Re: Denial of denial of service

[prev] [thread] [next] [lurker] [Date index for 2007/01/29]

From: Yossi Kreinin
Subject: Re: Denial of denial of service
Date: 14:22 on 29 Jan 2007
Andy Armstrong wrote:
> 
> I seem to remember that the Norcroft ARM C compiler did quite an  
> impressive job of register colouring and generally working out which  
> variables were worth keeping in registers. Certainly better than GCC  at 
> the time (about 10 years ago).

2 years ago, both the ARM and the Green Hills C++ compilers didn't do so well. 
The Green Hills compiler deals very well with MIPS or PowerPC, and it has a lot 
of ARM-specific hacks (like loading 4 coordinates of a rectangle structure with 
LDM to compute it's area, and emitting a lot of non-trivial shifter operands and 
conditioning).

> 
>> In fact I don't know if it's the compilers or the questionable  choice 
>> to save registers and instead add optional bells and  whistles to each 
>> and every instruction.
> 
> 
> Questionable my arse! That's what made it so great to program :)
> 

Well, I believe that if someone had enough fun with something to mention one's 
rear end in an argument about it, that something has probably got at least some 
important things right :)

I probably agree that ARM is the best assembly target among the RISC machines 
because of the design of it's ISA. I only claimed that it's a worse C target 
than other RISC machines. And of course I can't really prove that either - 
that's just my impression based on the work with certain compilers.
There's stuff above here

Generated at 23:01 on 06 Feb 2007 by mariachi 0.52