[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [avr-gcc-list] Improved loop patch

From: HutchinsonAndy
Subject: Re: [avr-gcc-list] Improved loop patch
Date: Tue, 01 Mar 2005 07:08:45 -0500

Marek Michalkiewicz <address@hidden> wrote:

>On Mon, Feb 28, 2005 at 06:39:17PM -0500, address@hidden wrote:
>> The loop change allows the SBRx instrcution to be use to skip over 1,2 or 3 
>> word instructions -previously it would only skip 1 word instructions.  So 
>> you will less SBRx Rx,n; RJMP nnnn combinations.
>Thanks for your work - I've just looked at it a little.  A few comments:
>1. Correct me if I am wrong, but what are these 3-word AVR instructions? :)

Ok you got me!
..but if there were 3 word instructions it would work too :-p

>2. Yes, 2-word instructions actually exist (CALL, JMP, LDS, STS), but you
>   have to be careful with them - some old (non-enhanced core) AVR devices
>   have an errata where skipping a 2-word opcode may not work correctly
>   (avr-as gives a warning when it detects such code).

I'll check on this errata - I think its out of date and can be handled at 
device level.

>3. Instead of checking for specific RTL patterns in avr.c (might be error
>   prone), I'd suggest to define a new insn attribute (say, "skip_ok"),
>   and add it to the few insns in avr.md which may be skipped.  Then, the
>   insn may be skipped if length == 1, or if the attribute is true and
>   the device is not affected by the errata mentioned above.

Yes I considered that. But as much of the code is produce by macros/functions 
it doesn't really help. 

>4. Please submit each logical change in a separate patch, with its own
>   ChangeLog entry.  While PR18251 is a regression and certainly should
>   be fixed, non-trivial improvements (which are not bug fixes) will
>   probably have to wait until after 4.0 is released.

Which I did. Roger Sayle was the only respondent and has offered to submit on 
my behalf. Unfortunately Roger has been busy.

For submitting here I combined the changes with "pr18251" on the basis of 
acceptance and the fact I have moved my baseline up. We dont need the log here 
and of course I'll separate for gcc.

>5. Have you done the FSF paperwork (copyright assignment or disclaimer)?
>   MODES_TIEABLE_P was a small change, big improvement but not significant
>   for copyrights.  Non-trivial changes like movmem* may be problematic
>   without the paperwork.  The FSF is very paranoid about these issues
>   (and the recent SCO problems show they may be right after all...).

I emailed maintainers and "assignments" to get the forms. As nothing hasppened 
(ahem!), Roger was kind enough to  give me the email form. I had an email 
yesterday saying the forms were on there way.

I think I will have to practice the special handshake a few more times and all 
will be well.

>AVR-GCC-list mailing list

Andy Hutchinson

Switch to Netscape Internet Service.
As low as $9.95 a month -- Sign up today at http://isp.netscape.com/register

Netscape. Just the Net You Need.

New! Netscape Toolbar for Internet Explorer
Search from anywhere on the Web and block those annoying pop-ups.
Download now at http://channels.netscape.com/ns/search/install.jsp

reply via email to

[Prev in Thread] Current Thread [Next in Thread]