On Mon, Apr 08, 2013 at 09:23:55AM +0200, Daniel Vetter wrote:
> With this at least the y-tiled test reliably fails on my machines, but
> x-tiled still passes on some. More ideas to tune this highly welcome.
>
> v2: Fill cpu caches with data for each newly allocated bo. This seems
> to do the trick on my snb here _really_ reliably. So apparently the
> backsnoop for llc gtt writes is the crucial ingredient here to make
> the test fail.
>
> While at it, also stop leaking mmap space.
>
> v3: Fixup commit message.
>
> Cc: Mika Kuoppala <mika.kuoppala at linux.intel.com>
> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
Yes, that is much more effective - even at weeding out incomplete fixes
which passed the original test case.
> ---
> tests/gem_fence_thrash.c | 23 ++++++++++++++++++++---
> 1 file changed, 20 insertions(+), 3 deletions(-)
>
> diff --git a/tests/gem_fence_thrash.c b/tests/gem_fence_thrash.c
> index 3fc72a1..01fd0f6 100644
> --- a/tests/gem_fence_thrash.c
> +++ b/tests/gem_fence_thrash.c
> @@ -69,6 +69,12 @@ bo_create (int fd, int tiling)
>
> handle = gem_create(fd, OBJECT_SIZE);
>
> + /* dirty cpu caches a bit ... */
> + ptr = gem_mmap__cpu(fd, handle, OBJECT_SIZE, PROT_READ | PROT_WRITE);
> + assert(ptr);
Pedagogical:+ gem_set_domain(fd, handle,
I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU);
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
