Request for review: 7022370 Launcher ergonomics doesn't need per-architecture implementations

[Mandy Chung]

  On 3/7/11 10:10 PM, David Holmes wrote:
> There was a glitch with the use of CTARGDIR (which isn't defined), so 
> now we check for the ergo file in LAUNCHER_PLATFORM_SRC. Webrev has 
> been updated - same location.
>

Looks good.  Sorry I didn't catch that CTARGDIR isn't defined as it 
exists in the existing makefile.

> Verified with i586 and non-i586 builds.
>

Thumbs up.
Mandy

> Thanks,
> David
>
> Gary Benson said the following on 03/07/11 19:57:
>> Hi David,
>>
>> I'm good with these changes, and I'm happy for ergo_zero.c to vanish.
>>
>> Thanks,
>> Gary
>>
>> David Holmes wrote:
>>> Hopefully all interested parties are addressed in the cc lists.
>>>
>>> webrev at:
>>>
>>> http://cr.openjdk.java.net/~dholmes/7022370/webrev/
>>>
>>> The launcher ergonomics (ergo.c) currently relies on
>>> per-architecture, eg ergo_sparc.c, ergo_i586.c, files to define the
>>> actual ergonomics operations. Only x86 is actually CPU specific,
>>> both sparc and zero share the same platform independent
>>> implementation. It will simplify things if we provide a platform
>>> independent default in ergo.c that is conditionally compiled, and
>>> modify the build system to cause that compilation if a platform
>>> specific ergo file is not found.
>>>
>>> We can potentially delete all the ergo_*.c files except for
>>> ergo_i586.c, and we no longer require that there be a
>>> per-architecture file, which makes additional porting easier.
>>>
>>> Gary: do you mind seeing ergo_zero.c go away, or would you prefer to
>>> leave it in case someone is doing a local customization? I suppose
>>> the some consideration could be given to ergo_sparc.c too. Is
>>> anybody aware of downstream distros that modify these files to
>>> change the default ergonomics policies?
>>>
>>> Thanks,
>>> David Holmes
>>