20 Feb
2015
20 Feb
'15
4:02 p.m.
-----Original Message-----
From: cctalk [mailto:cctalk-bounces at classiccmp.org] On Behalf Of Chuck
Guzis
Sent: 20 February 2015 23:00
To: General at classiccmp.org; Discussion at classiccmp.org:On-Topic and Off-
Topic Posts
Subject: Re: Pascal not considered harmful - was Re: Rich kids are into
COBOL
On 02/20/2015 02:32 PM, Ali wrote:
Almost all , so there exceptions, modern compliers are written in a high
level language. Most use LEXX and YACC or their modern equivalents to
generate the program fragments needed to generate the code.
Excuse my ignorance but then how is a compiler
written? I always
thought that you would need to know op codes and assembler to write a
compiler.
I.E.
you can't write and efficient compiler by
using a high level language.
You have a front-end, which does the lexical work and
translates the
program into some intermediate form, such as a tree. In the middle, you
have optimization and checking--and finally, you have the back end which
essentially emits code--but that doesn't necessarily imply that the
compiler
author knows the numeric opcodes or the precise
instruction format. Many
compiler backends feed into an existing assembler, which puts it all
together.
Knowing the numeric opcodes and instruction format isn't all that it would
seem to be. I never had much more than a passing familiarity with the
numeric opcodes of the CDC STAR--given that there was 8 bits for the
opcode and 8 more "modifier" bits, you arguably had a machine with
thousands of opcodes. What was hardest and very important was
committing to memory the *timings* of those instructions, within a
superscalar, segmented, pipelined vector architecture.
--Chuck