For texturing from GRFs, we now have payloads of arbitrary sizes up to the
message length limit.
v2 (Kenneth Graunke): Rebase on intel_context -> brw_context change.
v3: Add some comment text.
Acked-by: Kenneth Graunke
Acked-by: Matt Turner
---
src/mesa/drivers/dri/i965/brw_context.h
Looking at Lightsmark's shaders, the way we used MRFs (or in gen7's
case, GRFs) was bad in a couple of ways. One was that it prevented
compute-to-MRF for the common case of a texcoord that gets used
exactly once, but where the texcoord setup all gets emitted before the
texture calls (such as when
When we introduce texture-from-GRF, we'll have instructions where a src
means more than one virtual grf is being read. Rearrange things to allow
src accesses to unspill like partial dst accesses do (which already had to
do reading more than one vgrf).
---
src/mesa/drivers/dri/i965/brw_fs.h
Thanks to Ken and Matt for a bunch of nasty rebasing work, texture-grf is
working again. And this time, I've produced some reasonably solid
performance numbers indicating that it's a win, so I think we're ready to
commit.
Ken and Matt indicated that they were tired of reading this code (patch 2,
ir_instruction and its derived classes are only being allocated and
destroyed using ralloc_new/delete, which already keep track of the
allocation type of an object and will always call the correct
destructor. Having a virtual empty destructor in the definition of
ir_instruction is unnecessary (bec
This patch series tries out a different approach to allocating C++
objects from an ralloc context. It's independent to the last series I
sent earlier today [1] -- which probably makes sense on its own
because it's a more incremental improvement and is likely to take less
time to discuss and review
---
src/glsl/ast.h | 4 ---
src/glsl/glsl_parser_extras.h | 2 --
src/glsl/glsl_symbol_table.cpp | 2 --
src/glsl/glsl_symbol_table.h | 30 --
src/glsl/ir_function_detect_recur
---
src/glsl/ralloc.h | 57 +++
1 file changed, 57 insertions(+)
diff --git a/src/glsl/ralloc.h b/src/glsl/ralloc.h
index 31682d5..bc8d4de 100644
--- a/src/glsl/ralloc.h
+++ b/src/glsl/ralloc.h
@@ -404,6 +404,63 @@ bool ralloc_vasprintf_append(c
---
src/glsl/Makefile.am | 3 ++-
src/mesa/Makefile.am | 3 ++-
src/mesa/drivers/dri/i965/Makefile.am | 3 ++-
src/mesa/libdricore/Makefile.am | 2 ++
src/mesa/program/Makefile.am | 6 --
5 files changed, 12 insertions(+), 5 deletions(-)
diff -
On Wed, Oct 9, 2013 at 8:40 PM, Marek Olšák wrote:
bo_handles_mutex should also be renamed.
The next patch uses this mutex to protect both the bo_names map and a
new bo_handles map.
I could rename the mutex in this patch, and then unrename in the
subsequent patch if you'd prefer, but this
https://bugs.freedesktop.org/show_bug.cgi?id=70331
Priority: medium
Bug ID: 70331
Keywords: regression
CC: jfons...@vmware.com
Assignee: mesa-dev@lists.freedesktop.org
Summary: u_math.h:167:15: error: ‘long int lrint(double)’ was
Am 10.10.2013 02:01, schrieb jfons...@vmware.com:
> From: José Fonseca
>
> Unless the polygon fill mode is different PIPE_POLYGON_MODE_FILL, so
> checking the the polygon mode is sufficient.
>
> Testing done: no regression in polygon-mode-offset
> ---
> src/gallium/drivers/llvmpipe/lp_state_ras
From: José Fonseca
Unless the polygon fill mode is different PIPE_POLYGON_MODE_FILL, so
checking the the polygon mode is sufficient.
Testing done: no regression in polygon-mode-offset
---
src/gallium/drivers/llvmpipe/lp_state_rasterizer.c | 2 --
1 file changed, 2 deletions(-)
diff --git a/src
* Fix LLVM library and defines
* Only enable tracing when scons build=debug
---
src/gallium/targets/haiku-softpipe/GalliumContext.cpp | 3 +--
src/gallium/targets/haiku-softpipe/GalliumFramebuffer.cpp | 3 +--
src/gallium/targets/haiku-softpipe/SConscript | 5 +++--
src/gallium/tar
* /boot/common no longer exists in Haiku as of
a few days ago (and this is undefined)
---
src/gallium/targets/libgl-haiku/GLRendererRoster.cpp | 2 --
1 file changed, 2 deletions(-)
diff --git a/src/gallium/targets/libgl-haiku/GLRendererRoster.cpp
b/src/gallium/targets/libgl-haiku/GLRendererRo
On 10/09/2013 05:13 PM, Chad Versace wrote:
On 10/09/2013 03:14 PM, Ian Romanick wrote:
On 10/09/2013 02:49 PM, Chad Versace wrote:
On 10/08/2013 04:27 PM, Ben Widawsky wrote:
+ }
+ hz_height /= 2;
+ break;
+ default:
+ perf_debug("Unknown depthbuffer texture type (%d).",
On 10/09/2013 03:14 PM, Ian Romanick wrote:
On 10/09/2013 02:49 PM, Chad Versace wrote:
On 10/08/2013 04:27 PM, Ben Widawsky wrote:
+ }
+ hz_height /= 2;
+ break;
+ default:
+ perf_debug("Unknown depthbuffer texture type (%d).", mt->target);
I thought I commented on the
---
src/gallium/drivers/radeonsi/radeonsi_shader.c | 7 ---
1 file changed, 4 insertions(+), 3 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/radeonsi_shader.c
b/src/gallium/drivers/radeonsi/radeonsi_shader.c
index ab996cc..9d95997 100644
--- a/src/gallium/drivers/radeonsi/radeonsi_s
On 10/09/2013 02:49 PM, Chad Versace wrote:
> On 10/08/2013 04:27 PM, Ben Widawsky wrote:
>> + }
>> + hz_height /= 2;
>> + break;
>> + default:
>> + perf_debug("Unknown depthbuffer texture type (%d).", mt->target);
>
> I thought I commented on the perf_debug() before, but may
---
lib/Target/R600/AMDGPUCallingConv.td | 7 ++-
lib/Target/R600/SIISelLowering.cpp | 3 ++-
2 files changed, 8 insertions(+), 2 deletions(-)
diff --git a/lib/Target/R600/AMDGPUCallingConv.td
b/lib/Target/R600/AMDGPUCallingConv.td
index d26be32..a194e6d 100644
--- a/lib/Target/R600/AMDGPU
On 10/08/2013 04:27 PM, Ben Widawsky wrote:
The HiZ buffer's allocation requirements were relaxed on Ivybridge.
Following a "simple" formula in the bspec was all you needed to satisfy
the requirement.
To prepare the code for this, extract all places where the miptree was
used, when we really onl
On 10/08/2013 04:27 PM, Ben Widawsky wrote:
After the last patch, we can replace the region allocated in the miptree
creation with a more straightforward (and hopefully smaller resulting)
buffer based on the bspec's allocation formula.
Since I am relatively new to this part of the bspec, I would
Looks good to me!
Jose
- Original Message -
> Both the imul_hi and umul_hi are working with this patch.
>
> Signed-off-by: Zack Rusin
> ---
> src/gallium/auxiliary/gallivm/lp_bld_tgsi_action.c | 60
> ++
> 1 file changed, 60 insertions(+)
>
> diff --git a/src/gall
Updated series looks good to me.
Jose
- Original Message -
> From: Roland Scheidegger
>
> Not used since ages, and it wouldn't work at all with explicit derivatives
> now
> (not that it did before as it ignored them but now the code would just use
> the derivs pre-projected which would
Looks good to me. v2 changes seem the right ones.
Jose
- Original Message -
> From: Roland Scheidegger
>
> They need some special handling. Quite complicated.
> Additionally, use the same code for implicit derivatives too if no_rho_approx
> and no_quad_lod is set, because it seems while
R600_RESOURCE_FLAG_TRANSFER forces direct mapping, and reading from
VRAM is simply too slow. VDPAU GetBitsYCbCr is unusuable. Change to
the new PIPE_BIND_LINEAR and adjust r600_transfer_map so that it uses
a staging texture.
---
src/gallium/drivers/r600/r600_uvd.c | 6 +++---
src/gallium/dri
Add simple plain C routines for NV12<->YV12 and YUYV<->UYVY
conversions. The NV12->YV12 conversion is commonly used, for instance
by VLC.
---
src/gallium/state_trackers/vdpau/surface.c | 125 +++--
1 file changed, 117 insertions(+), 8 deletions(-)
diff --git a/src/gallium/
This patch makes sure that class destructors are called as they should
be when a C++ object allocated by ralloc is released.
Based on a previous patch by Kenneth Graunke, but it doesn't exhibit
the ~0.8% performance regression in shader compilation times because
we now use the HAS_TRIVIAL_DESTRUCT
Only implemented on GCC and Clang for now. Other compilers use a
dummy implementation that always returns false, which should be a safe
[but slightly inefficient] assumption in all cases.
---
src/mesa/main/compiler.h | 23 ++-
1 file changed, 22 insertions(+), 1 deletion(-)
d
No, I don't have commit access. I haven't even asked for it yet.
Marek
On Wed, Oct 9, 2013 at 8:01 PM, Tom Stellard wrote:
> On Wed, Oct 09, 2013 at 03:33:37PM +0200, Marek Olšák wrote:
>> From: Marek Olšák
>>
>> This fixes piglit:
>> - shaders/glsl-fs-texture2d-masked
>> - shaders/glsl-fs-text
On Wed, Oct 09, 2013 at 03:33:37PM +0200, Marek Olšák wrote:
> From: Marek Olšák
>
> This fixes piglit:
> - shaders/glsl-fs-texture2d-masked
> - shaders/glsl-fs-texture2d-masked-4
>
> Signed-off-by: Marek Olšák
> ---
> lib/Target/R600/SIISelLowering.cpp | 27 +++--
> test/CodeGen/R
On 27 September 2013 12:05, Paul Berry wrote:
> This patch adds an implementation of
> ir_dereference_record::update_max_array_access(), which ensures that
> ir_variable::max_ifc_array_access is properly updated to reflect the
> shader's use of arrays appearing within interface blocks.
> ---
> s
On 10/09/2013 10:54 AM, Topi Pohjolainen wrote:
Fixes GL2ExtensionTests/egl_image_external/TestSimpleUnassociated.test
which is part of gles2/3 conformance suite. Here image external
textures are switched to be treated the same as 2D textures. These
can be associated with the fallback texture pro
Khronos is proposing a change affecting EGL attribute lists, and they
are requesting feedback on this forum thread [1]. They have specifically
requested feedback from the opensource community.
[1]
http://www.khronos.org/message_boards/showthread.php/9138-Requesting-feedback-on-disallowing-handle
On Wed, Oct 9, 2013 at 10:32 AM, Christian König
wrote:
> From: Christian König
>
> Calling radeon_drm_cs_flush from multiple threads might cause deadlocks,
> fix this by immediately signaling the semaphore after waiting for it.
>
> This is a candidate for the stable branch(es).
Need to add:
CC
Fixes GL2ExtensionTests/egl_image_external/TestSimpleUnassociated.test
which is part of gles2/3 conformance suite. Here image external
textures are switched to be treated the same as 2D textures. These
can be associated with the fallback texture providing fixed sample
values of (0, 0, 0, 1).
The O
On 8 October 2013 23:25, Jordan Justen wrote:
> Patches 1, 2 & 4 - 10:
> Reviewed-by: Jordan Justen
>
> For 8, I replied with a non-essential question, and the same question
> is relevant in patch 10.
>
> For patch 3 I had that question (mentioned on irc) about ir having
> ast/parser knowledge.
On Mit, 2013-10-09 at 16:39 +0200, Marek Olšák wrote:
> No, I just added the test.
Which means you didn't address Tom's LLVM coding style comment,
requiring C++ formatting (// instead of /**/) for the comment below:
> >> +/* Lane means which subreg of %VGPRa_VGPRb_VGPRc_VGPRd is used.
> >>
On 8 October 2013 21:17, Jordan Justen wrote:
> On Fri, Sep 27, 2013 at 12:05 PM, Paul Berry
> wrote:
> > Unsized arrays appearing inside named interface blocks now get a
> > proper size assigned by the array_sizing_visitor.
> >
> > Fixes piglit tests:
> > - spec/glsl-1.50/execution/unsized-in-n
On 13-10-07 05:28 PM, Kenneth Graunke wrote:
> On 10/01/2013 02:07 PM, Kenneth Graunke wrote:
>> On 10/01/2013 06:31 AM, Gaetan Nadon wrote:
>>> On 13-09-28 04:23 PM, Emil Velikov wrote:
On 28/09/13 20:00, Gaetan Nadon wrote:
> The EGL library has some references to x11 but it gets the lin
No, I just added the test.
Marek
On Wed, Oct 9, 2013 at 4:28 PM, Christian König wrote:
> Are there any changes to the original patch you've send out?
>
> Anyway this version is: Reviewed-by: Christian König
>
>
> Am 09.10.2013 15:33, schrieb Marek Olšák:
>
>> From: Marek Olšák
>>
>> This fixe
From: Christian König
Calling radeon_drm_cs_flush from multiple threads might cause deadlocks,
fix this by immediately signaling the semaphore after waiting for it.
This is a candidate for the stable branch(es).
Partially fixes: https://bugs.freedesktop.org/show_bug.cgi?id=70123
v2: some fixes
https://bugs.freedesktop.org/show_bug.cgi?id=70322
Ian Romanick changed:
What|Removed |Added
Priority|high|medium
Assignee|mesa-dev@lists.
https://bugs.freedesktop.org/show_bug.cgi?id=70322
Ian Romanick changed:
What|Removed |Added
Status|NEW |NEEDINFO
--- Comment #1 from Ian Romanick
Are there any changes to the original patch you've send out?
Anyway this version is: Reviewed-by: Christian König
Am 09.10.2013 15:33, schrieb Marek Olšák:
From: Marek Olšák
This fixes piglit:
- shaders/glsl-fs-texture2d-masked
- shaders/glsl-fs-texture2d-masked-4
Signed-off-by: Marek Olš
On 10/08/2013 02:59 PM, Zack Rusin wrote:
As we're moving towards expanding the number of subpixel
bits and the width of the variables used in the computations
we need to make this code a bit more centralized.
Signed-off-by: Zack Rusin
---
src/gallium/drivers/llvmpipe/lp_rast.h | 9
On 10/08/2013 01:22 PM, Zack Rusin wrote:
only 8 and 32 bit integers were supported before.
Signed-off-by: Zack Rusin
---
src/gallium/auxiliary/gallivm/lp_bld_printf.c | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_printf.c
b/sr
https://bugs.freedesktop.org/show_bug.cgi?id=70322
wim changed:
What|Removed |Added
Hardware|Other |x86-64 (AMD64)
OS|All
https://bugs.freedesktop.org/show_bug.cgi?id=70322
Priority: medium
Bug ID: 70322
Assignee: mesa-dev@lists.freedesktop.org
Summary: glsl: uniform vars turns undefined in mesa 9.1.4; mesa
9.0 OK
Severity: normal
Classifica
https://bugs.freedesktop.org/show_bug.cgi?id=70123
--- Comment #7 from Christian König ---
(In reply to comment #5)
> Created attachment 87341 [details]
> Backtrace of compton deadlock
Are you sure that this is the whole output of "thread apply all bt"?
There is only one thread shown and that's
https://bugs.freedesktop.org/show_bug.cgi?id=70123
--- Comment #6 from Jeff Blake ---
I should have re-disabled compton after applying your patch, as I can start up
without it and crash when I run it. Before the patch disabling compton had no
effect and things froze anyway.
--
You are receiving
https://bugs.freedesktop.org/show_bug.cgi?id=70123
Shawn Starr changed:
What|Removed |Added
CC||shawn.st...@rogers.com
--
You are receivi
https://bugs.freedesktop.org/show_bug.cgi?id=70123
--- Comment #5 from Jeff Blake ---
Created attachment 87341
--> https://bugs.freedesktop.org/attachment.cgi?id=87341&action=edit
Backtrace of compton deadlock
Backtrace of compton deadlock
gdb attach
gdb thread apply all bt
--
You are rece
Otherwise the gen6 w/a in the kernel won't kick in and the write will
land nowhere.
Inspired by a patch Ken pointed me at which had the same issue (but
isn't yet merged and also for a gen7+ feature). An audit of the entire
driver didn't reveal any other case than the one in in the write_reg
helper
From: Marek Olšák
This fixes piglit:
- shaders/glsl-fs-texture2d-masked
- shaders/glsl-fs-texture2d-masked-4
Signed-off-by: Marek Olšák
---
lib/Target/R600/SIISelLowering.cpp | 27 +++--
test/CodeGen/R600/llvm.SI.sample-masked.ll | 93 ++
2 files changed
I reviewed and assumed that you don't have commit access so pushed the
whole series.
Thanks allot for the help,
Christian.
Am 09.10.2013 02:23, schrieb Grigori Goronzy:
UVD can only support NV12 in the case of hardware decoding, but we
can still use all other formats for software decoding. Use
Hi all,
I've been digging through the i965 driver implementation attempting to get my
head around it. I have a few questions which I hope can be answered:
* What is LIR? The comments say that Mesa GLSL IR is converted into LIR
which in turn is converted into GPU code. What is LIR and how i
On 10/08/2013 11:03 PM, Eric Anholt wrote:
Tapani Pälli writes:
Patch introduces ir_cache class that can serialize a gl_shader
to memory with help of memory_writer class and also unserialize
it back with help of memory_map class.
This can be used by the shader compiler to cache individual sha
bo_handles_mutex should also be renamed.
Marek
On Tue, Oct 8, 2013 at 12:48 AM,
wrote:
> From: Christopher James Halse Rogers
>
> It's a map of GEM name->bo, so identify it as such
> ---
> src/gallium/winsys/radeon/drm/radeon_drm_bo.c | 16
> 1 file changed, 8 insertions(+), 8
You should use --with-llvm-shared-libs if you don't use it already.
Marek
On Wed, Oct 9, 2013 at 1:54 AM, Mike Lothian wrote:
> Just had a failure because of this - should the default for shared libs not
> be yes now?
>
>
> On 7 October 2013 13:27, Tom Stellard wrote:
>>
>> On Mon, Oct 07, 2013
Haven't looked in detail at the patch, but I'm pretty sure that this
goes into the right direction.
Christian.
Am 08.10.2013 23:53, schrieb Marek Olšák:
Sorry, I didn't realize this. You are right. My bad. Regardless of my
comment, I still wonder what people's opinion on this patch is.
Marek
From: Chia-I Wu
Add fs_visitor::emit_dual_texture_gen7 that emulate SIMD16 sample_d with dual
SIMD8 sample_d on gen7+.
No piglit quick.tests regression on Ivy Bridge and Haswell.
Improved Xonotic with Ultra effects by 6.76479% +/- 0.619064% (at 95.0%
confidence) on Haswell.
Signed-off-by: Chia
From: Chia-I Wu
Add fs_visitor::emit_texture, which is used to emit the texture instruction
after the message payload has been set up.
Signed-off-by: Chia-I Wu
---
src/mesa/drivers/dri/i965/brw_fs.h | 10 +-
src/mesa/drivers/dri/i965/brw_fs_fp.cpp | 13 ++-
src/mesa/drivers/dr
From: Chia-I Wu
FS_OPCODE_OVERWRITE_DST is used to indicate that the destination register is
(completely) overwritten. No code is emitted, but the liveness analysis can
use it as a hint to add the destination register to DEF bitset. This is
needed because it is hard to figure out if some partia
Patches 1, 2 & 4 - 10:
Reviewed-by: Jordan Justen
For 8, I replied with a non-essential question, and the same question
is relevant in patch 10.
For patch 3 I had that question (mentioned on irc) about ir having
ast/parser knowledge. I guess I'd like to think about this one a
little more.
-Jord
On 09.10.2013 00:22, Chad Versace wrote:
On 10/08/2013 12:51 AM, Knut Petersen wrote:
[ 362.708791] thunderbird-bin[1888]: segfault at 8 ip b0f3f3aa sp bf93fd54
error 4 in i915_dri.so[b0a02000+fc]
[ 846.063129] firefox[2008]: segfault at 8 ip b0af63aa sp bff96c74 error 4 in
i915_dri.so[
This patch fixes this build error with Oracle Solaris Studio.
libtool: link: /opt/solarisstudio12.3/bin/cc -g -o glcpp/glcpp glcpp.o
prog_hash_table.o ./.libs/libglcpp.a
Undefined first referenced
symbol in file
sqrt
On Wed, Oct 9, 2013 at 3:35 AM, Eric Anholt wrote:
> Chia-I Wu writes:
>
>> From: Chia-I Wu
>>
>> FS_OPCODE_OVERWRITE_DST is used to indicate that the destination register is
>> (completely) overwritten. No code is emitted, but the liveness analysis can
>> use it as a hint to add the destinatio
68 matches
Mail list logo
