On Wed, 2015-10-21 at 23:24 -0700, Jordan Justen wrote:
> On 2015-10-20 00:43:13, Iago Toral wrote:
> > On Tue, 2015-10-20 at 00:12 -0700, Jordan Justen wrote:
> > > An untyped surface read is volatile because it might be affected by a
> > > write.
> > >
> > > In the ES31-CTS.compute_shader.resour
On 22/10/15 08:29, Timothy Arceri wrote:
> On Wed, 2015-10-21 at 12:18 +0200, Samuel Iglesias Gonsalvez wrote:
>> Commit f24e5e did not take into account arrays of named shader
>> storage blocks.
>>
>> Fixes 20 dEQP-GLES31.functional.ssbo.* tests:
>>
>> dEQP
>> -GLES31.functional.ssbo.layout.sing
On 2015-10-22 00:06:37, Iago Toral wrote:
> On Wed, 2015-10-21 at 23:24 -0700, Jordan Justen wrote:
> > On 2015-10-20 00:43:13, Iago Toral wrote:
> > > On Tue, 2015-10-20 at 00:12 -0700, Jordan Justen wrote:
> > > > An untyped surface read is volatile because it might be affected by a
> > > > write
On Thu, 2015-10-22 at 09:56 +0300, Tapani Pälli wrote:
> On 10/22/2015 09:41 AM, Timothy Arceri wrote:
> > On Thu, 2015-10-22 at 08:55 +0300, Tapani Pälli wrote:
> > > On 10/22/2015 08:29 AM, Timothy Arceri wrote:
> > > > Location has never been able to be a negative value because it
> > > > has
>
On Wed, Oct 21, 2015 at 10:34 PM, Marek Olšák wrote:
> On Wed, Oct 21, 2015 at 12:28 PM, Axel Davy wrote:
>> The PIPE_BIND_SHARED flag should be added whenever
>> the resource may be shared with another process.
>>
>> In particular if the resource is imported, or may
>> be exported, the flag shou
On Wed, Oct 21, 2015 at 9:56 AM, Axel Davy wrote:
> On 21/10/2015 00:10, Bas Nieuwenhuizen wrote:
>>
>>
>> DCC is disabled for textures that can be shared as sharing the
>> DCC buffers has not been implemented yet.
>>
>>
>> + surf->dcc_enabled = !(surf->flags & RADEON_SURF_Z_OR_SBUFFER) &&
>>
On Thu, Oct 22, 2015 at 10:22 AM, Erik Faye-Lund wrote:
> On Wed, Oct 21, 2015 at 10:34 PM, Marek Olšák wrote:
>> On Wed, Oct 21, 2015 at 12:28 PM, Axel Davy wrote:
>>> The PIPE_BIND_SHARED flag should be added whenever
>>> the resource may be shared with another process.
>>>
>>> In particular i
On Thu, Oct 22, 2015 at 10:54 AM, Marek Olšák wrote:
> On Thu, Oct 22, 2015 at 10:22 AM, Erik Faye-Lund wrote:
>> On Wed, Oct 21, 2015 at 10:34 PM, Marek Olšák wrote:
>>> On Wed, Oct 21, 2015 at 12:28 PM, Axel Davy wrote:
The PIPE_BIND_SHARED flag should be added whenever
the resource
From ARB_program_interface_query:
"For the property of BUFFER_DATA_SIZE, then the implementation-dependent
minimum total buffer object size, in basic machine units, required to hold
all active variables associated with an active uniform block, shader
storage block, or atomic counter buffer is w
From ARB_program_query_interface spec:
"uint GetProgramResourceIndex(uint program, enum programInterface,
const char *name);
[...]
If exactly matches the name string of one of the active resources
for , the index of the matched resource is returned.
Addition
Previously this extension was only enabled when blitting between two
multisampled buffers. However I don't think it does any harm to just
enable it all the time. The ‘enable’ option is used instead of
‘require’ so that the shader will still compile if the extension isn't
available in the cases wher
Previously there was a problem in i965 where if 16x MSAA is used then
some of the sample positions are exactly on the 0 x or y axis. When
the MSAA copy blit shader interpolates the texture coordinates at
these sample positions it was possible that it would jump to a
neighboring texel due to roundin
On 21 October 2015 at 22:44, Matt Turner wrote:
> On Wed, Oct 21, 2015 at 2:16 PM, Emil Velikov
> wrote:
>> On 21 October 2015 at 21:33, Kenneth Graunke wrote:
>>> On Monday, October 19, 2015 02:54:56 PM Emil Velikov wrote:
Ping on these two trivial patches ?
-Emil
>>>
>>> Oh, so
On 20 October 2015 at 16:43, Nanley Chery wrote:
> On Tue, Oct 20, 2015 at 8:16 AM, Marek Olšák wrote:
>>
>> Also, the FIXME comment should be on its own line.
>>
>
> I moved it aside to make editing the table easier. However, since the
> formatting of the
> table is unlikely to change much after
On 21/10/15 17:35, Gustaw Smolarczyk wrote:
I am just a bystander, but I have one suggestion to this patch.
2015-10-21 18:25 GMT+02:00 Jose Fonseca :
This should prevent disparity between features Mesa and LLVM
believe are supported by the CPU.
http://lists.freedesktop.org/archives/mesa-dev/20
On Wed, Oct 21, 2015 at 12:10 AM, Bas Nieuwenhuizen
wrote:
> Uses the DCC buffer instead of the CMASK buffer. The ELIMINATE_FAST_CLEAR
> still works. Furthermore, with DCC compression we can directly clear
> to a limited set of colors such that we do not need a postprocessing step.
>
> Signed-off-
On 19 October 2015 at 23:36, Nanley Chery wrote:
> From: Nanley Chery
>
> With this infrastructure set in place, we can now reuse the entries to
> generate useful code.
>
> Signed-off-by: Nanley Chery
> ---
> src/mesa/main/extensions.c | 321
> +--
> s
On 19 October 2015 at 23:36, Nanley Chery wrote:
> From: Nanley Chery
>
> Enable limiting advertised extension support by context version with
> finer granularity. GLuint is chosen over smaller datatypes because,
> when this field is eventually used, usage of this datatype provides
> the smallest
On 19 October 2015 at 23:36, Nanley Chery wrote:
> From: Nanley Chery
>
> Create a function which determines if an extension is supported in the
> current context.
>
> Signed-off-by: Nanley Chery
> ---
> src/mesa/main/extensions.c | 17 +
> 1 file changed, 17 insertions(+)
>
> d
On Thu, Oct 22, 2015 at 10:56 AM, Erik Faye-Lund wrote:
> On Thu, Oct 22, 2015 at 10:54 AM, Marek Olšák wrote:
>> On Thu, Oct 22, 2015 at 10:22 AM, Erik Faye-Lund wrote:
>>> On Wed, Oct 21, 2015 at 10:34 PM, Marek Olšák wrote:
On Wed, Oct 21, 2015 at 12:28 PM, Axel Davy wrote:
> The P
On 22 October 2015 at 07:43, Erik Faye-Lund wrote:
> On Tue, Oct 20, 2015 at 12:44 AM, Nanley Chery wrote:
>> From: Nanley Chery
>>
>> Since the version numbers being compared are integral and we don't ever
>> expect gl_context::Version to be equal to 0, subtract 1 from the rhs of
>> the equatio
On Thu, Oct 22, 2015 at 12:18 PM, Marek Olšák wrote:
> On Thu, Oct 22, 2015 at 10:56 AM, Erik Faye-Lund wrote:
>> On Thu, Oct 22, 2015 at 10:54 AM, Marek Olšák wrote:
>>> On Thu, Oct 22, 2015 at 10:22 AM, Erik Faye-Lund
>>> wrote:
On Wed, Oct 21, 2015 at 10:34 PM, Marek Olšák wrote:
On 19 October 2015 at 23:44, Nanley Chery wrote:
> From: Nanley Chery
>
> Aside from those modified in this commit, all gl_extensions structs are
> zero-initialized by default. There is therefore no need to memset the
> structs to 0. Also, remove the open-coded memset in
> _mesa_init_extensions()
On 19 October 2015 at 23:44, Nanley Chery wrote:
> From: Nanley Chery
>
> This variable existed to provide an unsigned error value for
> name_to_offset(). Since o(extension_sentinel) is also a valid unsigned
> error value, save space and replace this mysterious variable with
> a less mysterious o
On 21 October 2015 at 23:58, Matt Turner wrote:
> Add some instructions: illegal, movi, sends, sendsc.
>
> Remove some instructions with reused opcodes: msave, mrestore, push,
> pop, goto. I did have some gross code for disassembling opcodes
> per-generation, but there's very little meaningful ove
On Thu, 2015-10-22 at 11:01 +0200, Samuel Iglesias Gonsalvez wrote:
> From ARB_program_query_interface spec:
>
> "uint GetProgramResourceIndex(uint program, enum programInterface,
>const char *name);
> [...]
> If exactly matches the name string of one of the
On 10/22/2015 12:01 PM, Samuel Iglesias Gonsalvez wrote:
From ARB_program_query_interface spec:
"uint GetProgramResourceIndex(uint program, enum programInterface,
const char *name);
[...]
If exactly matches the name string of one of the active resources
On 21 October 2015 at 23:58, Matt Turner wrote:
> Initially just checks that sources are non-NULL, which would have
> alerted us to the problem fixed by commit 6c846dc5.
What are the chances of combining this with and/or removing
fs_visitor::validate() ? Afaict both of these validations are at
di
On 22/10/15 13:08, Tapani Pälli wrote:
> On 10/22/2015 12:01 PM, Samuel Iglesias Gonsalvez wrote:
>> From ARB_program_query_interface spec:
>>
>> "uint GetProgramResourceIndex(uint program, enum programInterface,
>> const char *name);
>> [...]
>> If exact
In the next commit we are going to allow CSE of SSBO loads, so we need
to make sure that we consider how SSBO stores/atomics invalidate previous
loads.
---
src/glsl/nir/nir_opt_cse.c | 142 +
1 file changed, 142 insertions(+)
diff --git a/src/glsl/nir/n
On 22/10/15 13:06, Timothy Arceri wrote:
> On Thu, 2015-10-22 at 11:01 +0200, Samuel Iglesias Gonsalvez wrote:
>> From ARB_program_query_interface spec:
>>
>> "uint GetProgramResourceIndex(uint program, enum programInterface,
>>const char *name);
>> [...]
>>
I implemented this first as a separate optimization pass in GLSL IR [1], but
Curro pointed out that this being pretty much a restricted form of a CSE pass
it would probably make more sense to do it inside CSE (and we no longer have
a CSE pass in GLSL IR).
Unlike other things we CSE in NIR, in the
So we can effectively CSE the ones that are safe to reuse. Makes a shader
such as this:
buffer SSBO {
mat4 sm4;
};
uniform mat4 um4;
void main() {
sm4 *= um4;
}
go from 16 SSBO loads to only 4.
---
src/glsl/nir/nir_instr_set.c | 24
1 file changed, 20 insertion
The same clamping already happens for glTexParameteri. This change
also fixes a bug in mipmap generation, see
https://bugs.freedesktop.org/show_bug.cgi?id=91993
piglit test cases have been submitted for review (as additions to
arb_texture_storage-texture-storage and arb_texture_view-max-level).
--
Ben Widawsky writes:
> Gen9 adds the ability to write out a stencil value, so we need to expand the
> virtual payload by one. Abstracting this now makes that change easier to read.
>
> I was admittedly confused early on about some of the hardcoding. If people
> believe the resulting code is infer
On Thu, Oct 22, 2015 at 7:21 AM, Iago Toral Quiroga wrote:
> I implemented this first as a separate optimization pass in GLSL IR [1], but
> Curro pointed out that this being pretty much a restricted form of a CSE pass
> it would probably make more sense to do it inside CSE (and we no longer have
>
On Thursday 22 October 2015, Nicolai Hähnle wrote:
> The same clamping already happens for glTexParameteri. This change
> also fixes a bug in mipmap generation, see
> https://bugs.freedesktop.org/show_bug.cgi?id=91993
I don't think this patch is correct. The ARB_texture_view specification
doesn't
On 18.10.2015 00:57, Emil Velikov wrote:
Add a list of driver descriptors and select one from the list, during
probe time.
As we'll need to have all the driver pipe_foo_screen_create() functions
provided externally (i.e. from another static lib) we need a separate
(non-inline) drm_helper, which
When both fadd and fmul instructions have at least one operand that is a
constant and it is only used once, the total number of instructions can
be reduced from 3 (1 ffma + 2 load_const) to 2 (1 fmul + 1 fadd); because
the constants will be progagated as immediate operands of fmul and fadd.
This p
Because the next patch will add an optimization that is specific to i965,
we want to move this loweing pass to that driver altogether.
This is safe because i965 is the only consumer.
---
src/glsl/Makefile.sources | 1 -
src/glsl/nir/nir.h
Hello,
Last month I was working on an optimization for nir_opt_peephole_ffma, and I
sent a request for comments to the list [1]. Then I went on holidays and this
work got stalled.
During the last days I resumed it, and have been experimenting with some ideas
Matt Turner suggested back then, in
On Thu, 2015-10-22 at 09:39 -0400, Connor Abbott wrote:
> On Thu, Oct 22, 2015 at 7:21 AM, Iago Toral Quiroga wrote:
> > I implemented this first as a separate optimization pass in GLSL IR [1], but
> > Curro pointed out that this being pretty much a restricted form of a CSE
> > pass
> > it would
Connor Abbott writes:
> On Thu, Oct 22, 2015 at 7:21 AM, Iago Toral Quiroga wrote:
>> I implemented this first as a separate optimization pass in GLSL IR [1], but
>> Curro pointed out that this being pretty much a restricted form of a CSE pass
>> it would probably make more sense to do it inside
On Thu, 2015-10-22 at 16:38 +0200, Iago Toral wrote:
> On Thu, 2015-10-22 at 09:39 -0400, Connor Abbott wrote:
> > On Thu, Oct 22, 2015 at 7:21 AM, Iago Toral Quiroga
> > wrote:
> > > I implemented this first as a separate optimization pass in GLSL IR [1],
> > > but
> > > Curro pointed out that
On 22 October 2015 at 15:07, Nicolai Hähnle wrote:
> On 18.10.2015 00:57, Emil Velikov wrote:
>>
>> Add a list of driver descriptors and select one from the list, during
>> probe time.
>>
>> As we'll need to have all the driver pipe_foo_screen_create() functions
>> provided externally (i.e. from a
On Thu, Oct 22, 2015 at 10:37 AM, Francisco Jerez wrote:
> Connor Abbott writes:
>
>> On Thu, Oct 22, 2015 at 7:21 AM, Iago Toral Quiroga
>> wrote:
>>> I implemented this first as a separate optimization pass in GLSL IR [1], but
>>> Curro pointed out that this being pretty much a restricted for
On Thu, Oct 22, 2015 at 7:12 AM, Eduardo Lima Mitev wrote:
> When both fadd and fmul instructions have at least one operand that is a
> constant and it is only used once, the total number of instructions can
> be reduced from 3 (1 ffma + 2 load_const) to 2 (1 fmul + 1 fadd); because
> the constant
On 22 October 2015 at 16:32, Emil Velikov wrote:
> On 22 October 2015 at 15:07, Nicolai Hähnle wrote:
>> On 18.10.2015 00:57, Emil Velikov wrote:
>>>
>>> Add a list of driver descriptors and select one from the list, during
>>> probe time.
>>>
>>> As we'll need to have all the driver pipe_foo_scr
On 22.10.2015 15:57, Fredrik Höglund wrote:
On Thursday 22 October 2015, Nicolai Hähnle wrote:
The same clamping already happens for glTexParameteri. This change
also fixes a bug in mipmap generation, see
https://bugs.freedesktop.org/show_bug.cgi?id=91993
I don't think this patch is correct.
Hi,
Thx for your review. I'll submit a new version of the patch. Just replying
here first to answer your questions:
On 19 October 2015 at 18:10, Emil Velikov wrote:
> On 17 October 2015 at 00:14, Julien Isorce
> wrote:
> > Improve following functions to support VA_PROFILE_NONE profile (vpp):
>
On 22.10.2015 17:32, Emil Velikov wrote:
On 22 October 2015 at 15:07, Nicolai Hähnle wrote:
On 18.10.2015 00:57, Emil Velikov wrote:
Add a list of driver descriptors and select one from the list, during
probe time.
As we'll need to have all the driver pipe_foo_screen_create() functions
provi
On Thu, Oct 22, 2015 at 9:39 AM, Connor Abbott wrote:
> 2. Add a special "load-combining" pass that does some dataflow
> analysis or similar (or, for now, only looks at things within a single
> block).
FYI nouveau does this at a local level (i.e. per-bb), if you're
interested take a look at:
http
From: Samuel Li
v2 (agd): rebase on mesa master, split pci ids to
separate commit
Signed-off-by: Samuel Li
---
src/gallium/drivers/radeon/r600_pipe_common.c | 2 ++
src/gallium/drivers/radeon/radeon_winsys.h| 1 +
src/gallium/drivers/radeonsi/si_state.c | 1 +
src/gallium/winsys/amdg
On Thu, Oct 22, 2015 at 4:21 AM, Iago Toral Quiroga wrote:
> I implemented this first as a separate optimization pass in GLSL IR [1], but
> Curro pointed out that this being pretty much a restricted form of a CSE pass
> it would probably make more sense to do it inside CSE (and we no longer have
>
From: Samuel Li
Signed-off-by: Samuel Li
---
include/pci_ids/radeonsi_pci_ids.h | 2 ++
1 file changed, 2 insertions(+)
diff --git a/include/pci_ids/radeonsi_pci_ids.h
b/include/pci_ids/radeonsi_pci_ids.h
index 52eada1..bcf15a1 100644
--- a/include/pci_ids/radeonsi_pci_ids.h
+++ b/include/pci
On 19 October 2015 at 18:10, Emil Velikov wrote:
> On 17 October 2015 at 00:14, Julien Isorce
> wrote:
> > +
> > + if (!(picture_width && picture_height) && !is_vpp)
> >return VA_STATUS_ERROR_INVALID_IMAGE_FORMAT;
> >
> > drv = VL_VA_DRIVER(ctx);
> > @@ -163,38 +177,48 @@ vlVaCreat
On Thu, Oct 22, 2015 at 12:03 PM, Nicolai Hähnle wrote:
> On 22.10.2015 15:57, Fredrik Höglund wrote:
>>
>> On Thursday 22 October 2015, Nicolai Hähnle wrote:
>>>
>>> The same clamping already happens for glTexParameteri. This change
>>> also fixes a bug in mipmap generation, see
>>> https://bugs.
The section for UVD 2 and older was not updated
when HEVC support was added. Reported by Kano
on irc.
Signed-off-by: Alex Deucher
Cc: mesa-sta...@lists.freedesktop.org
---
src/gallium/drivers/radeon/radeon_video.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/gallium/
Add support for VPP in the following functions:
vlVaCreateContext
vlVaDestroyContext
vlVaBeginPicture
vlVaRenderPicture
vlVaEndPicture
Add support for VAProcFilterNone in:
vlVaQueryVideoProcFilters
vlVaQueryVideoProcFilterCaps
vlVaQueryVideoProcPipelineCaps
Add handleVAProcPipelineParameterBuffer
Add support for VA_PROFILE_NONE and VAEntrypointVideoProc
in the 4 following functions:
vlVaQueryConfigProfiles
vlVaQueryConfigEntrypoints
vlVaCreateConfig
vlVaQueryConfigAttributes
Signed-off-by: Julien Isorce
---
src/gallium/state_trackers/va/config.c | 20
src/galliu
On Thu, Oct 22, 2015 at 12:12 PM, Marek Olšák wrote:
> On Wed, Oct 21, 2015 at 12:10 AM, Bas Nieuwenhuizen
> wrote:
>> Uses the DCC buffer instead of the CMASK buffer. The ELIMINATE_FAST_CLEAR
>> still works. Furthermore, with DCC compression we can directly clear
>> to a limited set of colors su
Am 22.10.2015 um 18:27 schrieb Alex Deucher:
> The section for UVD 2 and older was not updated
> when HEVC support was added. Reported by Kano
> on irc.
>
> Signed-off-by: Alex Deucher
> Cc: mesa-sta...@lists.freedesktop.org
> ---
> src/gallium/drivers/radeon/radeon_video.c | 3 ++-
> 1 file cha
The section for UVD 2 and older was not updated
when HEVC support was added. Reported by Kano
on irc.
v2: integrate the UVD2 and older checks into the
main switch statement.
Signed-off-by: Alex Deucher
Cc: mesa-sta...@lists.freedesktop.org
---
src/gallium/drivers/radeon/radeon_video.c | 46
On Mon 19 Oct 2015, Nanley Chery wrote:
> From: Nanley Chery
>
> Enable limiting advertised extension support by context version with
> finer granularity. GLuint is chosen over smaller datatypes because,
> when this field is eventually used, usage of this datatype provides
> the smallest .text si
On Thu 22 Oct 2015, Emil Velikov wrote:
> On 19 October 2015 at 23:36, Nanley Chery wrote:
> > From: Nanley Chery
> >
> > Create a function which determines if an extension is supported in the
> > current context.
> >
> > Signed-off-by: Nanley Chery
> > ---
> > src/mesa/main/extensions.c | 17 +
On Mon 19 Oct 2015, Nanley Chery wrote:
> From: Nanley Chery
>
> Replace open-coded checks for extension support with
> _mesa_extension_supported().
>
> Signed-off-by: Nanley Chery
> ---
> src/mesa/main/extensions.c | 54
>
> src/mesa/main/extens
On Mon, Oct 19, 2015 at 10:38 AM, Alejandro Piñeiro
wrote:
> vec4_live_variables tracks now each flag channel independently, so
> vec4_dead_code_eliminate can update the writemask of null registers,
> based on which component are alive at the moment. This would allow
> vec4_cmod_propagation to opt
On Mon 19 Oct 2015, Nanley Chery wrote:
> From: Nanley Chery
>
> The api_set field has no users outside of _mesa_extension_supported().
> Remove it and allow the version field to take its place.
>
> The brunt of the transformation was performed with the following vim commands:
> s/\(GL [^,]\+\),
On Thu, Oct 22, 2015 at 3:58 AM, Emil Velikov wrote:
> On 21 October 2015 at 23:58, Matt Turner wrote:
>> Add some instructions: illegal, movi, sends, sendsc.
>>
>> Remove some instructions with reused opcodes: msave, mrestore, push,
>> pop, goto. I did have some gross code for disassembling opco
On Thu, Oct 22, 2015 at 4:15 AM, Emil Velikov wrote:
> On 21 October 2015 at 23:58, Matt Turner wrote:
>> Initially just checks that sources are non-NULL, which would have
>> alerted us to the problem fixed by commit 6c846dc5.
>
> What are the chances of combining this with and/or removing
> fs_v
On Thu, Oct 22, 2015 at 11:24 AM, Chad Versace wrote:
> On Mon 19 Oct 2015, Nanley Chery wrote:
>> From: Nanley Chery
>>
>> The api_set field has no users outside of _mesa_extension_supported().
>> Remove it and allow the version field to take its place.
>>
>> The brunt of the transformation was
On Mon 19 Oct 2015, Nanley Chery wrote:
> From: Nanley Chery
>
> Generate functions which determine if an extension is supported in the
> current context. Initially, enums were going to be explicitly used with
> _mesa_extension_supported(). The idea to embed the function and enums
> into generate
On Thu 22 Oct 2015, Emil Velikov wrote:
> On 22 October 2015 at 07:43, Erik Faye-Lund wrote:
> > On Tue, Oct 20, 2015 at 12:44 AM, Nanley Chery
> > wrote:
> >> From: Nanley Chery
> >>
> >> Since the version numbers being compared are integral and we don't ever
> >> expect gl_context::Version to
On Thu, Oct 22, 2015 at 11:15 AM, Chad Versace wrote:
> On Mon 19 Oct 2015, Nanley Chery wrote:
>> From: Nanley Chery
>>
>> Replace open-coded checks for extension support with
>> _mesa_extension_supported().
>>
>> Signed-off-by: Nanley Chery
>> ---
>> src/mesa/main/extensions.c | 54
>>
We always set the mask to 0x, which is what it defaults to when no
header is present. Let's drop the header instead.
v2: Only remove header for untyped reads. Typed reads always need the
header.
Reviewed-by: Francisco Jerez
Reviewed-by: Jordan Justen
Signed-off-by: Kristian Høgsberg Kri
Write groups of enabled components together.
Signed-off-by: Kristian Høgsberg Kristensen
---
src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 56 +++-
1 file changed, 26 insertions(+), 30 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
b/src/mesa/driver
The scalar destination registers break copy propagation. Instead compute
the results to a regular register and then reference a component when we
later use the result as a source.
Signed-off-by: Kristian Høgsberg Kristensen
---
src/mesa/drivers/dri/i965/brw_fs_builder.h | 8
1 file chan
The emit_untyped_read and emit_untyped_write helpers already uniformize
the surface index argument. No need to do it before calling them.
Signed-off-by: Kristian Høgsberg Kristensen
---
src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 2 --
1 file changed, 2 deletions(-)
diff --git a/src/mesa/drivers
Signed-off-by: Kristian Høgsberg Kristensen
---
src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp | 15 +++
1 file changed, 15 insertions(+)
diff --git a/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp
b/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp
index 5589716..
Here's an updated and expanded ssbo optimization series. I found a bit
of low-hanging fruit around dynamic ssbo array indexing. I removed the
IMM shortcut in emit_uniformize() and added the constant propagation
for the read and write opcodes. The result is the same for constant
indexing, and it hel
Now that we don't read each component one-by-one, we don't need the
temoprary vgrf for the offset. More importantly, this register was type
UD while the nir source was type D. This broke copy propagation and left
a redundant MOV in the generated code.
Signed-off-by: Kristian Høgsberg Kristensen
-
The destination for SHADER_OPCODE_FIND_LIVE_CHANNEL is always a UD
register. When we replace the opcode with a MOV, make sure we use a UD
immediate 0 so copy propagation doesn't bail because of non-matching
types.
Signed-off-by: Kristian Høgsberg Kristensen
---
src/mesa/drivers/dri/i965/brw_fs.
Instead of looping through single-component reads, read all components
in one go.
Reviewed-by: Iago Toral Quiroga
Reviewed-by: Jordan Justen
Signed-off-by: Kristian Høgsberg Kristensen
---
src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 25 +++--
1 file changed, 7 insertions(+),
On Thu 22 Oct 2015, Matt Turner wrote:
> On Thu, Oct 22, 2015 at 11:24 AM, Chad Versace wrote:
> > On Mon 19 Oct 2015, Nanley Chery wrote:
> >> From: Nanley Chery
> >>
> >> The api_set field has no users outside of _mesa_extension_supported().
> >> Remove it and allow the version field to take it
Reviewed-by: Matt Turner
(though, there's really no reason for copy propagation to be this stupid :)
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
On Thu, Oct 22, 2015 at 11:19 AM, Matt Turner wrote:
> On Mon, Oct 19, 2015 at 10:38 AM, Alejandro Piñeiro
> wrote:
>> vec4_live_variables tracks now each flag channel independently, so
>> vec4_dead_code_eliminate can update the writemask of null registers,
>> based on which component are alive a
On Thu, Oct 22, 2015 at 11:36 AM, Kristian Høgsberg Kristensen
wrote:
> The emit_untyped_read and emit_untyped_write helpers already uniformize
> the surface index argument. No need to do it before calling them.
>
> Signed-off-by: Kristian Høgsberg Kristensen
> ---
> src/mesa/drivers/dri/i965/br
On Thu, Oct 22, 2015 at 12:11 PM, Matt Turner wrote:
> On Thu, Oct 22, 2015 at 11:36 AM, Kristian Høgsberg Kristensen
> wrote:
>> The emit_untyped_read and emit_untyped_write helpers already uniformize
>> the surface index argument. No need to do it before calling them.
>>
>> Signed-off-by: Krist
On Mon 19 Oct 2015, Nanley Chery wrote:
> From: Nanley Chery
>
> Stop leaks into the following contexts:
>* GLES in _mesa_base_tex_format() and lookup_view_class().
>* Pre-1.1 GL legacy contexts in all uses.
>
> Stop allowing compressed sRGB formats as valid formats in GLES3
> contexts.
On Tue 20 Oct 2015, Nanley Chery wrote:
> On Tue, Oct 20, 2015 at 8:37 AM, Marek Olšák wrote:
>
> > NAK. I'd like this extension in compatibility contexts. The fact the
> > spec requires OpenGL 3.1 was just authors' laziness.
> >
> >
> I had thought it might be the case that some specs may uneces
On Wed, Oct 21, 2015 at 10:05 AM, Matt Turner wrote:
> Not a functional difference, but register is loaded with a signed
> immediate (V) and added to a signed type (D) producing a signed result
> (D).
>
> Also change the type of g0 to allow for compaction.
> ---
> src/mesa/drivers/dri/i965/brw_fs
On 22.10.2015 19:50, Alex Deucher wrote:
The section for UVD 2 and older was not updated
when HEVC support was added. Reported by Kano
on irc.
v2: integrate the UVD2 and older checks into the
main switch statement.
Signed-off-by: Alex Deucher
Cc: mesa-sta...@lists.freedesktop.org
---
src/gal
On Thu, Oct 22, 2015 at 11:37 AM, Kristian Høgsberg Kristensen
wrote:
> The scalar destination registers break copy propagation. Instead compute
> the results to a regular register and then reference a component when we
> later use the result as a source.
It might be hairy to get it working, but
The section for UVD 2 and older was not updated
when HEVC support was added. Reported by Kano
on irc.
v2: integrate the UVD2 and older checks into the
main switch statement.
v3: handle encode checking as well. Encode is
already checked in the top case statement, so
drop encode checks in the lower
On 22.10.2015 21:35, Alex Deucher wrote:
The section for UVD 2 and older was not updated
when HEVC support was added. Reported by Kano
on irc.
v2: integrate the UVD2 and older checks into the
main switch statement.
v3: handle encode checking as well. Encode is
already checked in the top case st
Detected by Matt Turner while reviewing commit
a59359ecd22154cc2b3f88bb8c599f21af8a3934
---
src/mesa/drivers/dri/i965/brw_fs_dead_code_eliminate.cpp | 2 +-
src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/src/mesa/
On 22/10/15 00:43, Rowley, Timothy O wrote:
On Oct 20, 2015, at 5:58 PM, Jose Fonseca wrote:
Thanks for the explanations. It's closer now, but still a bit of gap:
$ KNOB_MAX_THREADS_PER_CORE=0 ./gloss
SWR create screen!
This processor supports AVX2.
--> numThreads = 3
1102 frames in 5.002 s
On Thu, Oct 22, 2015 at 12:24 PM, Matt Turner wrote:
> On Thu, Oct 22, 2015 at 11:37 AM, Kristian Høgsberg Kristensen
> wrote:
>> The scalar destination registers break copy propagation. Instead compute
>> the results to a regular register and then reference a component when we
>> later use the r
On Tue, Oct 20, 2015 at 11:56 AM, Francisco Jerez wrote:
> Kristian Høgsberg writes:
>
>> On Tue, Oct 20, 2015 at 3:16 AM, Francisco Jerez
>> wrote:
>>> Kristian Høgsberg writes:
>>>
On Mon, Oct 19, 2015 at 4:19 AM, Francisco Jerez
wrote:
> Neil Roberts writes:
>
>> Ju
On Thu, Oct 22, 2015 at 3:06 AM, Emil Velikov
wrote:
> On 20 October 2015 at 16:43, Nanley Chery wrote:
> > On Tue, Oct 20, 2015 at 8:16 AM, Marek Olšák wrote:
> >>
> >> Also, the FIXME comment should be on its own line.
> >>
> >
> > I moved it aside to make editing the table easier. However, s
1 - 100 of 151 matches
Mail list logo
