This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: GCC inline parameters (PR 10160 testcase)


> Hmm; I didn't know we'd changed our inlining heuristics to have this
> MIN_INLINE_INSNS concept.

I must say that I didn't know either until recently. I guess the new
heuristics was devised on powerful x86 boxes. The next step is probably to
fine-tune the MOORE_LAW_CONSTANT that we will need to add to 3.4 in order
not to outpace the aforementioned law in our hardware requirements :-)

> You can, however, set that to zero using --params.

Sure. We could also have done it the other way around: set it to zero by
default for safety concerns and document that you can change it, provided
that you do want to pay the price.

> But, I think the scheduler is your real problem here; there shouldn't be
> n^2 algorithms in there, unless they have clamps.

Apparently the scheduler is not the only problem, see Dave's last message.
Anyway, the regression comes from the tree inliner, not from anywhere else.

- Eric


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]