Bryan Cain [Sat, 20 Aug 2011 19:43:25 +0000 (14:43 -0500)]
glsl_to_tgsi: emit a MAD(b, -a, b) for !a && b
This is a port of commit
ff2cfb8989cd to glsl_to_tgsi.
Bryan Cain [Sat, 20 Aug 2011 19:17:52 +0000 (14:17 -0500)]
glsl_to_tgsi: fix typo
Bryan Cain [Sat, 20 Aug 2011 19:15:03 +0000 (14:15 -0500)]
glsl_to_tgsi: implement ir_binop_all_equal using DP4 w/SGE
This is a port of commit
ba01df11c4d0 to glsl_to_tgsi with integer support
added.
Bryan Cain [Sat, 20 Aug 2011 18:56:06 +0000 (13:56 -0500)]
glsl_to_tgsi: implement ir_binop_any_nequal using DP4 w/saturate or DP4 w/SLT
Implement the any() part of the operation the same way regular ir_unop_any
is implemented.
This is a port of commit
e7bf096e8b04 to glsl_to_tgsi, with added integer
support.
Bryan Cain [Sat, 20 Aug 2011 18:26:12 +0000 (13:26 -0500)]
glsl_to_tgsi: implement ir_unop_any using DP4 w/saturate or DP4 w/SLT
This is a port of commit
92ca560d68e8 to glsl_to_tgsi, with integer support
added.
Bryan Cain [Thu, 18 Aug 2011 01:34:19 +0000 (20:34 -0500)]
glsl_to_tgsi: make glsl_to_tgsi_visitor::emit_dp return the instruction
Bryan Cain [Wed, 17 Aug 2011 19:35:35 +0000 (14:35 -0500)]
glsl_to_tgsi: implement ir_binop_logic_or using an add w/saturate or add w/SLT
Logical-or is implemented using addition (followed by clamping to [0,1]) on
values of 0.0 and 1.0. Replacing the logical-or operators with addition gives
a + b which has a result on the range [0, 2].
Previously a SNE instruction was used to clamp the resulting logic value to
[0,1]. In a fragment shader, using a saturate on the add has the same effect.
Adding the saturate to the add is free, so (at least) one instruction is
saved. In a vertex shader, using an SLT on the negation of the add result has
the same effect. Many older shader architectures do not support the SNE
instruction. It must be emulated using two SLT instructions and an ADD. On
these architectures, the single SLT saves two instructions.
Note that SNE is still used when integers are used for boolean values, since
there is no such thing as an integer saturate, and older shader architectures
without SNE don't support integers.
This is a port of commit
41f8ffe5e07c to glsl_to_tgsi with integer support
added.
Bryan Cain [Wed, 17 Aug 2011 15:01:30 +0000 (10:01 -0500)]
glsl_to_tgsi: implement ir_unop_logic_not using 1-x
Since our logic values are 0.0 (false) and 1.0 (true), 1.0 - x accurately
implements logical not.
This is a port of commit
6ad08989d7c1 to glsl_to_tgsi.
Chia-I Wu [Tue, 2 Aug 2011 00:49:27 +0000 (09:49 +0900)]
st/egl: add a missing include
Reported by cwhuang.
Chia-I Wu [Thu, 4 Aug 2011 02:49:42 +0000 (11:49 +0900)]
st/egl: improve error logging
This helps diagnose problems in EGL initialization.
Chia-I Wu [Sun, 31 Jul 2011 02:16:53 +0000 (11:16 +0900)]
st/egl: add buffer preserving support to Android
Use a staging color buffer when buffer preserving is enabled.
Chia-I Wu [Sun, 31 Jul 2011 01:49:52 +0000 (10:49 +0900)]
st/egl: improve buffer cache for Android
There may be more than two back buffers. Clean up and prepare the
buffer cache for that.
Chia-I Wu [Wed, 20 Jul 2011 10:20:20 +0000 (18:20 +0800)]
st/egl: swapping without a buffer is not an error
This fixes Kwaak3.
Chia-I Wu [Sat, 30 Jul 2011 01:45:20 +0000 (10:45 +0900)]
st/egl: use HAL formats for Android backend
Native buffers use HAL formats, not UI formats.
Chia-I Wu [Sun, 31 Jul 2011 03:08:24 +0000 (12:08 +0900)]
winsys/sw/android: use HAL formats
Native buffers use HAL formats, not UI formats.
Chia-I Wu [Thu, 4 Aug 2011 08:50:51 +0000 (17:50 +0900)]
winsys/sw/android: set bo usage correctly
Since this is the software path, set GRALLOC_USAGE_SW_WRITE_OFTEN when
PIPE_BIND_RENDER_TARGET, and set GRALLOC_USAGE_SW_READ_OFTEN when
PIPE_BIND_SAMPLER_VIEW.
Chia-I Wu [Wed, 17 Aug 2011 04:19:23 +0000 (12:19 +0800)]
android: make libGLES_mesa real
libGLES_mesa with swrast should link in these libraries
libmesa_egl
libmesa_egl_gallium
libmesa_st_egl
libmesa_st_mesa
libmesa_glsl
libmesa_glsl_utils
libmesa_pipe_softpipe
libmesa_winsys_sw_android
libmesa_gallium
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Wed, 17 Aug 2011 04:10:12 +0000 (12:10 +0800)]
android: build shared glapi
This builds the shared library libglapi from shared glapi.
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Wed, 17 Aug 2011 03:56:43 +0000 (11:56 +0800)]
android: build glsl
This builds the static library libmesa_glsl and executable glsl_compiler
from glsl. glsl_compiler is only installed for engineering build.
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Wed, 17 Aug 2011 03:31:36 +0000 (11:31 +0800)]
android: build core mesa
This builds the static library libmesa_st_mesa from core mesa.
Acked-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Wed, 17 Aug 2011 03:24:11 +0000 (11:24 +0800)]
android: build core EGL
This builds the static library libmesa_egl from core EGL.
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Wed, 17 Aug 2011 03:51:08 +0000 (11:51 +0800)]
android: build softpipe
This builds the static library libmesa_pipe_softpipe from softpipe.
Chia-I Wu [Wed, 17 Aug 2011 03:47:42 +0000 (11:47 +0800)]
android: build android sw winsys
This builds the static library libmesa_winsys_sw_android from winsys/sw.
Chia-I Wu [Thu, 18 Aug 2011 16:28:22 +0000 (00:28 +0800)]
android: build targets/egl-static
This builds the static library libmesa_egl_gallium from
targets/egl-static.
Chia-I Wu [Wed, 17 Aug 2011 03:41:40 +0000 (11:41 +0800)]
android: build st/egl with android backend
This builds the static library libmesa_st_egl from st/egl.
Chia-I Wu [Wed, 17 Aug 2011 03:38:00 +0000 (11:38 +0800)]
android: build gallium auxiliaries
This builds the static library libmesa_gallium from gallium auxiliaries.
Chia-I Wu [Wed, 17 Aug 2011 03:07:01 +0000 (11:07 +0800)]
android: build libGLES_mesa
This is the first step to integrate Mesa into Android(-x86) build
system. You can git clone mesa under the external/ directory of Android
source tree and build Android with
$ make BOARD_GPU_DRIVERS=swrast
It will build libGLES_mesa that will be loaded by Android runtime.
libGLES_mesa is still a stub in this commit.
Chia-I Wu [Thu, 18 Aug 2011 02:24:25 +0000 (10:24 +0800)]
targets/egl-static: do not rely on libudev on Android
There is no libudev on Android. Use DRM to get the PCI ID directly.
Reviewed-by: Benjamin Franzke <benjaminfranzke@googlemail.com>
Chia-I Wu [Thu, 18 Aug 2011 02:28:29 +0000 (10:28 +0800)]
st/egl: add android backend
Both HW and SW rendering are supported for Android. For SW rendering,
we use the generic gralloc lock/unlock for mapping and unmapping color
buffers (in winsys/android).
For HW rendering, we need to know the real type of color buffers. This
backend works with drm_gralloc, where a color buffer is backed by a GEM
object.
Chia-I Wu [Thu, 18 Aug 2011 02:17:10 +0000 (10:17 +0800)]
winsys/android: new SW winsys for Android
On Android, color buffers are passed between server and clients as
opaque buffer_handle_t. This winsys makes use of gralloc, which
provides a generic way to map and unmap buffer_handle_t for CPU access.
Chia-I Wu [Fri, 1 Oct 2010 19:27:42 +0000 (15:27 -0400)]
egl: add Android-specific extensions
Add EGL_ANDROID_image_native_buffer and EGL_ANDROID_swap_rectangle.
There is no spec for them though.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Mon, 11 Oct 2010 08:09:52 +0000 (16:09 +0800)]
egl: add _EGL_PLATFORM_ANDROID
This is Android Gingerbread platform.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Fri, 5 Aug 2011 03:54:05 +0000 (12:54 +0900)]
mesa: android has no log2f nor ffs
Define log2f(v) to be logf(v) / M_LN2 and ffs to __builtin_ffs.
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Fri, 5 Aug 2011 03:52:49 +0000 (12:52 +0900)]
ralloc: include limits.h for SIZE_MAX on Android
Android does not define SIZE_MAX in stdint.h. We have to include
limits.h for it.
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Fri, 5 Aug 2011 03:50:12 +0000 (12:50 +0900)]
gallium: add PIPE_OS_ANDROID support
Android uses Linux kernel and its own C runtime. It resembles
PIPE_OS_LINUX a lot with some minor exceptions.
Reviewed-by: Brian Paul <brianp@vmware.com>
Chia-I Wu [Mon, 1 Aug 2011 02:14:18 +0000 (11:14 +0900)]
glsl: remove an unnecessary header include
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Chad Versace <chad@chad-versace.us>
Chia-I Wu [Fri, 8 Apr 2011 14:50:43 +0000 (22:50 +0800)]
mesa: fix !FEATURE_GL build
Move vbo_exec_FlushVertices_internal out of FEATURE_beginend.
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Chad Versace <chad@chad-versace.us>
Kenneth Graunke [Thu, 18 Aug 2011 09:15:56 +0000 (02:15 -0700)]
i965/gen7: Use align1 mode to set URB_WRITE_HWORD channel enables.
Makes the new vertex shader backend work on Ivybridge.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Kenneth Graunke [Thu, 11 Aug 2011 23:42:01 +0000 (16:42 -0700)]
i965/fs: Don't double-convert integer/boolean uniforms.
When ctx->Const.NativeIntegers is set, Core Mesa loads integer/boolean
uniforms directly, rather than loading the floating point equivalent.
So, when that's set, we don't need to perform any conversions.
Unfortunately, we can't properly support native integers with the old
vertex shader backend, so this patch leaves them disabled for now.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Kenneth Graunke [Mon, 15 Aug 2011 21:18:16 +0000 (14:18 -0700)]
mesa, glsl_to_tgsi: Add new gl_context::NativeIntegers flag.
Previously, native integer support was based on whether the driver
advertised GLSL 1.30 or not. However, drivers that natively support
integers may wish to do so for older GLSL versions as well. Adding this
new opt-in flag allows them to do so.
Currently disabled by default on all drivers, which was the existing
behavior (no drivers currently implement GLSL 1.30).
Fixes piglit tests on i965 with INTEL_GLSL_VERSION=130 set:
- spec/glsl-1.10/fs-uniform-int-110.shader_test
- spec/glsl-1.30/fs-uniform-int-130.shader_test
(it was doubly converting the data)
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Kenneth Graunke [Thu, 18 Aug 2011 21:08:06 +0000 (14:08 -0700)]
ir_to_mesa: Remove incorrect usage of the 'struct' keyword on classes.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Kenneth Graunke [Thu, 18 Aug 2011 21:05:11 +0000 (14:05 -0700)]
i965/fs: Change incorrect use of 'struct fs_reg' to simply 'fs_reg'.
It's actually a class.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Kenneth Graunke [Thu, 18 Aug 2011 21:03:26 +0000 (14:03 -0700)]
glsl: Remove unused variable.
Eric Anholt [Tue, 16 Aug 2011 22:28:53 +0000 (15:28 -0700)]
i965/vs: Implement proper register allocation instead of 1:1 mapping.
Fixes vs-atan-* and several others. This is not the real solution we
eventually want, which will pack floats, vec2s, and vec3s into vec4
registers, but this code should provide the framework for that.
Eric Anholt [Wed, 17 Aug 2011 17:50:17 +0000 (10:50 -0700)]
i965/vs: Add simple dead code elimination.
This is copied right from the fragment shader. It is needed for real
register allocation to work correctly.
Eric Anholt [Tue, 16 Aug 2011 22:09:48 +0000 (15:09 -0700)]
i965/vs: Copy the live intervals calculation over from the FS.
This is a rather pessimistic calculation, since it doesn't distinguish
individual channels of a vec4, or elements of an array, but should be
a minimum start for register allocation.
Eric Anholt [Tue, 16 Aug 2011 21:18:51 +0000 (14:18 -0700)]
i965/vs: Remove stale comment about compressed instructions.
This was copy'n'paste from the fragment shader, and didn't make sense
here.
Marek Olšák [Fri, 19 Aug 2011 20:57:56 +0000 (22:57 +0200)]
r600g: don't allocate separate depth and stencil for transfer textures on EG
The state tracker expects depth and stencil pixels interleaved.
Evergreen can bind an interleaved depth-stencil resource as a colorbuffer,
but not as a zbuffer.
The hardware can do the interleaving for us when decompressing.
Marek Olšák [Fri, 19 Aug 2011 20:43:08 +0000 (22:43 +0200)]
r600g: finally enable float depth buffers on evergreen
Marek Olšák [Fri, 19 Aug 2011 20:27:00 +0000 (22:27 +0200)]
r600g: rename resource -> view in create_sampler_view
The sampler view is not a resource.
Also remove the unused desc variable.
Marek Olšák [Fri, 19 Aug 2011 19:48:12 +0000 (21:48 +0200)]
r600g: hack around a problem with texture alignment
Marek Olšák [Fri, 19 Aug 2011 17:26:08 +0000 (19:26 +0200)]
r600g: simplify the conditionals determining array mode
Marek Olšák [Fri, 19 Aug 2011 17:21:40 +0000 (19:21 +0200)]
r600g: put depth and stencil into one backing buffer
For DRI2 sharing.
Marek Olšák [Thu, 18 Aug 2011 23:07:46 +0000 (01:07 +0200)]
r600g: fix depth-stencil on evergreen
Such that it actually works in apps which use both.
A separate buffer is allocated for stencil. The only exception is
the window-system-provided depth-stencil buffer, where depth and stencil
share the same buffer.
This fixes:
- fbo-depthstencil-GL_DEPTH24_STENCIL8-clear
- fbo-depthstencil-GL_DEPTH24_STENCIL8-drawpixels-FLOAT-and-USHORT
- fbo-depthstencil-GL_DEPTH24_STENCIL8-readpixels-24_8
- fbo-depthstencil-GL_DEPTH24_STENCIL8-readpixels-FLOAT-and-USHORT
Brian Paul [Thu, 18 Aug 2011 21:59:33 +0000 (15:59 -0600)]
mesa: handle array textures in GenerateMipmap(), FramebufferTexture1/2D()
This was an unfinished to-do item before.
With this patch and the two preceeding patches, piglit's
fbo-generatemipmap-array test runs and passes instead of generating
a GL error and dying on an assertion.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Brian Paul [Thu, 18 Aug 2011 21:58:00 +0000 (15:58 -0600)]
meta: use fallback mipmap generation for 1D/2D texture arrays
We could do 1D/2D arrays with textured quad rendering, but it'll take
some work (as with 3D textures).
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Brian Paul [Thu, 18 Aug 2011 21:54:53 +0000 (15:54 -0600)]
mesa: set Q=1 for OPCODE_TEX execution
Q should not be significant for OPCODE_TEX, but it winds up getting
passed to the compute_lambda() function. Make sure it's 1.0 to
prevent garbage values, which is effectively what we get when the
swizzle is coord.xyzz (which is what GLSL gives us).
Part of the fix for piglit's fbo-generatemipmap-array test.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Brian Paul [Thu, 18 Aug 2011 20:35:27 +0000 (14:35 -0600)]
mesa: restructure error checking in _mesa_FramebufferTexture1D/2DEXT()
In anticipation of adding more texture targets.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Chad Versace [Fri, 12 Aug 2011 23:50:27 +0000 (16:50 -0700)]
mesa: Declare _mesa_meta_begin()/end() as public
Declare _mesa_meta_begin()/end() in meta.h so that drivers can write
custom meta-ops (such as HiZ resolves for i965).
This necessitates moving the the META_* macros into meta.h. To prevent
naming collisions, this commit renames each macro to be MESA_META_*.
Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Chad Versace <chad@chad-versace.us>
Christoph Bumiller [Fri, 19 Aug 2011 14:57:30 +0000 (16:57 +0200)]
nv50: fix potential null deref in insn modifer optimization pass
Bug introduced in
34980cd153050111edc545545ddff11f5b68347e.
Brian Paul [Fri, 19 Aug 2011 14:36:22 +0000 (08:36 -0600)]
glx: use a block to fix declarations after code warning
Marek Olšák [Thu, 18 Aug 2011 20:59:22 +0000 (22:59 +0200)]
r600g: fix a possible crash in r600_adjust_gprs
Brian Paul [Thu, 18 Aug 2011 17:51:48 +0000 (11:51 -0600)]
mesa: fix incorrect error code in _mesa_FramebufferTexture1D/3DEXT()
The spec says GL_INVALID_OPERATION is generated when texture!=0 and
textarget is not a legal value. We had this right for the 2D function.
Paul Berry [Tue, 16 Aug 2011 17:23:52 +0000 (10:23 -0700)]
glapi: update .gitignore for generated ES dispatch headers
Commit
6eff33dc (glapi: generate ES dispatch headers from core mesa)
replaced the autogenerated files
src/mapi/es1api/main/{dispatch,remap_helper}.h with new autogenerated
files src/mesa/main/api_exec_es{1,2}_{dispatch,remap_helper}.h. This
patch updates the .gitignore files to properly ignore the new
autogenerated files, and stop ignoring the old autogenerated files.
Reviewed-by: Chia-I Wu <olv@lunarg.com>
José Fonseca [Thu, 18 Aug 2011 17:04:44 +0000 (18:04 +0100)]
llvmpipe: Don't build lp_test_arit on MSVC.
Several issues due to expf/logf/etc either not being declared,
or being defined as a macro.
José Fonseca [Thu, 18 Aug 2011 16:58:02 +0000 (17:58 +0100)]
llvmpipe: snprintf->util_snprintf.
For MSVC.
José Fonseca [Thu, 18 Aug 2011 15:06:00 +0000 (16:06 +0100)]
llvmpipe: Add u_math.h include.
Necessary on platforms with incomplete math.h
José Fonseca [Thu, 18 Aug 2011 15:00:03 +0000 (16:00 +0100)]
scons: Add support for LLVM-2.9 on Windows.
MinGW & MSVC, although I've only tested the former.
Kristian Høgsberg [Thu, 18 Aug 2011 12:46:02 +0000 (08:46 -0400)]
glx: Don't flush twice if we fallback to dri2CopySubBuffer
The flush extensions flush call indicates end of frame and should only
be called once per frame. However, in the dri2SwapBuffer fallback
path, we call flush and then call dri2CopySubBuffer, which also calls
flush. Refactor the code to only call flush once.
Marek Olšák [Tue, 16 Aug 2011 07:47:16 +0000 (09:47 +0200)]
r600g: implement NV_primitive_restart functionality (v2)
Needed for GL3.
v2: evergreen support
I don't set PA_SU_SC_MODE_CNTL.MULTI_PRIM_IB_ENA.
piglit/primitive-restart does pass though. Tested on RV730 and EG-REDWOOD.
Marek Olšák [Wed, 17 Aug 2011 20:53:10 +0000 (22:53 +0200)]
r600g: fix scons build
Eric Anholt [Tue, 16 Aug 2011 05:36:18 +0000 (22:36 -0700)]
i965/fs: Fix 32-bit integer multiplication.
The MUL opcode does a 16bit * 32bit multiply, and we need to do the
MACH to get the top 16bit * 32bit added in.
Fixes fs-op-mult-int-*, fs-op-mult-ivec*
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Christoph Bumiller [Wed, 17 Aug 2011 18:38:40 +0000 (20:38 +0200)]
nv50: don't drop flags definition when merging SAT with ADD/MAD
Brian Paul [Tue, 16 Aug 2011 22:30:52 +0000 (16:30 -0600)]
st/mesa: fix incorrect loop over instruction src regs
The array of src regs is of size 3, not 4.
Lauri Kasanen [Fri, 1 Jul 2011 09:57:06 +0000 (12:57 +0300)]
st/dri: Indent driconf options
Lauri Kasanen [Fri, 1 Jul 2011 10:01:00 +0000 (13:01 +0300)]
xmlpool.h: fix a typo
Lauri Kasanen [Fri, 1 Jul 2011 10:49:18 +0000 (13:49 +0300)]
xmlconfig: Make the error message more informative
Ian Romanick [Thu, 4 Aug 2011 00:12:29 +0000 (17:12 -0700)]
mesa: Bump instruction execution limit to 65536
Shader Model 3.0[1] requires that shaders be able to execute at least
65536 instructions. Bump Mesa maxExec to that limit. This allows
several vertex shaders in the OpenGL ES 2.0 conformance test suite to
run to completion.
1: http://en.wikipedia.org/wiki/High_Level_Shader_Language
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Thu, 10 Feb 2011 23:48:27 +0000 (15:48 -0800)]
mesa: Add partial constant propagation pass for Mesa IR
This cleans up some code generated by the IR-to-Mesa pass for i915.
In particular, some shaders involving arrays of constant matrices
result in really bad code.
v2: Silence several warnings from merging the gl_constant_value work.
Fix DP[23] folding. Add support for a bunch more opcodes that appear
in piglit runs on i915.
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Tue, 2 Aug 2011 19:17:20 +0000 (12:17 -0700)]
ir_to_mesa: Emit a MAD(b, -a, b) for !a && b
!a && b occurs frequently when nexted if-statements have been
flattened. It should also be possible use a MAD for (a && b) || c,
though that would require a MAD_SAT.
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Wed, 3 Aug 2011 22:42:05 +0000 (15:42 -0700)]
ir_to_mesa: Implement ir_binop_all_equal using DP4 w/SGE
The operation ir_binop_all_equal is !(a.x != b.x || a.y != b.y || a.z
!= b.z || a.w != b.w). Logical-or is implemented using addition
(followed by clampling to [0,1]) on values of 0.0 and 1.0. Replacing
the logical-or operators with addition gives !bool((int(a.x != b.x) +
int(a.y == b.y) + int(a.z == b.z) + int(a.w == b.w)). This can be
implemented using a dot-product with a vector of all 1.0. After the
dot-product, the value will be an integer on the range [0,4].
Previously a SEQ instruction was used to clamp the resulting logic
value to [0,1] and invert the result. Using an SGE instruction on the
negation of the dot-product result has the same effect. Many older
shader architectures do not support the SEQ instruction. It must be
emulated using two SGE instructions and a MUL. On these
architectures, the single SGE saves two instructions.
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Wed, 3 Aug 2011 22:35:01 +0000 (15:35 -0700)]
ir_to_mesa: Implement ir_binop_any_nequal using DP4 w/saturate or DP4 w/SLT
The operation ir_binop_any_nequal is (a.x != b.x) || (a.y != b.y) ||
(a.z != b.z) || (a.w != b.w), and that is the same as any(bvec4(a.x !=
b.x, a.y != b.y, a.z != b.z, a.w != b.w)). Implement the any() part
the same way the regular ir_unop_any is implemented.
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Wed, 3 Aug 2011 22:27:43 +0000 (15:27 -0700)]
ir_to_mesa: Implement ir_unop_any using DP4 w/saturate or DP4 w/SLT
This is just like the ir_binop_logic_or case. The operation
ir_unop_any is (a.x || a.y || a.z || a.w). Logical-or is implemented
using addition (followed by clampling to [0,1]) on values of 0.0 and
1.0. Replacing the logical-or operators with addition gives (a.x +
a.y + a.z + a.w). This can be implemented using a dot-product with a
vector of all 1.0.
Previously a SNE instruction was used to clamp the resulting logic
value to [0,1]. In a fragment shader, using a saturate on the
dot-product has the same effect. Adding the saturate to the
dot-product is free, so (at least) one instruction is saved.
In a vertex shader, using an SLT on the negation of the dot-product
result has the same effect. Many older shader architectures do not
support the SNE instruction. It must be emulated using two SLT
instructions and an ADD. On these architectures, the single SLT saves
two instructions.
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Sat, 30 Jul 2011 17:45:35 +0000 (10:45 -0700)]
ir_to_mesa: Make ir_to_mesa_visitor::emit_dp return the instruction
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Sat, 30 Jul 2011 17:49:49 +0000 (10:49 -0700)]
ir_to_mesa: Implement ir_binop_logic_or using an add w/saturate or add w/SLT
Logical-or is implemented using addition (followed by clampling to
[0,1]) on values of 0.0 and 1.0. Replacing the logical-or operators
with addition gives a + b which has a result on the range [0, 2].
Previously a SNE instruction was used to clamp the resulting logic
value to [0,1]. In a fragment shader, using a saturate on the add has
the same effect. Adding the saturate to the add is free, so (at
least) one instruction is saved.
In a vertex shader, using an SLT on the negation of the add result has
the same effect. Many older shader architectures do not support the
SNE instruction. It must be emulated using two SLT instructions and
an ADD. On these architectures, the single SLT saves two
instructions.
Reviewed-by: Eric Anholt <eric@anholt.net>
Ian Romanick [Sat, 30 Jul 2011 17:48:10 +0000 (10:48 -0700)]
ir_to_mesa: Implement ir_unop_logic_not using 1-x
Since our logic values are 0.0 (false) and 1.0 (true), 1.0 - x
accurately implements logical not.
Reviewed-by: Eric Anholt <eric@anholt.net>
Chad Versace [Mon, 15 Aug 2011 20:29:15 +0000 (13:29 -0700)]
mesa: Add Android to list of platforms that define fpclassify()
This is a fix for the Android build.
Signed-off-by: Chad Versace <chad@chad-versace.us>
Chad Versace [Mon, 15 Aug 2011 20:26:21 +0000 (13:26 -0700)]
mesa: Fix Android build by #ifdef'ing out locale support
Bionic does not support locales. This commit #ifdef's out the locale usage
in _mesa_strtof().
Signed-off-by: Chad Versace <chad@chad-versace.us>
Chad Versace [Mon, 15 Aug 2011 17:58:25 +0000 (10:58 -0700)]
mesa: Remove use of fpu_control.h
Remove the inclusion of fpu_control.h from compiler.h. Since Bionic lacks
fpu_control.h, this fixes the Android build.
Also remove the sole use of the fpu_control bits, which was in debug.c.
Those were brianp's debug bits, and he approved of their removal.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Chad Versace <chad@chad-versace.us>
Eric Anholt [Tue, 16 Aug 2011 04:02:10 +0000 (21:02 -0700)]
i965/vs: Fix multiplies to actually do 32-bit multiplies.
Fixes vs-op-mult-int-int and friends.
Eric Anholt [Tue, 16 Aug 2011 03:59:24 +0000 (20:59 -0700)]
i965/vs: Add support for conversion of FIXED_HW_REG src_reg to/from dst_reg.
This was quietly occurring in some emit code I produced, and failed.
Eric Anholt [Tue, 16 Aug 2011 03:43:42 +0000 (20:43 -0700)]
i965/vs: Fix memory leak of ralloc context for the visitor.
Eric Anholt [Tue, 16 Aug 2011 03:13:53 +0000 (20:13 -0700)]
i965/vs: Fix condition code for scalar expression all_equals.
Fixes vs-op-eq-bool-bool.
Eric Anholt [Fri, 12 Aug 2011 12:32:25 +0000 (05:32 -0700)]
i965/vs: Don't assertion fail on vertex texturing.
The linker will reject the program, but we need to survive until then.
Fixes abort in glsl1-2D Texture lookup with explicit lod (Vertex
shader)
Eric Anholt [Fri, 12 Aug 2011 12:28:53 +0000 (05:28 -0700)]
i965/gen6: Force WHILE exec size to 8.
We can't just look at the instruction that happens to appear at the
start of the loop, because it might be some other exec size and cause
us to only loop on the first N channels. We always want 8 in our
current code (since 16 doesn't work so we don't do 16-wide fragment in
that case).
Fixes loop-03.vert, which was triggering the assertions.
Eric Anholt [Fri, 12 Aug 2011 12:15:50 +0000 (05:15 -0700)]
i965/vs: Remove remaining use of foreach_iter.
Eric Anholt [Thu, 11 Aug 2011 23:27:41 +0000 (16:27 -0700)]
i965/vs: Fix abs/negate handling on attributes.
Fixes glsl-vs-neg-attribute and glsl-vs-abs-attribute.
Eric Anholt [Thu, 11 Aug 2011 15:09:10 +0000 (08:09 -0700)]
i965/vs: Avoid generating a MOV for most ir_assignment handling.
Removes an average of 11.5% of instructions in 54% of vertex shaders
in shader-db.
Eric Anholt [Thu, 11 Aug 2011 16:52:08 +0000 (09:52 -0700)]
i965/vs: Run the shader backend at link time and return compile failures.
Link failure is something that shouldn't happen, but we sometimes want
it during development. The precompile also allows analysis of shader
codegen with shader-db.
Eric Anholt [Tue, 16 Aug 2011 01:40:14 +0000 (18:40 -0700)]
i965: Fix assertion failure on a loop consisting of while (true) { break }.
On enabling the precompile step in the VS, we tripped over this
assertion failure in glsl-link-bug-30552.