GCC 4.3 Release Series
Changes, New Features, and Fixes

Caveats

• GCC requires the GMP and MPFR libraries for building all the various front-end languages it supports. See the prerequisites page for version requirements.

General Optimizer Improvements

• The GCC middle-end has been integrated with the MPFR library. This allows GCC to evaluate and replace at compile-time calls to built-in math functions having constant arguments with their mathematically equivalent results. In making use of MPFR, GCC can generate correct results regardless of the math library implementation or floating point precision of the host platform. This also allows GCC to generate identical results regardless of whether one compiles in native or cross-compile configurations to a particular target. The following built-in functions take advantage of this new capability: acos, acosh, asin, asinh, atan, atan2, atanh, cbrt, cos, cosh, erf, erfc, exp, exp10, exp2, expm1, fdim, fma, fmax, fmin, hypot, log, log10, log1p, log2, pow, pow10, sin, sincos, sinh, tan, tanh and tgamma. The float and long double variants of these functions (e.g. sinf and sinl) are also handled. The sqrt and cabs functions with constant arguments were already optimized in prior GCC releases. Now they also use MPFR.
• A new forward propagation pass on RTL was added. The new pass replaces several slower transformations, resulting in compile-time improvements as well as better code generation in some cases.
• A new command line switch -frecord-gcc-switches has been added to GCC, although it is only enabled for some targets. The switch causes the command line that was used to invoke the compiler to be recorded into the object file that is being created. The exact format of this recording is target and binary file format dependent, but it usually takes the form of a note section containing ASCII text. The switch is related to the -fverbose-asm switch, but that one only records the information in the assembler output file as comments, so the information never reaches the object file.
• The inliner heuristic is now aware of stack frame consumption. New command line parameters --param large-stack-frame and --param large-stack-frame-growth can be used to limit stack frame size growth caused by inlining.
• During feedback directed optimizations, the expected block size the memcpy, memset and bzero functions operate on is discovered and for cases of commonly used small sizes, specialized inline code is generated.
• __builtin_expect no longer requires its argument to be a compile time constant.
• Interprocedural optimization was reorganized to work on functions in SSA form. This enables more precise and cheaper dataflow analysis and makes writing interprocedural optimizations easier. The following improvements have been implemented on top of this framework:
  • Pre-inline optimization: Selected local optimization passes are run before the inliner (and other interprocedural passes) are executed. This significantly improves the accuracy of code growth estimates used by the inliner and reduces the overall memory footprint for large compilation units.
  • Early inlining (a simple bottom-up inliner pass inlining only functions whose body is smaller than the expected call overhead) is now executed with the early optimization passes, thus inlining already optimized function bodies into an unoptimized function that is subsequently optimized by early optimizers. This enables the compiler to quickly eliminate abstraction penalty in C++ programs.
  • Interprocedural constant propagation now operate on SSA form increasing accuracy of the analysis.
• A new internal representation for GIMPLE statements has been contributed, resulting in compile-time memory savings.

New Languages and Language specific improvements

C family

• Implicit conversions between generic vector types are now only permitted when the two vectors in question have the same number of elements and compatible element types. (Note that the restriction involves compatible element types, not implicitly-convertible element types: thus, a vector type with element type int may not be implicitly converted to a vector type with element type unsigned int.) This restriction, which is in line with specifications for SIMD architectures such as AltiVec, may be relaxed using the flag -flax-vector-conversions. This flag is intended only as a compatibility measure and should not be used for new code.
• -Warray-bounds has been added and is now enabled by default for -Wall . It produces warnings for array subscripts that can be determined at compile time to be always out of bounds. -Wno-array-bounds will disable the warning.
• The constructor and destructor function attributes now accept optional priority arguments which control the order in which the constructor and destructor functions are run.

C++

• Experimental support for the upcoming ISO C++ standard, C++0x, has been added. This support can be enabled with the -std=c++0x or -std=gnu++0x compiler options. The following C++0x features are enabled by this experimental mode:
  • Static assertions
  • C99 preprocessor
• -Wc++0x-compat has been added and is now enabled by default for -Wall. It produces warnings for constructs whose meaning differs between ISO C++ 1998 and C++0x.
• The -Wparentheses option now works for C++ as it does for C. It warns if parentheses are omitted when operators with confusing precedence are nested. It also warns about ambiguous else statements. Since -Wparentheses is enabled by -Wall, this may cause additional warnings with existing C++ code which uses -Wall. These new warnings may be disabled by using -Wall -Wno-parentheses.
• The -Wmissing-declarations now works for C++ as it does for C.

Fortran

• Due to the fact that the GMP and MPFR libraries are required for all languages, Fortran is no longer special in this regard and is available by default.
• The -fexternal-blas option has been added, which generates calls to BLAS routines for intrinsic matrix operations such as matmul rather than using the built-in algorithms.
• Fortran 2003 support has been extended:
  • Intrinsic statements VOLATILE and IMPORT
  • Intrinsic module ISO_ENV_FORTRAN

Java (GCJ)

• gcj now uses the Eclipse Java compiler for its Java parsing needs. This enables the use of all 1.5 language features, and fixes most existing front end bugs.
• libgcj now supports all 1.5 language features which require runtime support: foreach, enum, annotations, generics, and auto-boxing.
• The old jv-scan tool has been removed. gcjh has been replaced by the new gjavah, which has similar functionality but different command line options. grmic and grmiregistry have been rewritten. The new tools gjar, gjarsigner, gkeytool, gorbd, grmid, gserialver, and gtnameserv are now installed.

New Targets and Target Specific Improvements

IA-32/x86-64

• Tuning for Intel Core 2 processors is available via -mtune=core2 and -march=core2.
• Tuning for AMD Geode processors is available via -mtune=geode and -march=geode.
• Code generation of block move (memcpy) and block set (memset) was rewritten. GCC can now pick the best algorithm (loop, unrolled loop, instruction with rep prefix or a library call) based on the size of the block being copied and the CPU being optimized for. A new option -minline-stringops-dynamically has been added. With this option string operations of unknown size are expanded such that small blocks are copied by in-line code, while for large blocks a library call is used. This results in faster code than -minline-all-stringops when the library implementation is capable of using cache hierarchy hints. The heuristic choosing the particular algorithm can be overwritten via -mstringop-strategy. Newly also memset of values different from 0 is inlined.
• GCC no longer places the cld instruction before string operations. Both i386 and x86-64 ABI documents mandate the direction flag to be clear at the entry of a function. It is now invalid to set the flag in asm statement without reseting it afterward.
• Support for SSSE3 built-in functions and code generation are available via -mssse3.

ARM

• Compiler and Library support for Thumb-2 and the ARMv7 architecture has been added.

SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture (BEA)

• Support has been added for this new architecture.

Documentation improvements

Other significant improvements

• The compiler's --help command line option has been extended so that it now takes an optional set of arguments. These arguments restrict the information displayed to specific classes of command line options, and possibly only a subset of those options. It is also now possible to replace the descriptive text associated with each displayed option with an indication of its current value, or for binary options, whether it has been enabled or disabled.

  Here are some examples. The following will display all the options controlling warning messages:

        --help=warnings
      

  Whereas this will display all the undocumented, target specific options:

        --help=target,undocumented
      

  This sequence of commands will display the binary optimizations that are enabled by -O3:

        gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
        gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
        diff /tmp/O2-opts /tmp/O3-opts | grep enabled
      

Please send FSF & GNU inquiries & questions to [email protected]. There are also other ways to contact the FSF.

These pages are maintained by the GCC team.

For questions related to the use of GCC, please consult these web pages and the GCC manuals. If that fails, the [email protected] mailing list might help.
Please send comments on these web pages and the development of GCC to our developer mailing list at [email protected] or [email protected]. All of our lists have public archives.

Copyright (C) Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110, USA.

Verbatim copying and distribution of this entire article is permitted in any medium, provided this notice is preserved.

Last modified 2007-02-26