[m-rev.] for review: thread-local allocation

[Peter Wang]

On 2006-04-20, Julien Fischer <juliensf at cs.mu.OZ.AU> wrote:
> 
> On Thu, 20 Apr 2006, Peter Wang wrote:
> 
> > Estimated hours taken: 4
> > Branches: main
> 
> Should this go on the release branch?

I don't know.  It probably requires more testing, but it should help
hlc.par.gc.  But the parallel grades are not really supported yet,
right?

> > This patch enables the thread-local memory allocation feature of Boehm GC in
> > parallel grades on Linux hosts; without it, performance is terrible.
> 
> Can you quantify the improvement?

A naive quicksort program, sorting 1000000 ints, on saturn.

asm_fast.gc
./qsort  3.28s user 0.19s system 99% cpu 3.476 total

asm_fast.par.gc (before, one thread)
./qsort  6.17s user 0.17s system 99% cpu 6.345 total

asm_fast.par.gc (before, two threads)
MERCURY_OPTIONS=-P2 ./qsort  12.47s user 0.29s system 136% cpu 9.382 total

asm_fast.par.gc (after, one thread)
./qsort  3.64s user 0.17s system 99% cpu 3.815 total

asm_fast.par.gc (after, two threads)
MERCURY_OPTIONS=-P2 ./qsort  3.71s user 0.23s system 123% cpu 3.193 total

Unfortunately that's basically as good as it gets.

Peter

--------------------------------------------------------------------------
mercury-reviews mailing list
post:  mercury-reviews at cs.mu.oz.au
administrative address: owner-mercury-reviews at cs.mu.oz.au
unsubscribe: Address: mercury-reviews-request at cs.mu.oz.au Message: unsubscribe
subscribe:   Address: mercury-reviews-request at cs.mu.oz.au Message: subscribe
--------------------------------------------------------------------------