This will allow to monitor global performance counters through the
command stream of the GPU instead of using ioctls.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
src/gallium/drivers/nouveau/nv50/nv50_screen.c | 11 +++
src/gallium/drivers/nouveau/nv50/nv50_screen.h | 1
, Samuel Pitoiset
samuel.pitoi...@gmail.com wrote:
This may happen when nv50_query_create() fails to create a new query.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
src/gallium/drivers/nouveau/nv50/nv50_query.c | 15 ++-
1 file changed, 14 insertions(+), 1 deletion
This trivial patch makes thing more consistent since hardware signals
names are prefixed by 'pcXX'.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/nvkm/engine/pm/nv40.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drm/nouveau/nvkm/engine/pm/nv40.c
These signals and sources have been reverse engineered from CUPTI
(Linux). Graphics signals exposed by PerfKit (Windows only) will be
added later. I need to reverse engineer them and it's a bit painful.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/nvkm/engine/pm/gf100
gk104_pm_ctor() is equal to gf100_pm_ctor().
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/include/nvkm/engine/pm.h | 2 +-
drm/nouveau/nvkm/engine/device/gk104.c | 8 ++---
drm/nouveau/nvkm/engine/pm/gk104.c | 60 +-
3 files
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
This patch should be squashed with pm/nv50: add compute and graphics
signals/sources.
drm/nouveau/nvkm/engine/pm/gt200.c | 8
drm/nouveau/nvkm/engine/pm/gt215.c | 3 +--
2 files changed, 5 insertions(+), 6 deletions(-)
diff
On 06/14/2015 04:32 AM, Ben Skeggs wrote:
On 10 June 2015 at 07:53, Samuel Pitoiset samuel.pitoi...@gmail.com wrote:
On 06/09/2015 12:02 AM, Ben Skeggs wrote:
On 8 June 2015 at 06:40, Samuel Pitoiset samuel.pitoi...@gmail.com
wrote:
This commit introduces the NVIF_IOCTL_NEW_V0_PERFMON
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
This patch should be squashed with pm: allow the userspace to configure
sources.
drm/nouveau/nvkm/engine/pm/base.c | 21 -
1 file changed, 16 insertions(+), 5 deletions(-)
diff --git a/drm/nouveau/nvkm/engine/pm
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/include/nvkm/engine/pm.h | 2 +-
drm/nouveau/nvkm/engine/device/gf100.c | 18 +-
drm/nouveau/nvkm/engine/pm/gf100.c | 22 +-
drm/nouveau/nvkm/engine/pm/gf100.h | 11
These signals and sources have been reverse engineered from CUPTI
(Linux). Graphics signals exposed by PerfKit (Windows only) will be
added later. I need to reverse engineer them and it's a bit painful.
This commit also adds a new class for GF108 and GF117.
Signed-off-by: Samuel Pitoiset
Pushed.
On 05/27/2015 01:23 PM, Samuel Pitoiset wrote:
On 05/21/2015 06:40 AM, Dave Airlie wrote:
On 21 May 2015 at 03:26, Samuel Pitoiset samuel.pitoi...@gmail.com
wrote:
On 05/20/2015 07:13 PM, Ilia Mirkin wrote:
This is obviously a bug, but one that has been there for some time.
Please
On 06/09/2015 12:02 AM, Ben Skeggs wrote:
On 8 June 2015 at 06:40, Samuel Pitoiset samuel.pitoi...@gmail.com wrote:
This commit introduces the NVIF_IOCTL_NEW_V0_PERFMON class which will be
used in order to query domains, signals and sources. This separates the
querying and the counting
These functions will be also used by the interface for controlling
performance counters.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/nvkm/engine/sw/nv50.c | 56 ++-
1 file changed, 38 insertions(+), 18 deletions(-)
diff --git
16 bits is large enough to store the maximum number of signals available
for one domain (i.e. 256).
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
bin/nv_perfmon.c | 4 ++--
drm/nouveau/include/nvif/class.h | 7 ---
drm/nouveau/nvkm/engine/pm/base.c | 10
Signed-off-by: Samuel Pitoiset samuel.pitoiset at gmail.com
---
drm/nouveau/include/nvif/class.h | 1 +
drm/nouveau/nvkm/engine/pm/base.c | 102 ++
drm/nouveau/nvkm/engine/pm/priv.h | 2 +
3 files changed, 95 insertions(+), 10 deletions(-)
diff --git
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/include/nvif/class.h | 3 ++-
drm/nouveau/nvkm/engine/pm/base.c | 22 +++---
2 files changed, 21 insertions(+), 4 deletions(-)
diff --git a/drm/nouveau/include/nvif/class.h b/drm/nouveau/include/nvif
This allows to query the ID, the mask and the user-readable name of
sources for each signal.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/include/nvif/class.h | 12 ++
drm/nouveau/nvkm/engine/pm/base.c | 77 +++
2 files changed
nv_perfmon but this is going to be fixed
with the upcoming patch.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/include/nvif/class.h | 8 +++--
drm/nouveau/nvkm/engine/pm/base.c | 64 +--
drm/nouveau/nvkm/engine/pm/priv.h | 1 +
3 files
This commit introduces the NVIF_IOCTL_NEW_V0_PERFMON class which will be
used in order to query domains, signals and sources. This separates the
querying and the counting interface.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
bin/nv_perfmon.c | 12
This will allow to configure performance counters with hardware signal
indexes instead of user-readable names in an upcoming patch.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
bin/nv_perfmon.c | 185 --
drm/nouveau/include
This allows to query the number of available domains, including the
number of hardware counter and the number of signals per domain.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/include/nvif/class.h | 11 -
drm/nouveau/nvkm/engine/pm/base.c | 86
PDAEMON signals don't have to be exposed by the perfmon engine.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
Reviewed-by: Martin Peres martin.pe...@free.fr
---
drm/nouveau/include/nvkm/engine/pm.h | 4 --
drm/nouveau/nvkm/engine/pm/Kbuild| 1 -
drm/nouveau/nvkm/engine/pm
This software methods interface will allow the userspace to tie
monitoring of performance counters to the command stream.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/nvkm/engine/sw/nv50.c | 160 ++
drm/nouveau/nvkm/engine/sw/nv50.h
These signals and sources have been reverse engineered from NVIDIA
PerfKit (Windows) and CUPTI (Linux), they will be used to build complex
hardware events from the userspace.
This commit also adds a new class for GT200.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau
From: Samuel Pitoiset samuel.pitoisetœgmail.com
A source (or multiplexer) is a tuple addr+mask+shift which allows to
control a block of signals. The maximum number of sources that a signal
can define is arbitrary limited to 8 and this should be large enough.
This patch allows to define multi
a domain instead of
separate counters. This has the advantage to move all of the logic to
the userspace.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
bin/nv_perfmon.c | 193 ++
drm/nouveau/include/nvif/class.h | 30 +++---
drm/nouveau
This fixes a crash when multiple PM engine contexts are created.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
Reviewed-by: Martin Peres martin.pe...@free.fr
---
drm/nouveau/nvkm/engine/pm/base.c | 7 +--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drm/nouveau
/sources
Patches 19-20: expose a software methods interface for controlling PM
Feel free to make a review.
Thanks in advance!
Samuel Pitoiset (20):
pm: prevent freeing the wrong engine context
pm: fix a potential race condition when creating an engine context
pm: remove pmu signals
pm
There is always the possiblity that the ppm-context pointer would get
partially updated and accidentally would equal ctx. This would allow two
contexts to co-exist, which is not acceptable. Moving the test to the
critical section takes care of this problem.
Signed-off-by: Samuel Pitoiset
Since a new class has been introduced to query signals, we can now
return an error when the userspace wants to monitor unknown signals.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/nvkm/engine/pm/base.c | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
Reviewed-by: Martin Peres martin.pe...@free.fr
---
drm/nouveau/nvkm/engine/pm/base.c | 20
drm/nouveau/nvkm/engine/pm/priv.h | 4
2 files changed, 24 deletions(-)
diff --git a/drm/nouveau/nvkm/engine/pm/base.c
b
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
bin/nv_perfmon.c | 6 +++--
drm/nouveau/include/nvif/class.h | 11 ++---
drm/nouveau/nvkm/engine/pm/base.c | 52 +--
3 files changed, 23 insertions(+), 46 deletions(-)
diff
This signal index must be always allowed even if it's not clearly
defined in a domain in order to monitor a counter like 0x03020100
because it's the default value of signals.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/nouveau/nvkm/engine/pm/base.c | 4 ++--
drm/nouveau
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
src/util/list.h | 9 +
1 file changed, 9 insertions(+)
diff --git a/src/util/list.h b/src/util/list.h
index fb566f5..c782ef5 100644
--- a/src/util/list.h
+++ b/src/util/list.h
@@ -116,12 +116,21 @@ static inline void
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
src/gallium/drivers/nouveau/nv30/nv30_query.c | 3 ---
src/util/list.h | 3 +++
2 files changed, 3 insertions(+), 3 deletions(-)
diff --git a/src/gallium/drivers/nouveau/nv30/nv30_query.c
b/src/gallium
/libperfkit/
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
configure.ac | 30 ++
src/gallium/Makefile.am| 4 +
src/gallium/state_trackers/perfkit/Makefile.am | 41 ++
.../state_trackers/perfkit/Makefile.sources
and libperfkit.
Thanks!
Samuel Pitoiset (3):
util: move LIST_FIRST_ENTRY from nv30_query.c to list.h
util: add LIST_MOVE_TAIL to list.h
st/perfkit: add a state tracker for NVIDIA PerfKit
configure.ac | 30 ++
src/gallium/Makefile.am
On 05/21/2015 06:40 AM, Dave Airlie wrote:
On 21 May 2015 at 03:26, Samuel Pitoiset samuel.pitoi...@gmail.com wrote:
On 05/20/2015 07:13 PM, Ilia Mirkin wrote:
This is obviously a bug, but one that has been there for some time.
Please figure out what this is guarding, and confirm
On 05/26/2015 12:22 AM, Pierre Moreau wrote:
Signed-off-by: Pierre Moreau pierre.mor...@free.fr
---
drm/nouveau/nouveau_acpi.c | 10 --
1 file changed, 4 insertions(+), 6 deletions(-)
diff --git a/drm/nouveau/nouveau_acpi.c b/drm/nouveau/nouveau_acpi.c
index 36f4a40..073f7d7 100644
On 05/26/2015 12:22 AM, Pierre Moreau wrote:
Most _DSM will return an integer value of 0x8002 when given an unknown
UUID, revision ID or function ID. Checking locally allows us to differentiate
that case from other ACPI errors, and to not report a failed to evaluate _DSM
if 0x8002 is
On 05/26/2015 12:22 AM, Pierre Moreau wrote:
Signed-off-by: Pierre Moreau pierre.mor...@free.fr
---
drm/nouveau/nouveau_acpi.c | 53 --
drm/nouveau/nouveau_acpi.h | 2 ++
drm/nouveau/nouveau_drm.c | 6 --
drm/nouveau/nouveau_vga.c | 10
On 05/26/2015 02:46 PM, Pierre Moreau wrote:
I'm thinking of re-writing this patch to just OR the different returned retval
and test for individual bits directly in the final conditionals.
So this would give something like:
int retval = 0;
while ((pdev = pci_get_class(PCI_CLASS_DISPLAY_VGA
On 05/24/2015 06:58 AM, Ilia Mirkin wrote:
nv30_validate_clip depends on the rasterizer state. Also we should
upload all the new clip planes on change since next time the plane data
won't have changed, but the enables might.
Signed-off-by: Ilia Mirkin imir...@alum.mit.edu
---
play with it...
On Wed, May 20, 2015 at 1:11 PM, Tobias Klausmann
tobias.johannes.klausm...@mni.thm.de wrote:
looks good to me! :)
Feel free to add my R-b.
On 20.05.2015 17:08, Samuel Pitoiset wrote:
This is probably a typo error which has been introduced in 2009...
This fixes the following
-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
src/drmmode_display.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/drmmode_display.c b/src/drmmode_display.c
index 7c1d2bb..161bccd 100644
--- a/src/drmmode_display.c
+++ b/src/drmmode_display.c
@@ -904,7 +904,7
* bump *
Ben, could you take a look at the series ?
On 04/14/2015 10:43 PM, Samuel Pitoiset wrote:
There is always the possiblity that the ppm-context pointer would get
partially updated and accidentally would equal ctx. This would allow two
contexts to co-exist, which is not acceptable
This fixes a crash when multiple PM engine contexts are created.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
Reviewed-by: Martin Peres martin.pe...@free.fr
---
drm/nouveau/nvkm/engine/pm/base.c | 7 +--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/drm/nouveau
There is always the possiblity that the ppm-context pointer would get
partially updated and accidentally would equal ctx. This would allow two
contexts to co-exist, which is not acceptable. Moving the test to the
critical section takes care of this problem.
Signed-off-by: Samuel Pitoiset
Reviewed-by: Samuel Pitoiset samuel.pitoi...@gmail.com
But please, fix the comment in the nv50_query struct, now it is /* base
+ i * 32 */.
On 04/04/2015 06:00 AM, Ilia Mirkin wrote:
Commit 1a170980a09 started writing to q-data[4]/[5] but kept the
per-query space at 16, which meant
On 12/18/2014 11:25 AM, Ben Skeggs wrote:
On Thu, Dec 18, 2014 at 8:13 PM, Vince Hsu vin...@nvidia.com wrote:
Hello Ben,
On 12/18/2014 05:34 PM, Ben Skeggs wrote:
On Thu, Dec 18, 2014 at 4:28 PM, Vince Hsu vin...@nvidia.com wrote:
This patch creates a subdev for DVFS (Dynamic Voltage and
Hi folks,
This set of patches is just for clearing the way before introducing
the infrastructure for performance counters in perfmon.
I'll try to publish a new version of my kernel interface based on nvif
in a couple of days.
See you.
___
Nouveau
As discussed with Martin Peres, PDAEMON signals should be exposed
through sysfs and not managed by the perfmon engine which is only
used for performance counters.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
drm/Kbuild | 1 -
drm/core/engine/perfmon
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
nvkm/engine/perfmon/base.c | 20
nvkm/engine/perfmon/priv.h | 3 ---
2 files changed, 23 deletions(-)
diff --git a/nvkm/engine/perfmon/base.c b/nvkm/engine/perfmon/base.c
index 5fa45e1..b74734e 100644
--- a/nvkm
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
lib/kern.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/lib/kern.c b/lib/kern.c
index 930cd35..7e8f3e9 100644
--- a/lib/kern.c
+++ b/lib/kern.c
@@ -173,7 +173,7 @@ drm_client_init(const char *name, u64 device
This fixes a crash when we reload Nouveau DRM.
Signed-off-by: Samuel Pitoiset samuel.pitoi...@gmail.com
---
nvkm/engine/perfmon/base.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/nvkm/engine/perfmon/base.c b/nvkm/engine/perfmon/base.c
index e9c5e51..7481003 100644
--- a/nvkm/engine
101 - 155 of 155 matches
Mail list logo
