llvm-project/clang/lib/CodeGen
Eli Friedman db42a3e876 Make sure we or together the overflow flags of the multiply and add, so the
check is triggered appropriately.  Reported on cfe-dev.

llvm-svn: 129231
2011-04-09 19:54:33 +00:00
..
ABIInfo.h Fix Whitespace. 2010-10-19 06:39:39 +00:00
BackendUtil.cpp Added *hidden* flags -print-options and -print-all-options so 2011-04-05 18:56:55 +00:00
CGBlocks.cpp Whoops. 2011-03-31 09:19:20 +00:00
CGBlocks.h Remove a rather egregious use of getFunctionInfo. 2011-03-09 08:39:33 +00:00
CGBuilder.h Make CGBuilderTy a typedef again; its functionality has been rolled back 2010-07-06 18:43:48 +00:00
CGBuiltin.cpp add a __sync_swap builtin to fill out the rest of the __sync builtins. 2011-04-09 03:57:26 +00:00
CGCXX.cpp Convert the BaseOffset member of BaseSubobject to CharUnits from bits. No 2011-03-24 01:21:01 +00:00
CGCXXABI.cpp Use the "undergoes default argument promotion" bit on parameters to 2011-03-09 04:27:21 +00:00
CGCXXABI.h More capturing of 'this': implicit member expressions. Getting that 2011-02-03 08:15:49 +00:00
CGCall.cpp PR8369: make __attribute((regparm(0))) work correctly. Original patch by 2011-04-09 08:18:08 +00:00
CGCall.h PR8369: make __attribute((regparm(0))) work correctly. Original patch by 2011-04-09 08:18:08 +00:00
CGClass.cpp [Reapply r128776, modified so that it does not break debug info.] 2011-04-07 12:37:09 +00:00
CGCleanup.cpp Tame an assert; the scope depth of a jump destination does not 2011-02-25 04:19:13 +00:00
CGCleanup.h Move all the cleanups framework code into a single file. 2011-01-28 11:13:47 +00:00
CGDebugInfo.cpp Apply explicit braces to avoid ambiguous 'else' [-Wparentheses] 2011-04-09 00:25:15 +00:00
CGDebugInfo.h Emit debug info for function template parameters. 2011-04-05 22:54:11 +00:00
CGDecl.cpp Fix typo. 2011-04-05 20:28:21 +00:00
CGDeclCXX.cpp __cxa_atexit does not throw. 2011-03-20 20:52:32 +00:00
CGException.cpp Fix Objective-C++ exceptions (GNU runtime). 2011-03-20 21:35:39 +00:00
CGException.h Fix Objective-C++ exceptions (GNU runtime). 2011-03-20 21:35:39 +00:00
CGExpr.cpp Basic, untested implementation for an "unknown any" type requested by LLDB. 2011-04-07 08:22:57 +00:00
CGExprAgg.cpp Basic, untested implementation for an "unknown any" type requested by LLDB. 2011-04-07 08:22:57 +00:00
CGExprCXX.cpp Make sure we or together the overflow flags of the multiply and add, so the 2011-04-09 19:54:33 +00:00
CGExprComplex.cpp Remove PHINode::reserveOperandSpace(). Instead, add a parameter to 2011-03-30 11:28:58 +00:00
CGExprConstant.cpp Basic, untested implementation for an "unknown any" type requested by LLDB. 2011-04-07 08:22:57 +00:00
CGExprScalar.cpp Basic, untested implementation for an "unknown any" type requested by LLDB. 2011-04-07 08:22:57 +00:00
CGObjC.cpp Refine rules for atomic property api to 2011-04-06 16:05:26 +00:00
CGObjCGNU.cpp Fix copy-and-paste bug that I introduced while tidying up the code. 2011-04-05 17:15:18 +00:00
CGObjCMac.cpp Obj-C/NeXT: Update and reapply 108847, now that changes are more baked. 2011-03-25 20:09:09 +00:00
CGObjCRuntime.cpp Continuing work on ObjC tidyup: 2011-03-25 11:57:33 +00:00
CGObjCRuntime.h Continuing work on ObjC tidyup: 2011-03-25 11:57:33 +00:00
CGRTTI.cpp Eliminate a divide-by-8 in BuildVMIClassTypeInfo() by using CharUnits for 2011-04-09 01:09:56 +00:00
CGRecordLayout.h Perform zero-initialization of virtual base classes when emitting 2011-02-15 06:40:56 +00:00
CGRecordLayoutBuilder.cpp Fix typo in comment. 2011-03-22 17:35:47 +00:00
CGStmt.cpp Fix in r128471 is very broad. Some of the unconditional branches need line number information for better user experience. 2011-03-30 00:08:31 +00:00
CGTemporaries.cpp Fix some obvious bugs in the conditional-cleanup code and then make the 2011-01-26 19:15:39 +00:00
CGVTT.cpp We were emitting construction v-tables with internal linkage all the time. 2011-03-27 09:00:25 +00:00
CGVTables.cpp [Reapply r128776, modified so that it does not break debug info.] 2011-04-07 12:37:09 +00:00
CGVTables.h [Reapply r128776, modified so that it does not break debug info.] 2011-04-07 12:37:09 +00:00
CGValue.h More capturing of 'this': implicit member expressions. Getting that 2011-02-03 08:15:49 +00:00
CMakeLists.txt Unbreak CMake build. 2011-03-25 14:16:48 +00:00
CodeGenAction.cpp Re-instate r125819 and r125820 with no functionality change 2011-02-19 23:03:58 +00:00
CodeGenFunction.cpp Remove PHINode::reserveOperandSpace(). Instead, add a parameter to 2011-03-30 11:28:58 +00:00
CodeGenFunction.h After much contemplation, I've decided that we probably shouldn't "unique" 2011-03-31 08:03:29 +00:00
CodeGenModule.cpp fix indentation 2011-04-09 07:11:53 +00:00
CodeGenModule.h Basic, untested implementation for an "unknown any" type requested by LLDB. 2011-04-07 08:22:57 +00:00
CodeGenTBAA.cpp For consistency, use llvm::raw_ostream in the rest of the mangle api. 2011-02-11 02:52:17 +00:00
CodeGenTBAA.h Move name mangling support from CodeGen to AST. In the 2011-01-13 18:57:25 +00:00
CodeGenTypes.cpp Basic, untested implementation for an "unknown any" type requested by LLDB. 2011-04-07 08:22:57 +00:00
CodeGenTypes.h Use the "undergoes default argument promotion" bit on parameters to 2011-03-09 04:27:21 +00:00
GlobalDecl.h When IRgen refers to a function declaration that is not a definition, 2011-02-09 02:03:05 +00:00
ItaniumCXXABI.cpp Convert the PointerWidthInBytes variable in EmitMemberPointer() to CharUnits 2011-04-09 01:30:02 +00:00
Makefile BUILD_ARCHIVE is the default for libraries, no need to set it. 2010-07-18 00:14:47 +00:00
MicrosoftCXXABI.cpp Notes on dynamic array cookies in MSVC. 2011-01-27 02:46:02 +00:00
ModuleBuilder.cpp In addition to in-class member functions marked with the "used" 2011-02-19 21:54:50 +00:00
README.txt These IRgen improvements have been done. 2009-07-23 03:03:07 +00:00
TargetInfo.cpp Set AAPCS-VFP calling convention accordingly and hard float ABI command handling. 2011-04-05 00:23:47 +00:00
TargetInfo.h Re-instate r125819 and r125820 with no functionality change 2011-02-19 23:03:58 +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!

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