This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [boehm-gc] Define HBLKSIZE to 4096 for MIPS
- From: Richard Sandiford <rsandifo at nildram dot co dot uk>
- To: David Daney <ddaney at avtrex dot com>
- Cc: gcc-patches at gcc dot gnu dot org, hboehm at gcc dot gnu dot org
- Date: Sat, 26 Jan 2008 09:16:24 +0000
- Subject: Re: [boehm-gc] Define HBLKSIZE to 4096 for MIPS
- References: <87sl1dyi8i.fsf@firetop.home> <479A6483.7020304@avtrex.com>
David Daney <ddaney@avtrex.com> writes:
> Richard Sandiford wrote:
>> 4096 is the default (and smallest) page size for both 32-bit and 64-bit
>> MIPS Linux, so I'd like to use it as the default HBLKSIZE for both.
>> The definition in patch below overrides the target-independent LP64
>> default of 8k and copies similar definitions for other 64-bit Linux
>> targets.
>>
>> Tested on mips64-linux-gnu, where it fixes all remaining n64 libjava
>> failures on QEMU. (The o32 and n32 results are clean before and after
>> the patch.) OK to install?
>>
>> Richard
>>
>>
>> boehm-gc/
>> * include/private/gcconfig.h (HBLKSIZE): Define for MIPS.
>>
> Richard,
> This was approved here:
> http://gcc.gnu.org/ml/java-patches/2008-q1/msg00013.html
Yeah, sorry. Because Hans didn't seem to think this should make any
difference for Java, I wanted to go back and have another look before
committing, to see if there was some deeper issue. TBH, I never did
get a full trace of the problem. I stopped once I realised that
HBLKSIZE was smaller than a page, that this wasn't supposed to be so,
and that "fixing" it made things work.
I should have said so before, sorry. I kept thinking "I'll be able
to have a look tomorrow" and tomorrow kept throwing up something else...
Just committing was always a fallback position if I didn't get time.
Anyway, I think I really will have time to look at it today.
Richard