Re: avr-gcc making really bad inlining decisions.

Joern Rennecke <[email protected]>
Newsgroups gmane.comp.hardware.avr.gcc
Message-ID <CAMqJFCpMCUo5N6JHhPyEYRZXe9=Je4e37NZxc6+wscXW=L5R-A@mail.gmail.com>
On 9 April 2015 at 01:15, Bill Westfield <[email protected]> wrote:

> Is there a way to get gcc to dump the settings of all the parameters
> and option settings after it's done parsing the command line?

That would be -fverbose-asm combined with -S or --save-temps;
this dumps the settings in the assembly output.

And on the matter of inlining decisions, yes, it is done on the tree level.
Inlining used to be done on the rtl level, which offered an easy way
to get realistic
costings of the expansion, but had other drawbacks, e.g. it made it hard to
propagate function parameters effectively into the inlined function,
particularily
as other optimizers were shifted from rtl to tree as well.
The way the inlining costs are calculated reflect the ports that the
developers who coded and/or tested the tree passes cared about.
As the saying goes, he who pays the piper calls the tune.
Thus, properly tuning the port might also require contributing changes
to the optimizers to make their cost modesl more general.
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.