On Saturday, December 10, 2016 12:37:16 PM PST Matt Turner wrote:
> On Fri, Dec 9, 2016 at 8:28 PM, Kenneth Graunke wrote:
> > A number of games have large arrays of constants, which we promote to
> > uniforms. This introduces copies from the uniform array to the original
> > temporary array. No
On Thu, Dec 01, 2016 at 02:10:07PM -0800, Ben Widawsky wrote:
> From: Ben Widawsky
>
> Cc: Topi Pohjolainen
> Signed-off-by: Ben Widawsky
> ---
> src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 12 +++-
> 1 file changed, 3 insertions(+), 9 deletions(-)
>
> diff --git a/src/mesa/driver
On Mon, Dec 05, 2016 at 11:59:54AM -0800, Jason Ekstrand wrote:
> ---
> src/mesa/drivers/dri/i965/brw_defines.h| 6 ++
> src/mesa/drivers/dri/i965/brw_fs.h | 3 +++
> src/mesa/drivers/dri/i965/brw_fs_cse.cpp | 1 +
> src/mesa/drivers/dri/i965/brw_fs_generator.cpp |
On Tue, Dec 06, 2016 at 12:37:45PM -0800, Jason Ekstrand wrote:
> Cc: "13.0"
> ---
> src/mesa/drivers/dri/i965/intel_blit.c | 6 --
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/intel_blit.c
> b/src/mesa/drivers/dri/i965/intel_blit.c
> index 03
I don't see any difference in Deus Ex, but F1 2015 uses 64% less
spilled-temp-array memory with this on GCN.
Marek
On Sat, Dec 10, 2016 at 5:28 AM, Kenneth Graunke wrote:
> A number of games have large arrays of constants, which we promote to
> uniforms. This introduces copies from the uniform
This new buffer upload path enables to lock
faster than the normal path when using
DISCARD/NOOVERWRITE.
v2: Diverse cleanups and fixes.
Signed-off-by: Axel Davy
---
This patch needed a few cleanups, whih this new version adds.
With this version, I wasn't able to reproduce the memory leak.
I'm no
https://bugs.freedesktop.org/show_bug.cgi?id=99055
Bug ID: 99055
Summary: Games hang / freeze completely
Product: Mesa
Version: 13.0
Hardware: x86-64 (AMD64)
OS: Linux (All)
Status: NEW
Severity: major
Commit 929fcee47e46781c57f2a354ce0a013915c033d1 introduced code that
attempts to unlock an unlocked mutex which is undefined behaviour.
On OpenBSD this leads to an abort:
0 0x124dadfa96ba in thrkill () at :2
1 0x124dadf3da39 in *_libc_abort () at /usr/src/lib/libc/stdlib/abort.c:52
2 0
On Dec 11, 2016 12:28 AM, "Pohjolainen, Topi"
wrote:
On Tue, Dec 06, 2016 at 12:37:45PM -0800, Jason Ekstrand wrote:
> Cc: "13.0"
> ---
> src/mesa/drivers/dri/i965/intel_blit.c | 6 --
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/src/mesa/drivers/dri/i965/intel_blit.c
https://bugs.freedesktop.org/show_bug.cgi?id=98869
cosiek...@o2.pl changed:
What|Removed |Added
Priority|medium |high
--
You are receiving this mail be
anv_util.c and radv_util.c are essentially duplicates, so create a new
vk_util.c to be shared by both vulkan drivers. Some other content from
anv_private.h and radv_private.h was moved to vk_util.h too.
Signed-off-by: Grazvydas Ignotas
---
no commit access
configure.ac |
Emitted on release build in case vk_error() return is not used.
Signed-off-by: Grazvydas Ignotas
---
no commit access
src/vulkan/vk_util.c | 4
src/vulkan/vk_util.h | 13 -
2 files changed, 12 insertions(+), 5 deletions(-)
diff --git a/src/vulkan/vk_util.c b/src/vulkan/vk_uti
On Sun, Dec 11, 2016 at 07:56:59AM -0800, Jason Ekstrand wrote:
>On Dec 11, 2016 12:28 AM, "Pohjolainen, Topi"
><[1]topi.pohjolai...@gmail.com> wrote:
>
>On Tue, Dec 06, 2016 at 12:37:45PM -0800, Jason Ekstrand wrote:
>> Cc: "13.0" <[2]mesa-sta...@lists.freedesktop.org>
>> ---
Looks good.
Reviewed-by: Eduardo Lima Mitev
On 12/11/2016 04:42 PM, Jonathan Gray wrote:
> Commit 929fcee47e46781c57f2a354ce0a013915c033d1 introduced code that
> attempts to unlock an unlocked mutex which is undefined behaviour.
>
> On OpenBSD this leads to an abort:
>
> 0 0x124dadfa96ba
The patch looks good but now that we are at it, I would move the
definition and assignment of 'is_lossless_compressed' to the block where
it is used, right before the assert. Right now we are always defining
and assigning the variable, but it is consumed conditionally.
With that changed, patch is:
Looks good. I don't remember the MAYBE_UNUSED macro being used that much
in mesa opengl, but it seems it is trendy in vulkan code.
Reviewed-by: Eduardo Lima Mitev
On 12/10/2016 08:10 PM, Grazvydas Ignotas wrote:
> Signed-off-by: Grazvydas Ignotas
> ---
> no commit access, somebody please push
>
Looks good.
Reviewed-by: Eduardo Lima Mitev
Will push this and the other ones I reviewed.
Thanks!
Eduardo
On 12/10/2016 08:10 PM, Grazvydas Ignotas wrote:
> Doesn't look like this can work on 32bit, just rids of annoying
> warning.
>
> Signed-off-by: Grazvydas Ignotas
> ---
> no commit acce
Signed-off-by: Grazvydas Ignotas
---
v2: take Eduardo Lima's suggestion further - remove the variable completely
no commit access
src/mesa/drivers/dri/i965/brw_blorp.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/src/mesa/drivers/dri/i965/brw_blorp.c
b/src/mesa/drivers
https://bugs.freedesktop.org/show_bug.cgi?id=99055
--- Comment #1 from Etienne Bruines ---
Update: apitrace did not magically fix it, but it did delay the symptom by a
few hours.
--
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for the bug.___
Found by inspection, probably a typo because a surface store is
definitely not an atomic operation.
Signed-off-by: Samuel Pitoiset
---
src/gallium/drivers/nouveau/codegen/nv50_ir_target.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/gallium/drivers/nouveau/codegen/nv
I don't think we use SUSTB anyways... it was used in the old
handleSTORE logic in the tgsi -> nvir converter, but wtvr. I have a
feeling it's the sort of thing that might matter for nv50 (vs nvc0),
but for now this is fine. (Esp since the opclasses don't matter on
nv50.)
On Sun, Dec 11, 2016 at 5:
This new buffer upload path enables to lock
faster than the normal path when using
DISCARD/NOOVERWRITE.
v2: Diverse cleanups and fixes.
v3: Fix allocation size for 'lone' buffers and
add more debug info.
Signed-off-by: Axel Davy
---
This fixes the reported 'leak' (which was not a leak
but the co
We now print
START B15 <-B14 (42774 cycles)
indicating that we estimate B15 will take 42,774 cycles. Printing
this should make it easier where time is spent in the program.
Signed-off-by: Kenneth Graunke
---
src/mesa/drivers/dri/i965/intel_asm_annotation.c | 2 +-
1 file changed, 1 inserti
i965/vec4: handle 32 and 64 bit channels in liveness analysis
Please indent the returned multiline expressions in
var_from_reg() like we do elsewhere, so that the second line
begins on the same column as the first line.
*/ goes on its own line.
I'm having
Good plan.
Reviewed-by: Matt Turner
___
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
Section 2.2.2 (Data Conversions For State Query Commands) of the
OpenGL 4.5 October 24th 2016 specification says:
"If a command returning unsigned integer data is called, such as
GetSamplerParameterIuiv, negative values are clamped to zero."
Fixes GL44-CTS.gpu_shader_fp64.state_query.
Signed-of
https://bugs.freedesktop.org/show_bug.cgi?id=99055
--- Comment #2 from Timothy Arceri ---
(In reply to Etienne Bruines from comment #0)
>
> I'd be happy to provide any additional information.
The most helpful thing you could do is build mesa from source and bisect the
commit that introduces the
From: Roland Scheidegger
soa fetch so far always assumed that data was aligned. However, we want to
use this for vertex fetch, and data might not be aligned there, so handle
it in this path too (basically just pass through alignment through to other
functions). (It looks like it wouldn't work for
From: Roland Scheidegger
Now that there's some SoA fetch which never falls back, we should usually get
results which are better or at least not worse (something like rgba32f will
stay the same). I suppose though it might be worse in some cases where the
format doesn't require conversion (e.g. rg3
From: Roland Scheidegger
This can now handle rgtc (unorm) too - this path no longer handles plain
formats, but that's unnecessary they now all have their proper SoA unpack
(this will still be dog-slow though due to the actual fetch being per-pixel
util fallbacks).
---
src/gallium/auxiliary/galli
From: Roland Scheidegger
This previously always fell back to AoS conversion. Even for 4-float formats
(which is the optimal case by far for that fallback case) this was suboptimal,
since it meant the conversion couldn't be done with 256bit vectors. While this
may still only be partly possible for
From: Roland Scheidegger
We should do transpose, not extract/insert, at least with "sufficient" amount
of channels (for 4 channels, extract/insert shuffles generated otherwise look
truly terrifying). Albeit we shouldn't fallback to that so often in any case.
---
src/gallium/auxiliary/gallivm/lp_
From: Roland Scheidegger
By using a dst_type in the the gather interface, gather has some more
knowledge about how values should be fetched.
E.g. if this is a 3x32bit fetch and dst_type is 4x32bit vector gather
will no longer do a ZExt with a 96bit scalar value to 128bit, but
just fetch the 96bit
On Sun, 2016-12-11 at 00:00 -0800, Kenneth Graunke wrote:
> On Saturday, December 10, 2016 12:37:16 PM PST Matt Turner wrote:
> >
> > On Fri, Dec 9, 2016 at 8:28 PM, Kenneth Graunke > org> wrote:
> > >
> > > A number of games have large arrays of constants, which we
> > > promote to
> > > unifor
https://bugs.freedesktop.org/show_bug.cgi?id=98869
--- Comment #5 from Mark Janes ---
If you can't bisect mesa, then providing an apitrace of the workload will allow
someone else to bisect it.
--
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for t
Hold up..
Does this reintroduce the hang in glsl-fs-loop piglit test with
MESA_DEBUG=context though? Was that tested? I'm interested to know how
this got so muddled up in the first place.
Kind Regards,
Edward.
On 12/12/2016 05:21 AM, Eduardo Lima Mitev wrote:
> Looks good.
>
> Reviewed-by: Edua
Serge Martin writes:
> ---
> src/gallium/state_trackers/clover/llvm/codegen/common.cpp | 5 +
> 1 file changed, 5 insertions(+)
>
> diff --git a/src/gallium/state_trackers/clover/llvm/codegen/common.cpp
> b/src/gallium/state_trackers/clover/llvm/codegen/common.cpp
> index 13ccd59..aa6ca50 1
On Friday, December 9, 2016 11:03:27 AM PST Francisco Jerez wrote:
> We'll need roughly the same logic in other places and it would be
> annoying to duplicate it. Instead factor it out into a function-like
> macro that takes the number of dwords per block (which will prove more
> convenient than t
Previously there was no _mesa_unlock_debug_state() call at all,
one is still retained here for the case where _mesa_lock_debug_state()
is called and did not return NULL (which is documented as being
unlocked).
On Mon, Dec 12, 2016 at 11:40:32AM +1100, Edward O'Callaghan wrote:
> Hold up..
>
> Doe
Hello devs,
could someone please ack the patches from the Igalia guys.
It is really frustrating that their patches are still not in master.
By the way. I use this branch now for months...
https://github.com/Igalia/mesa/tree/i965-va64-gen7-scalar-vec4
and it also enables va64 for HSW and it works
Kenneth Graunke writes:
> On Friday, December 9, 2016 11:03:27 AM PST Francisco Jerez wrote:
>> We'll need roughly the same logic in other places and it would be
>> annoying to duplicate it. Instead factor it out into a function-like
>> macro that takes the number of dwords per block (which will
Serge Martin writes:
> ---
> src/gallium/state_trackers/clover/api/kernel.cpp | 92
> +-
> src/gallium/state_trackers/clover/core/module.cpp | 14
> src/gallium/state_trackers/clover/core/module.hpp | 11 +++
> .../state_trackers/clover/llvm/codegen/common.cpp | 11
Edward O'Callaghan writes:
> On 11/25/2016 03:38 PM, Francisco Jerez wrote:
>> Edward O'Callaghan writes:
>>
>>> Signed-off-by: Edward O'Callaghan
>>> ---
>>> src/gallium/state_trackers/clover/api/memory.cpp | 17 -
>>> src/gallium/state_trackers/clover/core/memory.cpp | 14 +
Edward O'Callaghan writes:
> Hi Francisco, thanks for the review..
>
> On 11/25/2016 03:39 PM, Francisco Jerez wrote:
>> Edward O'Callaghan writes:
>>
>>> Signed-off-by: Edward O'Callaghan
>>> ---
>>> src/gallium/state_trackers/clover/api/memory.cpp | 9 -
>>> src/gallium/state_trac
I have a patch in my tree that fixes that warning by making us actually
return the error code in that function rather than simply using it as a
fancy print function. I think that's the better fix.
On Dec 11, 2016 9:06 AM, "Grazvydas Ignotas" wrote:
> Emitted on release build in case vk_error() r
Hi,
On Fri, Dec 9, 2016 at 8:29 PM, Liu Zhiquan wrote:
> Some dri drivers will pass multiple bits in buffer_mask parameter
> to droid_image_get_buffer(), more than the actual supported buffer
> type combination. For such case, will go through all the bits, and
> will not return error when unsuppo
Brian/Roland ping?
On 12/07/2016 10:30 AM, Edward O'Callaghan wrote:
> As per the C spec, it is illegal to alias pointers to different
> types. This results in undefined behaviour after optimization
> passes, resulting in very subtle bugs that happen only on a
> full moon..
>
> Use a memcpy() as
On Mon, 2016-12-12 at 11:35 +1100, Timothy Arceri wrote:
> On Sun, 2016-12-11 at 00:00 -0800, Kenneth Graunke wrote:
> >
> > On Saturday, December 10, 2016 12:37:16 PM PST Matt Turner wrote:
> > >
> > >
> > > On Fri, Dec 9, 2016 at 8:28 PM, Kenneth Graunke > > e.
> > > org> wrote:
> > > >
> >
On Sunday, December 11, 2016 5:44:03 PM PST Francisco Jerez wrote:
> Kenneth Graunke writes:
>
> > On Friday, December 9, 2016 11:03:27 AM PST Francisco Jerez wrote:
> >> We'll need roughly the same logic in other places and it would be
> >> annoying to duplicate it. Instead factor it out into a
On 12/11/2016 08:57 PM, Grazvydas Ignotas wrote:
> Signed-off-by: Grazvydas Ignotas
> ---
> v2: take Eduardo Lima's suggestion further - remove the variable completely
> no commit access
>
I just pushed the patch. Thanks!
Eduardo
> src/mesa/drivers/dri/i965/brw_blorp.c | 4 +---
> 1 file chan
50 matches
Mail list logo
