fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-24 06:18:10 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1
47726 commits 66 branches 936 tags 985 MiB
Commit graph

14700 commits

Author SHA1 Message Date
Vadim Girlin
dcdc062dda r600g: fix op3 & write in merge_inst_groups 2011-11-14 11:11:31 -05:00
Alex Deucher
ebecbbc2e6 r600g: set max max tex/vtx instructions count to 16 for cayman 
Cayman is 16 as well.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
 2011-11-14 11:01:00 -05:00
Vadim Girlin
c32ca6d9c1 r600g: set max tex/vtx instructions count to 16 for evergreen 
According to evergreen-isa doc 16 is max value for evergreen.
More than 16 doesn't work for me.

Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
 2011-11-14 10:59:53 -05:00
José Fonseca
ab3ace56c4 llvmpipe: Silent warnings about undeclared llvmpipe_check_render_cond. 2011-11-14 13:50:33 +00:00
José Fonseca
6246c217ec util: Avoid signed/unsigned comparison in u_trim_pipe_prim(). 2011-11-14 10:06:01 +00:00
José Fonseca
d7edd5db31 llvmpipe: Remove unused variables. 
Reviewed-by: Brian Paul <brianp@vmware.com>
 2011-11-14 10:06:01 +00:00
José Fonseca
c88f3e0374 llvmpipe: Make more resilient to out-of-memory situations. 
Most of the code was alright, but we were missing a few paths.

Reviewed-by: Brian Paul <brianp@vmware.com>
 2011-11-14 10:06:01 +00:00
José Fonseca
9e29cdbe95 draw: Handle failure to allocate aligned_constant_storage. 
Also, actually update const_storage_size, therefore avoiding to
unnecessarily reallocate aligned_constant_storage every single time
draw_vs_set_constants() is called.

Reviewed-by: Brian Paul <brianp@vmware.com>
 2011-11-14 10:06:00 +00:00
José Fonseca
e21c5157b6 gallivm: Remove duplicate statement. 
ary_ge_arx_arz is already set earlier.

Reviewed-by: Brian Paul <brianp@vmware.com>
 2011-11-14 10:06:00 +00:00
José Fonseca
34930facfe gallivm: Include stddef.h before the LLVM C++ headers. 
Necessary with build against LLVM 2.6, with recent gcc, as LLVM headers
depend on ptrdiff_t but don't properly include stddef.h
 2011-11-14 10:06:00 +00:00
Dave Airlie
b04d19da10 llvmpipe: fix unswizzle of packed float types. 
I messed up adding the ubyte->float conversion.

This fixes getteximage-formats

https://bugs.freedesktop.org/show_bug.cgi?id=42837

Signed-off-by: Dave Airlie <airlied@redhat.com>
 2011-11-14 09:35:17 +00:00
Alex Deucher
579c04e42e r600g: properly handle cayman in is_alu_vec_unit_inst() 
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
 2011-11-13 17:46:01 -05:00
Vadim Girlin
3d441153a1 r600g: fix cb offset for flushed_depth_texture 
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
 2011-11-13 17:27:16 -05:00
Vadim Girlin
d5e91a2364 r600g: fix stencil buffer ref counting on evergreen 
Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
 2011-11-13 17:25:16 -05:00
Vadim Girlin
8e366dc365 r600g: lazy load for AR register 
Emit MOVA* instruction only when AR is used.

Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
 2011-11-13 17:10:02 -05:00
Vadim Girlin
88a140cd19 r600g: include INTERP_[XY|ZW] in is_alu_vec_unit_inst 
This will disallow moving them to the trans slot in merge_inst_groups

Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
 2011-11-13 17:09:59 -05:00
Brian Paul
871dc64869 svga: don't crash/assert if we fail to allocate a vertex buffer 
v2: check if pipe_buffer_map() returns NULL, and return NULL from
svga_vbuf_render_map_vertices().  Per Jose's suggestion.

Reviewed-by: José Fonseca <jfonseca@vmware.com>
 2011-11-11 07:10:56 -07:00
Brian Paul
aa66130682 draw: handle out of memory conditions 
If the vbuf backend fails to allocate a vertex buffer, don't crash
or assert.
 2011-11-11 07:10:56 -07:00
Brian Paul
7288bfb2a4 util: check for null vertex buffer object in blit code 
Don't crash if we fail to allocate a vertex buffer.
 2011-11-11 07:10:56 -07:00
Marek Olšák
b8428e6335 r600g: don't set the query result to 0 after a call to get_query_result 
get_query_result doesn't reset the result. Only begin_query does.
 2011-11-11 00:08:01 +01:00
Marek Olšák
72c1658554 r600g: the type of OCCLUSION_PREDICATE result should be boolean 2011-11-11 00:03:52 +01:00
Brian Paul
3da5196263 radeon: silence initializer warnings 2011-11-10 16:00:46 -07:00
Brian Paul
1462114475 draw/llvm: silence uninitialized variable warnings 2011-11-10 16:00:46 -07:00
Marek Olšák
b999be8374 r600g: implement PIPE_QUERY_OCCLUSION_PREDICATE 2011-11-10 23:11:57 +01:00
Marek Olšák
083482d493 r300g: fix query result of GPU_FINISHED 2011-11-10 22:58:34 +01:00
Marek Olšák
c5ae81652d r300g: implement PIPE_QUERY_GPU_FINISHED 2011-11-10 22:53:54 +01:00
Marek Olšák
014b3aa07d r300g: implement PIPE_QUERY_OCCLUSION_PREDICATE 2011-11-10 22:53:54 +01:00
Dave Airlie
4fd485666a llvmpipe/u_format: add support for EXT_texture_shared_exponent + EXT_packed_float 
These two are fairly unique types so add specific cases for decoding them.

Passes piglit fbo-clear-format and fbo-generatemipmap-format tests for these
two extensions.

Signed-off-by: Dave Airlie <airlied@redhat.com>
 2011-11-10 20:37:55 +00:00
Dave Airlie
ab14915dce llvmpipe: add NV_conditional_render support. 
This ports the softpipe NV_conditional_render support to llvmpipe.

This passes the nv_conditional_render-* piglit tests.

Signed-off-by: Dave Airlie <airlied@redhat.com>
 2011-11-10 20:32:13 +00:00
Marek Olšák
dbd24b5df6 r600g: remove useless texture barrier 2011-11-10 18:49:26 +01:00
Marek Olšák
ec3d2c7f08 r600g: flush the whole context when CS is full, not just hw_context 
Because we disable render condition in r600_flush, but not in r600_context_flush.
 2011-11-10 18:48:23 +01:00
Marek Olšák
6a490149d9 r600g: turn some CS overflow checks into assertions 
They're not really exhaustive and not so useful either.
 2011-11-10 18:09:10 +01:00
Marek Olšák
5222e16e2b r600g: put the rest of CS overflow checks in r600_need_cs_space 2011-11-10 18:09:10 +01:00
Marek Olšák
30a570e450 r600g: flush caches regardless of render condition 
What if somebody enables render condition just before we flush...
 2011-11-10 18:09:10 +01:00
Marek Olšák
578b211be6 r600g: disable render condition at the end of CS, re-enable at the beginning 2011-11-10 18:09:10 +01:00
Marek Olšák
9564185b0a r600g: don't suspend queries if they have already been suspended by blitter 
And add some assertions.
 2011-11-10 18:09:10 +01:00
Marek Olšák
cb7c6c30d0 r600g: properly reserve CS space for queries_suspend 2011-11-10 18:09:10 +01:00
Marek Olšák
1d09831253 r600g: reserve CS space for a draw command in begin_query and render_condition 
There's no point in emitting those if you can't emit a draw command too.
 2011-11-10 18:09:10 +01:00
Marek Olšák
11bdd28796 r600g: consolidate checking whether CS is full 
This adds a new function r600_need_cs_space. Currently, it's easy to overflow
the CS - queries are not counted in. I guess that's not the only case where
the driver may crap out.
 2011-11-10 18:09:09 +01:00
Marek Olšák
60302f83e2 r600g: clarify meaning of one variable in shader codegen 
It's easier to read now.
 2011-11-09 00:18:53 +01:00
Marek Olšák
1b375f8413 r600g: cosmetic changes in query code 
Mainly updating comments and removing one use of a magic number.
 2011-11-09 00:18:53 +01:00
Marek Olšák
8187ad0754 r600g: use modulo for computing index into query (ring) buffers 2011-11-09 00:18:53 +01:00
Marek Olšák
552e90bd6a r600g: make r600_query_result more generic 
We'll soon start adding new query types, maybe even querying more than
one value per query.
 2011-11-09 00:18:52 +01:00
José Fonseca
4eb3225b38 Remove tgsi_sse2. 
tgsi_exec is simple. llvm is fast. tgsi_sse2 ends up being neither.
 2011-11-08 22:57:34 +00:00
Morgan Armand
0a0d820f7b softpipe: don't clamp or do logical operations on floating-point buffers. 
Signed-off-by: Brian Paul <brianp@vmware.com>
 2011-11-08 08:36:39 -07:00
Vinson Lee
f4515d106c st/dri: Fix memory leak on error path. 
Fixes Coverity resource leak defect.

Reviewed-by: José Fonseca <jfonseca@vmware.com>
 2011-11-07 20:36:07 -08:00
Chia-I Wu
a69da5c0ce util: add log2f for Android 
It is needed for nv50's new shader backend.  With this change, both u_math.h
and imports.h in core mesa define the same function.  I have to #undef log2f
here to avoid the conflict.  Not sure if there is a better way to deal with
the situation.

Acked-by: José Fonseca <jfonseca@vmware.com>
 2011-11-07 17:04:11 -07:00
Morgan Armand
89d6044b7b softpipe: fix memory leaks 
This series of patches is a splitted version of my previous one, as suggested by Brian.

Signed-off-by: Brian Paul <brianp@vmware.com>
 2011-11-07 13:25:46 -07:00
Dave Airlie
0d8deb5bc9 llvmpipe: fix typo in the depth sampling aos code. 
Just found by reading llvmpipe code for no great reason.

Signed-off-by: Dave Airlie <airlied@redhat.com>
 2011-11-06 22:32:04 +00:00
Dave Airlie
73c6657596 llvmpipe: enable RGTC after u_format fix. 
The two piglit tests pass + render correctly.

Signed-off-by: Dave Airlie <airlied@redhat.com>
 2011-11-06 22:32:04 +00:00