hanchenye-llvm-project/clang/lib/CodeGen
Justin Lebar 9a2c0fbaf5 [CUDA] Don't crash when trying to printf a non-scalar object.
Summary:
We can't do the right thing, since there's no right thing to do, but at
least we can not crash the compiler.

Reviewers: majnemer, rnk

Subscribers: cfe-commits, jhen, tra

Differential Revision: http://reviews.llvm.org/D17103

llvm-svn: 260479
2016-02-11 02:00:52 +00:00
..
ABIInfo.h Support __builtin_ms_va_list. 2015-09-17 20:55:33 +00:00
Address.h Work around build failure due to GCC 4.8.1 bug. We don't completely understand 2016-02-02 23:11:49 +00:00
BackendUtil.cpp Eliminate an unnecessary enum, use the LLVM version. NFC 2016-02-05 23:23:25 +00:00
CGAtomic.cpp [MSVC Compat] Don't provide /volatile:ms semantics to types > pointer 2016-01-22 16:36:44 +00:00
CGBlocks.cpp Move DebugInfoKind into its own header to cut the cyclic dependency edge from Driver to Frontend. 2016-02-02 11:06:51 +00:00
CGBlocks.h Move BlockByrefHelpers back to CodeGenModule.h to placate MSVC. 2015-09-08 08:21:11 +00:00
CGBuilder.h Revert "Change memcpy/memset/memmove to have dest and source alignments." 2015-11-19 05:55:59 +00:00
CGBuiltin.cpp Add builtins for bitreverse intrinsic 2016-02-03 17:49:38 +00:00
CGCUDABuiltin.cpp [CUDA] Don't crash when trying to printf a non-scalar object. 2016-02-11 02:00:52 +00:00
CGCUDANV.cpp [CUDA] Invoke ptxas and fatbinary during compilation. 2016-01-14 21:41:27 +00:00
CGCUDARuntime.cpp Roll-back r250822. 2015-10-20 13:23:58 +00:00
CGCUDARuntime.h Revert r240270 ("Fixed/added namespace ending comments using clang-tidy"). 2015-06-22 23:07:51 +00:00
CGCXX.cpp Use CodeGenModule::addReplacement() instead of directly accessing Replacements[]. 2016-02-07 12:44:35 +00:00
CGCXXABI.cpp Add the `pass_object_size` attribute to clang. 2015-12-02 21:58:08 +00:00
CGCXXABI.h Fix use-after-free when a C++ thread_local variable gets replaced (because its 2015-12-01 01:10:48 +00:00
CGCall.cpp Fix undefined behavior when compiling in C++14 due to sized operator delete 2016-02-09 01:05:04 +00:00
CGCall.h Don't emit exceptional stackrestore cleanups around inalloca functions 2015-10-08 00:17:45 +00:00
CGClass.cpp Fix some Clang-tidy readability-redundant-control-flow warnings; other minor fixes. 2016-02-10 19:11:58 +00:00
CGCleanup.cpp Update for LLVM function name change. 2016-01-14 21:00:27 +00:00
CGCleanup.h Update for LLVM function name change. 2016-01-14 21:00:27 +00:00
CGDebugInfo.cpp Use llvm::TempDIScope instead of manually deleting a temporary MDNode. 2016-02-08 17:03:28 +00:00
CGDebugInfo.h Fix typo in comment. NFC 2016-02-07 06:39:23 +00:00
CGDecl.cpp Fix some Clang-tidy readability-redundant-control-flow warnings; other minor fixes. 2016-02-10 19:11:58 +00:00
CGDeclCXX.cpp [CUDA] Do not allow dynamic initialization of global device side variables. 2016-02-02 22:29:48 +00:00
CGException.cpp Reland "[SEH] Use llvm.x86.seh.recoverfp for SEH filters on x64 as well as 32bit" 2015-12-16 00:26:37 +00:00
CGExpr.cpp [OPENMP 4.0] Fixed support of array sections/array subscripts. 2016-02-04 11:27:03 +00:00
CGExprAgg.cpp [Bugfix] Fix ICE on constexpr vector splat. 2016-01-13 01:52:39 +00:00
CGExprCXX.cpp [MS ABI] Tolerate invokes of __RTDynamicCast 2015-11-23 03:01:14 +00:00
CGExprComplex.cpp [Bugfix] Fix ICE on constexpr vector splat. 2016-01-13 01:52:39 +00:00
CGExprConstant.cpp Update for LLVM function name change. 2016-01-14 21:00:27 +00:00
CGExprScalar.cpp Emit calls to objc_unsafeClaimAutoreleasedReturnValue when 2016-01-27 18:32:30 +00:00
CGLoopInfo.cpp Add new llvm.loop.unroll.enable metadata for use with "#pragma unroll". 2015-08-10 17:29:39 +00:00
CGLoopInfo.h Add new llvm.loop.unroll.enable metadata for use with "#pragma unroll". 2015-08-10 17:29:39 +00:00
CGObjC.cpp Emit calls to objc_unsafeClaimAutoreleasedReturnValue when 2016-01-27 18:32:30 +00:00
CGObjCGNU.cpp Fix some Clang-tidy readability-redundant-control-flow warnings; other minor fixes. 2016-02-10 19:11:58 +00:00
CGObjCMac.cpp Fix some Clang-tidy readability-redundant-control-flow warnings; other minor fixes. 2016-02-10 19:11:58 +00:00
CGObjCRuntime.cpp Update for LLVM function name change. 2016-01-14 21:00:27 +00:00
CGObjCRuntime.h Reapply "[CodeGen] Fix assignments of inline layouts into the byref structure" 2015-12-21 20:21:15 +00:00
CGOpenCLRuntime.cpp [OpenCL] Pipe type support 2016-01-09 12:53:17 +00:00
CGOpenCLRuntime.h [OpenCL] Pipe type support 2016-01-09 12:53:17 +00:00
CGOpenMPRuntime.cpp Fix some Clang-tidy readability-redundant-control-flow warnings; other minor fixes. 2016-02-10 19:11:58 +00:00
CGOpenMPRuntime.h Make the remaining headers self-contained. 2016-02-02 14:24:21 +00:00
CGOpenMPRuntimeNVPTX.cpp Re-apply for the 2nd-time r259977 - [OpenMP] Reorganize code to allow specialized code generation for different devices. 2016-02-08 15:59:20 +00:00
CGOpenMPRuntimeNVPTX.h Re-apply for the 2nd-time r259977 - [OpenMP] Reorganize code to allow specialized code generation for different devices. 2016-02-08 15:59:20 +00:00
CGRecordLayout.h Make CodeGen headers self-contained. 2016-02-02 16:05:18 +00:00
CGRecordLayoutBuilder.cpp Annotate dump() methods with LLVM_DUMP_METHOD, addressing Richard Smith r259192 post commit comment. 2016-01-29 19:38:18 +00:00
CGStmt.cpp [PGO] cc1 option name change for profile instrumentation 2016-02-04 18:39:09 +00:00
CGStmtOpenMP.cpp [OPENMP 4.0] Fixed support of array sections/array subscripts. 2016-02-04 11:27:03 +00:00
CGVTT.cpp Remove and forbid raw_svector_ostream::flush() calls. 2015-08-13 18:12:56 +00:00
CGVTables.cpp Use a consistent spelling for vtables. 2016-01-29 01:35:53 +00:00
CGVTables.h [CodeGen] Remove dead code. NFC. 2015-10-15 15:29:40 +00:00
CGValue.h Introduce __builtin_nontemporal_store and __builtin_nontemporal_load. 2015-09-08 23:52:33 +00:00
CMakeLists.txt Re-apply for the 2nd-time r259977 - [OpenMP] Reorganize code to allow specialized code generation for different devices. 2016-02-08 15:59:20 +00:00
CodeGenABITypes.cpp Add the `pass_object_size` attribute to clang. 2015-12-02 21:58:08 +00:00
CodeGenAction.cpp Add backend dignostic printer for unsupported features 2016-02-02 13:52:52 +00:00
CodeGenFunction.cpp [PGO] cc1 option name change for profile instrumentation 2016-02-04 18:39:09 +00:00
CodeGenFunction.h [PGO] cc1 option name change for profile instrumentation 2016-02-04 18:39:09 +00:00
CodeGenModule.cpp Re-apply for the 2nd-time r259977 - [OpenMP] Reorganize code to allow specialized code generation for different devices. 2016-02-08 15:59:20 +00:00
CodeGenModule.h [cfi] Safe handling of unaddressable vtable pointers (clang). 2016-02-03 22:18:55 +00:00
CodeGenPGO.cpp [PGO] code simplification: use existing VP annotation API /NFC 2016-02-04 19:54:17 +00:00
CodeGenPGO.h [PGO] Windows buildbot failure fix. [NFC] 2016-01-24 00:56:19 +00:00
CodeGenTBAA.cpp Roll-back r250822. 2015-10-20 13:23:58 +00:00
CodeGenTBAA.h Make the remaining headers self-contained. 2016-02-02 14:24:21 +00:00
CodeGenTypeCache.h Compute and preserve alignment more faithfully in IR-generation. 2015-09-08 08:05:57 +00:00
CodeGenTypes.cpp [MS ABI] Allow a member pointers' converted type to change 2016-01-26 19:30:26 +00:00
CodeGenTypes.h [MS ABI] Allow a member pointers' converted type to change 2016-01-26 19:30:26 +00:00
CoverageMappingGen.cpp [Coverage] Fix crash when handling certain macro expansions 2016-02-08 19:25:45 +00:00
CoverageMappingGen.h [Coverage] Reduce complexity of adding function mapping records 2016-01-21 19:25:35 +00:00
EHScopeStack.h Update clang to use the updated LLVM EH instructions 2015-12-12 05:39:21 +00:00
ItaniumCXXABI.cpp Fix Itanium RTTI emission so that we emit fundamental type information into the 2016-02-03 01:32:42 +00:00
MicrosoftCXXABI.cpp Revert r260388 "[MS ABI] Never reference dllimport'd vtables" 2016-02-10 22:18:37 +00:00
ModuleBuilder.cpp Check for frontend errors after releasing the Builder. 2016-01-28 23:29:02 +00:00
ObjectFilePCHContainerOperations.cpp Move DebugInfoKind into its own header to cut the cyclic dependency edge from Driver to Frontend. 2016-02-02 11:06:51 +00:00
README.txt
SanitizerMetadata.cpp [ASan] Initial support for Kernel AddressSanitizer 2015-06-19 12:19:07 +00:00
SanitizerMetadata.h Removing LLVM_DELETED_FUNCTION, as MSVC 2012 was the last reason for requiring the macro. NFC; Clang edition. 2015-02-15 22:54:08 +00:00
TargetInfo.cpp Add an ARC autoreleased-return-value caller marker on i386. 2016-02-05 21:37:38 +00:00
TargetInfo.h Fix Clang-tidy modernize-use-nullptr warnings in headers and generated files; other minor cleanups. 2015-09-29 20:56:43 +00:00

README.txt

IRgen optimization opportunities.

//===---------------------------------------------------------------------===//

The common pattern of
--
short x; // or char, etc
(x == 10)
--
generates an zext/sext of x which can easily be avoided.

//===---------------------------------------------------------------------===//

Bitfields accesses can be shifted to simplify masking and sign
extension. For example, if the bitfield width is 8 and it is
appropriately aligned then is is a lot shorter to just load the char
directly.

//===---------------------------------------------------------------------===//

It may be worth avoiding creation of alloca's for formal arguments
for the common situation where the argument is never written to or has
its address taken. The idea would be to begin generating code by using
the argument directly and if its address is taken or it is stored to
then generate the alloca and patch up the existing code.

In theory, the same optimization could be a win for block local
variables as long as the declaration dominates all statements in the
block.

NOTE: The main case we care about this for is for -O0 -g compile time
performance, and in that scenario we will need to emit the alloca
anyway currently to emit proper debug info. So this is blocked by
being able to emit debug information which refers to an LLVM
temporary, not an alloca.

//===---------------------------------------------------------------------===//

We should try and avoid generating basic blocks which only contain
jumps. At -O0, this penalizes us all the way from IRgen (malloc &
instruction overhead), all the way down through code generation and
assembly time.

On 176.gcc:expr.ll, it looks like over 12% of basic blocks are just
direct branches!

//===---------------------------------------------------------------------===//