Add 32-bit version of llvm-allocated-type.

Add 32-bit version of llvm-allocated-type.


BUG=162816

[Dai Mikurube (NOT FULLTIME), 2012-11-27 10:53:51]

Hi Nico,

Could you rubber-stamp?  I'm trying to profile memory usage of Linux 32-bit
Chromium (to compare 32-bit v.s. 64-bit).

[Dai Mikurube (NOT FULLTIME), 2012-11-27 11:05:34]

Also, if you have any ideas about http://crbug.com/162818, could you tell me
your ideas?

[Nico, 2012-11-27 16:11:27]

> Could you rubber-stamp?  I'm trying to profile memory usage of Linux 32-bit
> Chromium (to compare 32-bit v.s. 64-bit).

Do you need to check in a binary for this? I'd prefer not to add 32bit binaries,
as it increases the configuration matrix. Neither depot_tool's ninja nor
chromium's clang are available in 32bit, and neither has been an issue so far.
So I'd prefer if it was possible to measure this without checking in binaries.

> Also, if you have any ideas about http://crbug.com/162818, could you tell me
> your ideas?

I commented on the bug.

[Dai Mikurube (NOT FULLTIME), 2012-11-28 03:52:48]

On 2012/11/27 16:11:27, Nico wrote:
> > Could you rubber-stamp?  I'm trying to profile memory usage of Linux 32-bit
> > Chromium (to compare 32-bit v.s. 64-bit).
> 
> Do you need to check in a binary for this? I'd prefer not to add 32bit
binaries,
> as it increases the configuration matrix. Neither depot_tool's ninja nor
> chromium's clang are available in 32bit, and neither has been an issue so far.
> So I'd prefer if it was possible to measure this without checking in binaries.
> 
Makes sense.  I thought to compare them in buildbots and dashboards, but it's
not mandatory for now.  If we think continuous testing is required (in future),
I may reopen it just for testing without tree closing.

(btw, llvm-allocated-type 32-bit can be enabled with the following change:
https://codereview.chromium.org/11280181/)

> > Also, if you have any ideas about http://crbug.com/162818, could you tell me
> > your ideas?
> 
> I commented on the bug.

Thanks.