From mboxrd@z Thu Jan 1 00:00:00 1970 From: "j.glisse" Subject: Re: [PATCH 02/10] drm/radeon: add infrastructure for advanced ring synchronization Date: Thu, 24 May 2012 11:56:21 -0400 Message-ID: <20120524155620.GC3467@gmail.com> References: <1337845754-3718-1-git-send-email-deathsimple@vodafone.de> <1337845754-3718-2-git-send-email-deathsimple@vodafone.de> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Return-path: Received: from mail-vc0-f177.google.com (mail-vc0-f177.google.com [209.85.220.177]) by gabe.freedesktop.org (Postfix) with ESMTP id 29E6E9EF09 for ; Thu, 24 May 2012 08:56:37 -0700 (PDT) Received: by vcbf13 with SMTP id f13so1806451vcb.36 for ; Thu, 24 May 2012 08:56:37 -0700 (PDT) Content-Disposition: inline In-Reply-To: <1337845754-3718-2-git-send-email-deathsimple@vodafone.de> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dri-devel-bounces+sf-dri-devel=m.gmane.org@lists.freedesktop.org Errors-To: dri-devel-bounces+sf-dri-devel=m.gmane.org@lists.freedesktop.org To: Christian =?iso-8859-1?Q?K=F6nig?= Cc: dri-devel@lists.freedesktop.org List-Id: dri-devel@lists.freedesktop.org On Thu, May 24, 2012 at 09:49:06AM +0200, Christian K=F6nig wrote: > Signed-off-by: Christian K=F6nig Need a small improvement see below, otherwise Reviewed-by: Jerome Glisse > --- > drivers/gpu/drm/radeon/radeon.h | 23 ++++++++++- > drivers/gpu/drm/radeon/radeon_fence.c | 73 +++++++++++++++++++++++++++= ++---- > 2 files changed, 85 insertions(+), 11 deletions(-) > = > diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/rad= eon.h > index 5e259b4..4e232c3 100644 > --- a/drivers/gpu/drm/radeon/radeon.h > +++ b/drivers/gpu/drm/radeon/radeon.h > @@ -257,8 +257,8 @@ struct radeon_fence_driver { > uint32_t scratch_reg; > uint64_t gpu_addr; > volatile uint32_t *cpu_addr; > - /* seq is protected by ring emission lock */ > - uint64_t seq; > + /* sync_seq is protected by ring emission lock */ > + uint64_t sync_seq[RADEON_NUM_RINGS]; > atomic64_t last_seq; > unsigned long last_activity; > bool initialized; > @@ -288,6 +288,25 @@ int radeon_fence_wait_any(struct radeon_device *rdev, > struct radeon_fence *radeon_fence_ref(struct radeon_fence *fence); > void radeon_fence_unref(struct radeon_fence **fence); > unsigned radeon_fence_count_emitted(struct radeon_device *rdev, int ring= ); > +bool radeon_fence_need_sync(struct radeon_fence *fence, int ring); > +void radeon_fence_note_sync(struct radeon_fence *fence, int ring); > +static inline struct radeon_fence *radeon_fence_later(struct radeon_fenc= e *a, > + struct radeon_fence *b) > +{ > + if (!a) { > + return b; > + } > + > + if (!b) { > + return a; > + } Please add : BUG_ON(a->ring !=3D b->ring); So we can catch if someone badly use this function. > + > + if (a->seq > b->seq) { > + return a; > + } else { > + return b; > + } > +} > = > /* > * Tiling registers > diff --git a/drivers/gpu/drm/radeon/radeon_fence.c b/drivers/gpu/drm/rade= on/radeon_fence.c > index 401d346..7b55625 100644 > --- a/drivers/gpu/drm/radeon/radeon_fence.c > +++ b/drivers/gpu/drm/radeon/radeon_fence.c > @@ -72,7 +72,7 @@ int radeon_fence_emit(struct radeon_device *rdev, > } > kref_init(&((*fence)->kref)); > (*fence)->rdev =3D rdev; > - (*fence)->seq =3D ++rdev->fence_drv[ring].seq; > + (*fence)->seq =3D ++rdev->fence_drv[ring].sync_seq[ring]; > (*fence)->ring =3D ring; > radeon_fence_ring_emit(rdev, ring, *fence); > trace_radeon_fence_emit(rdev->ddev, (*fence)->seq); > @@ -449,7 +449,7 @@ int radeon_fence_wait_next_locked(struct radeon_devic= e *rdev, int ring) > * wait. > */ > seq =3D atomic64_read(&rdev->fence_drv[ring].last_seq) + 1ULL; > - if (seq >=3D rdev->fence_drv[ring].seq) { > + if (seq >=3D rdev->fence_drv[ring].sync_seq[ring]) { > /* nothing to wait for, last_seq is > already the last emited fence */ > return -ENOENT; > @@ -464,7 +464,7 @@ int radeon_fence_wait_empty_locked(struct radeon_devi= ce *rdev, int ring) > * activity can be scheduled so there won't be concurrent access > * to seq value. > */ > - return radeon_fence_wait_seq(rdev, rdev->fence_drv[ring].seq, > + return radeon_fence_wait_seq(rdev, rdev->fence_drv[ring].sync_seq[ring], > ring, false, false); > } > = > @@ -492,7 +492,8 @@ unsigned radeon_fence_count_emitted(struct radeon_dev= ice *rdev, int ring) > * but it's ok to report slightly wrong fence count here. > */ > radeon_fence_process(rdev, ring); > - emitted =3D rdev->fence_drv[ring].seq - atomic64_read(&rdev->fence_drv[= ring].last_seq); > + emitted =3D rdev->fence_drv[ring].sync_seq[ring] > + - atomic64_read(&rdev->fence_drv[ring].last_seq); > /* to avoid 32bits warp around */ > if (emitted > 0x10000000) { > emitted =3D 0x10000000; > @@ -500,6 +501,51 @@ unsigned radeon_fence_count_emitted(struct radeon_de= vice *rdev, int ring) > return (unsigned)emitted; > } > = > +bool radeon_fence_need_sync(struct radeon_fence *fence, int dst_ring) > +{ > + struct radeon_fence_driver *fdrv; > + > + if (!fence) { > + return false; > + } > + > + if (fence->ring =3D=3D dst_ring) { > + return false; > + } > + > + /* we are protected by the ring mutex */ > + fdrv =3D &fence->rdev->fence_drv[dst_ring]; > + if (fence->seq <=3D fdrv->sync_seq[fence->ring]) { > + return false; > + } > + > + return true; > +} > + > +void radeon_fence_note_sync(struct radeon_fence *fence, int dst_ring) > +{ > + struct radeon_fence_driver *dst, *src; > + unsigned i; > + > + if (!fence) { > + return; > + } > + > + if (fence->ring =3D=3D dst_ring) { > + return; > + } > + > + /* we are protected by the ring mutex */ > + src =3D &fence->rdev->fence_drv[fence->ring]; > + dst =3D &fence->rdev->fence_drv[dst_ring]; > + for (i =3D 0; i < RADEON_NUM_RINGS; ++i) { > + if (i =3D=3D dst_ring) { > + continue; > + } > + dst->sync_seq[i] =3D max(dst->sync_seq[i], src->sync_seq[i]); > + } > +} > + > int radeon_fence_driver_start_ring(struct radeon_device *rdev, int ring) > { > uint64_t index; > @@ -521,7 +567,7 @@ int radeon_fence_driver_start_ring(struct radeon_devi= ce *rdev, int ring) > } > rdev->fence_drv[ring].cpu_addr =3D &rdev->wb.wb[index/4]; > rdev->fence_drv[ring].gpu_addr =3D rdev->wb.gpu_addr + index; > - radeon_fence_write(rdev, rdev->fence_drv[ring].seq, ring); > + radeon_fence_write(rdev, rdev->fence_drv[ring].sync_seq[ring], ring); > rdev->fence_drv[ring].initialized =3D true; > dev_info(rdev->dev, "fence driver on ring %d use gpu addr 0x%016llx and= cpu addr 0x%p\n", > ring, rdev->fence_drv[ring].gpu_addr, rdev->fence_drv[ring].cpu_addr); > @@ -530,10 +576,13 @@ int radeon_fence_driver_start_ring(struct radeon_de= vice *rdev, int ring) > = > static void radeon_fence_driver_init_ring(struct radeon_device *rdev, in= t ring) > { > + int i; > + > rdev->fence_drv[ring].scratch_reg =3D -1; > rdev->fence_drv[ring].cpu_addr =3D NULL; > rdev->fence_drv[ring].gpu_addr =3D 0; > - rdev->fence_drv[ring].seq =3D 0; > + for (i =3D 0; i < RADEON_NUM_RINGS; ++i) > + rdev->fence_drv[ring].sync_seq[i] =3D 0; > atomic64_set(&rdev->fence_drv[ring].last_seq, 0); > rdev->fence_drv[ring].last_activity =3D jiffies; > rdev->fence_drv[ring].initialized =3D false; > @@ -579,7 +628,7 @@ static int radeon_debugfs_fence_info(struct seq_file = *m, void *data) > struct drm_info_node *node =3D (struct drm_info_node *)m->private; > struct drm_device *dev =3D node->minor->dev; > struct radeon_device *rdev =3D dev->dev_private; > - int i; > + int i, j; > = > for (i =3D 0; i < RADEON_NUM_RINGS; ++i) { > if (!rdev->fence_drv[i].initialized) > @@ -588,8 +637,14 @@ static int radeon_debugfs_fence_info(struct seq_file= *m, void *data) > seq_printf(m, "--- ring %d ---\n", i); > seq_printf(m, "Last signaled fence 0x%016llx\n", > (unsigned long long)atomic64_read(&rdev->fence_drv[i].last_seq)); > - seq_printf(m, "Last emitted 0x%016llx\n", > - rdev->fence_drv[i].seq); > + seq_printf(m, "Last emitted 0x%016llx\n", > + rdev->fence_drv[i].sync_seq[i]); > + > + for (j =3D 0; j < RADEON_NUM_RINGS; ++j) { > + if (i !=3D j && rdev->fence_drv[j].initialized) > + seq_printf(m, "Last sync to ring %d 0x%016llx\n", > + j, rdev->fence_drv[i].sync_seq[j]); > + } > } > return 0; > } > -- = > 1.7.9.5 > =