Re: HARD_REGNO_MODE_OK_FOR_CLASS Might Be Nice (tm)
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Tue, Jan 03, 2006 at 05:25:55PM +, Joern RENNECKE wrote: > In http://gcc.gnu.org/ml/gcc/2005-12/msg00642.html, Bernd Jendrissek > wrote: > > Which leads me to the subject. Would it be a win to have a macro > > HARD_REGNO_MODE_OK_FOR_CLASS (REGNO, MODE, CLASS) which would be the > > authoritative test for this loop in find_reg()? On my port, and I > > imagine on many others too, I think a default > > > > #ifndef HARD_REGNO_MODE_OK_FOR_CLASS > > #define HARD_REGNO_MODE_OK_FOR_CLASS(REGNO, MODE, CLASS) \ > > HARD_REGNO_MODE_OK ((REGNO), (MODE)) > > #endif > > > > would be okay. > > It's not that simple. For example, consider multi-word integer arithmetic. > If you want to allocate a 32 bit integer register on your 16 bit x86, > all the integer registers are suitable as parts of the allocation. > However, if you start with the last integer register, the second part > will end up in a hard register which is not an integer register. > So to make this work, you's have to say that the last integer register is > not suitable for SImode, SFmode or CHImode, the last three ones are not > suitable for DImode, DFmode, CSImode or SCmode etc. That's true, but it's also true of HARD_REGNO_MODE_OK already, isn't it? Maybe it's just a misunderstanding on my part but my definition for HARD_REGNO_MODE_OK already excludes the last (2^N)-1 registers from allocation to a integer of size N units. But... for other ports, it would be a problem. It *also* couldn't be #ifndef HARD_REGNO_MODE_OK_FOR_CLASS #define HARD_REGNO_MODE_OK_FOR_CLASS(REGNO, MODE, CLASS) \ (HARD_REGNO_MODE_OK ((REGNO), (MODE)) && HARD_REGNO_MODE_OK ((REGNO) + GET_MODE_SIZE ((MODE)) - 1, (MODE))) #endif since it would ignore boundaries between disjoint register classes which it may be illegal to cross. Hmm, how about this? #ifndef HARD_REGNO_MODE_OK_FOR_CLASS #define HARD_REGNO_MODE_OK_FOR_CLASS(REGNO, MODE, CLASS) \ (HARD_REGNO_MODE_OK ((REGNO), (MODE)) && REG_IN_CLASS_P ((REGNO), (CLASS)) && HARD_REGNO_MODE_OK ((REGNO) + GET_MODE_SIZE ((MODE)) - 1, (MODE)) && REG_IN_CLASS_P ((REGNO) + GET_MODE_SIZE ((MODE)) - 1, (CLASS))) #endif (Leaving REG_IN_CLASS_P as an exercise to the reader.) This does assume though that register classes consist of exactly one contiguous series of registers. Bleh, I think I'll just bite the bullet and use up all my remaining constraint characters for now. Thanks for the replies. - -- "If you lie to the compiler, it will get its revenge." - Henry Spencer -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.2 (GNU/Linux) Comment: Please fetch my new key 804177F8 from hkp://wwwkeys.eu.pgp.net/ iD8DBQFDv7THwyMv24BBd/gRAsLBAJ0X6LQuBoWJfNoD2+lLDAXYzvM7LwCbByid FFjZg0GAtEq8vmwbG6/nY7I= =kEtK -END PGP SIGNATURE-
gcc-4.2-20060107 is now available
Snapshot gcc-4.2-20060107 is now available on ftp://gcc.gnu.org/pub/gcc/snapshots/4.2-20060107/ and on various mirrors, see http://gcc.gnu.org/mirrors.html for details. This snapshot has been generated from the GCC 4.2 SVN branch with the following options: svn://gcc.gnu.org/svn/gcc/trunk revision 109451 You'll find: gcc-4.2-20060107.tar.bz2 Complete GCC (includes all of below) gcc-core-4.2-20060107.tar.bz2 C front end and core compiler gcc-ada-4.2-20060107.tar.bz2 Ada front end and runtime gcc-fortran-4.2-20060107.tar.bz2 Fortran front end and runtime gcc-g++-4.2-20060107.tar.bz2 C++ front end and runtime gcc-java-4.2-20060107.tar.bz2 Java front end and runtime gcc-objc-4.2-20060107.tar.bz2 Objective-C front end and runtime gcc-testsuite-4.2-20060107.tar.bz2The GCC testsuite Diffs from 4.2-20051231 are available in the diffs/ subdirectory. When a particular snapshot is ready for public consumption the LATEST-4.2 link is updated and a message is sent to the gcc list. Please do not use a snapshot before it has been announced that way.
Atomic operations on Sparc in libstdc++
Hello, why there is no cas-based version of __exchange_and_add for 32-bit mode Sparcs in libstdc++, namely in config/cpu/sparc/atomicity.h ? There is spinlock-based generic code for all Sparcs, but V9 does support 32-bit cas, which can be used instead of a spinlock. This spinlock _Atomicity_lock<__inst>::_S_atomicity_lock causes serious scalability problems with std::string. On a machine with 4 processors our 32-bit code is only about 2.3 times faster that on a uniprocessor. When I hack atomicity.h, providing my own cas-based __exchange_and_add, it works nearly 4 times faster compared to a UP. Is it possible to replace __exchange_and_add with my code without patching libstdc++? Unfortunately it is impossible to provide a generic patch for atomicity.h, because GCC does not define __sparcv9 in 32-bit mode, as described in this thread: http://gcc.gnu.org/ml/gcc/2005-11/msg00842.html Best regards Piotr Wyderski
successful bootstrap
Hey, just a note to let you guys know i bootstrapped gcc 3.4.5 on my SPARC64/OpenSolaris ( ultrasparc II / opensolaris nv28) box (after fighting with it, but that's my fault). I haven't tried to compile g77 or ada or any such thing, since i only need c and c++. Only one processor in the box, no idea how this'd work with SMP. evidently GNU make is a must. GNU binutils, similarly. The compiler i used to build GCC (and everything previous) was SunStudio10 (which is register-for-free-smokes free). also, the requirement that one passes CC="cc -xildoff -xarch=v9" is still in effect with studio10/solaris11 here's the shell output you guys want: [SPARCy] error404 :: objdir $ ./config.guess sparc-sun-solaris2.11 [SPARCy] error404 :: objdir $ gcc -v Reading specs from /opt/GNU/lib/gcc/sparc- sun-solaris2.11/3.4.5/specs Configured with: ../configure --with-gnu-as --with-as=/opt/GNU/bin/as --with-gnu-ld --with-ld=/opt/GNU/bin/ld --prefix=/opt/GNU --enable-languages=c,c++ --disable-nls --enable-shared Thread model: posix gcc version 3.4.5 Cheers! -john -- PGP public key availiable on wwwkeys.pgp.net
Re: successful bootstrap
> evidently GNU make is a must. That's documented. > GNU binutils, similarly. That would be a bug, the Sun tools are supported. > The compiler i > used to build GCC (and everything previous) was SunStudio10 (which is > register-for-free-smokes free). also, the requirement that one passes > CC="cc -xildoff -xarch=v9" is still in effect with studio10/solaris11 Actually -xildoff is unnecessary. -- Eric Botcazou
