Re: Register Files

Ali-Reza Adl-Tabatabai <ali@LARK.WARP.CS.CMU.EDU>
Thu, 22 Oct 1992 14:29:29 GMT

          From comp.compilers

Related articles
[2 earlier articles]
Re: Register Files ssimmons@convex.com (1992-10-06)
Re: Register Files cliffc@cs.rice.edu (1992-10-06)
Re: Register Files hobbs@steven.enet.dec.com (1992-10-06)
Re: Register Files kendall@centerline.com (1992-10-07)
Re: Register Files bart@cs.uoregon.edu (1992-10-07)
Re: Register Files henry@zoo.toronto.edu (1992-10-07)
Re: Register Files ali@LARK.WARP.CS.CMU.EDU (Ali-Reza Adl-Tabatabai) (1992-10-22)
Re: Register Files idacrd!desj@uunet.UU.NET (1992-10-23)
| List of all articles for this month |

Newsgroups: comp.compilers
From: Ali-Reza Adl-Tabatabai <ali@LARK.WARP.CS.CMU.EDU>
Organization: Compilers Central
Date: Thu, 22 Oct 1992 14:29:29 GMT
References: 92-10-019
Keywords: registers, architecture

Having split register files also makes superscalar dispatch simpler. The
dispatch stage need only check if the instructions are destined for
different FU classes (integer, FP, etc...), without checking for register
dependences (except for load and moves between register files). For
example in the IBM RS6000, there is the FP, integer and condition code
registers. Instructions in the instruction window that are destined for
different functional units can be issued in parallel (e.g. a decoupled
architecture).


Ali
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.