Re: Implementations of VLIW
- From: Eric Smith <eric@xxxxxxxxxxxx>
- Date: Tue, 08 Apr 2008 14:52:12 -0700
Ripunjay Tripathi wrote:
Different Implementations of the same VLIW architecture may not be
binary-compatible with each other.
I am looking for explaination on the above line.
Sounds like a homework problem to me. Did you read the textbook?
Anyhow, if the implementations aren't binary-compatible with each other,
they aren't the same architecture, by definition. Computer architecture
was defined by Brooks and Blaauw to be the characteristics of the
computer that are visible to the programmer, so two implementations of
the same architecture of necessity must be binary compatible.
The definition is not stictly adhered to in practice. For instance,
people talk about the x86 architecture in general, but not all x86
implementations have SSE3 instructions, as one example. Also, the
definition sometimes only refers to user-mode programs, so two
implementations of the DEC PDP-10 architecure, the KA10 and KI10 processors,
require some different code in the OS kernel.
Despite those laxities in the definition of architecture, I would still
expect two implementations of the same VLIW architecture to be binary
compatible.
A different VLIW implementation might have more (or fewer) functional
units, and thus a longer (or shorter) instruction word, but then it cannot
be said to be the *same* architecture.
.
- Follow-Ups:
- Re: Implementations of VLIW
- From: Ripunjay Tripathi
- Re: Implementations of VLIW
- References:
- Implementations of VLIW
- From: Ripunjay Tripathi
- Implementations of VLIW
- Prev by Date: Partial panel assembly
- Next by Date: DarknightX
- Previous by thread: Implementations of VLIW
- Next by thread: Re: Implementations of VLIW
- Index(es):
Relevant Pages
|