[M3commit] CVS Update: cm3

[Tony Hosking]

Cool.
Thanks for putting the time in on this.


On 4 Jun 2010, at 13:16, Jay K wrote:

> 
> Unclear. I've also worked very hard at this and I believe it is in better shape than before.
> Granted, some steps forward, some backward. But I think much more forward.
> 
> I'm using -O3.
> Sometimes what I do is in parse.c I explicitly enable everything but setting all the flags.
>  We should perhaps do that anyway, esp. for optimizations that don't take long and don't hurt debugging.
> 
> 
> I suspect these two bugs were present back then, you just didn't hit them because:
>  1) we didn't have the unused nested function? 
>  2) we didn't have cvsup? 
> 
> 
> I disabled very very little.
> In trade, we don't throw around volatile all over the place.
> 
> 
> I think if you look into it, things were broken back then just the same.
> But we didn't have the code to tickle the bugs.
> 
> 
> I would like to get back "unit at a time". That seems a good optimization wrt inlining and it is being disabled for something that should be easily fixable some other way.
> 
> 
> vrp/fre/pre I don't know.
> It'll take quite some debugging to fix those.
> vrp produces bad code.
> fre and/or pre crash in the compiler.
> 
> 
> I'd also like to fix the volatile float problem. I've worked on that quite a bit too.
> I tried to make loophole always go through a volatile temporary for example.
> I tried making loophole go through a union.
> No luck yet.
> I guess I should see what:
>  int a; 
>  float j;   
>  a = *(int*)&j;  
> 
> 
> does in C.
> 
> 
> Notice that some of this is in the release branch. Where all I've done is test more code.
> I didn't remove the volatility on all load/store for example.
> 
> 
> - Jay
> 
> 
> ________________________________
>> From: hosking at cs.purdue.edu
>> Date: Fri, 4 Jun 2010 12:36:24 -0400
>> To: jkrell at elego.de
>> CC: m3commit at elegosoft.com
>> Subject: Re: [M3commit] CVS Update: cm3
>> 
>> 
>> 
>> Are you planning to go through and fix things so we can re-enable optimisations?
>> 
>> I worked pretty hard to get everything to compile with -O2/-O3 when I worked on this with the current m3cg.
>> 
>> 
>> Antony Hosking | Associate Professor | Computer Science | Purdue University
>> 305 N. University Street | West Lafayette | IN 47907 | USA
>> Office +1 765 494 6001 | Mobile +1 765 427 5484
>> 
>> 
>> 
>> 
>> 
>> 
>> On 4 Jun 2010, at 18:15, Jay Krell wrote:
>> 
>> CVSROOT: /usr/cvs
>> Changes by: jkrell at birch. 10/06/04 18:15:09
>> 
>> Modified files:
>> cm3/m3-sys/m3cc/gcc/gcc/m3cg/: Tag: release_branch_cm3_5_8
>> parse.c
>> 
>> Log message:
>> turn off "pre" optimization, which crashes compiling
>> m3-tools/cvsup/server/FSServer.m3
>>