home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!sunic!dkuug!diku!torbenm
- From: torbenm@diku.dk (Torben AEgidius Mogensen)
- Newsgroups: comp.arch
- Subject: Re: DEC Alpha architecture issues
- Message-ID: <1992Nov20.180157.8811@odin.diku.dk>
- Date: 20 Nov 92 18:01:57 GMT
- References: <1992Nov18.112407.2518@doug.cae.wisc.edu> <1992Nov18.191730.1044@meiko.com> <lgnojjINN627@exodus.Eng.Sun.COM> <MOSS.92Nov20101727@CRAFTY.cs.cmu.edu>
- Sender: torbenm@freke.diku.dk
- Organization: Department of Computer Science, U of Copenhagen
- Lines: 19
-
- moss@cs.cmu.edu (Eliot Moss) writes:
-
- >In fact, conditional
- >instructions are easy to handle in a regular pipeline -- you just suppress the
- >store in the last step, and you certainly know the outcome of the comparison
- >by then. In a superscalar, where things get reordered, it is admittedly not
- >quite so simple.
-
- But neither are conditional branches. I would like to see a
- superscalar architecture that handles conditional branches by a
- mechanism that could not also be used (or simplifyed) for conditional
- moves or arithmetic. If need a conditional instruction on an ARM or
- Alpha, you certainly will need a conditional branch on "normal" RISCs.
- Looking for conditional branches in the pipeline is no easier than
- looking for conditional instructions in general, especially on the
- ARM, where the first 4 bits determine (uniformly) if the instruction
- is conditional.
-
- Torben Mogensen (torbenm@diku.dk)
-