> > >-----Original Message----- >From: Chris Wilson [mailto:chris@xxxxxxxxxxxxxxxxxx] >Sent: Tuesday, June 23, 2015 4:08 PM >To: Morton, Derek J >Cc: intel-gfx@xxxxxxxxxxxxxxxxxxxxx; Wood, Thomas >Subject: Re: [PATCH i-g-t] tests/gem_fence_thrash.c: Reduce memory usage > >On Tue, Jun 23, 2015 at 04:01:53PM +0100, Derek Morton wrote: >> On android platforms with 1Gb RAM gem_fence_thrash was failing with an >> out of memory error. >> This patch causes gem_close() to be called when a thread is finished >> with its handles rather than relying on the cleanup when the fd is >> closed. This greatly improves the memory footprint of the test >> allowing it to run on 1Mb systems. >> >> Also fixed a leak of the 'threads' variable. >> >> Signed-off-by: Derek Morton <derek.j.morton@xxxxxxxxx> >> --- >> tests/gem_fence_thrash.c | 33 ++++++++++++++++++++++----------- >> 1 file changed, 22 insertions(+), 11 deletions(-) >> >> diff --git a/tests/gem_fence_thrash.c b/tests/gem_fence_thrash.c index >> 6447e13..bfb2e6d 100644 >> --- a/tests/gem_fence_thrash.c >> +++ b/tests/gem_fence_thrash.c >> @@ -60,26 +60,25 @@ struct test { >> }; >> >> static void * >> -bo_create (int fd, int tiling) >> +bo_create (int fd, int tiling, uint32_t *handle) >> { >> void *ptr; >> - int handle; >> >> - handle = gem_create(fd, OBJECT_SIZE); >> + *handle = gem_create(fd, OBJECT_SIZE); >> >> /* dirty cpu caches a bit ... */ >> - ptr = gem_mmap__cpu(fd, handle, 0, OBJECT_SIZE, PROT_READ | PROT_WRITE); >> + ptr = gem_mmap__cpu(fd, *handle, 0, OBJECT_SIZE, PROT_READ | >> +PROT_WRITE); >> igt_assert(ptr); >> memset(ptr, 0, OBJECT_SIZE); >> munmap(ptr, OBJECT_SIZE); >> >> - gem_set_tiling(fd, handle, tiling, 1024); >> + gem_set_tiling(fd, *handle, tiling, 1024); >> >> - ptr = gem_mmap(fd, handle, OBJECT_SIZE, PROT_READ | PROT_WRITE); >> + ptr = gem_mmap(fd, *handle, OBJECT_SIZE, PROT_READ | PROT_WRITE); >> igt_assert(ptr); >> >> /* XXX: mmap_gtt pulls the bo into the GTT read domain. */ >> - gem_sync(fd, handle); >> + gem_sync(fd, *handle); > >Just call gem_close(fd, handle); the mmap will hold the reference for the test. Thanks for the suggestion. I did not realise that was the case. It simplifies the patch a lot. Will submit a new version. //Derek >-Chris > >-- >Chris Wilson, Intel Open Source Technology Centre > _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
