llvm-project/clang/lib/CodeGen
Duncan P. N. Exon Smith 38402dc917 PGO: Scale large counters down to 32-bits
PGO counters are 64-bit and branch weights are 32-bit.  Scale them down
when necessary, instead of just taking the lower 32 bits.

<rdar://problem/16276448>

llvm-svn: 203592
2014-03-11 18:18:10 +00:00
..
ABIInfo.h Sort all the #include lines with LLVM's utils/sort_includes.py which 2014-01-07 11:51:46 +00:00
BackendUtil.cpp [C++11] Replace OwningPtr include with <memory>. 2014-03-09 11:36:40 +00:00
CGAtomic.cpp Sema: produce error when invalid ordering is passed to atomic builtin 2014-03-11 10:49:14 +00:00
CGBlocks.cpp Replace OwningPtr with std::unique_ptr. 2014-03-07 20:03:18 +00:00
CGBlocks.h Remove useless 'llvm::' qualifier from names like StringRef and others that are 2013-01-12 19:30:44 +00:00
CGBuilder.h Rewrite #includes for llvm/Foo.h to llvm/IR/Foo.h as appropriate to 2013-01-02 11:45:17 +00:00
CGBuiltin.cpp Update for IR: add a second AtomicOrdering to cmpxchg insts. 2014-03-11 10:49:03 +00:00
CGCUDANV.cpp [Modules] Update to reflect the move of CallSite into the IR library in 2014-03-04 11:02:08 +00:00
CGCUDARuntime.cpp CodeGen: Initial instrumentation based PGO implementation 2014-01-06 22:27:43 +00:00
CGCUDARuntime.h CUDA: IR generation support for device stubs 2011-10-06 18:51:56 +00:00
CGCXX.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
CGCXXABI.cpp Sema: Do not assert when dereferencing member pointer using virtual inheritance with an incomplete class type 2014-02-20 23:22:07 +00:00
CGCXXABI.h Replace OwningPtr with std::unique_ptr. 2014-03-07 20:03:18 +00:00
CGCall.cpp [C++11] Update Clang for the change to LLVM's Use-Def chain iterators in 2014-03-09 03:16:50 +00:00
CGCall.h [ms-cxxabi] Use inalloca on win32 when passing non-trivial C++ objects 2014-02-01 00:04:45 +00:00
CGClass.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
CGCleanup.cpp [C++11] Update Clang for the change to LLVM's Use-Def chain iterators in 2014-03-09 03:16:50 +00:00
CGCleanup.h Fix leak in lib/CodeGen/CGException.cpp, PR18318 2014-01-09 09:22:32 +00:00
CGDebugInfo.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
CGDebugInfo.h [Layering] Update include of DebugInfo.h which moved to the IR library 2014-03-06 00:46:56 +00:00
CGDecl.cpp Use llvm.compiler.used instead of llvm.used for objc symbols. 2014-03-06 22:15:10 +00:00
CGDeclCXX.cpp [ms-cxxabi] Use x86_cdeclmethodcc for __cdecl methods on win32 2014-01-31 22:54:50 +00:00
CGException.cpp [C++11] Update Clang for the change to LLVM's Use-Def chain iterators in 2014-03-09 03:16:50 +00:00
CGExpr.cpp Sema: Do not assert when dereferencing member pointer using virtual inheritance with an incomplete class type 2014-02-20 23:22:07 +00:00
CGExprAgg.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
CGExprCXX.cpp [Modules] Update to reflect the move of CallSite into the IR library in 2014-03-04 11:02:08 +00:00
CGExprComplex.cpp Change PGO instrumentation to compute counts in a separate AST traversal. 2014-02-17 19:21:09 +00:00
CGExprConstant.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
CGExprScalar.cpp Update for IR: add a second AtomicOrdering to cmpxchg insts. 2014-03-11 10:49:03 +00:00
CGObjC.cpp [C++11] Replacing ObjCMethodDecl iterators param_begin() and param_end() with iterator_range params(). Updating all of the usages of the iterators with range-based for loops. 2014-03-07 17:50:17 +00:00
CGObjCGNU.cpp Remove dead return and simplify code. 2014-03-07 18:53:05 +00:00
CGObjCMac.cpp Objective-C IRGen. Fixes several regressions caused by changes made 2014-03-11 00:25:05 +00:00
CGObjCRuntime.cpp [Modules] Update to reflect the move of CallSite into the IR library in 2014-03-04 11:02:08 +00:00
CGObjCRuntime.h Refactor linkage computation in CGObjCMac.cpp. 2014-02-26 22:25:45 +00:00
CGOpenCLRuntime.cpp The OpenCL specification states that images are allocated 2014-01-09 13:37:30 +00:00
CGOpenCLRuntime.h Rewrite #includes for llvm/Foo.h to llvm/IR/Foo.h as appropriate to 2013-01-02 11:45:17 +00:00
CGRTTI.cpp RTTI symbols for visible local types may need weak linkage. 2014-03-10 22:27:33 +00:00
CGRecordLayout.h Correct hyphenations in comments and assert messages 2013-12-05 04:47:09 +00:00
CGRecordLayoutBuilder.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
CGStmt.cpp [C++11] Update Clang for the change to LLVM's Use-Def chain iterators in 2014-03-09 03:16:50 +00:00
CGVTT.cpp Remove the -fhidden-weak-vtables -cc1 option. It was dead, 2014-02-08 00:41:16 +00:00
CGVTables.cpp Replace OwningPtr with std::unique_ptr. 2014-03-07 20:03:18 +00:00
CGVTables.h MS ABI: Mangle member pointer template arguments 2014-02-05 17:27:08 +00:00
CGValue.h Simplify atomic load/store IRGen. 2013-07-11 01:32:21 +00:00
CMakeLists.txt [CMake] Use LINK_LIBS instead of target_link_libraries(). 2014-02-26 06:41:29 +00:00
CodeGenABITypes.cpp [ms-cxxabi] Use x86_cdeclmethodcc for __cdecl methods on win32 2014-01-31 22:54:50 +00:00
CodeGenAction.cpp [C++11] Replace OwningPtr include with <memory>. 2014-03-09 11:36:40 +00:00
CodeGenFunction.cpp [C++11] Replacing DeclBase iterators specific_attr_begin() and specific_attr_end() with iterator_range specific_attrs(). Updating all of the usages of the iterators with range-based for loops. 2014-03-10 17:08:28 +00:00
CodeGenFunction.h [Modules] Update to reflect ValueHandle moving to the IR library in LLVM 2014-03-04 11:18:19 +00:00
CodeGenModule.cpp [C++11] Replacing DeclBase iterators specific_attr_begin() and specific_attr_end() with iterator_range specific_attrs(). Updating all of the usages of the iterators with range-based for loops. 2014-03-10 17:08:28 +00:00
CodeGenModule.h Replace OwningPtr with std::unique_ptr. 2014-03-07 20:03:18 +00:00
CodeGenPGO.cpp PGO: Scale large counters down to 32-bits 2014-03-11 18:18:10 +00:00
CodeGenPGO.h [C++11] Replace OwningPtr include with <memory>. 2014-03-09 11:36:40 +00:00
CodeGenTBAA.cpp Add a mangler entry point for TBAA rather than using RTTI directly 2013-11-19 23:23:00 +00:00
CodeGenTBAA.h Fixed comment typo. 2013-06-08 00:27:19 +00:00
CodeGenTypes.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
CodeGenTypes.h [ms-cxxabi] Use inalloca on win32 when passing non-trivial C++ objects 2014-02-01 00:04:45 +00:00
EHScopeStack.h [C++11] Replace LLVM-style type traits with C++11 standard ones. 2014-03-07 14:30:23 +00:00
ItaniumCXXABI.cpp Sema: Do not assert when dereferencing member pointer using virtual inheritance with an incomplete class type 2014-02-20 23:22:07 +00:00
Makefile
MicrosoftCXXABI.cpp [-cxx-abi microsoft] Implement local manglings accurately 2014-03-05 08:57:59 +00:00
ModuleBuilder.cpp [C++11] Replace OwningPtr include with <memory>. 2014-03-09 11:36:40 +00:00
README.txt
TargetInfo.cpp [C++11] Replacing RecordDecl iterators field_begin() and field_end() with iterator_range fields(). Updating all of the usages of the iterators with range-based for loops. 2014-03-08 20:12:42 +00:00
TargetInfo.h Sort all the #include lines with LLVM's utils/sort_includes.py which 2014-01-07 11:51:46 +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!

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