Ping. This is a more accurate version than RSQ+MUL+CMP.
Marek
On Tue, May 31, 2016 at 1:19 AM, Marek Olšák wrote:
> From: Marek Olšák
>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94627
> (against nouveau)
> ---
> src/mesa/state_tracker/st_glsl_to_tgsi.cpp | 16 +---
>
Hi,
The shader-based resolve is slow.
With this series, one scene with 8xMSAA in HL2: Lost Coast goes from 8-9 FPS to
32 FPS on Evergreen & Wine/Nine.
Please review.
Marek
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedes
From: Marek Olšák
---
src/gallium/drivers/radeonsi/si_blit.c | 34 +-
1 file changed, 21 insertions(+), 13 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_blit.c
b/src/gallium/drivers/radeonsi/si_blit.c
index 1c2c538..80844ec 100644
--- a/src/gallium/d
From: Marek Olšák
This improves MSAA resolve performance.
---
src/gallium/drivers/r600/r600_blit.c | 60 +++-
1 file changed, 52 insertions(+), 8 deletions(-)
diff --git a/src/gallium/drivers/r600/r600_blit.c
b/src/gallium/drivers/r600/r600_blit.c
index c9d7823.
From: Marek Olšák
This improves MSAA resolve performance.
---
src/gallium/drivers/radeonsi/si_blit.c | 62 +-
1 file changed, 53 insertions(+), 9 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_blit.c
b/src/gallium/drivers/radeonsi/si_blit.c
index 3748
On Mon, May 30, 2016 at 7:19 PM, Marek Olšák wrote:
> From: Marek Olšák
>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94627
> (against nouveau)
> ---
> src/mesa/state_tracker/st_glsl_to_tgsi.cpp | 16 +---
> 1 file changed, 9 insertions(+), 7 deletions(-)
>
> diff --git
Images invalidation is a bit weird on Fermi and there is already a hack
which forces invalidating all images when launching a computer shader
to help in fixing 3D<->CP interaction.
However, we need to re-validate images for compute because
nvc0_compute_invalidate_surfaces() will destroy the previo
On Sun, Jun 5, 2016 at 12:13 PM, Samuel Pitoiset
wrote:
> Images invalidation is a bit weird on Fermi and there is already a hack
> which forces invalidating all images when launching a computer shader
> to help in fixing 3D<->CP interaction.
>
> However, we need to re-validate images for compute
Images invalidation is a bit weird on Fermi and there is already a hack
which forces invalidating all images when launching a computer shader
to help in fixing 3D<->CP interaction.
However, we need to re-validate images for compute because
nvc0_compute_invalidate_surfaces() will destroy the previo
if abs is not inserted in the source glsl, that's a wine bug, not a mesa
bug,
Thus I don't think doing this is a good idea.
On 05/06/2016 17:17, Ilia Mirkin wrote:
On Mon, May 30, 2016 at 7:19 PM, Marek Olšák wrote:
From: Marek Olšák
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94
On Sun, Jun 5, 2016 at 12:28 PM, Samuel Pitoiset
wrote:
> Images invalidation is a bit weird on Fermi and there is already a hack
> which forces invalidating all images when launching a computer shader
> to help in fixing 3D<->CP interaction.
>
> However, we need to re-validate images for compute
We should not call nouveau_bufctx_reset() inside a validate function.
This only affects Fermi where images are aliased between 3D and CP.
Signed-off-by: Samuel Pitoiset
Cc: "12.0"
---
src/gallium/drivers/nouveau/nvc0/nvc0_compute.c | 1 +
src/gallium/drivers/nouveau/nvc0/nvc0_tex.c | 6 +---
Reviewed-by: Ilia Mirkin
On Sun, Jun 5, 2016 at 12:56 PM, Samuel Pitoiset
wrote:
> We should not call nouveau_bufctx_reset() inside a validate function.
> This only affects Fermi where images are aliased between 3D and CP.
>
> Signed-off-by: Samuel Pitoiset
> Cc: "12.0"
> ---
> src/gallium/dr
On 2016.06.03 18:49, ⚛ wrote:
> On Fri, Jun 3, 2016 at 2:08 PM, Nicolai Hähnle wrote:
>> On 03.06.2016 13:12, wrote:
>>>
>>> Situation: Looking at the content displayed by the web browser for URL
>>> http://patchwork.freedesktop.org/project/mesa/series and sub-pages
>>> accessible via the links.
>
https://bugs.freedesktop.org/show_bug.cgi?id=94512
--- Comment #6 from Emil Velikov ---
Based of the backtrace there is no information if mesa/libglapi was build with
or w/o glx-tls. Following an IRC conversation it seems that the library is
built w/o glx-tls as i is missing the _glapi_tls_Contex
On Sun, Jun 5, 2016 at 6:34 PM, Axel Davy wrote:
> if abs is not inserted in the source glsl, that's a wine bug, not a mesa
> bug,
> Thus I don't think doing this is a good idea.
The old code used CMP, this code uses ABS instead. ABS is faster. I
don't see a problem with that considering the beha
From: Dave Airlie
The clamping code always clamps to 0..1, which for SNORM is
incorrect. This adds a bit to say that clamping is for
an snorm and to use the correct range.
This fixes a number of SNORM cases in:
GL33-CTS.texture_size_promotion.functional
v2: handle the case where the format is a
Please ignore this patch. It doesn't work.
Marek
On Sun, Jun 5, 2016 at 5:07 PM, Marek Olšák wrote:
> From: Marek Olšák
>
> ---
> src/gallium/drivers/radeonsi/si_blit.c | 34
> +-
> 1 file changed, 21 insertions(+), 13 deletions(-)
>
> diff --git a/src/gallium/
On 6 June 2016 at 05:40, Dave Airlie wrote:
> From: Dave Airlie
>
> The clamping code always clamps to 0..1, which for SNORM is
> incorrect. This adds a bit to say that clamping is for
> an snorm and to use the correct range.
>
> This fixes a number of SNORM cases in:
> GL33-CTS.texture_size_prom
From: Dave Airlie
GL43-CTS.compute_shader.work-group-size does
uniform uint g_uniform[gl_WorkGroupSize.z + 20] = {
1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24 };
The initializer triggers the GLSL 4.30/GLES3 tests
for constant sequence subexpressions, so it doesn't
happen unle
Build mesa 1516 failed
Commit c417c0c9c3 by Kenneth Graunke on 9/7/2014 3:26 AM:
mesa: Add MESA_SHADER_CAPTURE_PATH for writing .shader_test files.\n\nThis writes linked shader programs to .shader_test files to\n$MESA_SHADER_CAPTURE_PATH in the format used by s
From: Dave Airlie
This fixes a crash in
GL43-CTS.shader_subroutine.subroutines_not_allowed_as_variables_constructors_and_argument_or_return_types
If we can't find the func_name in one of these paths,
we have emitted an earlier error so just return here.
Signed-off-by: Dave Airlie
---
src/comp
On 5 June 2016 at 17:56, Samuel Pitoiset wrote:
> We should not call nouveau_bufctx_reset() inside a validate function.
This seems to contradict the changes introduced in nvc0_compute.c.
Worth explaining a bit better the dos and don'ts ?
-Emil
___
mesa-
On 5 June 2016 at 22:13, Emil Velikov wrote:
> On 5 June 2016 at 17:56, Samuel Pitoiset wrote:
>> We should not call nouveau_bufctx_reset() inside a validate function.
> This seems to contradict the changes introduced in nvc0_compute.c.
> Worth explaining a bit better the dos and don'ts ?
>
As th
On Sun, Jun 5, 2016 at 5:16 PM, Emil Velikov wrote:
> On 5 June 2016 at 22:13, Emil Velikov wrote:
>> On 5 June 2016 at 17:56, Samuel Pitoiset wrote:
>>> We should not call nouveau_bufctx_reset() inside a validate function.
>> This seems to contradict the changes introduced in nvc0_compute.c.
>>
On 06/05/2016 11:17 PM, Ilia Mirkin wrote:
On Sun, Jun 5, 2016 at 5:16 PM, Emil Velikov wrote:
On 5 June 2016 at 22:13, Emil Velikov wrote:
On 5 June 2016 at 17:56, Samuel Pitoiset wrote:
We should not call nouveau_bufctx_reset() inside a validate function.
This seems to contradict the c
On 5 June 2016 at 22:17, Ilia Mirkin wrote:
> On Sun, Jun 5, 2016 at 5:16 PM, Emil Velikov wrote:
>> On 5 June 2016 at 22:13, Emil Velikov wrote:
>>> On 5 June 2016 at 17:56, Samuel Pitoiset wrote:
We should not call nouveau_bufctx_reset() inside a validate function.
>>> This seems to cont
On Sun, Jun 5, 2016 at 5:34 PM, Emil Velikov wrote:
> On 5 June 2016 at 22:17, Ilia Mirkin wrote:
>> On Sun, Jun 5, 2016 at 5:16 PM, Emil Velikov
>> wrote:
>>> On 5 June 2016 at 22:13, Emil Velikov wrote:
On 5 June 2016 at 17:56, Samuel Pitoiset wrote:
> We should not call nouveau_bu
On 5 June 2016 at 22:36, Ilia Mirkin wrote:
> On Sun, Jun 5, 2016 at 5:34 PM, Emil Velikov wrote:
>> On 5 June 2016 at 22:17, Ilia Mirkin wrote:
>>> On Sun, Jun 5, 2016 at 5:16 PM, Emil Velikov
>>> wrote:
On 5 June 2016 at 22:13, Emil Velikov wrote:
> On 5 June 2016 at 17:56, Samuel
On 06/05/2016 11:50 PM, Emil Velikov wrote:
On 5 June 2016 at 22:36, Ilia Mirkin wrote:
On Sun, Jun 5, 2016 at 5:34 PM, Emil Velikov wrote:
On 5 June 2016 at 22:17, Ilia Mirkin wrote:
On Sun, Jun 5, 2016 at 5:16 PM, Emil Velikov wrote:
On 5 June 2016 at 22:13, Emil Velikov wrote:
On 5
On 5 June 2016 at 23:00, Samuel Pitoiset wrote:
>
>
> On 06/05/2016 11:50 PM, Emil Velikov wrote:
>>
>> On 5 June 2016 at 22:36, Ilia Mirkin wrote:
>>>
>>> On Sun, Jun 5, 2016 at 5:34 PM, Emil Velikov
>>> wrote:
On 5 June 2016 at 22:17, Ilia Mirkin wrote:
>
> On Sun, Jun 5, 20
On 06/06/2016 12:08 AM, Emil Velikov wrote:
On 5 June 2016 at 23:00, Samuel Pitoiset wrote:
On 06/05/2016 11:50 PM, Emil Velikov wrote:
On 5 June 2016 at 22:36, Ilia Mirkin wrote:
On Sun, Jun 5, 2016 at 5:34 PM, Emil Velikov
wrote:
On 5 June 2016 at 22:17, Ilia Mirkin wrote:
On S
On Sun, Jun 5, 2016 at 6:08 PM, Emil Velikov wrote:
> On 5 June 2016 at 23:00, Samuel Pitoiset wrote:
>>
>>
>> On 06/05/2016 11:50 PM, Emil Velikov wrote:
>>>
>>> On 5 June 2016 at 22:36, Ilia Mirkin wrote:
On Sun, Jun 5, 2016 at 5:34 PM, Emil Velikov
wrote:
>
> On 5 June
This fixes a problem with the CE preamble and restoring only stuff in the
preamble when needed.
To illustrate suppose we have two graphics IB's 1 and 2, which are submitted in
that order. Furthermore suppose IB 1 does not use CE ram, but IB 2 does, and we
have a context switch at the start of IB
On 5 June 2016 at 23:22, Ilia Mirkin wrote:
> On Sun, Jun 5, 2016 at 6:08 PM, Emil Velikov wrote:
>> On 5 June 2016 at 23:00, Samuel Pitoiset wrote:
>>>
>>>
>>> On 06/05/2016 11:50 PM, Emil Velikov wrote:
On 5 June 2016 at 22:36, Ilia Mirkin wrote:
>
> On Sun, Jun 5, 2016 at 5
On 06/06/2016 12:30 AM, Emil Velikov wrote:
On 5 June 2016 at 23:22, Ilia Mirkin wrote:
On Sun, Jun 5, 2016 at 6:08 PM, Emil Velikov wrote:
On 5 June 2016 at 23:00, Samuel Pitoiset wrote:
On 06/05/2016 11:50 PM, Emil Velikov wrote:
On 5 June 2016 at 22:36, Ilia Mirkin wrote:
On Sun
Change MESA into Mesa in CL_PLATFORM_VERSION and CL_DEVICE_VERSION. For
both, always append git version suffix from git_sha1.h.
v4: dropped #ifdef guards.
It would be very nice if this could be considered for 12.0 release.
---
src/gallium/state_trackers/clover/api/device.cpp | 5 -
src/gal
On 06/04/2016 10:13 AM, Francisco Jerez wrote:
Vedran Miletić writes:
OpenCL apps can quote arguments they pass to the OpenCL compiler, most
commonly include paths containing spaces. If the OpenCL compiler was
called via a shell, the shell would remove (single or double) quotes
before passing
ARB_shader_image_load_store only requires a very fixed list of formats
to be supported, while textures may be in all kinds of formats, like
BGRA which are presently not supported on at least Kepler.
Signed-off-by: Ilia Mirkin
---
src/mesa/state_tracker/st_cb_readpixels.c | 6 ++
1 file chang
On 06/04/2016 04:18 AM, Francisco Jerez wrote:
Serge Martin writes:
From: Vedran Miletić
Make sure that a struct argument did not get compiled into a pointer
type with the byval attribute. If we try to handle the pointer with
byval, we end up with the pointer size instead of the struct size.
This was disabled due to occasionally incorrect behavior when trying to
upload data. It later became apparent that nvc0 also had a similar but
slightly different issue, which was resolved in commit e50c01d5. This
takes the same logic as nvc0 and applies it to nv50 (which has somewhat
different inte
On 06/04/2016 04:27 AM, Francisco Jerez wrote:
Serge Martin writes:
This fix getting the size of a struct arg. vec3 types still work ok.
Only buit-in args need to have power of two alignment, getTypeAllocSize
reports the correct size.
Is there any guarantee that the alloc size of the type wi
On Sun, Jun 5, 2016 at 6:30 PM, Emil Velikov wrote:
> On 5 June 2016 at 23:22, Ilia Mirkin wrote:
>> You're arguing about clarity with the ~2 active developers/reviewers
>> who understand the code.
> What I'm saying is that with contradicting commit messages like this
> one there'll be little oth
On 06/06/2016 12:44 AM, Vedran Miletić wrote:
Change MESA into Mesa in CL_PLATFORM_VERSION and CL_DEVICE_VERSION. For
both, always append git version suffix from git_sha1.h.
v4: dropped #ifdef guards.
It would be very nice if this could be considered for 12.0 release.
---
src/gallium/state_tra
Build mesa 1517 completed
Commit f657a59d98 by Kenneth Graunke on 6/5/2016 11:31 PM:
mesa: Try to unbreak the MSVC build.\n\nPATH_MAX is apparently not a thing on Windows. Borrow the hack from\npipe_loader.c to try and make this work.\n\nSigned-off-by: Kenneth
Vedran Miletić writes:
> On 06/04/2016 10:13 AM, Francisco Jerez wrote:
>> Vedran Miletić writes:
>>
>>> OpenCL apps can quote arguments they pass to the OpenCL compiler, most
>>> commonly include paths containing spaces. If the OpenCL compiler was
>>> called via a shell, the shell would remove
Vedran Miletić writes:
> On 06/04/2016 04:27 AM, Francisco Jerez wrote:
>> Serge Martin writes:
>>
>>> This fix getting the size of a struct arg. vec3 types still work ok.
>>> Only buit-in args need to have power of two alignment, getTypeAllocSize
>>> reports the correct size.
>>>
>> Is there an
Vedran Miletić writes:
> Change MESA into Mesa in CL_PLATFORM_VERSION and CL_DEVICE_VERSION. For
> both, always append git version suffix from git_sha1.h.
>
> v4: dropped #ifdef guards.
>
> It would be very nice if this could be considered for 12.0 release.
> ---
> src/gallium/state_trackers/clo
On Sat, Jun 4, 2016 at 3:39 PM, Jason Ekstrand wrote:
> Signed-off-by: Jason Ekstrand
> Cc: Kristian Høgsberg Kristensen
Thanks for adding this, looks good to me. The copying of prog_data in
anv_pipeline_cache_load() is a little fiddly, but I got nothing
better...
Reviewed-by: Kristian Høgsber
From: Dave Airlie
With tessellation shaders we can have cases where we have
arrays of anon structs, so make sure we match using without_array().
Fixes:
GL45-CTS.tessellation_shader.tessellation_control_to_tessellation_evaluation.gl_in
Signed-off-by: Dave Airlie
---
src/compiler/glsl/link_vary
From: Dave Airlie
With tessellation shaders we can have cases where we have
arrays of anon structs, so make sure we match using without_array().
Fixes:
GL45-CTS.tessellation_shader.tessellation_control_to_tessellation_evaluation.gl_in
v2:
test lengths match as well (Ilia)
Signed-off-by: Dave A
From: Dave Airlie
With tessellation shaders we can have cases where we have
arrays of anon structs, so make sure we match using without_array().
Fixes:
GL45-CTS.tessellation_shader.tessellation_control_to_tessellation_evaluation.gl_in
v2:
test lengths match as well (Ilia)
v3:
descend array leng
On Sun, Jun 5, 2016 at 9:30 PM, Dave Airlie wrote:
> From: Dave Airlie
>
> With tessellation shaders we can have cases where we have
> arrays of anon structs, so make sure we match using without_array().
>
> Fixes:
> GL45-CTS.tessellation_shader.tessellation_control_to_tessellation_evaluation.gl_
ping
On Sun, May 29, 2016 at 2:01 PM, Ilia Mirkin wrote:
> Turns out Matt already did a bunch of this work last year, but since I
> only found out about that after I wrote my patches, I decided to keep
> mine in place (I like the "vote" naming better TBH). This passes the
> very simple piglit tes
Hi Kenneth,
Am 03.06.2016 um 10:08 schrieb Kenneth Graunke:
> Our previous code worked for desktop GL, and ES without geometry or
> tessellation shaders. But those features require fancier point size
> handling. Fortunately, we can use one rule for all APIs.
>
> Fixes a number of dEQP tests wit
On 6 June 2016 at 11:47, Ilia Mirkin wrote:
> ping
>
> On Sun, May 29, 2016 at 2:01 PM, Ilia Mirkin wrote:
>> Turns out Matt already did a bunch of this work last year, but since I
>> only found out about that after I wrote my patches, I decided to keep
>> mine in place (I like the "vote" naming
On Friday, June 3, 2016 4:40:46 PM PDT Ilia Mirkin wrote:
> On Fri, Jun 3, 2016 at 4:08 AM, Kenneth Graunke wrote:
> > Our previous code worked for desktop GL, and ES without geometry or
> > tessellation shaders. But those features require fancier point size
> > handling. Fortunately, we can use
From: Dave Airlie
Although the glsl_types.h stores this in a bitfield,
we should hide that from everyone else. Hide the cast
in an accessor method and use the enum everywhere.
This makes things a bit nicer in gdb, and improves type
safety.
v2: fix a few pieces of interface I missed that caused
With the const's gone on the return types of the new helpers, this is
Reviewed-by: Ilia Mirkin
On Mon, Jun 6, 2016 at 12:21 AM, Dave Airlie wrote:
> From: Dave Airlie
>
> Although the glsl_types.h stores this in a bitfield,
> we should hide that from everyone else. Hide the cast
> in an access
On 03/06/16 23:04, Dave Airlie wrote:
> On 4 June 2016 at 03:39, Alejandro Piñeiro wrote:
>> On 03/06/16 02:46, Dave Airlie wrote:
>>> From: Dave Airlie
>>>
>>> "all geometry shader output vertex count declarations in a
>>> program must declare the same count."
>> This spec quote lacks context. A
Hi,
Before reporting a possible invalid bug report. Does anyone knows why
radeaonfb is not configured the same way radeon is ? For instance on a
PowerPC machine, when Open Firmware Frame Buffer is used (OFfb), I
cannot `modprobe radeonfb` (but I can load `radeon`). It fails with:
[ 96.551486] r
61 matches
Mail list logo
