Скачать: ftp://ftp.gnu.org/gnu/gcc/gcc-4.3.1/

GCC 4.3 Release Series
Changes, New Features, and Fixes

The latest release in the 4.3 release series is GCC 4.3.1.

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.
• ColdFire targets now treat long double as having the same format as double. In earlier versions of GCC, they used the 68881 long double format instead.
• The m68k-uclinux target now uses the same calling conventions as m68k-linux-gnu. You can select the original calling conventions by configuring for m68k-uclinuxoldabi instead. Note that m68k-uclinuxoldabi also retains the original 80-bit long double on ColdFire targets.
• The -fforce-mem option has been removed because it has had no effect in the last few GCC releases.
• The i386 -msvr3-shlib option has been removed since it is no longer used.
• Fastcall for i386 has been changed not to pass aggregate arguments in registers, following Microsoft compilers.
• Support for the AOF assembler has been removed from the ARM back end; this affects only the targets arm-semi-aof and armel-semi-aof, which are no longer recognized. We removed these targets without a deprecation period because we discovered that they have been unusable since GCC 4.0.0.
• Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) has been removed. This support had been deprecated since GCC 4.0.0.

• Support for a number of older systems and recently unmaintained or untested target ports of GCC has been declared obsolete in GCC 4.3. Unless there is activity to revive them, the next release of GCC will have their sources permanently removed.

  All GCC ports for the following processor architectures have been declared obsolete:

  • Morpho MT (mt-*)

  The following aliases for processor architectures have been declared obsolete. Users should use the indicated generic target names instead, with compile-time options such as -mcpu or configure-time options such as --with-cpu to control the configuration more precisely.

  • strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* instead).
  • parisc*-*-* (use hppa*-*-* instead).
  • m680[012]0-*-* (use m68k-*-* instead).

  All GCC ports for the following operating systems have been declared obsolete:

  • BeOS (*-*-beos*)
  • kaOS (*-*-kaos*)
  • GNU/Linux using the a.out object format (*-*-linux*aout*)
  • GNU/Linux using version 1 of the GNU C Library (*-*-linux*libc1*)
  • Solaris versions before Solaris 7 (*-*-solaris2.[0-6], *-*-solaris2.[0-6].*)
  • Miscellaneous System V (*-*-sysv*)
  • WindISS (*-*-windiss*)

  Also, those for some individual systems on particular architectures have been obsoleted:

  • UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
  • CRIS with a.out object format (cris-*-aout)
  • BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
  • OSF/1 on PA-RISC (hppa1.1-*-osf*)
  • PRO on PA-RISC (hppa1.1-*-pro*)
  • Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, i[34567]86-sequent-sysv4*)
  • SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
  • UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host was previously removed in 2001, leaving only the support for UWIN as a target now being deprecated)
  • ChorusOS on PowerPC (powerpc-*-chorusos*)
  • All VAX configurations apart from NetBSD and OpenBSD (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
• The -Wconversion option has been modified. Its purpose now is to warn for implicit conversions that may alter a value. This new behavior is available for both C and C++. Warnings about conversions between signed and unsigned integers can be disabled by using -Wno-sign-conversion. In C++, they are disabled by default unless -Wsign-conversion is explicitly requested. The old behavior of -Wconversion, that is, warn for prototypes causing a type conversion that is different from what would happen to the same argument in the absence of a prototype, has been moved to a new option -Wtraditional-conversion, which is only available for C.
• The -m386, -m486, -mpentium and -mpentiumpro tuning options have been removed because they were deprecated for more than 3 GCC major releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or -mtune=pentiumpro as a replacement.
• The -funsafe-math-optimizations option now automatically turns on -fno-trapping-math in addition to -fno-signed-zeros, as it enables reassociation and thus may introduce or remove traps.
• The -ftree-vectorize option is now on by default under -O3. In order to generate code for a SIMD extension, it has to be enabled as well: use -maltivec for PowerPC platforms and -msse/-msse2 for i?86 and x86_64.
• More information on porting to GCC 4.3 from previous versions of GCC can be found in the porting guide for this release.

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, atan2, atan, atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, tan, tanh, tgamma, y0, y1 and yn. 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.
• The vectorizer was enhanced to support vectorization of outer loops, intra-iteration parallelism (loop-aware SLP), vectorization of strided accesses and loops with multiple data-types. Run-time dependency testing using loop versioning was added. The cost model, turned on by -fvect-cost-model, was developed.

New Languages and Language specific improvements

• We have added new command-line options -finstrument-functions-exclude-function-list and -finstrument-functions-exclude-file-list. They provide more control over which functions are annotated by the -finstrument-functions option.

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.
• New command-line options -Wtype-limits, -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body, -Wclobbered and -Wignored-qualifiers have been added for finer control of the diverse warnings enabled by -Wextra.
• A new function attribute alloc_size has been added to mark up malloc style functions. For constant sized allocations this can be used to find out the size of the returned pointer using the __builtin_object_size() function for buffer overflow checking and similar. This supplements the already built-in malloc and calloc constant size handling.
• Integer constants written in binary are now supported as a GCC extension. They consist of a prefix 0b or 0B, followed by a sequence of 0 and 1 digits.
• A new predefined macro __COUNTER__ has been added. It expands to sequential integral values starting from 0. In conjunction with the ## operator, this provides a convenient means to generate unique identifiers.
• A new command-line option -fdirectives-only has been added. It enables a special preprocessing mode which improves the performance of applications like distcc and ccache.
• Fixed-point data types and operators have been added. They are based on Chapter 4 of the Embedded-C specification (n1169.pdf). Currently, only MIPS targets are supported.
• Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, N1241, is now supported as a GCC extension to C for targets i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, and x86_64-*-linux-gnu. The feature introduces new data types _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, DD, and DL.

C++

• Experimental support for the upcoming ISO C++ standard, C++0x.
• -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.
• The -fvisibility-ms-compat flag was added, to make it easier to port larger projects using shared libraries from Microsoft's Visual Studio to ELF and Mach-O systems.
• C++ attribute handling has been overhauled for template arguments (ie dependent types). In particular, __attribute__((aligned(T))); works for C++ types.

Runtime Library (libstdc++)

• Experimental support for the upcoming ISO C++ standard, C++0x.
• Support for TR1 mathematical special functions and regular expressions. The implementation status for TR1 can be tracked in tr1.html
• Default what implementations give more elaborate exception strings for bad_cast, bad_typeid, bad_exception, and bad_alloc.
• Header dependencies have been streamlined, reducing unnecessary includes and pre-processed bloat.
• Variadic template implementations of items in and .
• An experimental parallel mode has been added. This is a parallel implementation of many C++ Standard library algorithms, like std::accumulate, std::for_each, std::transform, or std::sort, to give but four examples. These algorithms can be substituted for the normal (sequential) libstdc++ algorithms on a piecemeal basis, or all existing algorithms can be transformed via the -D_GLIBCXX_PARALLEL macro.
• Debug mode versions of classes in and .
• Formal deprecation of and , which are now and . This code:

      #include <ext/hash_set>
      __gnu_cxx::hash_set<int> s;
   

  Can be transformed (in order of preference) to:

      #include <tr1/unordered_set>
   
      std::tr1::unordered_set<int> s;
   

  or

      #include <backward/hash_set>
      __gnu_cxx::hash_set<int> s;
   

  Similar transformations apply to __gnu_cxx::hash_map, __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, __gnu_cxx::hash_multiset.

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.
• Support to give a backtrace (compiler flag -fbacktrace or environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a run-time error occured.
• GNU Fortran now defines __GFORTRAN__ when it runs the C preprocessor (CPP).
• The -finit-local-zero, -finit-real, -finit-integer, -finit-character, and -finit-logical options have been added, which can be used to initialize local variables.
• The intrinsic procedures GAMMA and LGAMMA have been added, which calculate the Gamma function and its logarithm. Use EXTERNAL gamma if you want to use your own gamma function.
• GNU Fortran now regards the backslash character as literal (as required by the Fortran 2003 standard); using -fbackslash GNU Fortran interprets backslashes as C-style escape characters.
• The interpretation of binary, octal and hexadecimal (BOZ) literal constants has been changed. Before they were always interpreted as integer; now they are bit-wise transferred as argument of INT, REAL, DBLE and CMPLX as required by the Fortran 2003 standard, and for real and complex variables in DATA statements or when directly assigned to real and complex variables. Everywhere else and especially in expressions they are still regarded as integer constants.
• Fortran 2003 support has been extended:
  • Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
  • Pointer intent
  • Intrinsic module ISO_ENV_FORTRAN
  • Interoperability with C (ISO C Bindings)
  • ABSTRACT INTERFACES and PROCEDURE statements (without POINTER attribute)
  • Fortran 2003 BOZ

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.
• We've made many changes to the tools shipped with gcj.
  • The old jv-scan tool has been removed. This tool never really worked properly. There is no replacement.
  • gcjh has been rewritten. Some of its more obscure options no longer work, but are still recognized in an attempt at compatibility. gjavah is a new program with similar functionality but different command-line options.
  • grmic and grmiregistry have been rewritten. grmid has been added.
  • gjar replaces the old fastjar.
  • gjarsigner (used for signing jars), gkeytool (used for key management), gorbd (for CORBA), gserialver (computes serialization UIDs), and gtnameserv (also for CORBA) are now installed.
• The ability to dump the contents of the java run time heap to a file for off-line analysis has been added. The heap dumps may be analyzed with the new gc-analyze tool. They may be generated on out-of-memory conditions or on demand and are controlled by the new run time class gnu.gcj.util.GCInfo.
• java.util.TimeZone can now read files from /usr/share/zoneinfo to provide correct, updated, timezone information. This means that packagers no longer have to update libgcj when a time zone change is published.

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.
• Support for SSE4.1 built-in functions and code generation are available via -msse4.1.
• Support for SSE4.2 built-in functions and code generation are available via -msse4.2.
• Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
• A new set of options -mpc32, -mpc64 and -mpc80 have been added to allow explicit control of x87 floating point precision.
• Support for __float128 (TFmode) IEEE quad type and corresponding TCmode IEEE complex quad type is available via the soft-fp library on x86_64 targets. This includes basic arithmetic operations (addition, subtraction, negation, multiplication and division) on __float128 real and TCmode complex values, the full set of IEEE comparisons between __float128 values, conversions to and from float, double and long double floating point types, as well as conversions to and from signed or unsigned integer, signed or unsigned long integer and signed or unsigned quad (TImode) integer types. Additionally, all operations generate the full set of IEEE exceptions and support the full set of IEEE rounding modes.
• GCC can now utilize the ACML library for vectorizing calls to a set of C99 functions on x86_64 if -mveclibabi=acml is specified and you link to an ACML ABI compatible library.

ARM

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

CRIS

New features

• Compiler and Library support for the CRIS v32 architecture, as found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been added.

Configuration changes

• The cris-*-elf target now includes support for CRIS v32, including libraries, through the -march=v32 option.
• A new crisv32-*-elf target defaults to generate code for CRIS v32.
• A new crisv32-*-linux* target defaults to generate code for CRIS v32.
• The cris-*-aout target has been obsoleted.

Improved support for built-in functions

• GCC can now use the lz and swapwbr instructions to implement the __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
• __builtin_bswap32 is now implemented using the swapwb instruction, when available.

m68k and ColdFire

New features

• Support for several new ColdFire processors has been added. You can generate code for them using the new -mcpu option.
• All targets now support ColdFire processors.
• m68k-uclinux targets have improved support for C++ constructors and destructors, and for shared libraries.
• It is now possible to set breakpoints on the first or last line of a function, even if there are no statements on that line.

Optimizations

• Support for sibling calls has been added.
• More use is now made of the ColdFire mov3q instruction.
• __builtin_clz is now implemented using the ff1 ColdFire instruction, when available.
• GCC now honors the -m68010 option. 68010 code now uses clr rather than move to zero volatile memory.
• 68020 targets and above can now use symbol(index.size*scale) addresses for indexed array accesses. Earlier compilers would always load the symbol into a base register first.

Configuration changes

• All m68k and ColdFire targets now allow the default processor to be set at configure time using --with-cpu.
• A --with-arch configuration option has been added. This option allows you to restrict a target to ColdFire or non-ColdFire processors.

Preprocessor macros

• An __mcfv*__ macro is now defined for all ColdFire targets. (Earlier versions of GCC only defined __mcfv4e__.)
• __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
• All targets now define __mc68010 and __mc68010__ when generating 68010 code.

Command-line changes

• New command-line options -march, -mcpu, -mtune and -mhard-float have been added. These options apply to both m68k and ColdFire targets.
• -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative versions of -mshort, etc.
• -fforce-addr has been removed. It is now ignored by the compiler.

Other improvements

• ColdFire targets now try to maintain a 4-byte-aligned stack where possible.
• m68k-uclinux targets now try to avoid situations that lead to the load-time error: BINFMT_FLAT: reloc outside program.

MIPS

Changes to existing configurations

• libffi and libjava now support all three GNU/Linux ABIs: o32, n32 and n64. Every GNU/Linux configuration now builds these libraries by default.
• GNU/Linux configurations now generate -mno-shared code unless overridden by -fpic, -fPIC, -fpie or -fPIE.
• mipsisa32*-linux-gnu configurations now generate hard-float code by default, just like other mipsisa32* and mips*-linux-gnu configurations. You can build a soft-float version of any mips*-linux-gnu configuration by passing --with-float=soft to configure.
• mips-wrs-vxworks now supports run-time processes (RTPs).

Changes to existing command-line options

• The -march and -mtune options no longer accept 24k as a processor name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
• The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 24kx, 24kex and 34kx.

New configurations

GCC now supports the following configurations:

• mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by default. Earlier releases also recognized this configuration, but they treated it in the same way as mipsisa32*-linux-gnu*. Note that you can customize any mips*-linux-gnu* configuration to a particular ISA or processor by passing an appropriate --with-arch option to configure.
• mipsisa*-sde-elf*, which provides compatibility with MIPS Technologies' SDE toolchains. The configuration uses the SDE libraries by default, but you can use it like other newlib-based ELF configurations by passing --with-newlib to configure. It is the only configuration besides mips64vr*-elf* to build MIPS16 as well as non-MIPS16 libraries.
• mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* configuration, but uses the o32 and o64 ABIs instead of the 32-bit and 64-bit forms of the EABI.

New processors and application-specific extensions

• Support for the SmartMIPS ASE is available through the new -msmartmips option.
• Support for revision 2 of the DSP ASE is available through the new -mdspr2 option. A new preprocessor macro called __mips_dsp_rev indicates the revision of the ASE in use.
• Support for the 4KS and 74K families of processors is available through the -march and -mtune options.

Improved support for built-in functions

• GCC can now use load-linked, store-conditional and sync instructions to implement atomic built-in functions such as __sync_fetch_and_add. The memory reference must be 4 bytes wide for 32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
• GCC can now use the clz and dclz instructions to implement the __builtin_ctz and __builtin_ffs families of functions.
• There is a new __builtin___clear_cache function for flushing the instruction cache. GCC expands this function inline on MIPS32 revision 2 targets, otherwise it calls the function specified by -mcache-flush-func.

MIPS16 improvements

• GCC can now compile objects that contain a mixture of MIPS16 and non-MIPS16 code. There are two new attributes, mips16 and nomips16, for specifying which mode a function should use.
• A new option called -minterlink-mips16 makes non-MIPS16 code link-compatible with MIPS16 code.
• After many bug fixes, the long-standing MIPS16 -mhard-float support should now work fairly reliably.
• GCC can now use the MIPS16e save and restore instructions.
• -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled with -G0 -fsection-anchors is often smaller than code compiled with -G8. However, please note that you must usually compile all objects in your application with the same -G option; see the documentation of -G for details.
• A new option called-mcode-readable specifies which instructions are allowed to load from the code segment. -mcode-readable=yes is the default and says that any instruction may load from the code segment. The other alternatives are -mcode-readable=pcrel, which says that only PC-relative MIPS16 instructions may load from the code segment, and -mcode-readable=no, which says that no instruction may do so. Please see the documentation for more details, including example uses.

Small-data improvements

There are three new options for controlling small data:

• -mno-extern-sdata, which disables small-data accesses for externally-defined variables. Code compiled with -Gn -mno-extern-sdata will be link-compatible with any -G setting between -G0 and -Gn inclusive.
• -mno-local-sdata, which disables the use of small-data sections for data that is not externally visible. This option can be a useful way of reducing small-data usage in less performance-critical parts of an application.
• -mno-gpopt, which disables the use of the $gp register while still honoring the -G limit when placing externally-visible data. This option implies -mno-extern-sdata and -mno-local-sdata and it can be useful in situations where $gp does not necessarily hold the expected value.

Miscellaneous improvements

• There is a new option called -mbranch-cost for tweaking the perceived cost of branches.
• If GCC is configured to use a version of GAS that supports the .gnu_attribute directive, it will use that directive to record certain properties of the output code. .gnu_attribute is new to GAS 2.18.
• There are two new function attributes, near and far, for overriding the command-line setting of -mlong-calls on a function-by-function basis.
• -mfp64, which previously required a 64-bit target, now works with MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and mipsisa*-sde-elf* configurations provide suitable library support.
• GCC now recognizes the -mdmx and -mmt options and passes them down to the assembler. It does nothing else with the options at present.

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

• Support has been added for this new architecture.

RS6000 (POWER/PowerPC)

• Support for the PowerPC 750CL paired-single instructions has been added with a new powerpc-*-linux*paired* target configuration. It is enabled by an associated -mpaired option and can be accessed using new built-in functions.
• Support for auto-detecting architecture and system configuration to auto-select processor optimization tuning.
• Support for VMX on AIX 5.3 has been added.
• Support for AIX Version 6.1 has been added.

S/390, zSeries and System z9

• Support for the IBM System z9 EC/BC processor (z9 GA3) has been added. When using the -march=z9-ec option, the compiler will generate code making use of instructions provided by the decimal floating point facility and the floating point conversion facility (pfpo). Besides the instructions used to implement decimal floating point operations these facilities also contain instructions to move between general purpose and floating point registers and to modify and copy the sign-bit of floating point values.
• When the -march=z9-ec option is used the new -mhard-dfp/-mno-hard-dfp options can be used to specify whether the decimal floating point hardware instructions will be used or not. If none of them is given the hardware support is enabled by default.
• The -mstack-guard option can now be omitted when using stack checking via -mstack-size in order to let GCC choose a sensible stack guard value according to the frame size of each function.
• Various changes to improve performance of generated code have been implemented, including:
  • The condition code set by an add logical with carry instruction is now available for overflow checks like: a + b + carry < b.
  • The test data class instruction is now used to implement sign-bit and infinity checks of binary and decimal floating point numbers.

Xtensa

• Stack unwinding for exception handling now uses by default a specialized version of DWARF unwinding. This is not binary-compatible with the setjmp/longjmp (sjlj) unwinding used for Xtensa with previous versions of GCC.
• For Xtensa processors that include the Conditional Store option, the built-in functions for atomic memory access are now implemented using S32C1I instructions.
• If the Xtensa NSA option is available, GCC will use it to implement the __builtin_ctz and __builtin_clz functions.

Documentation improvements

• Existing libstdc++ documentation has been edited and restructured into a single DocBook XML manual. The results can be viewed online here.

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
   

• The configure options --with-pkgversion and --with-bugurl have been added. These allow distributors of GCC to include a distributor-specific string in manuals and --version output and to specify the URL for reporting bugs in their versions of GCC.

GCC 4.3.1

This is the list of problem reports (PRs) from GCC's bug tracking system that are known to be fixed in the 4.3.1 release. This list might not be complete (that is, it is possible that some PRs that have been fixed are not listed here).

Target Specific Changes

IA-32/x86-64

ABI changes

• Starting with GCC 4.3.1, decimal floating point variables are aligned to their natural boundaries when they are passed on the stack for i386.

Command-line changes

• Starting with GCC 4.3.1, the -mcld option has been added to automatically generate a cld instruction in the prologue of functions that use string instructions. This option is used for backward compatibility on some operating systems and can be enabled by default on 32-bit x86 targets by configuring GCC with the --enable-cld configure option.