I am working on a new version of this patch. I have one version which does
away with all the bitfields, and uses
functions to update the index.
Another approach would be to change the union to a struct, and use a
function to get the index.
Yet another approach would be to keep the contents of the
On Mon, Apr 9, 2018 at 4:26 AM, Michel Dänzer wrote:
> On 2018-04-06 07:14 PM, Marek Olšák wrote:
> > On Fri, Apr 6, 2018 at 11:41 AM, Michel Dänzer
> wrote:
> >
> >> On 2018-04-06 03:25 PM, Marek Olšák wrote:
> >>> On Thu, Apr 5, 2018, 3:09 AM Michel
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeonsi/si_get.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/gallium/drivers/radeonsi/si_get.c
b/src/gallium/drivers/radeonsi/si_get.c
index 4483ca766d..c8aacfe182 100644
---
Signed-off-by: Leo Liu
---
src/gallium/state_trackers/va/config.c | 2 +-
src/gallium/state_trackers/va/va_private.h | 4
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/gallium/state_trackers/va/config.c
b/src/gallium/state_trackers/va/config.c
index
On 9 April 2018 at 15:58, Juan A. Suarez Romero wrote:
> Fixes: 042ee4bea2 "(spirv: Move SPIR-V building to Makefile.spirv.am and
> spirv/meson.build")
This and the radeonsi patch are
Reviewed-by: Emil Velikov
-Emil
On Mon, Apr 9, 2018 at 10:25 AM, Jason Ekstrand wrote:
> On Mon, Apr 9, 2018 at 5:35 AM, Rob Clark wrote:
>>
>> On Mon, Apr 9, 2018 at 1:35 AM, Jason Ekstrand
>> wrote:
>> > Rather lively discussion we've got going here...
>> >
>>
For the series:
Reviewed-by: Samuel Pitoiset
On 04/09/2018 04:56 PM, Bas Nieuwenhuizen wrote:
The packet can sometimes be skipped, but we still think the change takes effect.
This just makes the packet always take effect.
Fixes: ad11fc3571 "radv: don't emit
https://bugs.freedesktop.org/show_bug.cgi?id=105960
Christian Gmeiner changed:
What|Removed |Added
Status|NEW
Fixes: 042ee4bea2 "(spirv: Move SPIR-V building to Makefile.spirv.am and
spirv/meson.build")
---
src/compiler/Makefile.spirv.am | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/compiler/Makefile.spirv.am b/src/compiler/Makefile.spirv.am
index a9f90c26e7b..2ac97f30acf
Signed-off-by: Leo Liu
---
src/gallium/auxiliary/Makefile.sources| 3 +-
src/gallium/auxiliary/meson.build | 1 +
src/gallium/auxiliary/vl/vl_probs_table.h | 585 ++
3 files changed, 588 insertions(+), 1 deletion(-)
create mode 100644
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.c | 52 +++--
src/gallium/drivers/radeon/radeon_vcn_dec.h | 3 ++
2 files changed, 37 insertions(+), 18 deletions(-)
diff --git a/src/gallium/drivers/radeon/radeon_vcn_dec.c
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.h | 134
1 file changed, 134 insertions(+)
diff --git a/src/gallium/drivers/radeon/radeon_vcn_dec.h
b/src/gallium/drivers/radeon/radeon_vcn_dec.h
index 0a8c343e58..3ae04a1087
Signed-off-by: Leo Liu
---
src/gallium/include/pipe/p_video_state.h | 94
1 file changed, 94 insertions(+)
diff --git a/src/gallium/include/pipe/p_video_state.h
b/src/gallium/include/pipe/p_video_state.h
index 2533ba4fab..05855a36e2 100644
---
The current FW has restricted the size to the worse case,
and the new dynamic dpb buffer support is on the way from
firmware side, we will change accordingly.
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.c | 6 ++
1 file changed, 6 insertions(+)
Signed-off-by: Leo Liu
---
src/gallium/auxiliary/util/u_video.h | 3 +++
src/gallium/include/pipe/p_video_enums.h | 6 --
2 files changed, 7 insertions(+), 2 deletions(-)
diff --git a/src/gallium/auxiliary/util/u_video.h
b/src/gallium/auxiliary/util/u_video.h
index
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.c | 3 +++
src/gallium/drivers/radeon/radeon_vcn_dec.h | 1 +
2 files changed, 4 insertions(+)
diff --git a/src/gallium/drivers/radeon/radeon_vcn_dec.c
b/src/gallium/drivers/radeon/radeon_vcn_dec.c
index
This series will enable VP9 support for profile0 and profile2 on VCN.
It will support players with VP9 VA-API enabled.
Leo Liu (22):
vl: add VP9 profile0 and format
vl: add VP9 picture description
radeon/vcn: add VP9 stream type for decoder
radeon/vcn: add VP9 dpb buffer size
vl: add
Okay, this needs more work, looks like almost everything got reset
somewhere else already, and num_instances wasn't but should not be
needed.
On Mon, Apr 9, 2018 at 4:37 PM, Bas Nieuwenhuizen
wrote:
> Fixes: ad11fc3571 "radv: don't emit unneeded vertex state."
>
https://bugs.freedesktop.org/show_bug.cgi?id=105960
--- Comment #2 from Emil Velikov ---
https://patchwork.freedesktop.org/patch/215989/
should fix it
--
You are receiving this mail because:
You are the QA Contact for the
Lionel Landwerlin writes:
> Signed-off-by: Lionel Landwerlin
> ---
> src/intel/tools/gen_batch_decoder.c | 22 +-
> 1 file changed, 13 insertions(+), 9 deletions(-)
>
> diff --git
Emil you are only minutes faster then me :) Will push this change in
some minutes.
2018-04-09 18:57 GMT+02:00 Emil Velikov :
> From: Emil Velikov
>
> Otherwise building the driver will fail with unresolved symbols.
>
> Bugzilla:
https://bugs.freedesktop.org/show_bug.cgi?id=105918
Jonas Platte changed:
What|Removed |Added
CC|
https://bugs.freedesktop.org/show_bug.cgi?id=105960
Bug ID: 105960
Summary: [bisected] meson build test fails with: undefined
reference to `etna_pm_create_query'
Product: Mesa
Version: git
Hardware: Other
From: Emil Velikov
Otherwise building the driver will fail with unresolved symbols.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105960
Fixes: 72d2043be06 ("etnaviv: add perfmon query implementation")
Cc: Christian Gmeiner
Cc:
Fixes: 94f726c36d ("radeonsi: move r600_cs.h contents into si_pipe.h,
si_build_pm4.h")
---
src/gallium/drivers/radeonsi/Makefile.sources | 1 +
1 file changed, 1 insertion(+)
diff --git a/src/gallium/drivers/radeonsi/Makefile.sources
b/src/gallium/drivers/radeonsi/Makefile.sources
index
Signed-off-by: Leo Liu
---
src/gallium/state_trackers/va/picture_vp9.c | 52 -
1 file changed, 51 insertions(+), 1 deletion(-)
diff --git a/src/gallium/state_trackers/va/picture_vp9.c
b/src/gallium/state_trackers/va/picture_vp9.c
index
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeonsi/si_get.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/src/gallium/drivers/radeonsi/si_get.c
b/src/gallium/drivers/radeonsi/si_get.c
index 761ca6f4cd..4483ca766d 100644
--- a/src/gallium/drivers/radeonsi/si_get.c
Signed-off-by: Leo Liu
---
src/gallium/state_trackers/va/Makefile.sources | 1 +
src/gallium/state_trackers/va/meson.build | 4 +--
src/gallium/state_trackers/va/picture.c| 11
src/gallium/state_trackers/va/picture_vp9.c| 38 ++
To get some of UVD required parameters.
Signed-off-by: Leo Liu
---
src/gallium/state_trackers/va/picture.c | 2 +-
src/gallium/state_trackers/va/picture_vp9.c | 237
src/gallium/state_trackers/va/va_private.h | 1 +
3 files changed, 239
Signed-off-by: Leo Liu
---
src/gallium/auxiliary/util/u_video.h | 1 +
src/gallium/include/pipe/p_video_enums.h | 3 ++-
2 files changed, 3 insertions(+), 1 deletion(-)
diff --git a/src/gallium/auxiliary/util/u_video.h
b/src/gallium/auxiliary/util/u_video.h
index
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.c | 16
1 file changed, 16 insertions(+)
diff --git a/src/gallium/drivers/radeon/radeon_vcn_dec.c
b/src/gallium/drivers/radeon/radeon_vcn_dec.c
index b4cfba1713..046b371384 100644
---
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.c | 176 +++-
src/gallium/drivers/radeon/radeon_vcn_dec.h | 1 +
2 files changed, 176 insertions(+), 1 deletion(-)
diff --git a/src/gallium/drivers/radeon/radeon_vcn_dec.c
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.c | 26 ++
1 file changed, 26 insertions(+)
diff --git a/src/gallium/drivers/radeon/radeon_vcn_dec.c
b/src/gallium/drivers/radeon/radeon_vcn_dec.c
index 170cc3fa41..b4cfba1713 100644
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeonsi/si_get.c | 4
1 file changed, 4 insertions(+)
diff --git a/src/gallium/drivers/radeonsi/si_get.c
b/src/gallium/drivers/radeonsi/si_get.c
index 85cfb11338..761ca6f4cd 100644
--- a/src/gallium/drivers/radeonsi/si_get.c
Signed-off-by: Leo Liu
---
src/gallium/state_trackers/va/context.c| 4
src/gallium/state_trackers/va/va_private.h | 1 +
2 files changed, 5 insertions(+)
diff --git a/src/gallium/state_trackers/va/context.c
b/src/gallium/state_trackers/va/context.c
index
Signed-off-by: Leo Liu
---
src/gallium/state_trackers/va/config.c | 2 +-
src/gallium/state_trackers/va/va_private.h | 4
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/gallium/state_trackers/va/config.c
b/src/gallium/state_trackers/va/config.c
index
Signed-off-by: Leo Liu
---
src/gallium/state_trackers/va/picture_vp9.c | 25 -
1 file changed, 24 insertions(+), 1 deletion(-)
diff --git a/src/gallium/state_trackers/va/picture_vp9.c
b/src/gallium/state_trackers/va/picture_vp9.c
index
Signed-off-by: Leo Liu
---
src/gallium/drivers/radeon/radeon_vcn_dec.c | 38 +
1 file changed, 38 insertions(+)
diff --git a/src/gallium/drivers/radeon/radeon_vcn_dec.c
b/src/gallium/drivers/radeon/radeon_vcn_dec.c
index b29ba37b3c..bcfdd33d31
https://bugs.freedesktop.org/show_bug.cgi?id=105464
Samuel Pitoiset changed:
What|Removed |Added
Resolution|--- |FIXED
https://bugs.freedesktop.org/show_bug.cgi?id=105918
--- Comment #2 from Jonas Platte ---
Played around a bit more and found out this only happens with compton. And it
seems to have been reported a while ago already:
compton bug:
On Mon, Apr 9, 2018 at 10:51 AM, Bas Vermeulen wrote:
> I am working on a new version of this patch. I have one version which does
> away with all the bitfields, and uses
> functions to update the index.
> Another approach would be to change the union to a struct, and use a
>
Hi Mark,
We need add a dmabuf export test case in piglit, unfortunately I don't have
time to work on it at the moment, it's on my to-do list.
Thanks,
James
-Original Message-
From: Janes, Mark A
Sent: Monday, April 9, 2018 11:17 AM
To: Xiong, James ;
I pushed both patches, thanks.
Marek
On Mon, Apr 9, 2018 at 7:06 AM, Bas Vermeulen wrote:
> The parameters for the compute engine are wrong when using
> an E8860 on a big endian machine.
> To fix this, convert the contents of struct dispatch_packet
> to little endian.
>
>
Have you tried using the meson build if you're that concerned about build times?
The plan remains to remove the autotools build at some point, so creating a
massive amount of churn in the autotools build that may or may not break
something to get some amount of speedup when you could just change
Quoting Emil Velikov (2018-04-09 11:02:45)
> From: Emil Velikov
>
> Dylan has kindly stepped up to help with 18.1.0, while I've taken the
> liberty to nominate Andres for 18.2.0 ;-)
>
> As always, people are welcome to swap/adjust where needed.
>
> Cc: Dylan Baker
Do you need break statements?
Marek
On Thu, Apr 5, 2018 at 1:34 AM, Timothy Arceri
wrote:
> V2: add missing intrinsics (Spotted-by: Samuel Pitoiset)
> ---
> src/gallium/drivers/radeonsi/si_shader_nir.c | 13 -
> 1 file changed, 12 insertions(+), 1
If you add break statements into patch 5, the series is:
Reviewed-by: Marek Olšák
Marek
On Thu, Apr 5, 2018 at 1:34 AM, Timothy Arceri
wrote:
> This makes it easier to follow the code, and also initialises
> dynamic_index which will be useful for
https://bugs.freedesktop.org/show_bug.cgi?id=105960
Clayton Craft changed:
What|Removed |Added
Resolution|--- |FIXED
meson fails to link LLVM on my setup, so I can't use it, therefore all my
meson changes are untested.
Even if meson worked, I have to use make, because that's what users use.
This change simplifies the meson build too.
Marek
On Mon, Apr 9, 2018 at 2:21 PM, Dylan Baker
For the series:
Reviewed-by: Marek Olšák
Marek
On Thu, Apr 5, 2018 at 3:37 AM, Timothy Arceri
wrote:
> We don't create variants of the NIR so here we finalise it before
> caching to avoid unnecessary processing when restoring it.
> ---
>
https://bugs.freedesktop.org/show_bug.cgi?id=105918
Timothy Arceri changed:
What|Removed |Added
Status|NEW |RESOLVED
Originally the "each" variable was just a part of the "drivers"
variable. It's not anymore so it's a bit ambiguous.
Signed-off-by: Dylan Baker
---
bin/install_megadrivers.py | 16
1 file changed, 8 insertions(+), 8 deletions(-)
diff --git
This fixes -Ddri-drivers-path, -Dvdpau-libs-path, etc. with DESTDIR when
those paths are absolute. Currently due to the way python's os.path.join
handles absolute paths these will ignore DESTDIR, which is bad. This
fixes them to be relative to DESTDIR if that is set.
Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=105567
--- Comment #10 from Dylan Baker ---
My suspicion was correct, we're not handling absolute paths and DESTDIR
correctly. I think that this patch should resolve it:
https://patchwork.freedesktop.org/patch/216020/
--
You
Quoting Marek Olšák (2018-04-09 13:44:27)
> meson fails to link LLVM on my setup, so I can't use it, therefore all my
> meson
> changes are untested.
>
> Even if meson worked, I have to use make, because that's what users use.
>
> This change simplifies the meson build too.
>
> Marek
>
What
This removes the emission of unnecessary states, for example
when performing a fast depth stencil clear (ie. clearing htile),
we don't need to save/restore the graphics pipeline.
For GFX9 chips that have the scissor bug workaround, that
should also reduce the number of partial flushes.
Am Montag, den 09.04.2018, 14:03 -0400 schrieb Marek Olšák:
> On Mon, Apr 9, 2018 at 10:51 AM, Bas Vermeulen
> wrote:
Which solution is better depends on what is done more often: reading
the index or writing to the bit fields.
> > I am working on a new version of this patch.
On Mon, Apr 9, 2018 at 5:19 PM, Gert Wollny wrote:
> Am Montag, den 09.04.2018, 14:03 -0400 schrieb Marek Olšák:
> > On Mon, Apr 9, 2018 at 10:51 AM, Bas Vermeulen
> > wrote:
> Which solution is better depends on what is done more often: reading
> the
See:
https://cgit.freedesktop.org/mesa/mesa/commit/?id=f55d1f806e6b6c33af559de166d08ec8fa3ebe90
Marek
On Mon, Apr 9, 2018 at 5:08 PM, Dylan Baker wrote:
> Quoting Marek Olšák (2018-04-09 13:44:27)
> > meson fails to link LLVM on my setup, so I can't use it, therefore all
>
Are you building LLVM yourself, or is that a build that comes with your distro?
Also, what is your distro?
Quoting Marek Olšák (2018-04-09 14:27:10)
> See:
> https://cgit.freedesktop.org/mesa/mesa/commit/?id=
> f55d1f806e6b6c33af559de166d08ec8fa3ebe90
>
> Marek
>
> On Mon, Apr 9, 2018 at 5:08
On Mon, Apr 9, 2018 at 1:35 AM, Jason Ekstrand wrote:
> Rather lively discussion we've got going here...
>
> On Sun, Apr 8, 2018 at 3:23 PM, Rob Clark wrote:
>>
>> On Sun, Apr 8, 2018 at 5:54 PM, Bas Nieuwenhuizen
>> wrote:
>>
On Wed, Mar 7, 2018 at 11:24 PM, Keith Packard wrote:
>
> +/*
> + * Implement vkGetPhysicalDeviceDisplayPropertiesKHR (VK_KHR_display)
> + */
> +VkResult
> +wsi_display_get_physical_device_display_properties(VkPhysicalDevice
>physical_device,
> +
From: "Xiong, James"
On non-LLC platforms, we malloc shadow batch/state buffers
of the same sizes as our batch/state buffers' GEM allocations.
However the buffer allocator reuses similar-sized gem objects,
it returns a buffer larger than we asked for in some cases
and we
Hi Marek,
On Saturday, 7 April 2018 01:53:58 CEST Marek Olšák wrote:
> So interleaved attribs are unsupported, right?
>
> is_interleaved_arrays was probably slowing things down, so I'm OK with that.
I am currently away from all the source code and be back at about the 22.4.
But out of my head:
Sorry for the multitude of replies. :-(
On Wed, Mar 7, 2018 at 11:24 PM, Keith Packard wrote:
> This adds support for the KHR_display extension support to the vulkan
> WSI layer. Driver support will be added separately.
>
> v2:
> * fix double ;; in
Hi,
> typedef struct {
> - nir_parameter_type param_type;
> - const struct glsl_type *type;
> + uint8_t num_components;
> + uint8_t bit_size;
> } nir_parameter;
(...)
> @@ -683,18 +692,12 @@ validate_tex_instr(nir_tex_instr *instr, validate_state
> *state)
> static void
>
On Wed, Mar 7, 2018 at 11:25 PM, Keith Packard wrote:
> This adds support for the KHR_display extension to the anv Vulkan
> driver. The driver now attempts to open the master DRM node when the
> KHR_display extension is requested so that the common winsys code can
> perform
On Tue, Apr 10, 2018 at 12:37 AM, Rob Clark wrote:
> On Mon, Apr 9, 2018 at 1:35 AM, Jason Ekstrand wrote:
>> Rather lively discussion we've got going here...
>>
>> On Sun, Apr 8, 2018 at 3:23 PM, Rob Clark wrote:
>>>
>>> On Sun,
On Wed, Mar 7, 2018 at 11:25 PM, Keith Packard wrote:
> Add support for the EXT_direct_mode_display extension. This just
> provides the vkReleaseDisplayEXT function.
>
> Signed-off-by: Keith Packard
> ---
> src/vulkan/wsi/wsi_common_display.c | 17
On Wed, Mar 7, 2018 at 11:25 PM, Keith Packard wrote:
> Add support for the EXT_direct_mode_display extension. This just
> provides the vkReleaseDisplayEXT function.
>
> Signed-off-by: Keith Packard
> ---
> src/intel/vulkan/anv_extensions.py | 1 +
>
Hi,
Given the fixes you already made based on my comments. Patches 1-20,
22-27, 29-43, and 61 (multiview!) are
Reviewed-by: Caio Marcelo de Oliveira Filho
Patches 46-47 and 49 seem to be valid regardless the rest of the code,
so I'd consider getting them in
On 09/04/18 17:04, Scott D Phillips wrote:
Lionel Landwerlin writes:
Signed-off-by: Lionel Landwerlin
---
src/intel/tools/gen_batch_decoder.c | 22 +-
1 file changed, 13 insertions(+), 9 deletions(-)
diff
Hi,
> >> Question: nir_deref_instr_get_variable will walk the deref instr
> >> chain, but does it even make sense if there's a array or struct in
> >> this deref chain? Should this be asserted?
> >>
> >
> > That's an interesting question. Certainly, at this point in the patch
> > series, we
unsigneds are needed by ARB_bindless_texture 64 bit vertex attribs, both for
NV_vertex_attrib_integer64.
Fixes the new piglit sampler-vertex-attrib-input-output test I sent some days
ago for bindless_texture.
The change inside vbo_attrtype_to_double_flag is what I am most concerned
about. Maybe
On Mon, Apr 9, 2018 at 8:39 PM, Karol Herbst wrote:
> unsigneds are needed by ARB_bindless_texture 64 bit vertex attribs, both for
> NV_vertex_attrib_integer64.
>
> Fixes the new piglit sampler-vertex-attrib-input-output test I sent some days
> ago for bindless_texture.
>
>
On Mon, Apr 9, 2018 at 1:35 AM, Jonathan Gray wrote:
> What happened with this patch? It seems the problem it is fixing got
> cherry-picked into 17.3 but the fix for master (and now 17.3) is still
> not merged?
I think Eric's on holiday now, so I wouldn't expect a speedy
On 10/04/18 06:29, Marek Olšák wrote:
Do you need break statements?
Whoops the first one does. I'll add a /* fall through */ comment to the
second one. Thanks.
Marek
On Thu, Apr 5, 2018 at 1:34 AM, Timothy Arceri > wrote:
V2:
On Tue, Apr 10, 2018 at 2:43 AM, Ilia Mirkin wrote:
> On Mon, Apr 9, 2018 at 8:39 PM, Karol Herbst wrote:
>> unsigneds are needed by ARB_bindless_texture 64 bit vertex attribs, both for
>> NV_vertex_attrib_integer64.
>>
>> Fixes the new piglit
As with patch 1, I've gone through and made a pile of style changes to
bring things back under 80 characters. You can find it on this branch:
https://gitlab.freedesktop.org/jekstrand/mesa/commits/wip/drm-lease-v3-whitespace
I also fixed a couple of the comments below.
On Wed, Mar 7, 2018 at
On Mon, Apr 9, 2018 at 4:58 PM, Caio Marcelo de Oliveira Filho <
caio.olive...@intel.com> wrote:
> Hi,
>
> Given the fixes you already made based on my comments. Patches 1-20,
> 22-27, 29-43, and 61 (multiview!) are
>
> Reviewed-by: Caio Marcelo de Oliveira Filho
>
>
On Mon, Apr 9, 2018 at 4:41 PM, Bas Nieuwenhuizen
wrote:
> On Tue, Apr 10, 2018 at 12:37 AM, Rob Clark wrote:
> > On Mon, Apr 9, 2018 at 1:35 AM, Jason Ekstrand
> wrote:
> >> Rather lively discussion we've got going here...
>
For nv50 we coalesce the srcs and defs into a single node (some ops
can't take a separate src and dst, while others can only handle one reg
in the short encoding mode).
[Side-note: the ops for which there is an encoding that can take a src
should instead use the RA's register preference
On Mon, Apr 9, 2018 at 10:23 PM, Ilia Mirkin wrote:
> For nv50 we coalesce the srcs and defs into a single node (some ops
> can't take a separate src and dst, while others can only handle one reg
> in the short encoding mode).
>
> [Side-note: the ops for which there is an
+ A bunch of potentially interested parties.
On Mon, Apr 9, 2018 at 4:25 PM, Caio Marcelo de Oliveira Filho <
caio.olive...@intel.com> wrote:
> Hi,
>
> > typedef struct {
> > - nir_parameter_type param_type;
> > - const struct glsl_type *type;
> > + uint8_t num_components;
> > + uint8_t
On Mon, Apr 9, 2018 at 6:20 PM, Jason Ekstrand wrote:
> On Mon, Apr 9, 2018 at 4:58 PM, Caio Marcelo de Oliveira Filho <
> caio.olive...@intel.com> wrote:
>
>> Hi,
>>
>> Given the fixes you already made based on my comments. Patches 1-20,
>> 22-27, 29-43, and 61
On Mon, Apr 9, 2018 at 5:21 PM, Caio Marcelo de Oliveira Filho <
caio.olive...@intel.com> wrote:
> Hi,
>
> > >> Question: nir_deref_instr_get_variable will walk the deref instr
> > >> chain, but does it even make sense if there's a array or struct in
> > >> this deref chain? Should this be
For nv50 we coalesce the srcs and defs into a single node. As such, we
can end up with impossible constraints if the source is referenced
after the tex operation (which, due to the coalescing of values, will
have overwritten it).
This logic already exists for inserting moves for MERGE/UNION
---
src/compiler/glsl/ast_to_hir.cpp | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/src/compiler/glsl/ast_to_hir.cpp b/src/compiler/glsl/ast_to_hir.cpp
index 168ab7eec2f..4d7383c580f 100644
--- a/src/compiler/glsl/ast_to_hir.cpp
+++ b/src/compiler/glsl/ast_to_hir.cpp
@@
This series is part of an effort to reduce the regression in compile
times when switching radeonsi from TGIS -> NIR. But it also turns
out to be quite handy for i965 too.
The idea is to make better use of GLSLOptimizeConservatively.
Currently TGSI must ignore the flag until all functions have
---
src/compiler/glsl/ir_optimization.h | 2 ++
1 file changed, 2 insertions(+)
diff --git a/src/compiler/glsl/ir_optimization.h
b/src/compiler/glsl/ir_optimization.h
index 81049a479e8..f5d2bea0cd5 100644
--- a/src/compiler/glsl/ir_optimization.h
+++ b/src/compiler/glsl/ir_optimization.h
@@
Otherwise we can end up with IR that looks like this:
(
(declare (temporary ) vec4 f@8)
(assign (xyzw) (var_ref f@8) (var_ref f) )
(call f16 ((swiz y (var_ref f@8) )))
(assign (xyzw) (var_ref f) (var_ref f@8) )
))
When we really need:
(declare
NIR validation will failing without this once we start using
glsl_to_nir() for functions other than main. i.e. once we stop
lowering all functions in GLSL IR.
---
src/compiler/glsl/glsl_to_nir.cpp | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git
We now call this for all drivers in glsl_to_nir() instead.
---
src/mesa/drivers/dri/i965/brw_program.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/src/mesa/drivers/dri/i965/brw_program.c
b/src/mesa/drivers/dri/i965/brw_program.c
index fc77926d6e0..8c3ac70280f 100644
---
Shader-db compile times on IVB:
With this patch:
Thread 1 took 368.68 seconds and compiled 16434 shaders
Thread 2 took 372.83 seconds and compiled 16930 shaders
Thread 3 took 370.27 seconds and compiled 16891 shaders
Thread 0 took 377.69 seconds and compiled 16753 shaders
Without this patch:
Fixes new piglit test:
tests/spec/glsl-1.20/execution/qualifiers/vs-out-conversion-int-to-float-vec4-index.shader_test
---
src/compiler/glsl/ast_function.cpp | 94 ++
1 file changed, 94 insertions(+)
diff --git a/src/compiler/glsl/ast_function.cpp
This support was missing because currently we always let GLSL IR
inline the functions.
---
src/compiler/glsl/glsl_to_nir.cpp | 76 +++
1 file changed, 68 insertions(+), 8 deletions(-)
diff --git a/src/compiler/glsl/glsl_to_nir.cpp
b/src/compiler/glsl/glsl_to_nir.cpp
Otherwise we create unused conditional return flags and things
get unnecessarily ugly fast when lowering nested functions.
---
src/compiler/nir/nir_lower_returns.c | 8
1 file changed, 8 insertions(+)
diff --git a/src/compiler/nir/nir_lower_returns.c
---
src/compiler/glsl/glsl_to_nir.cpp | 20
1 file changed, 20 insertions(+)
diff --git a/src/compiler/glsl/glsl_to_nir.cpp
b/src/compiler/glsl/glsl_to_nir.cpp
index 5a36963607e..55c01024669 100644
--- a/src/compiler/glsl/glsl_to_nir.cpp
+++
Reviewed-by: Marek Olšák
Marek
On Mon, Apr 9, 2018 at 10:57 AM, Juan A. Suarez Romero
wrote:
> Fixes: 94f726c36d ("radeonsi: move r600_cs.h contents into si_pipe.h,
> si_build_pm4.h")
> ---
> src/gallium/drivers/radeonsi/Makefile.sources | 1 +
> 1
From: Emil Velikov
Dylan has kindly stepped up to help with 18.1.0, while I've taken the
liberty to nominate Andres for 18.2.0 ;-)
As always, people are welcome to swap/adjust where needed.
Cc: Dylan Baker
Cc: Andres Gomez
1 - 100 of 160 matches
Mail list logo