PATCH: tune ARM's rtx_costs function
Richard Earnshaw
rearnsha@arm.com
Mon Sep 22 16:00:00 GMT 2003
> I am contributing the following patch on behalf of Jim Lemke. The
> patch tunes the behaviour of arm_rtx_costs to improve the selection of
> multiply instructions for XScale. Regression tested for xscale-elf.
>
This is two independent changes rolled into one patch. In future, if
changes are independent, please submit separate patches for them. It
makes review easier.
> Okay to commit?
> Ben
>
>
> 2003-09-13 Jim Lemke <jim@wasabisystems.com>
>
> * config/arm/arm.c (arm_rtx_costs): Corrections for XScale
> multiply instructions.
This needs to be reformatted according to GNU coding standards.
> * config/arm/arm.md (addsi3): Remove check of
> preserve_subexpressions_p() so that a new pseudo is used. This
> allows better pattern matching, improving code at -O.
>
I need evidence that this is the right thing to do for -O1. -O1 is a
compromise between generated code and compilation time, the test is a
heuristic to determine when generating additional pseudos might be
worthwile. To make a case for this change you need to show either that:
1) Compilation times do not increase noticeably for -O1 (or at all for
-O0) and
2) The resulting performance of executables warrants that increase in
compilation time.
Or:
3) Compilation time unconditionally decreases by not doing the check at
both -O0 and -O1.
Either way these need to be for realistic functions, not trivial test
cases.
R.
More information about the Gcc-patches
mailing list