) {
+ int memory_scope = 0;
+ if (flag & CLK_GLOBAL_MEM_FENCE)
+ memory_scope |= __MEMORY_SCOPE_DEVICE;
+ if (flag & CLK_LOCAL_MEM_FENCE)
+ memory_scope |= __MEMORY_SCOPE_WRKGRP;
+ return memory_scope;
+}
+
+#endif // __CLC_OPENCL_SYNCHRONIZATION_UTILS_H__
diff --git a/libclc/opencl/lib/amdgcn/mem_fence/fence.cl b/libclc/opencl/lib/amdgcn/mem_fence/fence.cl
index 88b953005aae6..81216d6a26cf2 100644
--- a/libclc/opencl/lib/amdgcn/mem_fence/fence.cl
+++ b/libclc/opencl/lib/amdgcn/mem_fence/fence.cl
@@ -6,34 +6,15 @@
//
//===----------------------------------------------------------------------===//
+#include <clc/mem_fence/clc_mem_fence.h>
#include <clc/opencl/explicit_fence/explicit_memory_fence.h>
-
-void __clc_amdgcn_s_waitcnt(unsigned flags);
-
-// s_waitcnt takes 16bit argument with a combined number of maximum allowed
-// pending operations:
-// [12:8] LGKM -- LDS, GDS, Konstant (SMRD), Messages
-// [7] -- undefined
-// [6:4] -- exports, GDS, and mem write
-// [3:0] -- vector memory operations
-
-// Newer clang supports __builtin_amdgcn_s_waitcnt
-#if __clang_major__ >= 5
-#define __waitcnt(x) __builtin_amdgcn_s_waitcnt(x)
-#else
-#define __waitcnt(x) __clc_amdgcn_s_waitcnt(x)
-_CLC_DEF void __clc_amdgcn_s_waitcnt(unsigned) __asm("llvm.amdgcn.s.waitcnt");
-#endif
+#include <clc/opencl/synchronization/utils.h>
_CLC_DEF _CLC_OVERLOAD void mem_fence(cl_mem_fence_flags flags) {
- if (flags & CLK_GLOBAL_MEM_FENCE) {
- // scalar loads are counted with LGKM but we don't know whether
- // the compiler turned any loads to scalar
- __waitcnt(0);
- } else if (flags & CLK_LOCAL_MEM_FENCE)
- __waitcnt(0xff); // LGKM is [12:8]
+ int memory_scope = getCLCMemoryScope(flags);
+ int memory_order = __ATOMIC_SEQ_CST;
+ __clc_mem_fence(memory_scope, memory_order);
}
-#undef __waitcnt
// We don't have separate mechanism for read and write fences
_CLC_DEF _CLC_OVERLOAD void read_mem_fence(cl_mem_fence_flags flags) {
diff --git a/libclc/opencl/lib/amdgcn/synchronization/barrier.cl b/libclc/opencl/lib/amdgcn/synchronization/barrier.cl
index 5203db72f484c..c8322e602302c 100644
--- a/libclc/opencl/lib/amdgcn/synchronization/barrier.cl
+++ b/libclc/opencl/lib/amdgcn/synchronization/barrier.cl
@@ -6,10 +6,12 @@
//
//===----------------------------------------------------------------------===//
-#include <clc/opencl/explicit_fence/explicit_memory_fence.h>
#include <clc/opencl/synchronization/barrier.h>
+#include <clc/opencl/synchronization/utils.h>
+#include <clc/synchronization/clc_work_group_barrier.h>
_CLC_DEF _CLC_OVERLOAD void barrier(cl_mem_fence_flags flags) {
- mem_fence(flags);
- __builtin_amdgcn_s_barrier();
+ int memory_scope = getCLCMemoryScope(flags);
+ int memory_order = __ATOMIC_SEQ_CST;
+ __clc_work_group_barrier(memory_scope, memory_order);
}
diff --git a/libclc/opencl/lib/ptx-nvidiacl/mem_fence/fence.cl b/libclc/opencl/lib/ptx-nvidiacl/mem_fence/fence.cl
index d24569ecda1bc..e22ed870a7e6b 100644
--- a/libclc/opencl/lib/ptx-nvidiacl/mem_fence/fence.cl
+++ b/libclc/opencl/lib/ptx-nvidiacl/mem_fence/fence.cl
@@ -6,11 +6,14 @@
//
//===----------------------------------------------------------------------===//
+#include <clc/mem_fence/clc_mem_fence.h>
#include <clc/opencl/explicit_fence/explicit_memory_fence.h>
+#include <clc/opencl/synchronization/utils.h>
_CLC_DEF _CLC_OVERLOAD void mem_fence(cl_mem_fence_flags flags) {
- if (flags & (CLK_GLOBAL_MEM_FENCE | CLK_LOCAL_MEM_FENCE))
- __nvvm_membar_cta();
+ int memory_scope = getCLCMemoryScope(flags);
+ int memory_order = __ATOMIC_SEQ_CST;
+ __clc_mem_fence(memory_scope, memory_order);
}
// We do not have separate mechanism for read and write fences.
diff --git a/libclc/opencl/lib/ptx-nvidiacl/synchronization/barrier.cl b/libclc/opencl/lib/ptx-nvidiacl/synchronization/barrier.cl
index 7c57478795dda..c8322e602302c 100644
--- a/libclc/opencl/lib/ptx-nvidiacl/synchronization/barrier.cl
+++ b/libclc/opencl/lib/ptx-nvidiacl/synchronization/barrier.cl
@@ -7,7 +7,11 @@
//===----------------------------------------------------------------------===//
#include <clc/opencl/synchronization/barrier.h>
+#include <clc/opencl/synchronization/utils.h>
+#include <clc/synchronization/clc_work_group_barrier.h>
_CLC_DEF _CLC_OVERLOAD void barrier(cl_mem_fence_flags flags) {
- __syncthreads();
+ int memory_scope = getCLCMemoryScope(flags);
+ int memory_order = __ATOMIC_SEQ_CST;
+ __clc_work_group_barrier(memory_scope, memory_order);
}
diff --git a/llvm/cmake/modules/LLVMProcessSources.cmake b/llvm/cmake/modules/LLVMProcessSources.cmake
index cf358a88f5fb6..0670d60bf2afd 100644
--- a/llvm/cmake/modules/LLVMProcessSources.cmake
+++ b/llvm/cmake/modules/LLVMProcessSources.cmake
@@ -58,21 +58,6 @@ function(llvm_process_sources OUT_VAR)
set(sources ${ARG_UNPARSED_ARGUMENTS})
llvm_check_source_file_list(${sources})
- # Don't generate __SHORT_FILE__ on VS builds as it can prevent build parallelisation.
- if(NOT CMAKE_GENERATOR MATCHES "Visual Studio")
- foreach(fn ${sources})
- get_filename_component(suf ${fn} EXT)
- if("${suf}" STREQUAL ".cpp" OR "${suf}" STREQUAL ".c")
- get_filename_component(short_name ${fn} NAME)
- set_property(
- SOURCE ${fn}
- APPEND
- PROPERTY COMPILE_DEFINITIONS __SHORT_FILE__="${short_name}")
- endif()
- endforeach()
- endif()
-
-
# This adds .td and .h files to the Visual Studio solution:
add_td_sources(sources)
find_all_header_files(hdrs "${ARG_ADDITIONAL_HEADER_DIRS}")
diff --git a/llvm/include/llvm/CodeGen/TargetLowering.h b/llvm/include/llvm/CodeGen/TargetLowering.h
index 52729e9e7cee7..01f8fb5ed061f 100644
--- a/llvm/include/llvm/CodeGen/TargetLowering.h
+++ b/llvm/include/llvm/CodeGen/TargetLowering.h
@@ -3553,6 +3553,11 @@ class LLVM_ABI TargetLoweringBase {
return Libcalls.getLibcallName(Call);
}
+ /// Get the libcall routine name for the specified libcall implementation
+ const char *getLibcallImplName(RTLIB::LibcallImpl Call) const {
+ return Libcalls.getLibcallImplName(Call);
+ }
+
const char *getMemcpyName() const { return Libcalls.getMemcpyName(); }
/// Get the comparison predicate that's to be used to test the result of the
diff --git a/llvm/include/llvm/IR/RuntimeLibcalls.td b/llvm/include/llvm/IR/RuntimeLibcalls.td
index 553a302aa47b6..df472d4b9cfee 100644
--- a/llvm/include/llvm/IR/RuntimeLibcalls.td
+++ b/llvm/include/llvm/IR/RuntimeLibcalls.td
@@ -25,7 +25,8 @@ def isNotOSMSVCRT : RuntimeLibcallPredicate<"!TT.isOSMSVCRT()">;
def isPS : RuntimeLibcallPredicate<"TT.isPS()">;
def isNotOSWindowsOrIsCygwinMinGW
: RuntimeLibcallPredicate<"!TT.isOSWindows() || TT.isOSCygMing()">;
-
+def isWindowsMSVCEnvironment : RuntimeLibcallPredicate<
+ [{TT.isWindowsMSVCEnvironment()}]>;
def isGNUEnvironment : RuntimeLibcallPredicate<"TT.isGNUEnvironment()">;
def darwinHasSinCosStret : RuntimeLibcallPredicate<"darwinHasSinCosStret(TT)">;
@@ -369,6 +370,8 @@ def STACK_SMASH_HANDLER : RuntimeLibcall;
// Safe stack
def SAFESTACK_POINTER_ADDRESS : RuntimeLibcall;
+def SECURITY_CHECK_COOKIE : RuntimeLibcall;
+
// Deoptimization
def DEOPTIMIZE : RuntimeLibcall;
@@ -1009,6 +1012,10 @@ def __stack_smash_handler : RuntimeLibcallImpl<STACK_SMASH_HANDLER>;
def __riscv_flush_icache : RuntimeLibcallImpl<RISCV_FLUSH_ICACHE>;
+def __security_check_cookie : RuntimeLibcallImpl<SECURITY_CHECK_COOKIE>;
+def __security_check_cookie_arm64ec : RuntimeLibcallImpl<SECURITY_CHECK_COOKIE,
+ "#__security_check_cookie_arm64ec">;
+
//===----------------------------------------------------------------------===//
// F128 libm Runtime Libcalls
//===----------------------------------------------------------------------===//
@@ -1111,6 +1118,9 @@ defvar DarwinSinCosStret = LibcallImpls<(add __sincosf_stret, __sincos_stret),
darwinHasSinCosStret>;
defvar DarwinExp10 = LibcallImpls<(add __exp10f, __exp10), darwinHasExp10>;
+defvar SecurityCheckCookieIfWinMSVC =
+ LibcallImpls<(add __security_check_cookie), isWindowsMSVCEnvironment>;
+
defvar LibmHasSinCosF32 = LibcallImpls<(add sincosf), hasSinCos>;
defvar LibmHasSinCosF64 = LibcallImpls<(add sincos), hasSinCos>;
defvar LibmHasSinCosF80 = LibcallImpls<(add sincos_f80), hasSinCos>;
@@ -1233,7 +1243,8 @@ def AArch64SystemLibrary : SystemRuntimeLibrary<
DarwinExp10, DarwinSinCosStret,
LibmHasSinCosF32, LibmHasSinCosF64, LibmHasSinCosF128,
DefaultLibmExp10,
- DefaultStackProtector)
+ DefaultStackProtector,
+ SecurityCheckCookieIfWinMSVC)
>;
// Prepend a # to every name
@@ -1252,7 +1263,9 @@ def arm64ec___stack_chk_fail : DuplicateLibcallImplWithPrefix<__stack_chk_fail,
def WindowsARM64ECSystemLibrary
: SystemRuntimeLibrary<isWindowsArm64EC,
(add WinArm64ECDefaultRuntimeLibcallImpls,
- arm64ec___stack_chk_fail)>;
+ arm64ec___stack_chk_fail,
+ LibcallImpls<(add __security_check_cookie_arm64ec),
+ isWindowsMSVCEnvironment>)>;
//===----------------------------------------------------------------------===//
// AMDGPU Runtime Libcalls
@@ -1500,6 +1513,7 @@ def ARMSystemLibrary
LibmHasFrexpF128, LibmHasLdexpF128,
WindowARMDivRemCalls,
WindowARMFPIntCasts,
+ SecurityCheckCookieIfWinMSVC,
AEABIDivRemCalls,
DarwinSinCosStret, DarwinExp10,
LibmHasSinCosF32, LibmHasSinCosF64, LibmHasSinCosF128,
@@ -2158,6 +2172,7 @@ defvar X86CommonLibcalls =
DefaultRuntimeLibcallImpls_f80,
LibmHasExp10F32, LibmHasExp10F64, LibmHasExp10F80,
LibcallImpls<(add MostPowI), isNotOSMSVCRT>,
+ SecurityCheckCookieIfWinMSVC,
// FIXME: MSVCRT doesn't have powi. The f128 case is added as a
// hack for one test relying on it.
__powitf2_f128,
diff --git a/llvm/include/llvm/Support/DebugLog.h b/llvm/include/llvm/Support/DebugLog.h
index a3312950da94e..386cb7f1f9716 100644
--- a/llvm/include/llvm/Support/DebugLog.h
+++ b/llvm/include/llvm/Support/DebugLog.h
@@ -41,81 +41,32 @@ namespace llvm {
//
#define LDBG(...) _GET_LDBG_MACRO(__VA_ARGS__)(__VA_ARGS__)
-// Helper macros to choose the correct macro based on the number of arguments.
-#define LDBG_FUNC_CHOOSER(_f1, _f2, ...) _f2
-#define LDBG_FUNC_RECOMPOSER(argsWithParentheses) \
- LDBG_FUNC_CHOOSER argsWithParentheses
-#define LDBG_CHOOSE_FROM_ARG_COUNT(...) \
- LDBG_FUNC_RECOMPOSER((__VA_ARGS__, LDBG_LOG_LEVEL, ))
-#define LDBG_NO_ARG_EXPANDER() , LDBG_LOG_LEVEL_1
-#define _GET_LDBG_MACRO(...) \
- LDBG_CHOOSE_FROM_ARG_COUNT(LDBG_NO_ARG_EXPANDER __VA_ARGS__())
-
-// Dispatch macros to support the `level` argument or none (default to 1)
-#define LDBG_LOG_LEVEL(LEVEL) \
- DEBUGLOG_WITH_STREAM_AND_TYPE(llvm::dbgs(), LEVEL, DEBUG_TYPE)
-#define LDBG_LOG_LEVEL_1() LDBG_LOG_LEVEL(1)
-
-#define DEBUGLOG_WITH_STREAM_TYPE_FILE_AND_LINE(STREAM, LEVEL, TYPE, FILE, \
- LINE) \
- for (bool _c = \
- (::llvm::DebugFlag && ::llvm::isCurrentDebugType(TYPE, LEVEL)); \
- _c; _c = false) \
- for (::llvm::impl::RAIINewLineStream NewLineStream{(STREAM)}; _c; \
- _c = false) \
- ::llvm::impl::raw_ldbg_ostream{ \
- ::llvm::impl::computePrefix(TYPE, FILE, LINE, LEVEL), NewLineStream} \
- .asLvalue()
-
-#define DEBUGLOG_WITH_STREAM_TYPE_AND_FILE(STREAM, LEVEL, TYPE, FILE) \
- DEBUGLOG_WITH_STREAM_TYPE_FILE_AND_LINE(STREAM, LEVEL, TYPE, FILE, __LINE__)
-// When __SHORT_FILE__ is not defined, the File is the full path,
-// otherwise __SHORT_FILE__ is defined in CMake to provide the file name
-// without the path prefix.
-#if defined(__SHORT_FILE__)
-#define DEBUGLOG_WITH_STREAM_AND_TYPE(STREAM, LEVEL, TYPE) \
- DEBUGLOG_WITH_STREAM_TYPE_AND_FILE(STREAM, LEVEL, TYPE, __SHORT_FILE__)
-#else
-#define DEBUGLOG_WITH_STREAM_AND_TYPE(STREAM, LEVEL, TYPE) \
- DEBUGLOG_WITH_STREAM_TYPE_AND_FILE(STREAM, LEVEL, TYPE, \
- ::llvm::impl::getShortFileName(__FILE__))
-#endif
+#define DEBUGLOG_WITH_STREAM_AND_TYPE(STREAM, TYPE) \
+ for (bool _c = (::llvm::DebugFlag && ::llvm::isCurrentDebugType(TYPE)); _c; \
+ _c = false) \
+ ::llvm::impl::LogWithNewline(TYPE, __FILE__, __LINE__, (STREAM))
namespace impl {
-
-/// A raw_ostream that tracks `\n` and print the prefix after each
-/// newline.
-class LLVM_ABI raw_ldbg_ostream final : public raw_ostream {
- std::string Prefix;
- raw_ostream &Os;
- bool HasPendingNewline;
-
- /// Split the line on newlines and insert the prefix before each
- /// newline. Forward everything to the underlying stream.
- void write_impl(const char *Ptr, size_t Size) final {
- auto Str = StringRef(Ptr, Size);
- // Handle the initial prefix.
- if (!Str.empty())
- writeWithPrefix(StringRef());
-
- auto Eol = Str.find('\n');
- while (Eol != StringRef::npos) {
- StringRef Line = Str.take_front(Eol + 1);
- if (!Line.empty())
- writeWithPrefix(Line);
- HasPendingNewline = true;
- Str = Str.drop_front(Eol + 1);
- Eol = Str.find('\n');
- }
- if (!Str.empty())
- writeWithPrefix(Str);
+class LogWithNewline {
+public:
+ LogWithNewline(const char *debug_type, const char *file, int line,
+ raw_ostream &os)
+ : os(os) {
+ if (debug_type)
+ os << "[" << debug_type << "] ";
+ os << file << ":" << line << " ";
}
- void emitPrefix() { Os.write(Prefix.c_str(), Prefix.size()); }
- void writeWithPrefix(StringRef Str) {
- flushEol();
- Os.write(Str.data(), Str.size());
+ ~LogWithNewline() { os << '\n'; }
+ template <typename T> raw_ostream &operator<<(const T &t) && {
+ return os << t;
}
+ // Prevent copying, as this class manages newline responsibility and is
+ // intended for use as a temporary.
+ LogWithNewline(const LogWithNewline &) = delete;
+ LogWithNewline &operator=(const LogWithNewline &) = delete;
+ LogWithNewline &operator=(LogWithNewline &&) = delete;
+
public:
explicit raw_ldbg_ostream(std::string Prefix, raw_ostream &Os,
bool HasPendingNewline = true)
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 2b6ea86ee1af5..018c16d61b12d 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -28609,14 +28609,16 @@ Value *AArch64TargetLowering::getIRStackGuard(IRBuilderBase &IRB) const {
void AArch64TargetLowering::insertSSPDeclarations(Module &M) const {
// MSVC CRT provides functionalities for stack protection.
- if (Subtarget->getTargetTriple().isWindowsMSVCEnvironment()) {
+ RTLIB::LibcallImpl SecurityCheckCookieLibcall =
+ getLibcallImpl(RTLIB::SECURITY_CHECK_COOKIE);
+ if (SecurityCheckCookieLibcall != RTLIB::Unsupported) {
// MSVC CRT has a global variable holding security cookie.
M.getOrInsertGlobal("__security_cookie",
PointerType::getUnqual(M.getContext()));
// MSVC CRT has a function to validate security cookie.
FunctionCallee SecurityCheckCookie =
- M.getOrInsertFunction(Subtarget->getSecurityCheckCookieName(),
+ M.getOrInsertFunction(getLibcallImplName(SecurityCheckCookieLibcall),
Type::getVoidTy(M.getContext()),
PointerType::getUnqual(M.getContext()));
if (Function *F = dyn_cast<Function>(SecurityCheckCookie.getCallee())) {
@@ -28637,8 +28639,10 @@ Value *AArch64TargetLowering::getSDagStackGuard(const Module &M) const {
Function *AArch64TargetLowering::getSSPStackGuardCheck(const Module &M) const {
// MSVC CRT has a function to validate security cookie.
- if (Subtarget->getTargetTriple().isWindowsMSVCEnvironment())
- return M.getFunction(Subtarget->getSecurityCheckCookieName());
+ RTLIB::LibcallImpl SecurityCheckCookieLibcall =
+ getLibcallImpl(RTLIB::SECURITY_CHECK_COOKIE);
+ if (SecurityCheckCookieLibcall != RTLIB::Unsupported)
+ return M.getFunction(getLibcallImplName(SecurityCheckCookieLibcall));
return TargetLowering::getSSPStackGuardCheck(M);
}
diff --git a/llvm/lib/Target/AArch64/AArch64Subtarget.h b/llvm/lib/Target/AArch64/AArch64Subtarget.h
index 061ed611e5e47..d00e4471e107d 100644
--- a/llvm/lib/Target/AArch64/AArch64Subtarget.h
+++ b/llvm/lib/Target/AArch64/AArch64Subtarget.h
@@ -451,12 +451,6 @@ class AArch64Subtarget final : public AArch64GenSubtargetInfo {
return "__chkstk";
}
- const char* getSecurityCheckCookieName() const {
- if (isWindowsArm64EC())
- return "#__security_check_cookie_arm64ec";
- return "__security_check_cookie";
- }
-
/// Choose a method of checking LR before performing a tail call.
AArch64PAuth::AuthCheckMethod
getAuthenticatedLRCheckMethod(const MachineFunction &MF) const;
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 18766c8c6befa..74c7c97e6e927 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -21358,7 +21358,9 @@ bool ARMTargetLowering::useLoadStackGuardNode(const Module &M) const {
}
void ARMTargetLowering::insertSSPDeclarations(Module &M) const {
- if (!Subtarget->getTargetTriple().isWindowsMSVCEnvironment())
+ RTLIB::LibcallImpl SecurityCheckCookieLibcall =
+ getLibcallImpl(RTLIB::SECURITY_CHECK_COOKIE);
+ if (SecurityCheckCookieLibcall == RTLIB::Unsupported)
return TargetLowering::insertSSPDeclarations(M);
// MSVC CRT has a global variable holding security cookie.
@@ -21367,23 +21369,32 @@ void ARMTargetLowering::insertSSPDeclarations(Module &M) const {
// MSVC CRT has a function to validate security cookie.
FunctionCallee SecurityCheckCookie = M.getOrInsertFunction(
- "__security_check_cookie", Type::getVoidTy(M.getContext()),
- PointerType::getUnqual(M.getContext()));
+ getLibcallImplName(SecurityCheckCookieLibcall),
+ Type::getVoidTy(M.getContext()), PointerType::getUnqual(M.getContext()));
if (Function *F = dyn_cast<Function>(SecurityCheckCookie.getCallee()))
F->addParamAttr(0, Attribute::AttrKind::InReg);
}
Value *ARMTargetLowering::getSDagStackGuard(const Module &M) const {
- // MSVC CRT has a global variable holding security cookie.
- if (Subtarget->getTargetTriple().isWindowsMSVCEnvironment())
+ RTLIB::LibcallImpl SecurityCheckCookieLibcall =
+ getLibcallImpl(RTLIB::SECURITY_CHECK_COOKIE);
+ if (SecurityCheckCookieLibcall != RTLIB::Unsupported) {
+ // MSVC CRT has a global variable holding security cookie.
+ //
+ // FIXME: We have a libcall entry for the correlated check function, but not
+ // the global name.
return M.getGlobalVariable("__security_cookie");
+ }
+
return TargetLowering::getSDagStackGuard(M);
}
Function *ARMTargetLowering::getSSPStackGuardCheck(const Module &M) const {
// MSVC CRT has a function to validate security cookie.
- if (Subtarget->getTargetTriple().isWindowsMSVCEnvironment())
- return M.getFunction("__security_check_cookie");
+ RTLIB::LibcallImpl SecurityCheckCookie =
+ getLibcallImpl(RTLIB::SECURITY_CHECK_COOKIE);
+ if (SecurityCheckCookie != RTLIB::Unsupported)
+ return M.getFunction(getLibcallImplName(SecurityCheckCookie));
return TargetLowering::getSSPStackGuardCheck(M);
}
diff --git a/llvm/test/CodeGen/AArch64/stacksmash-arm64ec.ll b/llvm/test/CodeGen/AArch64/stacksmash-arm64ec.ll
index 0960133d7d054..bd4110173f013 100644
--- a/llvm/test/CodeGen/AArch64/stacksmash-arm64ec.ll
+++ b/llvm/test/CodeGen/AArch64/stacksmash-arm64ec.ll
@@ -1,8 +1,10 @@
-; RUN: llc -mtriple=arm64ec-unknown-windows-gnu < %s | FileCheck %s
+; RUN: llc -mtriple=arm64ec-unknown-windows < %s | FileCheck -check-prefixes=CHECK,NONGNU %s
+; RUN: llc -mtriple=arm64ec-unknown-windows-gnu < %s | FileCheck -check-prefixes=CHECK,GNU %s
; CHECK-LABEL: func = "#func"
; CHECK: bl "#other"
-; CHECK: bl "#__stack_chk_fail"
+; NONGNU: bl "#__security_check_cookie_arm64ec"
+; GNU: bl "#__stack_chk_fail"
define void @func() #0 {
entry:
%buf = alloca [10 x i8], align 1
More information about the llvm-branch-commits
mailing list