On Mar 31, 2016 12:09 PM, "Samuel Pitoiset"
wrote:
>
> For Maxwell, the ATOMS instruction can be used to perform atomic
> operations on shared memory instead of this load/store lowering pass.
>
> Signed-off-by: Samuel Pitoiset
> ---
>
On Mar 31, 2016 12:09 PM, "Samuel Pitoiset"
wrote:
>
> Make sure to avoid out of bounds access in presence of indirect
> array indexing by loading the size from the driver constant buffer.
>
> Signed-off-by: Samuel Pitoiset
> ---
>
On Mar 31, 2016 12:09 PM, "Samuel Pitoiset"
wrote:
>
> The grid size is stored as three 32-bits integers in the indirect
> buffer but the launch descriptor uses a 32-bits integer for both
> griddim_y and griddim_z like this (z << 16) | y. To make it work,
> the 16 high
Please see if you can't rejigger some stuff to make it all fit into 1024...
I know it will have to grow for images, but let's try to limit the damage...
On Mar 31, 2016 12:09 PM, "Samuel Pitoiset"
wrote:
> The compute launch descriptor only allows to set up 8 CBs, but
As I recall, we already upload ms offsets for fs. Perhaps reuse the same
spot for these CP ones? What are they used for, anyways? Can't think of
anything offhand, but perhaps opencl needs something funky?
On Mar 31, 2016 12:09 PM, "Samuel Pitoiset"
wrote:
> Instead of
2016-04-01 2:40 GMT+08:00 Jaap Jan Meijer :
> Op donderdag 31 maart 2016 13:02:36 UTC+2 schreef Chih-Wei Huang:
> Thanks for the patch, I tested it on both x86 and x64 kernels and it doesn't
> crash anymore, previously it did on both.
>
>> Thanks for the fix. I'll apply it.
Hi Ian and Roland,
I am new to mesa development, so please bare with my questions:
- Is there some protocol for testing against the private shader-db?
(I assume private means I don't have access)
- Would the person checking PlaneShift be someone who owns the
game and can test it against
Op donderdag 31 maart 2016 13:02:36 UTC+2 schreef Chih-Wei Huang:
>
> 2016-03-31 0:09 GMT+08:00 Rob Herring :
> > On Tue, Mar 29, 2016 at 2:49 PM, Jaap Jan Meijer > wrote:
> >>
> >> First of all, thanks for the hard work!
> >>
> >> Only one problem I
Reviewed-by: Edward O'Callaghan
On 2016-03-31 18:03, Dave Airlie wrote:
From: Dave Airlie
This refactors out the code and fixes it up to be used
for images later. It uses the code in the current RAT binding
for compute.
Signed-off-by: Dave
Reviewed-by: Edward O'Callaghan
On 2016-03-31 18:03, Dave Airlie wrote:
From: Dave Airlie
This moves the code to create CB info out into
a separate function so it can be reused in images
code to create RATs.
Signed-off-by: Dave Airlie
Reviewed-by: Edward O'Callaghan
On 2016-03-31 18:03, Dave Airlie wrote:
From: Dave Airlie
This takes the code required to initialise a buffer resource
out of the texture buffer code, into it's own function.
This is going to be used for the
Reviewed-by: Edward O'Callaghan
On 2016-03-31 18:03, Dave Airlie wrote:
From: Dave Airlie
This refactors out the code to setup a texture resource
so we can reuse it later from the images code.
Signed-off-by: Dave Airlie
str is a builtin in python, although python doesn't prevent builtins
from being shadowed, its still a bad idea.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_proto_send.py | 14 +++---
1 file changed, 7 insertions(+), 7 deletions(-)
diff --git
This is a pretty minor refactor, but it avoids having an else statement
with a nested if. It instead makes it one big if statement.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 13 ++---
1 file changed, 6 insertions(+), 7 deletions(-)
diff
Rather than using a loop, and augmenting a variable use a generator and
sum to achieve the same result.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_x86_asm.py | 10 ++
1 file changed, 2 insertions(+), 8 deletions(-)
diff --git
This just saves a bit of typing.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_server_table.py | 12 ++--
1 file changed, 2 insertions(+), 10 deletions(-)
diff --git a/src/mapi/glapi/gen/glX_server_table.py
b/src/mapi/glapi/gen/glX_server_table.py
This method is a no-op, but it shuts up tools that get upset that Print
calls a non-existent method.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 3 +++
1 file changed, 3 insertions(+)
diff --git a/src/mapi/glapi/gen/gl_XML.py
This code has for a long time returned 0 on entry. This makes the code
dead, and the one place it is invoked a no-op if statement (it will
always go down the if path).
Thanks to version control it's rather simple to restore it if someday
someone decides that they want it.
Signed-off-by: Dylan
Rather than using a variable called junk, use _ for unused unpacked
values. This is much more idiomatic.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_proto_common.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
There are two ways to handle this. One is to use the _ placeholder value
when multiple values are needed. The second is to use list indexing to
get only the values actually needed when only one value is desired.
Signed-off-by: Dylan Baker
---
Don't assign variables that aren't actually used, just put them in _,
the python unused variable.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_x86-64_asm.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/mapi/glapi/gen/gl_x86-64_asm.py
The pass keyword is the solution for python's whitespace significant
code. It's used when something is needed for whitespace rules, but not
for programming. Don't add it after actual code.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_proto_send.py | 1 -
1
The variable has the same value as the explicit None return, but makes
it clearer what's going on.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_proto_common.py | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
This makes the function a little simpler by using a special case for the
initial "ARB", and just walking the str as a sequence rather than
indexing into it.
It also uses a list to join rather than overwritting a str over and over
again.
Signed-off-by: Dylan Baker
---
This uses the collections.defaultdict to remove the need to check and
add an empty dict, since defaultdict will call the default factory for
us whenever a KeyError would be raised (by a missing key value).
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py |
Use a boolean instead of 0 or 1, and use an assert rather than a runtime
error.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 12 +---
1 file changed, 5 insertions(+), 7 deletions(-)
diff --git a/src/mapi/glapi/gen/gl_XML.py
Remove parens that don't do anything. Mostly these are extra grouping
parens in if statements, that aren't needed. Unlike C python doesn't
require parens around the conditions of if statements, except to group
them when using logical operators.
It also removes them around the assert keyword.
rather than a list.
In python tuples are mainly for heterogenous collections, especially
when position has relevance. Which makes them the right type for the
job.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 4 ++--
1 file changed, 2 insertions(+), 2
This avoids recompiles of the re each time the function is called.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/src/mapi/glapi/gen/gl_XML.py b/src/mapi/glapi/gen/gl_XML.py
index
This implementation is somewhat more efficient than the previous, and is
less code. It works by replacing one of the list building functions with
a generator, so that the last list can be avoided. It also decreases the
size of the list that remains by using sorting instead of a sparse list.
This can easily be restored with a revert if it's ever actually needed.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_proto_common.py | 1 -
1 file changed, 1 deletion(-)
diff --git a/src/mapi/glapi/gen/glX_proto_common.py
Rather than using the string module, use the same method for the str
instance (join, split, replace, etc).
It's more forward looking, since it will make smoothing over the
str/bytes/unicode problem between python 2 and python 3. It's also more
standard, using the string module isn't very common.
This reworks this method to use the sorted() built-in rather than taking
the dictionary keys (which involves creating a brand new copy of the
list in memory), sorting it, then iterating the new sorted list and
doing a lookup into the dictionary. Then it does away with building a
list and returning
The python code has no standardization on how many newlines to put
anywere. Sometimes top level items have one space between them, sometimes
two, and sometimes three. Sometiems methods and other sub items had two,
sometimes one, sometimes 3.
This patches uses PEP8 styling (2 lines between top
Python has what are often referred to as "magic methods". __iter__,
__str__, __add__, and countless more. All of these methods exist to be
called by another function or by an operator. The __add__ method
implements the + operator, for example.
There are times to call the magic methods directly,
Originally I planned to make it more efficient, but then realized that
it wasn't actually used anywhere, so lets just delete it.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 10 --
1 file changed, 10 deletions(-)
diff --git
It's a bit odd to call IBM out in particular, especially since Brian
Paul is listed as a copyright holder on chronologically before IBM.
Either way this should probably have the standard "authors and/or
copyright holders", in keeping with the standard MIT license.k
Signed-off-by: Dylan Baker
This allows large blocks of C code that is going to be printed to be
nested nicely in the python functions that print them, but still be
printed at the same level they were previously. This helps to make the
code visually more readable, and fixes syntax folding in vim.
Signed-off-by: Dylan Baker
In python 2 range returns a new list, which is okay for very small
values, but can get quite expensive for larger lists. xrange on the
other hand is an iterator, which requires less memory and processing
power.
Signed-off-by: Dylan Baker
---
Python's range and xrange functions start at 0 by default anyway, so
don't bother setting it explicitly to 0.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_proto_size.py | 2 +-
src/mapi/glapi/gen/glX_server_table.py | 4 ++--
src/mapi/glapi/gen/gl_XML.py
Use the max function and a generator expression instead of a loop and
if. This is certainly no less efficient, and saves several lines of
code.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 5 +
1 file changed, 1 insertion(+), 4 deletions(-)
diff
RuntimeError is a very specific type of error. In almost very case it's
being raised where an assert is the right choice, and in the rest using
a plain Exception is better, at least there it's obvious it's not that
the python runtime hit an internal error and died.
Signed-off-by: Dylan Baker
Python has two equality comparitors, 'is' and '=='. The difference is
that 'is' determines whether instance A *is* instance B, while '=='
determines if instance A *has the same value as* instance B.
Comparisons with singletons (the notable ones being None, True, and
False) should always by done
One line if statements (if foo: break) are pretty uncommon, and are not
nearly as readable as their counterparts in other languages.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_XML.py| 3 ++-
src/mapi/glapi/gen/glX_proto_send.py | 9 ++---
This method (and it's sibling has_* methods) are deprecated in favor of
the 'item in container' syntax. There are countless advantages to this
for python, but for us this is necessary to get python 3 support.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_XML.py
All versions of python that are even worth considering for support
(2.6+) have booleans (actually, so do a lot of versions not worth
supporting), so don't use 1 and 0 for emulating them, which also helps
to clarify the code, since returning an actual value of 1 or 0 happens).
Signed-off-by: Dylan
Don't call and store a value that might not be used unless it's know
it's needed.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/gl_XML.py | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/src/mapi/glapi/gen/gl_XML.py
This allows, among other things, the proper use of the Copyright symbol
in the copyright header (which this patch also does).
It cannot be used in the printed templates however, because they are
printed, not directly written to a file. If a user doesn't have a UTF-8
locale (or has an operating
In python functions return None by default, and it's uncommon to use an
explicit return statement without a return value unless it's to return
early. It's also uncommon to return None explicitly unless it's an
alternate to another possible return value.
This patch changes the code to only use
This does basically the same thing as the previous implementation,
although it uses sorted() to act on the dictionary without first copying
the keys out into a list, and then indexing back into the dict; and uses
yield to make the method a generator, removing the need to make another
list.
They don't do anything, they're not required, and they look weird. The
only use for semi-colons in python is to put two statements in a single
line, which is quite frowned upon by the python community.
Signed-off-by: Dylan Baker
---
src/mapi/glapi/gen/glX_proto_size.py
This is the first part of a much larger project I'm working on to
overhaul the generators. The end goal is to use mako templates rather
than a massive series of print statements, be python 2.7 (the current)
and python 3.3+ (I might be talked into support 3.2 if someone has a
really good case)
Completely clean the imports:
- Split so that one module is imported per line
- Remove unused imports
- Group stdlib imports, then 3rd party modules, and finally local
modules
- sort alphabetically within those groups
Signed-off-by: Dylan Baker
---
On 01/04/16 00:04, Roland Scheidegger wrote:
Am 01.04.2016 um 00:42 schrieb Jose Fonseca:
This is an old patch I had around.
Vector selects seem to work well from LLVM 3.4. Using them might
improve code quality, as it might make constant propagation pass more
effective.
Tested lp_test_*
---
Am 01.04.2016 um 00:42 schrieb Jose Fonseca:
> This is an old patch I had around.
>
> Vector selects seem to work well from LLVM 3.4. Using them might
> improve code quality, as it might make constant propagation pass more
> effective.
>
> Tested lp_test_*
> ---
>
This is an old patch I had around.
Vector selects seem to work well from LLVM 3.4. Using them might
improve code quality, as it might make constant propagation pass more
effective.
Tested lp_test_*
---
src/gallium/auxiliary/gallivm/lp_bld_logic.c | 8 +---
1 file changed, 5 insertions(+),
On Thursday, March 31, 2016 12:37:31 AM PDT Jordan Justen wrote:
> On 2016-03-30 02:03:29, Alejandro Piñeiro wrote:
> > On 30/03/16 09:44, Alejandro Piñeiro wrote:
> > > On 30/03/16 06:16, Kenneth Graunke wrote:
> >
> > >> I was just looking at that today...adding an ordinary global variable
> >
Starting from C++11, several math functions, like isinf, moved into the std
namespace. Since cmath undefines those functions before redefining them inside
the namespace, and glibc 2.23 defines the C variants as macros, the C variants
in global namespace are not accessible any longer.
v2: Move the
From: Marek Olšák
---
src/mesa/drivers/dri/common/drirc | 8
1 file changed, 8 insertions(+)
diff --git a/src/mesa/drivers/dri/common/drirc
b/src/mesa/drivers/dri/common/drirc
index 183a1dc..3912d8b 100644
--- a/src/mesa/drivers/dri/common/drirc
+++
> My take is (and always has been) that the point of GL3.txt is to coordinate
> amongst developers, so clarity is paramount.
I completely understand and I'm sure we can do both. The problem here is
simply that we're mixing dev comments and feature statuses.
I'm working on a way to simply edit
On 03/31/2016 06:27 PM, Ilia Mirkin wrote:
On Thu, Mar 31, 2016 at 12:08 PM, Samuel Pitoiset
wrote:
Hi,
This series adds support for ARB_compute_shader on GK104 and GM107+, except on
GK110 where one test miserably fail (related to texelFetch) for really weird
On Thu, Mar 31, 2016 at 2:53 PM, Kenneth Graunke wrote:
> System values are just built-in input variables that we've opted to
> special-case out of convenience. We need to consider all inputs,
> regardless of how we've classified them.
>
> Unfortunately, there's one
A program will either have gl_VertexID or gl_VertexIDMESA (the lowered
zero-based version), not both. Just spoof it in the resource list so
the hacks are done in a single place.
Signed-off-by: Kenneth Graunke
---
src/compiler/glsl/linker.cpp | 12 ++--
With SSO, the GL_PROGRAM_INPUT and GL_PROGRAM_OUTPUT interfaces refer to
the first and last shader stage linked into a program. This may not be
the vertex and fragment shader stages.
So, subtracting VERT_ATTRIB_GENERIC0 and FRAG_RESULT_DATA0 is bogus.
We need to subtract VERT_ATTRIB_GENERIC0 for
This makes no sense. If the stage being considered is the vertex
shader, then we'll add inputs and system values appropriately.
If we're not considering the vertex shader, then we absolutely should
not do anything with it.
Signed-off-by: Kenneth Graunke
---
We were recording locations for all variables, even ones without an
explicit location set. Implement the rules from the spec, and record
-1 in the resource list accordngly. Make program_resource_location
stop doing math on negative values. Remove hacks that are no longer
necessary now that
System values are just built-in input variables that we've opted to
special-case out of convenience. We need to consider all inputs,
regardless of how we've classified them.
Unfortunately, there's one exception: we shouldn't add gl_BaseVertex
unless ARB_shader_draw_parameters is enabled, because
According to Timothy, block_index is an index into the
BufferInterfaceBlocks array. We want to return an index into the
UniformBlocks or ShaderStorageBlocks array, as that's what everything
else expects.
Fixes several dEQP-GLES31.functional tests:
-
add_interface_variables() is supposed to add variables for the inputs of
the first shader stage linked into a program, and the outputs of the
last shader stage linked into a program.
From the ARB_program_interface_query specification:
"* PROGRAM_INPUT corresponds to the set of active input
add_interface_variables is supposed to add variables from either the
first or last stage of a linked shader. But it has no way of knowing
the stage it's being asked to process, which makes it impossible to
produce correct stagerefs.
Signed-off-by: Kenneth Graunke
---
stages is always 1 << stage now.
Signed-off-by: Kenneth Graunke
---
src/compiler/glsl/linker.cpp | 6 +-
1 file changed, 1 insertion(+), 5 deletions(-)
diff --git a/src/compiler/glsl/linker.cpp b/src/compiler/glsl/linker.cpp
index 48aa395..7d6c60a 100644
---
This is a bitfield of which stages refer to a variable. It is not used
to mask off bits. In fact, it's used to contribute additional bits.
Rename it and tidy a bit of the logic.
Signed-off-by: Kenneth Graunke
---
src/compiler/glsl/linker.cpp | 10 +-
1 file
On Wednesday, March 30, 2016 4:23:30 PM PDT Ian Romanick wrote:
> From: Ian Romanick
>
> The KIL instruction doesn't have a destination, so ptn_kil never uses
> dest.
>
> program/prog_to_nir.c: In function ‘ptn_kil’:
> program/prog_to_nir.c:547:38: warning: unused
On Thu, Mar 31, 2016 at 2:10 PM, Romain Failliot
wrote:
> Hi!
>
> Two extensions have been recently flagged "DONE (core only)".
> From what we understood [1], it means that it's done in mesa, but drivers
> need to implement it.
>
> So, first: is that it?
Yes. (More
Hi!
Two extensions have been recently flagged "DONE (core only)".
>From what we understood [1], it means that it's done in mesa, but drivers
need to implement it.
So, first: is that it?
And, second: if it is, could we change for the syntax made for it: "DONE
()"?
I'll be happy to do the patch
Am 31.03.2016 um 03:06 schrieb Brian Paul:
> It was kind of overloaded, returning two different things. Now get
> the index of the shadow reference src register with a new
> tgsi_util_get_shadow_ref_src_index() function.
>
> To verify the new code, I added some temp/debug code which looped
>
Hi,
This series adds support for ARB_compute_shader on GK104 and GM107+, except on
GK110 where one test miserably fail (related to texelFetch) for really weird
reasons. Anyway, this is not going to break anything because NVF0_COMPUTE is
still required for using compute on GK110. I will have a
On Thu, Mar 31, 2016 at 12:08 PM, Samuel Pitoiset
wrote:
> Hi,
>
> This series adds support for ARB_compute_shader on GK104 and GM107+, except on
> GK110 where one test miserably fail (related to texelFetch) for really weird
> reasons. Anyway, this is not going to break
Reduce likelihood of collision with real buffers by placing the
hole at the top of the 4G area. This fixes some indirect draw+compute
tests with large buffers.
Suggested by Ilia Mirkin.
Signed-off-by: Samuel Pitoiset
---
Signed-off-by: Samuel Pitoiset
---
.../drivers/nouveau/codegen/nv50_ir_driver.h | 1 +
src/gallium/drivers/nouveau/nvc0/nvc0_context.h| 3 +++
src/gallium/drivers/nouveau/nvc0/nvc0_program.c| 1 +
src/gallium/drivers/nouveau/nvc0/nve4_compute.c| 26
https://bugs.freedesktop.org/show_bug.cgi?id=94752
--- Comment #4 from Guilherme ---
(In reply to Guilherme from comment #3)
> Just to add that I could also reproduce the bug with direct rendering. Of
> course, with the patch that forces the second dri_drawable to be allocated
Compute support on GK110 is still unstable for weird reasons, but
this can be fixed later as the NVF0_COMPUTE envvar prevent using
compute.
Note that GL3.txt is not updated yet because GL_ARB_compute_shader is
a bit useless without ARB_shader_image_load_store.
Signed-off-by: Samuel Pitoiset
Make sure to avoid out of bounds access in presence of indirect
array indexing by loading the size from the driver constant buffer.
Signed-off-by: Samuel Pitoiset
---
.../nouveau/codegen/nv50_ir_lowering_nvc0.cpp | 55 +-
This fixes 84b9b8f (nvc0/ir: add missing emission of locked load
predicate).
Signed-off-by: Samuel Pitoiset
---
src/gallium/drivers/nouveau/codegen/nv50_ir_emit_nvc0.cpp | 5 -
1 file changed, 4 insertions(+), 1 deletion(-)
diff --git
The grid size is stored as three 32-bits integers in the indirect
buffer but the launch descriptor uses a 32-bits integer for both
griddim_y and griddim_z like this (z << 16) | y. To make it work,
the 16 high bits of griddim_y are overwritten by griddim_z.
Changes from v2:
- upload the whole
For Maxwell, the ATOMS instruction can be used to perform atomic
operations on shared memory instead of this load/store lowering pass.
Signed-off-by: Samuel Pitoiset
---
.../nouveau/codegen/nv50_ir_lowering_nvc0.cpp| 20 +---
1 file changed, 13
Signed-off-by: Samuel Pitoiset
---
.../nouveau/codegen/nv50_ir_lowering_nvc0.cpp | 114 -
.../nouveau/codegen/nv50_ir_lowering_nvc0.h| 1 +
2 files changed, 114 insertions(+), 1 deletion(-)
diff --git
The maximum number of uniform blocks (MAX_COMPUTE_UNIFORM_BLOCKS)
per compute program must be at least 12.
Signed-off-by: Samuel Pitoiset
---
src/gallium/drivers/nouveau/nvc0/nvc0_screen.c | 2 --
src/gallium/drivers/nouveau/nvc0/nvc0_screen.h | 1 -
2 files changed,
Signed-off-by: Samuel Pitoiset
---
src/gallium/drivers/nouveau/nvc0/nvc0_program.c | 4 +--
src/gallium/drivers/nouveau/nvc0/nve4_compute.c | 38 +
2 files changed, 39 insertions(+), 3 deletions(-)
diff --git
Uniform buffer objects will be sticked to the driver constant buffer
like buffers because the launch descriptor only allows 8 CBs.
Input kernel parameters for OpenCL are still uploaded to screen->parm
which is bound on c0, but this will be changed later with a new series.
Signed-off-by: Samuel
The compute launch descriptor only allows to set up 8 CBs, but OpenGL
requires at least 12 UBOs. To bypass this limitation, we store the ubos
info into the driver constant buffer. This will be only used for Kepler.
This bumps the size of each shader area from 1K to 2K.
Signed-off-by: Samuel
Instead of using the screen->parm buffer object which will be removed,
upload auxiliary constants to uniform_bo to be consistent regarding
what we already do for Fermi.
This breaks surfaces support (for compute only) but this will be
properly re-introduced later for ARB_shader_image_load_store.
On 03/31/2016 02:54 AM, Samuel Pitoiset wrote:
The number of channels must be 4 for all RGBA components.
Fixes: 22d129601 ("tgsi: add support for image operations to tgsi_exec. (v2.1)")
Signed-off-by: Samuel Pitoiset
Cc: Dave Airlie
---
On 03/31/2016 07:02 AM, Marek Olšák wrote:
On Thu, Mar 31, 2016 at 3:15 AM, Dave Airlie wrote:
On 31 March 2016 at 11:06, Brian Paul wrote:
It was kind of overloaded, returning two different things. Now get
the index of the shadow reference src register
From: Marek Olšák
Instead of failing an assertion, disable DCC and CMASK on the first export
that needs it, and merge the external usage flags.
---
src/gallium/drivers/radeon/r600_texture.c | 55 ++-
1 file changed, 33 insertions(+), 22
This seems a little ugly to me. Since these ops are just moving bits
around, maybe we should just convert the float64 src/dest to uint64?
We do this for other opcodes that just shuffle bits around.
On Mon, Mar 21, 2016 at 8:06 AM, Samuel Iglesias Gonsálvez
wrote:
> From:
Am 31.03.2016 um 04:17 schrieb Ian Romanick:
> Arg. :(
>
> It turns out that shaders in the game PlaneShift use float suffixes in
> GLSL 1.10 shaders... at least the version of the shaders in our private
> shader-db repo do. Once some one check to make sure the game still does
> this, we will
On Thu, Mar 31, 2016 at 5:07 AM, Alejandro Piñeiro wrote:
> On 31/03/16 09:37, Jordan Justen wrote:
>> On 2016-03-30 02:03:29, Alejandro Piñeiro wrote:
>>> On 30/03/16 09:44, Alejandro Piñeiro wrote:
On 30/03/16 06:16, Kenneth Graunke wrote:
> I was just looking at
On Thu, Mar 31, 2016 at 6:13 AM, Daniel Stone wrote:
> Hi,
>
> On 31 March 2016 at 04:21, Rob Herring wrote:
>> diff --git a/include/GL/internal/dri_interface.h
>> b/include/GL/internal/dri_interface.h
>> index 6bbd3fa..b059112 100644
>> ---
Presumably you also want to remove the enable from Consts.Extensions? (mtypes.h)
On Thu, Mar 31, 2016 at 6:58 AM, Timothy Arceri
wrote:
> ---
> src/compiler/glsl/glsl_parser_extras.cpp | 2 +-
> src/mesa/main/extensions_table.h | 2 +-
> 2 files changed, 2
1 - 100 of 201 matches
Mail list logo