home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!spool.mu.edu!agate!stanford.edu!enterpoop.mit.edu!eff!ssd.intel.com!ichips!ichips!glew
- From: glew@pdx007.intel.com (Andy Glew)
- Subject: Re: branch-and-link
- In-Reply-To: dswartz@sw.stratus.com's message of 27 Dec 92 17:54:11 GMT
- Message-ID: <GLEW.92Dec31162920@pdx007.intel.com>
- Sender: news@ichips.intel.com (News Account)
- Organization: Intel Corp., Hillsboro, Oregon
- References: <1hkqk3INNsns@transfer.stratus.com>
- Date: Fri, 1 Jan 1993 00:29:20 GMT
- Lines: 30
-
- Here's a question for the hardware types: is there any reason either of
- design complexity or performance to not have *all* branches (conditional
- and unconditional) save their origin somewhere (probably in a fast
- dedicated register)? I know various cpu's (the rs6000?) have done this
- in the past, but only as an option. If there is no negative impact to
- doing this, it would be a real win for debugging crashes where a program
- (application or kernel) branch off to nowhere and die - and the poor S/W
- type is left guessing...
-
- Make it an option, specific to a model.
-
- Some microarchitectures can do it easily.
-
- Other, quite reasonable, microarchitectures can't do it at all without
- slowing down (or requiring a PC to be attached to every instruction as
- it flows through the pipe).
-
- When you can do it, great! But don't slow down other implementations
- for a debug feature that is not in use most of the time.
-
- --
-
- Andy Glew, glew@ichips.intel.com
- Intel Corp., M/S JF1-19, 5200 NE Elam Young Pkwy,
- Hillsboro, Oregon 97124-6497
-
- This is a private posting; it does not indicate opinions or positions
- of Intel Corp.
-
- Intel Inside (tm)
-