This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[tree-ssa] Do alias analysis after SSA. Improvements to PR8361
On Wed, 7 Jan 2004, Diego Novillo wrote:
> Although the patch adds one additional call to DOM and another to DCE,
> compile times and memory usage are down. For instance, on
> generate-3.4.ii (PR 8361), we are now 8% faster (55s down from 60s) and
> allocate 13% less memory (236Mb down from 270Mb). We also generate
> smaller code (about 1% smaller, mostly because the additional DOM/DCE
> run I suspect).
>
> On the collection of .i/.ii files from a typical bootstrap, the tree
> optimizers are about 15% faster (which translates to a 2% improvement
> overall). The passes where we are now significantly faster are:
> [...]
On Wed, 7 Jan 2004, Diego Novillo wrote:
> I should've used 'top', not -fmem-report. The total memory consumption
> is down to 289Mb from 308Mb (6%).
Still, also 6% is quite a difference, and 8% runtime really is
impressive!
(Much of the runtime improvement might be due to less swapping
and iowait going on, unless your machine has a lot of memory.)
Given that you also report a 1% reduction in the size of the
generated binary, I really will try to perform a round of benchmarks
this weekend, comparing 2.95, 3.3.2 and mainline against tree-ssa.
Gerald