Tree-ssa status with POOMA

[Richard Guenther]

On Thu, 5 Feb 2004, Richard Guenther wrote:

> Hi!
>
> I promised to give some updates on tree-ssa and POOMA.  Tree-ssa is
> currently failing three regression tests that 3.4 succeeds in.  One
> is a clear miscompilation resulting in wrong results, two look like the
> same bug causing a SIGSEGV.  As usual the testcases would be incredibly
> large and hard to debug, so this is mainly a heads-up (though 3.3.3 won't
> be any better here).
>
> For my real-world application I'm quite happy with tree-ssa:
>
> -O2 -funroll-loops -ffast-math -march=athlon -fno-exceptions
>
> 		memory	compile-time	run-time	stripped static size
> tree-ssa	450MB	3m57s		46s		1729680
> gcc3.4		530MB	3m16s		44s		1734596

With -ftree-points-to=andersen I get:

		890MB	5m14s	41s		1725584

a nice 10% speedup (but that memory usage!).

Richard.

--
Richard Guenther <richard dot guenther at uni-tuebingen dot de>
WWW: http://www.tat.physik.uni-tuebingen.de/~rguenth/