I see, we would be using the presence of this parameter to drive lipo
stress-testing in general. I made the change and added a comment to
that effect and will commit it.
Thanks,
Teresa
On Thu, Oct 10, 2013 at 12:03 PM, Xinliang David Li wrote:
> The option/parameter in theory should not have any
The option/parameter in theory should not have any effect to
profile-use build, but it is usually passed to both profile-gen and
profile-use compilation.
The reason to turn it on is to surface the type of error you have
fixed recently (it is triggered only with trivial module group).
David
On T
On Thu, Oct 10, 2013 at 11:42 AM, Xinliang David Li wrote:
> It is better to turn it on when random stress test is also on: param
> value of PARAM_LIPO_RANDOM_GROUP_SIZE is not the default value (0).
I don't understand why - the random group size is used at dyn-ipa time
to compute module groups,
It is better to turn it on when random stress test is also on: param
value of PARAM_LIPO_RANDOM_GROUP_SIZE is not the default value (0).
if (!L_IPO_COMP_MODE || (num_in_fnames == 1 && PARAM_VALUE
(PARAM_LIPO_RANDOM_GROUP_SIZE) == 0))
ok with this change.
David
On Thu, Oct 10, 2013 at 11:27 AM,
The following patch fixes an inefficiency whereby type unification
was being attempted unnecessarily in LIPO even when there were no
aux modules.
Tested with regression tests and internal LIPO benchmark.
Ok for google/4_8?
Thanks,
Teresa
2013-10-10 Teresa Johnson
* l-ipo.c (cgraph_u