rachidc@syma.sussex.ac.uk (Rachid Chalabi) (02/26/90)
I have managed to compile gcc-1.37.1 on our SEQUENT symmetry without any problems. I am quite happy with it as the code generated by the compiler is quite compact. The only minor problem is that currently the distribution available supports only the INTEL 80387 co-processor as a target. Has anyone added the support for the WEITEK 1167 co-processor and if yes could he make it available? Could you please reply to the net or directly to myself. Many thanks. Rachid Chalabi, Physics Department, University of Sussex, Brighton, UK JANET:rachidc@uk.ac.sussex.syma INTERNET: rachidc%syma.sussex.ac.uk@nsfnet-relay.ac.uk BITNET: rachidc@syma.sussex.ac.uk UUCP: rachidc@syma.uucp -- Rachid Chalabi, Physics Department, University of Sussex, Brighton, UK JANET:rachidc@uk.ac.sussex.syma INTERNET: rachidc%syma.sussex.ac.uk@nsfnet-relay.ac.uk BITNET: rachidc@syma.sussex.ac.uk UUCP: rachidc@syma.uucp
james@bigtex.cactus.org (James Van Artsdalen) (02/28/90)
In <2268@syma.sussex.ac.uk>, rachidc@syma.UUCP (Rachid Chalabi) wrote: > I am quite happy with it as the code generated by the compiler is > quite compact. The only minor problem is that currently the > distribution available supports only the INTEL 80387 co-processor as a > target. Has anyone added the support for the WEITEK 1167 co-processor > and if yes could he make it available? I have been working on this for the last few days. gcc does not yet bootstrap itself with Weitek support, but sometime this week it should I think. Send mail if you would like to help test & debug. There are two major problems with the port. First, gcc requires that one be able to move from memory to any register (register reloads). The Weitek can't do this. gcc needs to be changed so that it can reload through one of the 386 registers. At the moment, each move to and from the Weitek & memory generates a very slow code sequence. Secondly, where should functions return FP values? AT&T returns them in the 387. My natural inclination is to avoid that and return them in the Weitek, but the problem arises of how to handle system libraries. At the moment gcc knows of both the 387 and the Weitek, and receives and returns floats through the 387, but does all other arithmetic through the Weitek when it is enabled. Maybe I need yet another switch, -mweitek-returns-in-weitek. -- James R. Van Artsdalen james@bigtex.cactus.org "Live Free or Die" Dell Computer Co 9505 Arboretum Blvd Austin TX 78759 512-338-8789