From d.j.walker.slh0101@oasis.icl.co.uk Mon Mar 11 18:04:24 1996 Received: from relay2.pipex.net (relay2.pipex.net [158.43.192.4]) by cip.informatik.uni-erlangen.de with SMTP id SAA11230 (8.6.12/7.4h-FAU); for ; Mon, 11 Mar 1996 18:04:05 +0100 Received: from Q.icl.co.uk by bath.pipex.net with SMTP (PP); Mon, 11 Mar 1996 17:03:05 +0000 Received: from x400-gw0.pipex.net by Q.icl.co.uk (4.1/icl-2.12-server) id AB27085; Mon, 11 Mar 96 17:02:38 GMT Received: from getafix.oasis.icl.co.uk by obelix over ESMTP id QAA23221; Mon, 11 Mar 1996 16:45:40 GMT From: d.j.walker.slh0101@oasis.icl.co.uk Date: Mon, 11 Mar 1996 16:48:06 GMT Message-Id: <199603111648.28924.0@getafix.oasis.icl.co.uk> Received: by getafix.oasis.icl.co.uk id QAA28924; Mon, 11 Mar 1996 16:48:06 GMT Reply-To: d.j.walker.slh0101@oasis.icl.co.uk Subject: 420dREAD.ME To: rdzidlic@cip.informatik.uni-erlangen.de Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="PART.BOUNDARY.1041.28884.unix.826562879.1" Status: RO --PART.BOUNDARY.1041.28884.unix.826562879.1 Content-ID: <1041_28884_826562879_1@unix> Content-type: application/octet-stream; name="16474201.1.513_pt" Content-Description: External file C68 EXPERIMENTAL RELEASE 2 - 10th March 1996 ============================================ There have been major internal changes to the C68 program that is part of the "C68 for QDOS" release. In addition there are new versions of many of the assembler support routines that are called by the generated code. This new version has passed its initial validation. However due to the widespread nature of the changes I felt that it was advisable to make this experimental version more widely available so that users could test it on their own source before it becomes the mainstream release. The advantage of this release is that it should generate more efficient code than the earlier releases. This will be particularly noticeable when generating code that uses floating point arithmetic, or uses a lot of 'short' data types. This is the second release of this experimental version of C68. The first release went reasonably well, but there were enough errors found (particularily for programs that used floating point) that I felt it was advisable to put out this second release of the experimental version. This second release fixes the problems reported on the first experimental release. It also restores some code optimisations that had got lost in re-working the compiler, so those who reported a slow-down on some programs with the first experimental release should now find that (hopefully) the speed is now the same or better than the 4.20a and 4.20b releases of C68. However note that it is important to use the -O option to get maximum speed (as well as minimum size). The files for this release have been split into two, and each set is designed to fit onto a single 720Kb disk. The RUN1 files are all the files that are needed to compile (but not link) a program. The RUN2 files are those files needed for linking. This split has been made to accomodate those who only have 720Kb disk drives. Those who have HD or ED disks are advised to put all these files onto a single disk. If you only have 720Kb disk drives, then please provide feedback on whether you find this new split of files an improvement or not. All programs and libraries supplied have been re-compiled with this release of the compiler even when there is no change to the source. Please, therefore, provide feedback if you have any programs that either do not run or which produce unexpected behaviour when using this release. A new feature at this release is the inclusion of the GWASS (George Gwilt's Assembler) assembler. The special feature of this release is support for the extra addressing modes of the 68020, 68030, 68040 and 68060 processors as well as for the FPU related isntructions of the 68040 and 68060. This assembler needs a 68020 to run, however, so only those with suitable hardware can invoke it (CC will give an error message if you try to run it on unsutiable hardware). It is invoked by including the -gwass option to the CC program. This assembler has only just had the facility to produce S-ROFF output added to it and has had limited testing to date, so it is quite likely that there may be problems in using it. Please pass back any problem reports. Users who have hardware Floating Point Units can also try out the new -fpu runtime option to CC. This will cause the C68 phase to generate inline hardware FPU instructions, and will also invoke the GWASS assembler (there is no need to include the explicit -gwass option in this case) to assemble the resulting code. You will find that this gives a significant performance boost over all earlier releases for such code (at the expense of loosing portability of the binary versions of programs to systems that do not have the FPU hardware fitted). Where this portability matters, then the previous method of detecting FPU hardware within the C68 library support routines will still work, but will not provide as dramatic a speed increase. As before, however, please try and initially use limited programs in such testing as the GWASS assembler may not yet be up to handling larger programs. Please keep me informed with feedback on this release. Dave Walker 10 Mar 1996 --PART.BOUNDARY.1041.28884.unix.826562879.1--