| Message ID | 20180505105549.14351-1-chris@chris-wilson.co.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Sat, May 05, 2018 at 11:55:49AM +0100, Chris Wilson wrote: > When waiting for a single fence beneath a syncobj, forgo our open coding > for waiting over multiple fences and call the driver specific > dma_fence_wait_timeout(). This gives the opportunity for the driver to > handle it more efficiently then the lcd of signal callbacks, and avoids > the temporary allocations and array iterations. > > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Aside: Do the igts test both the single and multi-case? -Daniel > --- > drivers/gpu/drm/drm_syncobj.c | 8 ++++++++ > 1 file changed, 8 insertions(+) > > diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c > index d4f4ce484529..c0f787f7a298 100644 > --- a/drivers/gpu/drm/drm_syncobj.c > +++ b/drivers/gpu/drm/drm_syncobj.c > @@ -695,6 +695,14 @@ static signed long drm_syncobj_array_wait_timeout(struct drm_syncobj **syncobjs, > signed long ret; > uint32_t signaled_count, i; > > + /* KISS for the common case of waiting for a single submitted fence. */ > + if (count == 1 && (fence = drm_syncobj_fence_get(syncobjs[0]))) { > + ret = dma_fence_wait_timeout(fence, true, timeout); > + dma_fence_put(fence); > + *idx = 0; > + return ret; > + } > + > entries = kcalloc(count, sizeof(*entries), GFP_KERNEL); > if (!entries) > return -ENOMEM; > -- > 2.17.0 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
diff --git a/drivers/gpu/drm/drm_syncobj.c b/drivers/gpu/drm/drm_syncobj.c index d4f4ce484529..c0f787f7a298 100644 --- a/drivers/gpu/drm/drm_syncobj.c +++ b/drivers/gpu/drm/drm_syncobj.c @@ -695,6 +695,14 @@ static signed long drm_syncobj_array_wait_timeout(struct drm_syncobj **syncobjs, signed long ret; uint32_t signaled_count, i; + /* KISS for the common case of waiting for a single submitted fence. */ + if (count == 1 && (fence = drm_syncobj_fence_get(syncobjs[0]))) { + ret = dma_fence_wait_timeout(fence, true, timeout); + dma_fence_put(fence); + *idx = 0; + return ret; + } + entries = kcalloc(count, sizeof(*entries), GFP_KERNEL); if (!entries) return -ENOMEM;
When waiting for a single fence beneath a syncobj, forgo our open coding for waiting over multiple fences and call the driver specific dma_fence_wait_timeout(). This gives the opportunity for the driver to handle it more efficiently then the lcd of signal callbacks, and avoids the temporary allocations and array iterations. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> --- drivers/gpu/drm/drm_syncobj.c | 8 ++++++++ 1 file changed, 8 insertions(+)