Article View: comp.arch
Article #109Re: What's wrong with stack machines?
From: Josh Osborne
Date: Sat, 31 Oct 1992 15:29
Date: Sat, 31 Oct 1992 15:29
41 lines
2657 bytes
2657 bytes
In article <1729@idacrd.UUCP> desj@ccr-p.ida.org (David desJardins) writes:
>Jan Vorbrueggen <jan@pallas.neuroinformatik.ruhr-uni-bochum.de> writes:
>>So this loop requires 42 ops, ie, on the Alpha 164 bytes.
>>[...]
>>[On the transputer] The whole loop thus requires 91 ops and 131 bytes.
>
>Who cares how many bytes it is?
The cache, the external bus, and the disk.
> It seems like the obvious tendency for
>microprocessors is to go to longer and longer instructions, because with
>modern technologies the overhead of fetching and decoding the
>instructions is rapidly disappearing. In 1960 or so it might have made
>a big difference whether you could pack the instructions into one byte
>instead of four, but I don't see the argument for it today.
With everything else being equal code density is a big win (for example
dubling the code density is the same as dubling the I-cache size, and
the bus width during I-fetches, as long as everything else remains the
same). However everything else does not remin the same. (oh, yeah
it would halve TLB misses due to I-fetch, and it would...)
>Obviously if the length of the instructions were important, the Alpha
>could have been designed with variable-length instructions. So the
>evidence is that the cost of 4-byte instructions instead of 1-byte
>instructions is quite small, less than the extra complexity of dealing
>with variable-length and unaligned instructions.
No, it's if the Alpha designers had thought the code density was more
important then a simple fast decoder the Alpha would have variable
length instructions. I am sure the Alpha designers know that code
density is important.
> David desJardins
--
stripes@pix.com "Security for Unix is like
Josh_Osborne@Real_World,The Multitasking for MS-DOS"
"The dyslexic porgramer" - Kevin Lockwood
We all agree on the necessity of compromise. We just can't agree on
when it's necessary to compromise. - Larry Wall
Technical Details
Message-ID:
<BwzsD0.KEM@pix.com>Path:
arc1-ca.pugleaf.net!sparky!uunet!wupost!darwin.sura.net!mojo.eng.umd.edu!pandora.pix.com!stripesReferences:
<Bwx4pF.8ur@rice.edu> <JAN.92Oct30175815@pallas.neuroinformatik.ruhr-uni-bochum.de> <1729@idacrd.UUCP>Show Full Headers
Complete Article Headers
Newsgroups: comp.arch Path: arc1-ca.pugleaf.net!sparky!uunet!wupost!darwin.sura.net!mojo.eng.umd.edu!pandora.pix.com!stripes From: stripes@pix.com (Josh Osborne) Subject: Re: What's wrong with stack machines? Message-ID: <BwzsD0.KEM@pix.com> Sender: news@pix.com (The News Subsystem) Nntp-Posting-Host: pandora.pix.com Organization: Pix Technologies -- The company with no adult supervision References: <Bwx4pF.8ur@rice.edu> <JAN.92Oct30175815@pallas.neuroinformatik.ruhr-uni-bochum.de> <1729@idacrd.UUCP> Date: Sat, 31 Oct 1992 15:29:23 GMT Lines: 41