gmane.comp.video.mesa3d.devel

Re: [PATCH] i830_state:i830Enable(): Fix crash for GL_STENCIL_TEST

Kurt Roeckx — 2012-05-25T23:26:04

Can someone please review and commit this?  A lot of the piglit
tests fail because of this.


Kurt

On Fri, May 11, 2012 at 12:19:42AM +0200, Kurt Roeckx wrote:

Re: [PATCH 4/6] meta: Cleanup the resources we allocate.

Ian Romanick — 2012-05-25T22:28:08

Reviewed-by: Ian Romanick <ian.d.romanick< at >intel.com>

This should also be tagged with:

NOTE: This is a candidate for the 8.0 branch.

Re: [PATCH 2/6] glx: Handle a null reply in QueryVersion.

Ian Romanick — 2012-05-25T22:21:34

Reviewed-by: Ian Romanick <ian.d.romanick< at >intel.com>

This should also be tagged with:

NOTE: This is a candidate for the 8.0 branch.

Re: [PATCH 6/6] mesa: Fix hash table leak

Ian Romanick — 2012-05-25T22:20:44

This looks good, but it should return bool instead of int.

Re: [PATCH 5/6] mesa: Free uniforms correclty.

Ian Romanick — 2012-05-25T22:19:45

Reviewed-by: Ian Romanick <ian.d.romanick< at >intel.com>

This should also be tagged with:

NOTE: This is a candidate for the 8.0 branch.

Re: [PATCH 2/2] mesa: glsl: allow ES2 to report support for GLSL versions >= 1.20

Ian Romanick — 2012-05-25T22:14:32

Do we even need that?  It looks like the shader uses 'out' so that it 
can have an ivec4 output for clearing an integer render target.  We 
don't have those in ES2 contexts.  Right?

Re: [PATCH] mesa: Remove the generated glapi from source control, and just build it.

Ian Romanick — 2012-05-25T22:12:05

I'm not going to block these changes from landing.

I'm putting together a branch with a big pile of unit tests that I think 
should catch most of catastrophic cases that I'm worried about.  I 
already have tests that verify

   src/glx/indirect_init.c
   src/glx/indirect_size.c
   src/mapi/glapi/glapitable.h
   src/mapi/glapi/glprocs.h
   src/mesa/main/enums.c

I even found some (really minor) bugs along the way.

The main things left that worry me are src/glx/indirect.c and the 
various generated assembly files.  indirect.c is a pain in the ass for 
unit tests.  As soon as you link with it, it has to pull in glxcmds.c 
and glxext.c, and the world explodes.

Re: [PATCH 2/2] mesa: glsl: allow ES2 to report supportfor GLSL versions >= 1.20

Eric Anholt — 2012-05-25T20:13:13

I don't think we should do this -- aren't you letting someone do
#version 130 in their ES2 context?

In the past, for "let core Mesa do desktop GL things in an ES context",
we've put some flag in place across linking.  But in this case, I think
the right answer is to just have a separate shader string in the meta
code for the ES2 case.
_______________________________________________
mesa-dev mailing list
mesa-dev< at >lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

Re: [PATCH] - main/ff_fragment_shader.cpp

Eric Anholt — 2012-05-25T19:45:45

I think we just need to turn off that gcc warning.  We're not using
c++11, so it shouldn't be griping about it.

_______________________________________________
mesa-dev mailing list
mesa-dev< at >lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

Re: [PATCH] r600g: handle R16G16B16_FLOAT andR32G32B32_FLOAT in translate_colorswap

Marek Olšák — 2012-05-25T19:29:32

Pushed, thanks.

Marek

On Fri, May 25, 2012 at 4:27 PM, Kai Wasserbäch
<kai< at >dev.carbon-project.org> wrote:
_______________________________________________
mesa-dev mailing list
mesa-dev< at >lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

[Bug 50318] [r600g] Piglit: spec/ARB_texture_float/get-renderbuffer-internalformat: unsupported colorswap formats 30 and 93

2012-05-25T19:13:12

https://bugs.freedesktop.org/show_bug.cgi?id=50318

Kai <kai< at >dev.carbon-project.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED

--- Comment #3 from Kai <kai< at >dev.carbon-project.org> 2012-05-25 12:13:12 PDT ---
Fixed by
http://cgit.freedesktop.org/mesa/mesa/commit/?id=2df2c31087770a6e4e79f90fe096e370cd632e00

[PATCH] - main/ff_fragment_shader.cpp

Han Shen(沈涵 — 2012-05-25T17:51:31

Hi I am working on chromium, and while I was compiling mesa trunk
using gcc 4.7, I had to add explicit type cast for narrower conversion
inside the initialization list.

Could you take a look at the patch and upstream it if it is ok to you? Thanks,
-Han

diff --git a/src/mesa/main/ff_fragment_shader.cpp
b/src/mesa/main/ff_fragment_shader.cpp
index 0233f38..5a6041a 100644
--- a/src/mesa/main/ff_fragment_shader.cpp
+++ b/src/mesa/main/ff_fragment_shader.cpp
< at >< at > -887,10 +887,10 < at >< at > emit_texenv(struct texenv_fragment_program *p,
GLuint unit)
       }
       else {
  float const_data[4] = {
-    1 << rgb_shift,
-    1 << rgb_shift,
-    1 << rgb_shift,
-    1 << alpha_shift
+    static_cast<float>(1 << rgb_shift),
+    static_cast<float>(1 << rgb_shift),
+    static_cast<float>(1 << rgb_shift),
+    static_cast<float>(1 << alpha_shift)
  };
  shift = new(p->mem_ctx) ir_constant(glsl_type::vec4_type,
      (ir_constant_data *)const_data);
_______________________________________________

Re: No more make clean!

Dan Nicholson — 2012-05-25T16:51:07

for a few months now and I'm mostly happy with it.
to the command line, but the -O2 flag is still there and that sometimes
causes trouble in gdb.
--enable-debug option to remove -O2?

In my own automake branch that I stalled on I had the idea of turning the
static configs into configure wrapper scripts that just help you do known
things like set debug flags. Any interest?

Dan
_______________________________________________
mesa-dev mailing list
mesa-dev< at >lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

[PATCH] draw: simplify index buffer specification

Brian Paul — 2012-05-25T16:35:44

Replace draw_set_index_buffer() and draw_set_mapped_index_buffer() with
draw_set_indexes() which simply takes a pointer and an index size.
---
 src/gallium/auxiliary/draw/draw_context.c       |   30 ++++++++++------------
 src/gallium/auxiliary/draw/draw_context.h       |    7 +----
 src/gallium/auxiliary/draw/draw_private.h       |    2 -
 src/gallium/auxiliary/draw/draw_pt.c            |   22 ++++------------
 src/gallium/auxiliary/draw/draw_pt_vsplit_tmp.h |    6 +---
 src/gallium/drivers/i915/i915_context.c         |    6 +++-
 src/gallium/drivers/i915/i915_state.c           |    3 --
 src/gallium/drivers/llvmpipe/lp_draw_arrays.c   |    8 ++++--
 src/gallium/drivers/llvmpipe/lp_state_vertex.c  |    2 -
 src/gallium/drivers/nv30/nv30_draw.c            |    7 +++--
 src/gallium/drivers/r300/r300_state.c           |   12 ++++----
 src/gallium/drivers/softpipe/sp_draw_arrays.c   |    7 ++++-
 src/gallium/drivers/softpipe/sp_state_vertex.c  |    2 -
 src/gallium/drivers/svga/svga_swtnl_draw.c      |    9 +++

Re: [PATCH] gallium: add st_api feature mask to prevent advertising MS visuals

Roland Scheidegger — 2012-05-25T16:12:37

Am 25.05.2012 13:27, schrieb Christoph Bumiller:
You're right indeed. That is odd I don't think any driver from ATI ever
exposed this. Maybe it doesn't quite work right (or the cost is too
high) or it just wasn't deemed too useful. But in any case checking for
odd sizes shouldn't hurt much neither.

Roland

Re: [PATCH] gallium/targets: pass ldflags parameter to MKLIB

Brian Paul — 2012-05-25T15:36:41

Looks good.  I'll push it shortly.

-Brian
_______________________________________________
mesa-dev mailing list
mesa-dev< at >lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev

[Bug 50318] [r600g] Piglit: spec/ARB_texture_float/get-renderbuffer-internalformat: unsupported colorswap formats 30 and 93

2012-05-25T14:30:16

https://bugs.freedesktop.org/show_bug.cgi?id=50318

--- Comment #2 from Kai <kai< at >dev.carbon-project.org> 2012-05-25 07:30:16 PDT ---
(In reply to comment #1)

Thanks for the pointer, I've sent a patch for review to mesa-dev:
<http://thread.gmane.org/gmane.comp.video.mesa3d.devel/40804>

[PATCH] r600g: handle R16G16B16_FLOAT andR32G32B32_FLOAT in translate_colorswap

Kai Wasserbäch — 2012-05-25T14:27:08

Fixes https://bugs.freedesktop.org/show_bug.cgi?id=50318

Signed-off-by: Kai Wasserbäch <kai< at >dev.carbon-project.org>
---
 Verified with a full Piglit run on my HW (see the bug).
 Thanks to Marek for pointing out b2d63860 to me!

 In case you accept this patch: please commit it for me, I don't have access.

 src/gallium/drivers/r600/r600_state.c |    2 ++
 1 file changed, 2 insertions(+)

diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c
index ed08fd6..124eba2 100644
--- a/src/gallium/drivers/r600/r600_state.c
+++ b/src/gallium/drivers/r600/r600_state.c
< at >< at > -281,6 +281,8 < at >< at > static uint32_t r600_translate_colorswap(enum pipe_format format)
 case PIPE_FORMAT_R16G16_FLOAT:
 case PIPE_FORMAT_R16G16_UINT:
 case PIPE_FORMAT_R16G16_SINT:
+case PIPE_FORMAT_R16G16B16_FLOAT:
+case PIPE_FORMAT_R32G32B32_FLOAT:
 case PIPE_FORMAT_R32_UINT:
 case PIPE_FORMAT_R32_SINT:
 case PIPE_FORMAT_R32_FLOAT:

[PATCH 13/13] r600g: compute support for evergreen v2

Tom Stellard — 2012-05-25T13:34:05

From: Adam Rak <adam.rak< at >streamnovation.com>

Tom Stellard:
  - Updated for gallium interface changes
  - Fixed a few bugs:
    + Set the loop counter
    + Calculate the correct number of pipes
  - Added hooks into the LLVM compiler

v2: Tom Stellard
  - Rebase on top of other changes
---
 configure.ac                                       |    7 +-
 src/gallium/drivers/r600/Makefile.am               |   18 +-
 src/gallium/drivers/r600/Makefile.sources          |    6 +-
 src/gallium/drivers/r600/compute_memory_pool.c     |  391 +++++++++
 src/gallium/drivers/r600/compute_memory_pool.h     |   78 ++
 src/gallium/drivers/r600/compute_resource.def      |   38 +
 src/gallium/drivers/r600/evergreen_compute.c       |  808 +++++++++++++++++++
 src/gallium/drivers/r600/evergreen_compute.h       |   69 ++
 .../drivers/r600/evergreen_compute_internal.c      |  830 ++++++++++++++++++++
 .../drivers/r600/evergreen_compute_internal.h      |  118 +++
 src/gallium/drivers/r600/evergreen_state.c         |    3 +
 src/gall

Re: No more make clean!

Brian Paul — 2012-05-25T14:22:50

I'm fine with removing the old/static configs.  I've been using 
autoconf for a few months now and I'm mostly happy with it.

One thing I wish was different is the --enable-debug option.  It adds 
-g to the command line, but the -O2 flag is still there and that 
sometimes causes trouble in gdb.

To fix that I have to do "CFLAGS=-g ./autogen ...."   Could we fix the 
--enable-debug option to remove -O2?

-Brian

[PATCH 12/13] clover: Add function for building aclover::module for non-TGSI targets v4

2012-05-25T13:31:42

From: Tom Stellard <thomas.stellard< at >amd.com>

v2:
  -Separate IR type and LLVM triple
  -Do the OpenCL C->LLVM IR and linking steps for all PIPE_SHADER_IR
   types.

v3:
  - Coding style fixes
  - Removed compatibility code for LLVM < 3.1
  - Split build_module_llvm() into three functions:
    compile(), link(), and build_module_llvm()

v4:
  - Use struct pipe_compute_program
---
 .../state_trackers/clover/core/compiler.hpp        |    2 +
 src/gallium/state_trackers/clover/core/program.cpp |    9 +-
 .../state_trackers/clover/llvm/invocation.cpp      |  167 ++++++++++++++++++--
 3 files changed, 164 insertions(+), 14 deletions(-)

diff --git a/src/gallium/state_trackers/clover/core/compiler.hpp b/src/gallium/state_trackers/clover/core/compiler.hpp
index 686c7d8..a43050a 100644
--- a/src/gallium/state_trackers/clover/core/compiler.hpp
+++ b/src/gallium/state_trackers/clover/core/compiler.hpp
< at >< at > -25,6 +25,7 < at >< at >
 
 #include "core/compat.hpp"
 #include "core/module.hpp"
+#include "pipe/p_defines.h"
 
 namespace