Hi Yogesh:
http://oorja.iind.intel.com/mediawiki/index.php/Flatland
can you also try the flatland on this page.
For AOSP flatland, yes, the EGL patch may solve the issue.
However, I met one case that the batch buffer is empty just at the swapbuffer
(glfush is just called before that), then
Reviewed-by: Tapani Pälli
On 08/03/2017 07:52 PM, Jason Ekstrand wrote:
The make_shareable function deletes the aux buffer and then whacks
aux_usage to ISL_AUX_USAGE_NONE but not unsetting supports_fast_clear.
Since we only look at supports_fast_clear to decide whether
This can be dropped. I'm running with egl patch alone and things seem fine.
Zhongmin, please comment if you don’t think so.
> -Original Message-
> From: mesa-dev [mailto:mesa-dev-boun...@lists.freedesktop.org] On Behalf
> Of Marathe, Yogesh
> Sent: Friday, August 4, 2017 9:18 PM
>
>
Tomasz,
> -Original Message-
> From: Tomasz Figa [mailto:tf...@chromium.org]
> Sent: Saturday, August 5, 2017 8:47 AM
>
> Hi Yogesh,
>
> On Sat, Aug 5, 2017 at 1:22 AM, Marathe, Yogesh
> wrote:
> >> -Original Message-
> >> From: Tomasz Figa
https://bugs.freedesktop.org/show_bug.cgi?id=101334
--- Comment #35 from John ---
Created attachment 133274
--> https://bugs.freedesktop.org/attachment.cgi?id=133274=edit
radv trace after 2nd patch
I've just tested with the hacky patch and the one from #28 and it
Previous behavior was inconsistent with other texture targets so this has been
fixed in OpenGL 4.6.
Fixes:
KHR-GL45.direct_state_access.textures_storage_errors
---
src/mesa/main/texstorage.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/mesa/main/texstorage.c
The OpenGL 4.6 specs have been updated so that GetTextureParameter*
with a texture object with an incompatible TEXTURE_TARGET should now
report INVALID_OPERATION instead of INVALID_ENUM.
Fixes:
KHR-GL45.direct_state_access.textures_parameter_errors
---
src/mesa/main/texparam.c | 2 +-
1 file
https://bugs.freedesktop.org/show_bug.cgi?id=101334
--- Comment #34 from Dave Airlie ---
Created attachment 133273
--> https://bugs.freedesktop.org/attachment.cgi?id=133273=edit
hacky patch?
does this patch make any difference? (on top of the one from the list).
--
For the series:
Tested-by: Dieter Nützel
on RX580
09/11 didn't apply clean (fixed by hand)
Dieter
Am 07.08.2017 00:20, schrieb Marek Olšák:
The last series didn't get any attention, so this is a full re-send.
These are tiny adjustements and cleanups, and the last one
The Deus Ex: Mankind Divided shaders go from spending ~20 seconds
in the GLSL IR compilers front-end down to ~18.5 seconds on a
Ryzen 1800X.
Tested by compiling once with shader-db then deleting the index file
from the shader cache and compiling again.
---
src/compiler/glsl/glcpp/glcpp-parse.y |
f81ede469910d fixed a problem with shaders including IR that was
owned by builtins. However the approach of cloning the whole
function each time we referenced it lead to a significant
reduction in the GLSL IR compiler performance.
Everything was already cloned when inlining the function, as
far
This function differs from ralloc_strcat() and ralloc_strncat()
in that it does not do any strlen() calls which can become
costly on large strings.
---
src/util/ralloc.c | 17 +
src/util/ralloc.h | 18 ++
2 files changed, 35 insertions(+)
diff --git
The cloning was introduced in f81ede469910d to fixed a problem with
shaders including IR that was owned by builtins.
However the approach of cloning the whole function each time we
reference a builtin lead to a significant reduction in the GLSL
IR compilers performance.
The previous patch fixes
I've sent all but the last 2 already but they haven't received much
feedback yet.
This series reduces compile times of the Deus Ex shaders on my
Ryzen 1800X from 2m27s -> 2m8s with a cold cache on radeonsi.
Note: The above times are from compiling the shaders with shader-db
on a single thread,
Also add a comment that this should only be used by the ir_reader
interface for testing purposes.
---
src/compiler/glsl/ir.cpp | 8 ++--
src/compiler/glsl/ir.h | 14 +++---
2 files changed, 9 insertions(+), 13 deletions(-)
diff --git a/src/compiler/glsl/ir.cpp
Extra validation is added to ir_validate to make sure this is
always updated to the correct numer of operands, as passes like
lower_instructions modify the instructions directly rather then
generating a new one.
---
src/compiler/glsl/glsl_to_nir.cpp | 4 +--
Reviewed-by: Thomas Helland
Tested-by: Dieter Nützel
---
src/compiler/glsl/ir.h | 5 -
1 file changed, 5 deletions(-)
diff --git a/src/compiler/glsl/ir.h b/src/compiler/glsl/ir.h
index 40f3338470..d425b97aca 100644
---
We are currently copying the name for each member dereference
but we can just share a single instance of the string provided
by the type.
This change also stops us recalculating the field index
repeatedly.
---
src/compiler/glsl/ast_array_index.cpp | 14 -
hello, I am currently pursuing my undergraduate degree in
Theoretical computer science.I came to know about The X.Org Endless
Vacation of Code(EVoC) while searching for the GSoc, since it is already
became too late to apply for GSoc , I found an interesting opportunity
while reading the
On Mon, 2017-07-31 at 23:15 +0100, Chris Wilson wrote:
> Quoting Chris Wilson (2017-07-31 22:51:25)
> > Luminance/Intensity when converted to RGB should be replicated to
> > fill
> > the RGB channels, but they differ on how the alpha channel is
> > filled, as
> > luminance is set to 1 (unless
This patch is still missing a review. Any takers?
Iago
On Fri, 2017-07-21 at 10:26 +0200, Iago Toral Quiroga wrote:
> Mesa will map user defined vertex input attributes to slots
> starting at VERT_ATTRIB_GENERIC0 which gives us room for only 16
> slots (up to GL_VERT_ATTRIB_MAX). This sufficient
Hi,
Yeah, I have quite a few patches that have been stuck
for a while, and many of them are also reviewed.
Just need to get around to get commit-access,
so that I don't have to bother everyone else.
I was hoping to get that done one of the coming days.
It's basically hanging on me at the moment
From: Marek Olšák
---
src/gallium/drivers/radeonsi/si_cp_dma.c | 80 +---
1 file changed, 63 insertions(+), 17 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_cp_dma.c
b/src/gallium/drivers/radeonsi/si_cp_dma.c
index 21202b3..91a6aff
From: Marek Olšák
Process most new SET packets in parallel with previous draw calls, then
flush caches and wait, start the draw, and do L2 prefetches last.
This decreases the [CP busy / SPI busy] ratio (verified with GRBM perf
counters). In other words, the time window when
From: Marek Olšák
---
src/gallium/drivers/radeon/r600_gpu_load.c| 6 +++---
src/gallium/drivers/radeon/r600_pipe_common.h | 2 +-
src/gallium/drivers/radeon/r600_query.c | 6 +++---
src/gallium/drivers/radeon/r600_query.h | 2 +-
4 files changed, 8
From: Marek Olšák
---
src/gallium/drivers/radeonsi/si_compute.c | 2 +-
src/gallium/drivers/radeonsi/si_descriptors.c | 24
src/gallium/drivers/radeonsi/si_pipe.h| 2 +-
src/gallium/drivers/radeonsi/si_state.h | 6 +++---
From: Marek Olšák
---
src/gallium/drivers/radeon/r600_pipe_common.h | 1 +
src/gallium/drivers/radeon/r600_query.c | 7 +++
src/gallium/drivers/radeon/r600_query.h | 1 +
src/gallium/drivers/radeonsi/si_state_draw.c | 18 +++---
4 files
From: Marek Olšák
---
src/gallium/drivers/radeonsi/si_cp_dma.c| 20 +---
src/gallium/drivers/radeonsi/si_state.h | 3 +++
src/gallium/drivers/radeonsi/si_state_shaders.c | 12 ++--
3 files changed, 18 insertions(+), 17 deletions(-)
From: Marek Olšák
---
src/gallium/drivers/radeonsi/si_state.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_state.c
b/src/gallium/drivers/radeonsi/si_state.c
index c9bb108..11dee49 100644
---
From: Marek Olšák
also set max_alignment on amdgpu.
---
src/gallium/drivers/radeon/r600_pipe_common.c | 16 ++--
src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c | 2 +-
2 files changed, 15 insertions(+), 3 deletions(-)
diff --git
From: Marek Olšák
so that we don't rely on si_pm4_state_enabled_and_changed, allowing us
to move prefetches after draw calls.
---
src/gallium/drivers/radeonsi/si_cp_dma.c| 16
src/gallium/drivers/radeonsi/si_descriptors.c | 3 +--
From: Marek Olšák
---
src/gallium/drivers/radeonsi/si_descriptors.c | 25 +
1 file changed, 17 insertions(+), 8 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_descriptors.c
b/src/gallium/drivers/radeonsi/si_descriptors.c
index
From: Marek Olšák
I'd like to be able to move the prefetch call site around.
---
src/gallium/drivers/radeonsi/si_cp_dma.c| 7 +++
src/gallium/drivers/radeonsi/si_descriptors.c | 2 +-
src/gallium/drivers/radeonsi/si_hw_context.c| 2 +-
The last series didn't get any attention, so this is a full re-send.
These are tiny adjustements and cleanups, and the last one is
an interesting micro-optimization.
Please review.
Thanks,
Marek
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
Reviewed-by: Bas Nieuwenhuizen
On Fri, Aug 4, 2017 at 7:55 AM, Dave Airlie wrote:
> From: Dave Airlie
>
> This mirrors what Marek has done for radeonsi, and uses
> a separate counter to handle the fmask surface for MSAA
> MRTs.
>
Reviewed-by: Bas Nieuwenhuizen
On Fri, Aug 4, 2017 at 7:46 AM, Dave Airlie wrote:
> From: Dave Airlie
>
> When Marek enabled mipmapped swizzle, radv didn't
> have the code in place to handle it. This fixes the
> regression.
>
>
For the series:
Tested-by: Dieter Nützel
on RX580
Dieter
Am 04.08.2017 09:25, schrieb Timothy Arceri:
---
src/compiler/glsl/ir.h | 5 -
1 file changed, 5 deletions(-)
diff --git a/src/compiler/glsl/ir.h b/src/compiler/glsl/ir.h
index 40f3338470..d425b97aca 100644
2017-07-28 16:05 GMT+02:00 Wladimir J. van der Laan :
> R8_UNORM textures can be emulated by means of L8 and a swizzle.
>
> Signed-off-by: Wladimir J. van der Laan
Reviewed-by: Christian Gmeiner
> ---
>
Great,
then I only can point to the 'prehash work' from Thomas Helland
'[Mesa-dev] [PATCH 0/6] Prehash all the things'
https://lists.freedesktop.org/archives/mesa-dev/2017-May/156564.html
with which I'm running during the last weeks.
No one commented other then me...
It has my 'Tested-by:'
On Thursday, August 3, 2017 9:52:37 AM PDT Jason Ekstrand wrote:
> The make_shareable function deletes the aux buffer and then whacks
> aux_usage to ISL_AUX_USAGE_NONE but not unsetting supports_fast_clear.
> Since we only look at supports_fast_clear to decide whether or not to do
> fast clears,
Hi Dieter,
Yes, all of my GLSL compilation speedup work landed.
I'm glad glibc got a better allocator.
Marek
On Sun, Aug 6, 2017 at 7:48 PM, Dieter Nützel wrote:
> Hello Marek,
>
> with latest glibc-2.26 results on Phoronix, I remember your GREAT GLSL
> compilation
Hello Marek,
with latest glibc-2.26 results on Phoronix, I remember your GREAT GLSL
compilation speedup work. Did anything of this landed? --- I think not?
(Can't find related commits.)
Some other ralloc related optimizations landed during October 2016...
ralloc: don't memset ralloc_header,
We already expose glMultiDrawElementsBaseVertexEXT as part of the
EXT_draw_elements_base_vertex chunk, so this one can just be removed.
Signed-off-by: Ilia Mirkin
---
src/mapi/glapi/gen/es_EXT.xml | 10 --
1 file changed, 10 deletions(-)
diff --git
There was a previous error in the gl.xml and generated files that
referenced glMultiDrawElementsBaseVertexOES. This function should not
exist, only the EXT-suffixed version should.
Leaving the other headers alone to avoid conflicts with GL 4.6 work.
Signed-off-by: Ilia Mirkin
https://bugs.freedesktop.org/show_bug.cgi?id=102057
Bug ID: 102057
Summary: Enabling DSA in COMPATIBILITY PROFILE
Product: Mesa
Version: git
Hardware: All
OS: All
Status: NEW
Severity: enhancement
45 matches
Mail list logo