12Operation
-- -
965000.0 0.016 10x10 wide rectangle outline
Something is still not quite right here. This should be mostly CPU bound,
and even my Atom gets 734k.
Can you check that (a) it is CPU bound and (b) the worst offenders
according to
On Thu, 12 May 2011 09:19:39 +0200, Knut Petersen knut_peter...@t-online.de
wrote:
12Operation
-- -
965000.0 0.016 10x10 wide rectangle outline
Something is still not quite right here. This should be mostly CPU bound,
and even my Atom
Please do something like 'perf record -f -g -a x11perf -d :0 -worect10;
perf report | head -150' and paste the output.
-Chris
Attached find the perf log
Knut
# Events: 19K cycles
#
# Overhead CommandShared Object
On Thu, 12 May 2011 10:24:00 +0200, Knut Petersen knut_peter...@t-online.de
wrote:
Please do something like 'perf record -f -g -a x11perf -d :0 -worect10;
perf report | head -150' and paste the output.
-Chris
Attached find the perf log
Oh, damage. A compositing WM? If you turn off
Oh, damage. A compositing WM? If you turn off compositing, do you see
similar performance levels to xorg-1.6?
-Chris
That makes difference 16.300 reps speed up to 1.280.000 reps ... 78.5
times faster.
I think I will rerun the tests.
cu,
Knut
On Wed, 2011-05-11 at 23:22 +0200, Knut Petersen wrote:
Yes, there is
15400.00.54 GetProperty
15500.00.54 QueryPointer
but we also see
815.01.21 X protocol NoOperation
NoOp isn't a round trip, it does not generate a reply. That test
measures how fast
From: Daniel Vetter daniel.vet...@ffwll.ch
Signed-off-by: Daniel Vetter daniel.vet...@ffwll.ch
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/i915_debugfs.c |2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git
Convert our open coded offset_in_page() to the common macro.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/i915_gem.c | 21 ++---
1 files changed, 10 insertions(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c
Rather than proceed on and silently return false by default, mention why
we rejected the presence of an EDID as implying the presence of a VGA
monitor. (The question arises whether there is a broken EDID which falsely
reports a digital connection when attached by VGA.)
Signed-off-by: Chris Wilson
The computation of the first-level watermarks for g4x and gen5+ are
based on the same algorithm, so we can refactor those code paths to
use a single function.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/intel_display.c | 88 --
Superseded by the tracking the render generation in the chipset
capabiltiies struct.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/i915_drv.h |7 ---
1 files changed, 0 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h
From: Daniel Vetter daniel.vet...@ffwll.ch
A tile on gen2 has a size of 2kb, stride of 128 bytes and 16 rows.
Userspace was broken and assumed 8 rows. Chris Wilson noted that the
kernel unfortunately can't reliable check that because libdrm rounds
up the size to the next bucket.
Signed-off-by:
Make the audio property creation routine common and share the single
property between the connectors.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/i915_drv.h|1 +
drivers/gpu/drm/i915/intel_dp.c| 15 ++-
drivers/gpu/drm/i915/intel_drv.h
As we cannot wait upon an object to be released by the GPU once we have
disabled pagefaults, process any pending retirements first in the hope
that we move any potential relocations off the active list.
References: https://bugs.freedesktop.org/show_bug.cgi?id=35733
Signed-off-by: Chris Wilson
Toggle the Software Clear Interrupt bit which resets the controller to
clear any prior BUS_ERROR condition before we begin to use the
controller in earnest.
Suggested-by: Ben Widawsky b...@bwidawsk.net
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/intel_i2c.c |
... as they only had a single PCI-ID each, and so using the pci-id is
easier than using a capability bit.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/i915_debugfs.c |2 --
drivers/gpu/drm/i915/i915_drv.c |9 +++--
drivers/gpu/drm/i915/i915_drv.h
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
---
drivers/gpu/drm/i915/i915_gem.c |4 +++-
1 files changed, 3 insertions(+), 1 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index afdbbd9..b92e8ea 100644
---
Keith complained that GMBUSx + reg_offset was ugly. An alternative
naming scheme which is more consistent with the reset of the code base
is to store the address of the GMBUS0 and then reference each of the
GMBUSx registers as an offset from GMBUS0.
Signed-off-by: Chris Wilson
Replace the three nearly identical copies of the code with a single
function. And take advantage of the opportunity to do some
micro-optimisation: avoid the vmalloc if at all possible and also avoid
dropping the lock unless we are forced to acquire the mm semaphore.
Measuring the impact of the
The read back of the available FIFO entries is vital for system
stability, but extremely costly. However, we only need a guide so as to
avoid eating into the reserved entries and since we are the only
consumer we can cache the read of the count from the last write.
Signed-off-by: Chris Wilson
Instead of scanning through the static array using strcmp to find our
matching tv_mode, just keep a direct link around. The historical reason
for the strcmp was to match the connection property tv format, but now
that property is translated for us into an index by the drm core.
Signed-off-by:
Am Freitag, den 29.04.2011, 22:16 +0200 schrieb Paul Menzel:
Am Montag, den 25.04.2011, 22:13 +0100 schrieb Chris Wilson:
On Mon, 25 Apr 2011 18:58:01 +0200, Paul Menzel
paulepan...@users.sourceforge.net wrote:
On Wed, Mar 23, 2011 at 14:08:24 +, Chris Wilson wrote:
commit
Signed-off-by: Keith Packard kei...@keithp.com
---
drivers/gpu/drm/i915/intel_tv.c |6 ++
1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c
index 6b22c1d..876064c 100644
--- a/drivers/gpu/drm/i915/intel_tv.c
Do not use this bit to indicate that load detection has completed,
instead just wait for vblank, at which point the load registers will
have been updated.
Signed-off-by: Keith Packard kei...@keithp.com
---
drivers/gpu/drm/i915/intel_tv.c | 40 +++---
1 files
On Thu, 12 May 2011 22:17:10 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
+ pages = kmalloc(n*sizeof(struct page *),
+ GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN);
+ if (pages == NULL) {
+ pages = drm_malloc_ab(n, sizeof(struct page *));
+
On Thu, 12 May 2011 22:17:11 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
Convert our open coded offset_in_page() to the common macro.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
Reviewed-by: Keith Packard kei...@keithp.com
--
keith.pack...@intel.com
pgpjf0bivf3pS.pgp
On Thu, 12 May 2011 22:17:17 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
Toggle the Software Clear Interrupt bit which resets the controller to
clear any prior BUS_ERROR condition before we begin to use the
controller in earnest.
Looks reasonable, except for the bad register offsets
On Thu, 12 May 2011 22:17:13 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
Instead of scanning through the static array using strcmp to find our
matching tv_mode, just keep a direct link around. The historical reason
for the strcmp was to match the connection property tv format, but now
On Thu, 12 May 2011 22:17:19 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
- return -ENOSPC;
+ return -EDEADLK;
Would be nice to have the kernel print out a debugging message at this
point -- the only way to hit this is from a bug in user or kernel code.
--
On Thu, 12 May 2011 22:17:16 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
- I915_WRITE(GMBUS0 + reg_offset, 0);
+ intel_i2c_reset(dev_priv-dev);
This looks like an unrelated change. Please split this out.
--
keith.pack...@intel.com
pgp2uVPK5zOxO.pgp
Description: PGP
On Thu, 12 May 2011 22:17:22 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
... as they only had a single PCI-ID each, and so using the pci-id is
easier than using a capability bit.
This doesn't seem useful to me; it only saves a couple of bits in the
struct and replaces that with
On Thu, 12 May 2011 22:17:23 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
Can you justify this change with performance data?
--
keith.pack...@intel.com
pgpwnyNWMT1aq.pgp
Description: PGP signature
On Thu, 12 May 2011 22:17:18 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
As we cannot wait upon an object to be released by the GPU once we have
disabled pagefaults, process any pending retirements first in the hope
that we move any potential relocations off the active list.
This
On Thu, 12 May 2011 22:17:24 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
Make the audio property creation routine common and share the single
property between the connectors.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
Reviewed-by: Keith Packard kei...@keithp.com
--
On Thu, 12 May 2011 22:17:20 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
From: Daniel Vetter daniel.vet...@ffwll.ch
A tile on gen2 has a size of 2kb, stride of 128 bytes and 16 rows.
Userspace was broken and assumed 8 rows. Chris Wilson noted that the
kernel unfortunately can't
On Thu, 12 May 2011 22:17:21 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
Superseded by the tracking the render generation in the chipset
capabiltiies struct.
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
Yes please.
Reviewed-by: Keith Packard kei...@keithp.com
--
On Thu, 12 May 2011 22:17:14 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
The computation of the first-level watermarks for g4x and gen5+ are
based on the same algorithm, so we can refactor those code paths to
use a single function.
g4x_compute_wm0 takes a plane. ironlake_compute_wm0
On Thu, 12 May 2011 22:17:12 +0100, Chris Wilson ch...@chris-wilson.co.uk
wrote:
From: Daniel Vetter daniel.vet...@ffwll.ch
Signed-off-by: Daniel Vetter daniel.vet...@ffwll.ch
Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk
This change seems correct. Would be nice if this code printed
38 matches
Mail list logo
