Dave Airlie [Fri, 10 Sep 2010 01:27:31 +0000 (11:27 +1000)]
r600g: align flushing of cb/db with DDX/r600c.
the DDX and r600c both flush cb/db after the draw is emitted,
as long as they do that, r600g can't be different, as it races.
We end up with r600g flush, set CB, DDX set CB, flush. This
was causing misrendering on my evergreen, where sometimes the drawing
would go to an old CB.
Dave Airlie [Fri, 10 Sep 2010 01:22:41 +0000 (11:22 +1000)]
r600g: don't need 3 bos here.
the code should reloc correctly a single BO 3 times.
Brian Paul [Fri, 10 Sep 2010 01:18:45 +0000 (19:18 -0600)]
graw: added test for finding shader mem leaks
Brian Paul [Fri, 10 Sep 2010 00:52:27 +0000 (18:52 -0600)]
draw: minor reformatting
Brian Paul [Fri, 10 Sep 2010 00:52:10 +0000 (18:52 -0600)]
graw: emit warnings when context/surface creation failes
Brian Paul [Fri, 10 Sep 2010 00:50:49 +0000 (18:50 -0600)]
winsys: emit warning in null_sw_displaytarget_create()
Brian Paul [Thu, 9 Sep 2010 21:11:37 +0000 (15:11 -0600)]
graw: fix array size, indentation,
Jakob Bornecrantz [Thu, 9 Sep 2010 23:16:19 +0000 (01:16 +0200)]
glsl2: Fix scons build for all platforms
Kristian Høgsberg [Thu, 9 Sep 2010 22:59:49 +0000 (18:59 -0400)]
Revert "glapi: Implement optional dispatch logging"
This reverts commit
b9abc6139a310677a37754ea7172d976dbf56979 and the
follow on fixes (
7aae704 and
6fe1b47). It's changing the glapi/driver
ABI and causes a number of problems for debug/non-debug builds.
Ian Romanick [Thu, 2 Sep 2010 04:23:52 +0000 (21:23 -0700)]
glsl2: Implement noise[1234] built-in functions using ir_unop_noise
Ian Romanick [Thu, 9 Sep 2010 22:25:32 +0000 (15:25 -0700)]
glsl2: Add EmitNoNoise flag, use it to remove noise opcodes
Ian Romanick [Thu, 9 Sep 2010 22:20:09 +0000 (15:20 -0700)]
glsl2: Add lowering pass to remove noise opcodes
Ian Romanick [Thu, 2 Sep 2010 04:12:10 +0000 (21:12 -0700)]
glsl2: Add ir_unop_noise
Kenneth Graunke [Thu, 9 Sep 2010 22:15:31 +0000 (15:15 -0700)]
glsl/builtins: normalize of a negative scalar should be -1.0.
Kristian Høgsberg [Thu, 9 Sep 2010 22:00:44 +0000 (18:00 -0400)]
mesa: Only reference logging symbols in debug builds
Kristian Høgsberg [Thu, 9 Sep 2010 21:08:12 +0000 (17:08 -0400)]
mesa: Don't reuse DummyFramebuffer as the incomplete framebuffer
Binding framebuffer 0 on a context that doesn't have a winsys drawable
will try to bind the incomplete framebuffer. That fails when that's
also the dummy framebuffer.
Jakob Bornecrantz [Thu, 9 Sep 2010 20:46:49 +0000 (22:46 +0200)]
glapi: Fix non-debug builds
ben [Thu, 26 Aug 2010 15:32:32 +0000 (17:32 +0200)]
st/dri: support EGL_MESA_image_drm: queryImage
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
ben [Thu, 26 Aug 2010 15:31:42 +0000 (17:31 +0200)]
st/dri: support EGL_MESA_image_drm: createImage
Signed-off-by: Jakob Bornecrantz <wallbraker@gmail.com>
Kristian Høgsberg [Thu, 9 Sep 2010 19:00:02 +0000 (15:00 -0400)]
eglglx: Convert glx visuals/fbconfigs straight to EGL configs
In other words, skip the __GLcontextModes middle man.
Kristian Høgsberg [Thu, 9 Sep 2010 18:05:30 +0000 (14:05 -0400)]
egl: Remove old egldri driver
This driver doesn't work with any of the DRI drivers in the source tree.
Kristian Høgsberg [Thu, 9 Sep 2010 16:59:14 +0000 (12:59 -0400)]
glapi: Implement optional dispatch logging
There's a useful feature buried in glapi to log all API calls to stderr.
Unfortunately it requires editing the code and then it's enabled
unconditionally for that build. This patch builds in API logging for
debug builds and makes it run-time switchable by setting MESA_DEBUG=dispatch.
Kristian Høgsberg [Thu, 9 Sep 2010 17:18:40 +0000 (13:18 -0400)]
glx: Optimize out no-op make current calls
This make a lot more sense now that we might have to recreate the
glx drawables for legacy code paths.
Chia-I Wu [Thu, 9 Sep 2010 16:06:32 +0000 (00:06 +0800)]
egl: Use _EGL_CHECK_DISPLAY in eglCreateContext.
_EGL_CHECK_DISPLAY checks the display and returns from eglCreateContext
on error.
Chia-I Wu [Thu, 9 Sep 2010 16:02:47 +0000 (00:02 +0800)]
egl: Display may be NULL in _eglLookup*.
This fixes several NULL dereferences.
Brian Paul [Thu, 9 Sep 2010 16:03:46 +0000 (10:03 -0600)]
llvmpipe: remove redundant tgsi_dup_tokens() call
We were calling this twice so the first allocation was orphaned/leaked.
Francisco Jerez [Thu, 9 Sep 2010 12:07:38 +0000 (14:07 +0200)]
dri/nouveau: Expose EXT_texture_env_combine.
Francisco Jerez [Thu, 9 Sep 2010 12:14:48 +0000 (14:14 +0200)]
dri/nv10-nv20: Add support for NV_texture_env_combine4.
Francisco Jerez [Thu, 9 Sep 2010 12:14:14 +0000 (14:14 +0200)]
dri/nv04: Add support for NV_texture_env_combine4.
Francisco Jerez [Thu, 2 Sep 2010 00:18:02 +0000 (02:18 +0200)]
dri/nouveau: Minor cleanup.
Kristian Høgsberg [Thu, 9 Sep 2010 12:06:40 +0000 (08:06 -0400)]
glx: Fix another use-after-free problem
Andre Maasikas [Thu, 2 Sep 2010 15:16:42 +0000 (18:16 +0300)]
r600: add TXL instruction and note about TXB
Andre Maasikas [Thu, 2 Sep 2010 08:37:23 +0000 (11:37 +0300)]
r600: remove depth exports move, just set to output x <- z in the export intruction
Andre Maasikas [Thu, 2 Sep 2010 08:09:52 +0000 (11:09 +0300)]
r600: protect cleanup instructions from double free
We might get the cleanup when we have not translated the shader yet
e.g 2 programstringnotifys in a row
Andre Maasikas [Thu, 2 Sep 2010 08:03:38 +0000 (11:03 +0300)]
r600: remove mask from output intructions
in case of relative addressing we never get to know which comps
were really written.
Andre Maasikas [Thu, 2 Sep 2010 07:49:30 +0000 (10:49 +0300)]
r600: allow relative addressing of temps, inputs, outputs
Andre Maasikas [Wed, 1 Sep 2010 07:57:35 +0000 (10:57 +0300)]
r600: handle LIT writemask
Andre Maasikas [Wed, 1 Sep 2010 07:38:07 +0000 (10:38 +0300)]
r600: fix rsq from negative input
arbfp specifies rsq of abs value
Chia-I Wu [Thu, 9 Sep 2010 05:51:59 +0000 (13:51 +0800)]
docs: Update egl.html.
Mention that EGL_DRIVERS_PATH should be specified for uninstalled build.
Update TODOs.
Luca Barbieri [Sun, 5 Sep 2010 20:29:58 +0000 (22:29 +0200)]
glsl: add several EmitNo* options, and MaxUnrollIterations
This increases the chance that GLSL programs will actually work.
Note that continues and returns are not yet lowered, so linking
will just fail if not supported.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Luca Barbieri [Sun, 5 Sep 2010 16:49:54 +0000 (18:49 +0200)]
glsl: make compiler options per-target
This allows us to specify different options, especially useful for chips
without unified shaders.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Luca Barbieri [Sun, 5 Sep 2010 22:56:07 +0000 (00:56 +0200)]
mesa: add PIPE_SHADER_* like constants and conversions to/from enums (v2)
Changes in v2:
- No longer adds tessellation enums
Eric Anholt [Thu, 9 Sep 2010 01:04:35 +0000 (18:04 -0700)]
glsl: Add info about talloc and optimization passes to the README.
Eric Anholt [Thu, 9 Sep 2010 00:49:10 +0000 (17:49 -0700)]
glsl: Update README talking about multi-instruction operations.
The previous thing taking multiple instructions ended up being handled
at the IR level, as we suggested would be the common result. Pick a
new one.
Kristian Høgsberg [Thu, 9 Sep 2010 00:55:02 +0000 (20:55 -0400)]
glx: Destroy pixmap after destroying glx and dri drawables
Now that we suppress BadDrawable from DRI2DestroyDrawable, this doesn't
matter, but we would get that error before when destroying pbuffers.
Kristian Høgsberg [Thu, 9 Sep 2010 00:34:07 +0000 (20:34 -0400)]
configure: Enable -fvisibility=hidden for g++ as well
Kenneth Graunke [Wed, 8 Sep 2010 23:03:33 +0000 (16:03 -0700)]
glsl/builtins: Set the API in the fake context.
Otherwise it gets used uninitialized.
Ian Romanick [Thu, 9 Sep 2010 00:13:00 +0000 (17:13 -0700)]
glsl2: Clear out profile pointers in _mesa_glsl_release_functions
Otherwise builtin_profiles contains dangling pointers the next time
_mesa_read_profile is called. I suspect this may fix bugzilla #29847,
but I was never able to reproduce it.
Ian Romanick [Wed, 8 Sep 2010 17:46:10 +0000 (10:46 -0700)]
mesa: Refactor parameter processing in set_combiner_(operand|source)
The enum values were chosen to have sequential values for a reason.
Use that to compact and simplify the code.
Ian Romanick [Wed, 8 Sep 2010 17:37:54 +0000 (10:37 -0700)]
mesa: Fix handling of texenv operands for EXT vs ARB version
GL_EXT_texture_env_combine has slightly more restrictive limits on the
valid sources for some operands. This wasn't caught before because
almost every driver in Mesa that supports the EXT version also
supports the ARB version.
Inspired by a patch posted the the mesa-dev mailing list by Andrew
Randrianasulu.
Kristian Høgsberg [Wed, 8 Sep 2010 22:54:30 +0000 (18:54 -0400)]
glx: Ignore DRI2 event for drawables we've destroyed
Since we now actually destroy GLX drawables, we get into situations where
we get events for drawables that no longer exist. Just ignore the
event in that case.
Kristian Høgsberg [Wed, 8 Sep 2010 18:51:18 +0000 (14:51 -0400)]
glx: Fix use after free problem
Kenneth Graunke [Wed, 8 Sep 2010 22:38:09 +0000 (15:38 -0700)]
glsl: Fix for scalar float built-in definitions.
These need abs, and we need more tests.
Eric Anholt [Wed, 8 Sep 2010 22:01:02 +0000 (15:01 -0700)]
glsl: regenerate builtins
Eric Anholt [Thu, 9 Sep 2010 05:25:12 +0000 (22:25 -0700)]
glsl: Fix typo in builtin step() using a wrong channel.
Kenneth Graunke [Wed, 8 Sep 2010 06:22:10 +0000 (23:22 -0700)]
ir_validate: Ensure ir_binop_dot is only used on vector types.
Kenneth Graunke [Wed, 8 Sep 2010 06:23:03 +0000 (23:23 -0700)]
glsl: Refresh automatically generated file builtin_function.cpp.
Kenneth Graunke [Wed, 8 Sep 2010 06:21:22 +0000 (23:21 -0700)]
glsl/builtins: Don't use ir_binop_dot on floating point values.
ir_binop_dot is only defined for vector types. Use ir_binop_mul.
Kenneth Graunke [Wed, 8 Sep 2010 06:16:26 +0000 (23:16 -0700)]
glsl/builtins: Simplify degenerate scalar float cases.
The code being generated was just stupid, considering that:
- normalize(x) = 1.0
- length(x) = x
- distance(x, y) = x - y
Ian Romanick [Wed, 8 Sep 2010 18:33:12 +0000 (11:33 -0700)]
glsl2: Make sure _mesa_glsl_parse_state constructor gets a context
Fix an major regression in
dc754586. Too bad that change was
obviously never tested.
Kristian Høgsberg [Fri, 27 Aug 2010 16:40:11 +0000 (12:40 -0400)]
glx: Drop broken drawable garbage collection
Doesn't work for pixmaps, was looking up the GLX XID and was never thread
safe. Instead, just destroy the client side structures when the
drawable is no long current for a context.
Kristian Høgsberg [Wed, 8 Sep 2010 12:52:54 +0000 (08:52 -0400)]
dri: Unset current context and dispatch table when unbinding
Otherwise, when we switch to an indirect glx context and then back, it looks
like we're still current.
https://bugs.freedesktop.org/show_bug.cgi?id=29977#c7
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Chia-I Wu [Wed, 8 Sep 2010 10:52:27 +0000 (18:52 +0800)]
glsl: Support GLSL ES in the standalone compile.
GLSL ES mode is enabled when --glsl-es is passed to glsl_compiler.
Chia-I Wu [Wed, 8 Sep 2010 10:48:12 +0000 (18:48 +0800)]
glsl: Require a context in _mesa_glsl_parse_state.
Create a dummy context in the standalone compiler and pass it to
_mesa_glsl_parse_state.
Chia-I Wu [Wed, 8 Sep 2010 10:20:37 +0000 (18:20 +0800)]
st/dri: Call dri_init_extensions only for API_OPENGL.
libmesagallium.a that this state tracker will be linked to expects
OpenGL's _glapi_table. That is, it expects libGL.so instead of
libGLESv1_CM.so or libGLESv2.so. As there is no clean way to know the
shared library the app links to, use the api as a simple check. It
might be as well to simply remove this function call though.
Chia-I Wu [Wed, 8 Sep 2010 09:41:43 +0000 (17:41 +0800)]
st/dri: Use enum st_api_type internally.
nobled [Mon, 30 Aug 2010 20:23:54 +0000 (20:23 +0000)]
st/dri: Add multi-api support
Make st/dri screens capable of creating OpenGL ES and
OpenGL ES2 contexts.
TODO: Figure out the "get_current" problem with multiple
st_api's for real.
(s/API_OPENGLES1/API_OPENGLES/ by Chia-I Wu)
nobled [Tue, 7 Sep 2010 16:10:45 +0000 (12:10 -0400)]
st/dri: Make clear which API's are supported
If the caller requests a GLES context, don't silently create
a desktop GL context in its place.
Dave Airlie [Wed, 8 Sep 2010 04:09:40 +0000 (14:09 +1000)]
r600g: add initial bank swizzle support.
this is ported from r600c mostly, bank swizzling is real messy and I don't think
I got enough sleep last night to fully understand it.
Kenneth Graunke [Wed, 8 Sep 2010 01:01:33 +0000 (18:01 -0700)]
glcpp: Fix build on non-GCC compilers.
Kenneth Graunke [Tue, 7 Sep 2010 09:59:38 +0000 (02:59 -0700)]
ast_to_hir: Mark arrays as lvalues in GLSL ES, but prohibit assignment.
This allows them to be passed as out/inout parameters, but still
prevents them from being used as the target of an assignment. This is
per section 5.8 of the GLSL ES 1.00 specification.
Kenneth Graunke [Sun, 5 Sep 2010 08:51:28 +0000 (01:51 -0700)]
glsl: Allow overloading of built-ins without hiding in GLSL ES.
The rules are explicitly different from desktop GLSL.
Kenneth Graunke [Sun, 5 Sep 2010 08:48:11 +0000 (01:48 -0700)]
glsl: Move is_builtin flag back to ir_function_signature.
This effectively reverts
b6f15869b324ae64a00d0fe46fa3c8c62c1edb6c.
In desktop GLSL, defining a function with the same name as a built-in
hides that built-in function completely, so there would never be
built-in and user function signatures in the same ir_function.
However, in GLSL ES, overloading built-ins is allowed, and does not
hide the built-in signatures - so we're back to needing this.
Kenneth Graunke [Mon, 16 Aug 2010 21:02:25 +0000 (14:02 -0700)]
ast_to_hir: Reject embedded structure definitions in GLSL ES 1.00.
Kenneth Graunke [Sat, 7 Aug 2010 09:56:01 +0000 (02:56 -0700)]
ast_to_hir: Reject unsized array declarations in GLSL ES 1.00.
Kenneth Graunke [Wed, 1 Sep 2010 20:12:10 +0000 (13:12 -0700)]
ast_to_hir: Allow matrix-from-matrix constructors in GLSL ES.
Everything but 1.10 supports this, so just change the check to ==.
Kenneth Graunke [Mon, 30 Aug 2010 19:05:05 +0000 (12:05 -0700)]
i965: Enable EXT_framebuffer_blit internally.
Otherwise, ES2's BindFramebuffer entrypoint hits this assertion:
main/fbobject.c:1323: _mesa_BindFramebufferEXT: Assertion
`ctx->Extensions.EXT_framebuffer_blit' failed.
Kenneth Graunke [Mon, 30 Aug 2010 19:27:07 +0000 (12:27 -0700)]
mesa: Enable GL_MAX_VERTEX_ATTRIBS in ES2.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Kenneth Graunke [Tue, 31 Aug 2010 16:33:58 +0000 (09:33 -0700)]
linker: Fix assertion and cross-version checks for version 100.
Fixes an assert (min_version >= 110) which was no longer correct, and
also prohibits linking ES2 shaders with non-ES2 shaders. I'm not
positive this is correct, but the specification doesn't seem to say.
Kenneth Graunke [Mon, 16 Aug 2010 20:04:09 +0000 (13:04 -0700)]
glsl: Add built-in function profiles for GLSL ES 1.00.
Kenneth Graunke [Sat, 7 Aug 2010 09:45:33 +0000 (02:45 -0700)]
glsl: Add built-in variables for GLSL ES 1.00.
Kenneth Graunke [Sat, 7 Aug 2010 09:03:00 +0000 (02:03 -0700)]
glsl: Split out types that are in 1.10 but not GLSL ES 1.00.
Kenneth Graunke [Sun, 5 Sep 2010 08:32:32 +0000 (01:32 -0700)]
glsl: Recognize GLSL ES 1.00 keywords.
Kenneth Graunke [Sun, 5 Sep 2010 08:27:02 +0000 (01:27 -0700)]
glsl: Define GL_ES preprocessor macro if API is OpenGL ES 2.0.
Also define it if #version 100 is encountered.
Kenneth Graunke [Mon, 16 Aug 2010 19:34:53 +0000 (12:34 -0700)]
glsl: Accept language version 100 and make it the default on ES2.
Kenneth Graunke [Sun, 5 Sep 2010 07:31:28 +0000 (00:31 -0700)]
glsl: Set default language version in mesa_glsl_parse_state constructor.
This should make it easier to change the default version based on the
API (say, version 1.00 for OpenGL ES).
Also, synchronize the symbol table's version with the parse state's
version just before doing AST-to-HIR. This way, it will be set when
it matters, but the main initialization code doesn't have to care about
the symbol table.
Eric Anholt [Tue, 7 Sep 2010 23:25:22 +0000 (16:25 -0700)]
ir_to_mesa: Add support for gl_NormalScale.
Bug #30040.
Kenneth Graunke [Tue, 7 Sep 2010 23:25:53 +0000 (16:25 -0700)]
glsl: Make sure shader source isn't NULL.
This should only occur if glCompileShader is called without a prior call
to glShaderSource. An empty source program should be the empty string.
Tilman Sauerbeck [Tue, 7 Sep 2010 09:38:18 +0000 (11:38 +0200)]
r600g: Added missing r600_bc_add_literal() calls to XPD implementation.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Tilman Sauerbeck [Tue, 7 Sep 2010 09:37:56 +0000 (11:37 +0200)]
r600g: Added preliminary support for the LOG opcode.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Tilman Sauerbeck [Tue, 7 Sep 2010 19:29:43 +0000 (21:29 +0200)]
r600g: Added missing abs-ification to RSQ.
This makes the 'fp1-RSQ test 2' piglit test work.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Dave Airlie [Tue, 7 Sep 2010 22:49:00 +0000 (08:49 +1000)]
r600g: split opcodes out and add wrapper around usage.
This splits the r600 opcodes out of the sq file and adds a wrapper
so we can convert to evergreen opcodes later without touching these functions
too much.
Dave Airlie [Tue, 7 Sep 2010 22:41:57 +0000 (08:41 +1000)]
r600g: add support for constants in memory buffers.
DX9 constants were in the constant file, and evergreen no longer support
cfile. r600/700 can also use constants in memory buffers, so add the code
(disabled for now) to enable that as precursor for evergreen.
Dave Airlie [Tue, 7 Sep 2010 22:30:59 +0000 (08:30 +1000)]
r600g: abstract the hw states out behind a vtbl.
this is step one towards evergreen support, it lets us plug in whole
new hw level states.
Ian Romanick [Tue, 7 Sep 2010 21:30:06 +0000 (14:30 -0700)]
glsl2: Forbid array-types in ?: operator in GLSL 1.10
Fixes bugzilla #30039.
Eric Anholt [Tue, 7 Sep 2010 21:31:22 +0000 (14:31 -0700)]
mesa: Set the base format of GL_ALPHA FBOs and teach swrast about it.
Fixes assertion failures in fbo-alpha with a debug build of Mesa.
Bug #29781.
Eric Anholt [Tue, 7 Sep 2010 21:02:39 +0000 (14:02 -0700)]
ir_to_mesa: Fix warning in last commit.
I swear there was some git option for "don't push things when you've
got uncommitted changes", but I can't find it now.
Eric Anholt [Wed, 1 Sep 2010 23:06:32 +0000 (16:06 -0700)]
ir_to_mesa: Move the STATE_VAR elements of a builtin uniform to a temp (v2).
Like the constant handling and the handling of other uniforms, we add
the whole thing to the Parameters, avoiding messy, incomplete logic
for adding just the elements of a builting uniform that get used.
This means that a driver that relies only on ParameterValues[] for its
parameters will have an increased parameter load, but drivers
generally don't do that (since they have other params they need to
handle, too).
Fixes glsl-fs-statevar-call (testcase for Ember). Bug #29687.
v2: Continue referencing the STATE_VAR[] file directly when the
uniform will land in STATE_VAR[] formatted exactly as we'd put into a
temporary. When there's array dereferencing, we don't copy-propagate
in Mesa IR (not knowing where the array is in register space), so
smarts here are required or we'll massively increase the temp count.
Ian Romanick [Tue, 7 Sep 2010 20:33:48 +0000 (13:33 -0700)]
glsl2: Early return with visit_continue in loop_analysis::visit(ir_dereference_variable *)
Returning early with visit_continue_with_parent prevented the
then-statements and else-statements of if-statements such as the
following from being processed:
if (some_var) { ... } else { ... }
Fixes piglit test case glsl-fs-loop-nested-if and bugzilla #30030.
Kenneth Graunke [Tue, 7 Sep 2010 19:56:18 +0000 (12:56 -0700)]
glsl: Change grammar rules for selection statements to match the spec.
Fixes piglit test case loop-06.vert.
Unfortunately, causes 1 shift/reduce conflict.