Mailinglist Archive: radeonhd (364 mails)
| < Previous | Next > |
[radeonhd] [Bug 20593] radeonhd driver generates unaligned traps on Alpha
- From: bugzilla-daemon@xxxxxxxxxxxxxxx
- Date: Sat, 14 Mar 2009 18:43:36 -0700 (PDT)
- Message-id: <20090315014336.5CA1A130057@xxxxxxxxxxxxxxxxxxxxxxxx>
http://bugs.freedesktop.org/show_bug.cgi?id=20593
--- Comment #6 from Michael Cree <mcree@xxxxxxxxxxxx> 2009-03-14 18:43:35 PST
---
(In reply to comment #5)
Yeah, that is what I was expecting. Coupled with your statement following, it
explains the lack of interest in these issues.
x86 does have alignment issues. I am lead to believe that accesses to a
misaligned datum takes extra CPU clock cycles if it requires more than one
memory access. In time critical code that may be important. Admittedly that's
not an issue for us.
Are people not wanting to run radeonhd on PPC? Isn't that big endian?
Isn't Xorg 'sposed to be a multiplatform project? If so, then porting access
to the AtomBIOS, at some point in the development cycle, will become an issue.
I'm happy to accept that it cannot be priority at the moment, when there are
other much more important issues to focus on, nevertheless I think it is good
to list these issues at this stage, and get a feel for how many people are
impacted by them. That will determine implementation focus at a later stage.
Michael.
--
Configure bugmail: http://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
_______________________________________________
xorg-team mailing list
xorg-team@xxxxxxxxxxx
http://lists.x.org/mailman/listinfo/xorg-team
--
To unsubscribe, e-mail: radeonhd+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: radeonhd+help@xxxxxxxxxxxx
--- Comment #6 from Michael Cree <mcree@xxxxxxxxxxxx> 2009-03-14 18:43:35 PST
---
(In reply to comment #5)
The AtomBIOS data is not aligned at all. 16 and 32 bit words can appear
on any byte boundary and they do.
Yeah, that is what I was expecting. Coupled with your statement following, it
explains the lack of interest in these issues.
AtomBios was written for x86
architecture which is of course little endian and has no alignment
issues.
x86 does have alignment issues. I am lead to believe that accesses to a
misaligned datum takes extra CPU clock cycles if it requires more than one
memory access. In time critical code that may be important. Admittedly that's
not an issue for us.
I started wrapping all these accesses in a macro that will do the endian
byte swapping for me. It also solves the alignment problem at the same
time since to do the byte swapping it is convenient to read the AtomBIOS
data out one byte at a time anyway.
All this work was done just for my benefit so far and I haven't even
considered whether to return any of this to X.org and whether they would
even want it, since most people don't have a problem with it.
Are people not wanting to run radeonhd on PPC? Isn't that big endian?
Isn't Xorg 'sposed to be a multiplatform project? If so, then porting access
to the AtomBIOS, at some point in the development cycle, will become an issue.
I'm happy to accept that it cannot be priority at the moment, when there are
other much more important issues to focus on, nevertheless I think it is good
to list these issues at this stage, and get a feel for how many people are
impacted by them. That will determine implementation focus at a later stage.
Michael.
--
Configure bugmail: http://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
_______________________________________________
xorg-team mailing list
xorg-team@xxxxxxxxxxx
http://lists.x.org/mailman/listinfo/xorg-team
--
To unsubscribe, e-mail: radeonhd+unsubscribe@xxxxxxxxxxxx
For additional commands, e-mail: radeonhd+help@xxxxxxxxxxxx
| < Previous | Next > |