mesa.git
9 years agor600g: use the tgsi opcode from parse.FullToken.FullInstruction
Emil Velikov [Mon, 16 Mar 2015 14:47:07 +0000 (14:47 +0000)]
r600g: use the tgsi opcode from parse.FullToken.FullInstruction

... rather than the local one in inst_info->tgsi_opcode.

This will allow us to simplify struct r600_shader_tgsi_instruction.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
9 years agoi965/fs: Apply gl_FrontFacing ? -1 : 1 optimization only for floats
Ian Romanick [Sat, 28 Feb 2015 16:32:57 +0000 (08:32 -0800)]
i965/fs: Apply gl_FrontFacing ? -1 : 1 optimization only for floats

At the very least, unreal4/sun-temple/102.shader_test uses this pattern
for a signed integer result.  However, that shader did not hit the
optimization in the first place because it uses !gl_FrontFacing.  I
changed the shader to use remove the logical-not and reverse the other
operands.  I verified that incorrect code is generated before this
change and correct code is generated after.

Fixes fs-frontfacing-ternary-1-neg-1.shader_test.

No shader-db changes.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965/fs: Change try_opt_frontfacing_ternary to eliminate asserts
Ian Romanick [Sat, 28 Feb 2015 16:26:37 +0000 (08:26 -0800)]
i965/fs: Change try_opt_frontfacing_ternary to eliminate asserts

If we check for the case that is actually necessary, the asserts
become superfluous.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965/fs: Handle CMP.nz ... 0 and AND.nz ... 1 similarly in cmod propagation
Ian Romanick [Tue, 3 Feb 2015 19:12:28 +0000 (21:12 +0200)]
i965/fs: Handle CMP.nz ... 0 and AND.nz ... 1 similarly in cmod propagation

Espically on platforms that do not natively generate 0u and ~0u for
Boolean results, we generate a lot of sequences where a CMP is
followed by an AND with 1.  emit_bool_to_cond_code does this, for
example.  On ILK, this results in a sequence like:

    add(8)          g3<1>F          g8<8,8,1>F      -g4<0,1,0>F
    cmp.l.f0(8)     g3<1>D          g3<8,8,1>F      0F
    and.nz.f0(8)    null            g3<8,8,1>D      1D
    (+f0) iff(8)    Jump: 6

The AND.nz is obviously redundant.  By propagating the cmod, we can
instead generate

    add.l.f0(8)     null            g8<8,8,1>F      -g4<0,1,0>F
    (+f0) iff(8)    Jump: 6

Existing code already handles the propagation from the CMP to the ADD.

Shader-db results:

GM45 (0x2A42):
total instructions in shared programs: 3550829 -> 3550788 (-0.00%)
instructions in affected programs:     10028 -> 9987 (-0.41%)
helped:                                24

Iron Lake (0x0046):
total instructions in shared programs: 4993146 -> 4993105 (-0.00%)
instructions in affected programs:     9675 -> 9634 (-0.42%)
helped:                                24

Ivy Bridge (0x0166):
total instructions in shared programs: 6291870 -> 6291794 (-0.00%)
instructions in affected programs:     17914 -> 17838 (-0.42%)
helped:                                48

Haswell (0x0426):
total instructions in shared programs: 5779256 -> 5779180 (-0.00%)
instructions in affected programs:     16694 -> 16618 (-0.46%)
helped:                                48

Broadwell (0x162E):
total instructions in shared programs: 6823088 -> 6823014 (-0.00%)
instructions in affected programs:     15824 -> 15750 (-0.47%)
helped:                                46

No chage on Sandy Bridge or on any platform when NIR is used.

v2: Add unit tests suggested by Matt.  Remove spurious writes_flag()
check on scan_inst when scan_inst is known to be BRW_OPCODE_CMP (also
suggested by Matt).

v3: Fix some comments and remove some explicit int() casts in fs_reg
constructors in the unit tests.  Both suggested by Matt.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965: Mark paths in linear <-> tiled functions as unreachable().
Matt Turner [Thu, 5 Mar 2015 01:27:21 +0000 (17:27 -0800)]
i965: Mark paths in linear <-> tiled functions as unreachable().

text    data     bss     dec     hex filename
9663       0       0    9663    25bf intel_tiled_memcpy.o   before
8215       0       0    8215    2017 intel_tiled_memcpy.o   after

Reviewed-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
9 years agoegl: Remove eglQueryString virtual dispatch.
Matt Turner [Sat, 14 Mar 2015 00:00:26 +0000 (17:00 -0700)]
egl: Remove eglQueryString virtual dispatch.

Reviewed-by: Chad Versace <chad.versace@intel.com>
9 years agomain: Correct _mesa_error with no format in bufferobj.c.
Laura Ekstrand [Tue, 17 Mar 2015 20:27:31 +0000 (13:27 -0700)]
main: Correct _mesa_error with no format in bufferobj.c.

This fixes Bug 89616, a build failure due to line 1639 of bufferobj.c:
_mesa_error(ctx, GL_INVALID_OPERATION, func);

Trivial.

9 years agomain: Cosmetic changes to GetBufferSubData.
Laura Ekstrand [Thu, 12 Feb 2015 00:53:46 +0000 (16:53 -0800)]
main: Cosmetic changes to GetBufferSubData.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Add entry point for GetNamedBufferSubData.
Laura Ekstrand [Tue, 20 Jan 2015 23:24:53 +0000 (15:24 -0800)]
main: Add entry point for GetNamedBufferSubData.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Cosmetic updates to GetBufferPointerv.
Laura Ekstrand [Mon, 16 Mar 2015 23:08:36 +0000 (16:08 -0700)]
main: Cosmetic updates to GetBufferPointerv.

v3: Review from Fredrik Hoglund
   -Split cosmetic refactor of GetBufferPointerv out into a separate commit

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Add entry point for GetNamedBufferPointerv.
Laura Ekstrand [Tue, 20 Jan 2015 22:32:35 +0000 (14:32 -0800)]
main: Add entry point for GetNamedBufferPointerv.

v3: Review from Fredrik Hoglund
   -Split cosmetic refactor of GetBufferPointerv out into a separate commit

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Add entry points for GetNamedBufferParameteri[64]v.
Laura Ekstrand [Thu, 12 Feb 2015 00:10:20 +0000 (16:10 -0800)]
main: Add entry points for GetNamedBufferParameteri[64]v.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Refactor GetBufferParameteri[64]v.
Laura Ekstrand [Thu, 12 Feb 2015 00:07:45 +0000 (16:07 -0800)]
main: Refactor GetBufferParameteri[64]v.

v2: Split into a refactor commit and an entry point commit.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Add entry point for FlushMappedNamedBufferRange.
Laura Ekstrand [Thu, 12 Feb 2015 00:06:52 +0000 (16:06 -0800)]
main: Add entry point for FlushMappedNamedBufferRange.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Refactor FlushMappedBufferRange.
Laura Ekstrand [Thu, 15 Jan 2015 01:01:20 +0000 (17:01 -0800)]
main: Refactor FlushMappedBufferRange.

v2:-Remove "_mesa" from in front of static software fallback.
   -Split out the refactor from the addition of the DSA entry points.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Add entry point for UnmapNamedBuffer.
Laura Ekstrand [Wed, 14 Jan 2015 22:52:01 +0000 (14:52 -0800)]
main: Add entry point for UnmapNamedBuffer.

v2: review from Ian Romanick
   - Restore VBO_DEBUG and BOUNDS_CHECK
   - Remove _mesa from static software fallback unmap_buffer.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Add entry points for MapNamedBuffer[Range].
Laura Ekstrand [Wed, 11 Feb 2015 22:09:52 +0000 (14:09 -0800)]
main: Add entry points for MapNamedBuffer[Range].

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Refactor MapBuffer[Range].
Laura Ekstrand [Wed, 14 Jan 2015 20:44:39 +0000 (12:44 -0800)]
main: Refactor MapBuffer[Range].

v2: review from Jason Ekstrand
   - Split refactor from addition of DSA entry points.
    review from Ian Romanick
   - Remove "_mesa" from static software fallback map_buffer_range
   - Restore VBO_DEBUG and BOUNDS_CHECK

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Minor whitespace fixes in ClearNamedBuffer[Sub]Data.
Laura Ekstrand [Wed, 11 Feb 2015 19:45:57 +0000 (11:45 -0800)]
main: Minor whitespace fixes in ClearNamedBuffer[Sub]Data.

Reviewed-by: Fredrik Höglund <fredrik@kde.org>
9 years agomain: Add entry points for ClearNamedBuffer[Sub]Data.
Laura Ekstrand [Wed, 11 Feb 2015 20:17:38 +0000 (12:17 -0800)]
main: Add entry points for ClearNamedBuffer[Sub]Data.

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agomain: Refactor ClearBuffer[Sub]Data.
Laura Ekstrand [Tue, 13 Jan 2015 23:20:19 +0000 (15:20 -0800)]
main: Refactor ClearBuffer[Sub]Data.

v2: review by Jason Ekstrand
   - Split refactor of clear buffer sub data from addition of DSA entry
     points.

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agomain: Add entry point for CopyNamedBufferSubData.
Laura Ekstrand [Tue, 13 Jan 2015 21:28:08 +0000 (13:28 -0800)]
main: Add entry point for CopyNamedBufferSubData.

v2: remove _mesa in front of static software fallback.

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agomain: Improve errors and style in BufferSubData.
Laura Ekstrand [Wed, 11 Feb 2015 19:06:42 +0000 (11:06 -0800)]
main: Improve errors and style in BufferSubData.

- More explicit error reporting.
- Removed legacy style.

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agomain: Add entry point for NamedBufferSubData.
Laura Ekstrand [Tue, 13 Jan 2015 19:28:17 +0000 (11:28 -0800)]
main: Add entry point for NamedBufferSubData.

v2: review by Ian Romanick
   - Remove "_mesa" from name of static software fallback buffer_sub_data.
   - Remove mappedRange from _mesa_buffer_sub_data.
   - Removed some cosmetic changes to a separate commit.

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agomain: Add entry point for NamedBufferData.
Laura Ekstrand [Tue, 10 Feb 2015 01:57:46 +0000 (17:57 -0800)]
main: Add entry point for NamedBufferData.

v2: review from Ian Romanick
   - Fix space in ARB_direct_state_access.xml.
   - Remove "_mesa" from the name of buffer_data static fallback.
   - Restore VBO_DEBUG and BOUNDS_CHECK.
   - Fix beginning of comment to start on same line as /*

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agomain: Add entry point for NamedBufferStorage.
Laura Ekstrand [Sat, 10 Jan 2015 00:17:10 +0000 (16:17 -0800)]
main: Add entry point for NamedBufferStorage.

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agomain: Add entry point for CreateBuffers.
Laura Ekstrand [Fri, 19 Dec 2014 01:10:06 +0000 (17:10 -0800)]
main: Add entry point for CreateBuffers.

Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
9 years agoRevert "main: _mesa_cube_level_complete checks NumLayers."
Laura Ekstrand [Tue, 17 Mar 2015 16:43:52 +0000 (09:43 -0700)]
Revert "main: _mesa_cube_level_complete checks NumLayers."

This reverts commit 1ee000a0b6737d6c140d4f07b6044908b8ebfdc7.
Failures with the GLES3 conformance suite and Synmark2 OGLHdrBloom revealed
that this commit was in error.

Extensive testing with Piglit prior to patch review and upstreaming did not
reveal this problem because, in the few Piglit tests that test for cube
completeness, NumLayers = 6.  This is because all of the existing tests use
TextureStorage to initialize the texture, which sets NumLayers.

A new Piglit test has been sent to the mailing list that reproduces the bug
related to this patch ("texturing: Testing
glGenerateMipmap(GL_TEXTURE_CUBE_MAP) without glTexStorage2D").

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/skl: Send a message header when doing constant loads SIMD4x2
Neil Roberts [Thu, 12 Mar 2015 17:41:07 +0000 (17:41 +0000)]
i965/skl: Send a message header when doing constant loads SIMD4x2

Commit 0ac4c272755c7 made it add a header for the send message when
using SIMD4x2 on Skylake because without this it will end up using
SIMD8D. However the patch missed the case when a sampler is being used
to implement constant loads from a buffer surface in a SIMD4x2 vertex
shader.

This fixes 29 Piglit tests, mostly related to the ARL instruction in
vertex programs.

Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
9 years agoi965/fs: in MAD optimizations, switch last argument to be immediate
Tapani Pälli [Mon, 16 Mar 2015 08:08:08 +0000 (10:08 +0200)]
i965/fs: in MAD optimizations, switch last argument to be immediate

Commit bb33a31 introduced optimizations that transform cases of MAD
in to simpler forms but it did not take in to account that src[0]
can not be immediate and did not report progress. Patch switches
src[0] and src[1] if src[0] is immediate and adds progress
reporting. If both sources are immediates, this is taken care of by
the same opt_algebraic pass on later run.

v2: Fix for all cases, use temporary fs_reg (Matt, Kenneth)

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89569
Reviewed-by: Francisco Jerez <currojerez@riseup.net> (v1)
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "10.5" <mesa-stable@lists.freedesktop.org>
9 years agocommon.py: Fix PEP 8 issues.
Vinson Lee [Sat, 14 Mar 2015 08:45:03 +0000 (01:45 -0700)]
common.py: Fix PEP 8 issues.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agogallivm: abort properly when running out of buffer space in lp_disassembly
Roland Scheidegger [Fri, 13 Mar 2015 22:45:20 +0000 (23:45 +0100)]
gallivm: abort properly when running out of buffer space in lp_disassembly

Before this actually ran into an infinite loop printing out "invalid"...

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
9 years agodocs/GL3: also mark GLES3/GS5 for radeonsi as done
Marek Olšák [Mon, 16 Mar 2015 22:24:15 +0000 (23:24 +0100)]
docs/GL3: also mark GLES3/GS5 for radeonsi as done

9 years agost/dri: remove unused include from the automake/scons build
Emil Velikov [Mon, 16 Mar 2015 15:00:19 +0000 (15:00 +0000)]
st/dri: remove unused include from the automake/scons build

st/dri/common hasn't been around for a while.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agoauxiliary/os: fix the android build - s/drm_munmap/os_munmap/
Emil Velikov [Mon, 16 Mar 2015 15:00:18 +0000 (15:00 +0000)]
auxiliary/os: fix the android build - s/drm_munmap/os_munmap/

Squash this silly typo introduced with commit c63eb5dd5ec(auxiliary/os: get
the mmap/munmap wrappers working with android)

Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agogallium/sw/kms: trivial cleanups
Emil Velikov [Mon, 16 Mar 2015 11:50:47 +0000 (11:50 +0000)]
gallium/sw/kms: trivial cleanups

Remove the forward declaration and make use of the DEBUG_PRINT macro for
debug builds.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
9 years agoloader: include <sys/stat.h> for non-sysfs builds
Emil Velikov [Wed, 11 Mar 2015 19:12:35 +0000 (19:12 +0000)]
loader: include <sys/stat.h> for non-sysfs builds

Required by fstat(), otherwise we'll error out due to implicit function
declaration.

Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89530
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reported-by: Vadim Rutkovsky <vrutkovs@redhat.com>
Tested-by: Vadim Rutkovsky <vrutkovs@redhat.com>
9 years agoc11/threads: Use PTHREAD_MUTEX_RECURSIVE by default
Felix Janda [Mon, 2 Feb 2015 19:04:16 +0000 (20:04 +0100)]
c11/threads: Use PTHREAD_MUTEX_RECURSIVE by default

Previously PTHREAD_MUTEX_RECURSIVE_NP had been used on linux for
compatibility with old glibc. Since mesa defines __GNU_SOURCE__
on linux PTHREAD_MUTEX_RECURSIVE is also available since at least
1998. So we can unconditionally use the portable version
PTHREAD_MUTEX_RECURSIVE.

Cc: "10.5" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88534
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agoradeonsi: implement TGSI_OPCODE_BFI (v2)
Marek Olšák [Sat, 28 Feb 2015 13:31:45 +0000 (14:31 +0100)]
radeonsi: implement TGSI_OPCODE_BFI (v2)

v2: Don't use the intrinsics, the shader backend can recognize these
    patterns and generates optimal code automatically.

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
9 years agoradeonsi: add a helper for extracting bitfields from parameters (v2)
Marek Olšák [Fri, 27 Feb 2015 18:09:30 +0000 (19:09 +0100)]
radeonsi: add a helper for extracting bitfields from parameters (v2)

This will be used a lot (especially by tessellation).

v2: don't use the bfe intrinsic

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
9 years agoi965: Emit IF/ELSE/ENDIF/WHILE JIP with type W on Gen7
Antia Puentes [Thu, 12 Mar 2015 12:59:17 +0000 (13:59 +0100)]
i965: Emit IF/ELSE/ENDIF/WHILE JIP with type W on Gen7

IvyBridge and Haswell PRM say that the JIP should be emitted
with type W but we were using UD. The previous implementation
did not show adverse effects, but IMHO it is safer to follow
the specification thoroughly.

Reviewed-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Antia Puentes <apuentes@igalia.com>
9 years agoradeonsi: move scratch reloc state setup
Marek Olšák [Sun, 15 Mar 2015 19:13:52 +0000 (20:13 +0100)]
radeonsi: move scratch reloc state setup

- move it to its own function
- do it after all states are emitted
- bump SI_MAX_DRAW_CS_DWORDS

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: don't emit PA_SC_LINE_STIPPLE if not rendering lines
Marek Olšák [Sun, 15 Mar 2015 18:24:13 +0000 (19:24 +0100)]
radeonsi: don't emit PA_SC_LINE_STIPPLE if not rendering lines

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: don't emit PA_SC_LINE_STIPPLE after every rasterizer state change
Marek Olšák [Sun, 15 Mar 2015 18:21:31 +0000 (19:21 +0100)]
radeonsi: don't emit PA_SC_LINE_STIPPLE after every rasterizer state change

Do it only when the line stipple state is changed.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: move PA_SU_SC_MODE_CNTL to rasterizer state
Marek Olšák [Sun, 15 Mar 2015 17:53:50 +0000 (18:53 +0100)]
radeonsi: move PA_SU_SC_MODE_CNTL to rasterizer state

This requires enabling the optional GL provoking vertex behavior for quads.

+ some cosmetic changes, so that the register is set exactly the same as
on r600.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: implement line and polygon smoothing
Marek Olšák [Sun, 15 Mar 2015 17:20:19 +0000 (18:20 +0100)]
radeonsi: implement line and polygon smoothing

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: add shader code for smoothing
Marek Olšák [Sun, 15 Mar 2015 17:11:19 +0000 (18:11 +0100)]
radeonsi: add shader code for smoothing

The fragment shader multiplies the alpha channel with gl_SampleMaskIn.
If blending is enabled, it looks like MSAA.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: split sample locations into its own state atom
Marek Olšák [Sun, 15 Mar 2015 16:54:29 +0000 (17:54 +0100)]
radeonsi: split sample locations into its own state atom

Sample locations are not updated as often as framebuffers.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: add basic code for overrasterization
Marek Olšák [Sun, 15 Mar 2015 16:14:53 +0000 (17:14 +0100)]
radeonsi: add basic code for overrasterization

This will be used for line and polygon smoothing.
This is GCN-only even though it's in shared code.

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: small cleanup in si_shader_selector_key
Marek Olšák [Sat, 28 Feb 2015 16:22:54 +0000 (17:22 +0100)]
radeonsi: small cleanup in si_shader_selector_key

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: simplify accessing alpha pointer in si_llvm_emit_fs_epilogue
Marek Olšák [Sat, 28 Feb 2015 16:16:57 +0000 (17:16 +0100)]
radeonsi: simplify accessing alpha pointer in si_llvm_emit_fs_epilogue

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
9 years agoradeonsi: add support for easy opcodes from ARB_gpu_shader5
Marek Olšák [Fri, 13 Mar 2015 15:21:11 +0000 (16:21 +0100)]
radeonsi: add support for easy opcodes from ARB_gpu_shader5

I have to use the BFE instrinsics, because BFE is one of the most complex
instructions that can't be matched easily. BFE has 3 conditional branches
and one of them is quite big.

In the isel DAG, lowered BFE has 27 nodes (including leafs).

9 years agoradeonsi: implement bit-finding opcodes from ARB_gpu_shader5
Marek Olšák [Sat, 28 Feb 2015 13:01:43 +0000 (14:01 +0100)]
radeonsi: implement bit-finding opcodes from ARB_gpu_shader5

Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agoradeonsi: implement gl_SampleMaskIn
Marek Olšák [Fri, 27 Feb 2015 23:30:26 +0000 (00:30 +0100)]
radeonsi: implement gl_SampleMaskIn

Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agoradeonsi: add support for SQRT
Marek Olšák [Mon, 2 Mar 2015 01:40:57 +0000 (02:40 +0100)]
radeonsi: add support for SQRT

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agoradeonsi: add support for FMA
Marek Olšák [Fri, 27 Feb 2015 23:44:19 +0000 (00:44 +0100)]
radeonsi: add support for FMA

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
Reviewed-by: Glenn Kennard <glenn.kennard@gmail.com>
9 years agogallium/radeon: don't use LLVMReadOnlyAttribute for ALU
Marek Olšák [Fri, 27 Feb 2015 17:39:40 +0000 (18:39 +0100)]
gallium/radeon: don't use LLVMReadOnlyAttribute for ALU

None of the instructions use a pointer argument.
(+ small cosmetic changes)

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
9 years agotgsi: handle bitwise opcodes in tgsi_opcode_infer_type (v2)
Marek Olšák [Fri, 27 Feb 2015 23:34:53 +0000 (00:34 +0100)]
tgsi: handle bitwise opcodes in tgsi_opcode_infer_type (v2)

v2: set the same types as the destination type in tgsi_exec

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agogallium: add FMA and DFMA opcodes (v3)
Marek Olšák [Fri, 27 Feb 2015 23:26:31 +0000 (00:26 +0100)]
gallium: add FMA and DFMA opcodes (v3)

Needed by ARB_gpu_shader5.

v2: select DMAD for FMA with double precision
v3: add and select DFMA

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
9 years agofreedreno: update generated headers
Rob Clark [Sun, 15 Mar 2015 21:59:01 +0000 (17:59 -0400)]
freedreno: update generated headers

Fix a3xx texture layer-size.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>
9 years agofreedreno/ir3: remove old compiler
Rob Clark [Wed, 11 Mar 2015 19:10:25 +0000 (15:10 -0400)]
freedreno/ir3: remove old compiler

Now that piglit is no longer falling back to old compiler for any tests,
we can remove it.  Hurray \o/

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/ir3: avoid scheduler deadlock
Rob Clark [Wed, 11 Mar 2015 17:21:42 +0000 (13:21 -0400)]
freedreno/ir3: avoid scheduler deadlock

Deadlock can occur if we schedule an address register write, yet some
instructions which depend on that address register value also depend on
other unscheduled instructions that depend on a different address
register value.  To solve this, before scheduling an address register
write, ensure that all the other dependencies of the instructions which
consume this address register are already scheduled.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/ir3: bit of cleanup
Rob Clark [Wed, 11 Mar 2015 16:36:26 +0000 (12:36 -0400)]
freedreno/ir3: bit of cleanup

Add an array_insert() macro to simplify inserting into dynamically sized
arrays, add a comment, and remove unused prototype inherited from the
original freedreno.git/fdre-a3xx test code, etc.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
9 years agoi965: De-duplicate is_expression_commutative() functions.
Kenneth Graunke [Fri, 13 Mar 2015 21:34:06 +0000 (14:34 -0700)]
i965: De-duplicate is_expression_commutative() functions.

Create a backend_inst::is_commutative() method to replace two static
functions that did the exact same thing.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
9 years agoi965/gen4-5: Cope with immutable-format texture revalidation
Chris Forbes [Mon, 8 Dec 2014 07:37:00 +0000 (20:37 +1300)]
i965/gen4-5: Cope with immutable-format texture revalidation

This is unfortunately sometimes necessary due to rebasing levels when
rendering into them.

16 piglits crash -> pass, when building mesa with debug enabled.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agodocs: add news item and link release notes for mesa 10.5.1
Emil Velikov [Fri, 13 Mar 2015 23:36:33 +0000 (23:36 +0000)]
docs: add news item and link release notes for mesa 10.5.1

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agodocs: Add sha256 sums for the 10.5.1 release
Emil Velikov [Fri, 13 Mar 2015 23:32:12 +0000 (23:32 +0000)]
docs: Add sha256 sums for the 10.5.1 release

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit 2abba086ca84f200fae940129c0a5342c3748f00)

9 years agoAdd release notes for the 10.5.1 release
Emil Velikov [Fri, 13 Mar 2015 22:32:57 +0000 (22:32 +0000)]
Add release notes for the 10.5.1 release

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
(cherry picked from commit 11c0ff60ef19cca84452aa989fb8bb25127473e0)

9 years agofreedreno: fix slice pitch calculations
Ilia Mirkin [Fri, 13 Mar 2015 05:36:57 +0000 (01:36 -0400)]
freedreno: fix slice pitch calculations

For example if width were 65, the first slice would get 96 while the
second would get 32. However the hardware appears to expect the second
pitch to be 64, based on halving the 96 (and aligning up to 32).

This fixes texelFetch piglit tests on a3xx below a certain size. Going
higher they break again, but most likely due to unrelated reasons.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agofreedreno/a3xx: use the same layer size for all slices
Ilia Mirkin [Fri, 13 Mar 2015 04:53:49 +0000 (00:53 -0400)]
freedreno/a3xx: use the same layer size for all slices

We only program in one layer size per texture, so that means that all
levels must share one size. This makes the piglit test

bin/texelFetch fs sampler2DArray

have the same breakage as its non-array version instead of being
completely off, and makes

bin/ext_texture_array-gen-mipmap

start passing.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
9 years agoi965/vs: Add missing resolve_bool_comparison calls on GEN4 and GEN5
Ian Romanick [Wed, 25 Feb 2015 01:57:18 +0000 (20:57 -0500)]
i965/vs: Add missing resolve_bool_comparison calls on GEN4 and GEN5

The ir_unop_any problem was discovered by some later optimization passes
that generate ir_triop_csel.  I was also able to reproduce it by
modifying the gl-2.0-vertexattribpointer vertex shader to generate its
result using

   color = mix(vec4(0, 1, 0, 0),
               vec4(1, 0, 0, 0),
               bvec4(any(greaterThan(diff, vec4(tolerance)))));

instead of an if-statement.  This also required using #version 130 and
MESA_GLSL_VERSION_OVERRIDE=130.

I have not nominated this for stable releases because I don't think
there's any way to trigger the problem without GLSL 1.30 or
optimizations that don't exist in stable.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Abdiel Janulgue <abdiel.janulgue@intel.com>
9 years agoi965/disasm: Fix format strings
Chris Forbes [Fri, 13 Mar 2015 18:10:11 +0000 (07:10 +1300)]
i965/disasm: Fix format strings

Most of the brw_inst_* api returns 64bit values. This fixes disassembly
of sampler messages, etc.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agoi965/disasm: Mark format() as being printf-style.
Chris Forbes [Fri, 13 Mar 2015 18:10:10 +0000 (07:10 +1300)]
i965/disasm: Mark format() as being printf-style.

This allows us to get warnings from GCC when we mess up the format
strings.

Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agodocs: List ARB_shading_language_packing/EXT_shader_integer_mix.
Matt Turner [Thu, 12 Mar 2015 01:43:56 +0000 (18:43 -0700)]
docs: List ARB_shading_language_packing/EXT_shader_integer_mix.

Reviewed-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
9 years agoglsl: Expose built-in packing functions under GLSL 4.2.
Matt Turner [Thu, 12 Mar 2015 01:14:28 +0000 (18:14 -0700)]
glsl: Expose built-in packing functions under GLSL 4.2.

ARB_shading_language_packing is part of GLSL 4.2, not 4.0 as I
mistakenly believed. The following functions are available only with
ARB_shading_language_packing, GLSL 4.2 (not GLSL 4.0), or ES 3.0:

   - packSnorm2x16
   - unpackSnorm2x16
   - packHalf2x16
   - unpackHalf2x16

Reviewed-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
9 years agoegl: Create queryable strings in eglInitialize().
Matt Turner [Tue, 10 Mar 2015 18:41:57 +0000 (11:41 -0700)]
egl: Create queryable strings in eglInitialize().

Creating/recreating the strings in eglQueryString() is extra work and
isn't thread-safe, as exhibited by shader-db's run.c using libepoxy.

Multiple threads in run.c call eglReleaseThread() around the same time.
libepoxy calls eglQueryString() to determine whether eglReleaseThread()
exists, and our EGL implementation passes a pointer to the version
string to libepoxy while simultaneously overwriting the string, leading
to a failure in libepoxy.

Moreover, the EGL spec says (emphasis mine):

"eglQueryString returns a pointer to a *static*, zero-terminated string"

This patch moves some auxiliary functions from eglmisc.c to eglapi.c so
that they may be used to create the extension, API, and version strings
once during eglInitialize(). The auxiliary functions are renamed from
_eglUpdate* to _eglCreate*, and some checks made unnecessary by calling
the functions from eglInitialize() are removed.

Reviewed-by: Chad Versace <chad.versace@intel.com>
9 years agoglsl: optimize (0 cmp x + y) into (-x cmp y).
Samuel Iglesias Gonsalvez [Tue, 24 Feb 2015 18:02:57 +0000 (19:02 +0100)]
glsl: optimize (0 cmp x + y) into (-x cmp y).

The optimization done by commit 34ec1a24d did not take it into account.

Fixes:

dEQP-GLES3.functional.shaders.random.all_features.fragment.20

Signed-off-by: Samuel Iglesias Gonsalvez <siglesias@igalia.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>
9 years agomesa: Check for valid PBO access in gl(Compressed)Tex(Sub)Image calls
Eduardo Lima Mitev [Thu, 12 Mar 2015 07:16:09 +0000 (08:16 +0100)]
mesa: Check for valid PBO access in gl(Compressed)Tex(Sub)Image calls

This patch adds two types of checks to the gl(Compressed)Tex(Sub)Imgage family
of functions when a pixel buffer object is bound to GL_PIXEL_UNPACK_BUFFER:

- That the buffer is not mapped.
- The total data size is within the boundaries of the buffer size.

It does so by calling auxiliary validations functions from PBO API:
_mesa_validate_pbo_source() for non-compressed texture calls, and
_mesa_validate_pbo_source_compressed() for compressed texture calls.

The first check is defined in Section 6.3.2 'Effects of Mapping Buffers
on Other GL Commands' of the GLES 3.1 spec, page 57:

    "Any GL command which attempts to read from, write to, or change the
     state of a buffer object may generate an INVALID_OPERATION error if all
     or part of the buffer object is mapped. However, only commands which
     explicitly describe this error are required to do so. If an error is not
     generated, using such commands to perform invalid reads, writes, or
     state changes will have undefined results and may result in GL
     interruption or termination."

Similar wording exists in GL 4.5 spec, page 76.

In the case of gl(Compressed)Tex(Sub)Image(2,3)D, the specification doesn't force
implemtations to throw an error. However since Mesa don't currently implement
checks to determine when it is safe to read/write from/to a mapped PBO, we
should always return the error if all or parts of it are mapped.

The 2nd check is defined in Section 8.5 'Texture Image Specification' of the
OpenGL 4.5 spec, page 203:

    "An INVALID_OPERATION error is generated if a pixel unpack buffer object
     is bound and storing texture data would access memory beyond the end of
     the pixel unpack buffer."

Fixes 4 dEQP tests:
* dEQP-GLES3.functional.negative_api.texture.compressedteximage2d_invalid_buffer_target
* dEQP-GLES3.functional.negative_api.texture.compressedtexsubimage2d_invalid_buffer_target
* dEQP-GLES3.functional.negative_api.texture.compressedteximage3d_invalid_buffer_target
* dEQP-GLES3.functional.negative_api.texture.compressedtexsubimage3d_invalid_buffer_target

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Separate PBO validation checks from buffer mapping, to allow reuse
Eduardo Lima Mitev [Thu, 12 Mar 2015 07:14:03 +0000 (08:14 +0100)]
mesa: Separate PBO validation checks from buffer mapping, to allow reuse

Internal PBO functions such as _mesa_map_validate_pbo_source() and
_mesa_validate_pbo_compressed_teximage() perform validation and buffer mapping
within the same call.

This patch takes out the validation into separate functions to allow reuse
of functionality by other code (i.e, gl(Compressed)Tex(Sub)Image).

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agomesa: Set the correct image size in _mesa_validate_pbo_access()
Eduardo Lima Mitev [Thu, 5 Mar 2015 08:20:11 +0000 (09:20 +0100)]
mesa: Set the correct image size in _mesa_validate_pbo_access()

_mesa_validate_pbo_access() provides a generic way to check that a
requested pixel transfer operation on a PBO falls within the
boundaries of the buffer. It is used in various other places, and
depending on the caller, some arguments are used or not.

In particular, the 'clientMemSize' argument is used only by calls
that are knowledgeable of the total size of the user data involved
in a pixel transfer, such as the case of compressed texture image
calls. Other calls don't provide 'clientMemSize' directly since it
is made implicit from the size and format of the texture, and its
data type. In these cases, a sufficiently big value is passed to
'clientMemSize' (INT_MAX) to avoid an incorrect constrain.

The problem is that _mesa_validate_pbo_access() use uint
pointers to make the calculations, which are 64 bits long in 64
bits platforms, meanwhile the dummy INT_MAX passed in 'clientMemSize'
is just 32 bits. This causes a constrain that is not desired.

This patch fixes that by checking that if 'clientMemSize' is MAX_INT,
then UINTPTR_MAX is assumed instead.

This is an ugly workaround to the fact that _mesa_validate_pbo_access()
intends to be a one function fits all. The clean solution here would
be to break it into different functions that provide the adequate API
for each of the possible code paths and validation needs.

Since there are callers relying on passing INT_MAX to 'clientMemSize',
this patch is necessary to deal with the problem above while a cleaner
implementation of the PBO API is not implemented.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agometa: Remove error checks for texture <-> pixel-buffer transfers that don't belong...
Eduardo Lima Mitev [Tue, 10 Mar 2015 18:33:30 +0000 (19:33 +0100)]
meta: Remove error checks for texture <-> pixel-buffer transfers that don't belong in driver code

The implementation of texture <-> pixel-buffer transfers in drivers common layer
includes certain error checks and argument validation that don't belong there,
considering how the Mesa codebase is laid out. These are higher level
validations that, if necessary, should be performed earlier (i.e, in GL API
entry points).

This patch simply removes these error checks from driver code.

For more information, see discussion at
http://lists.freedesktop.org/archives/mesa-dev/2015-February/077417.html.

Reviewed-by: Laura Ekstrand <laura@jlekstrand.net>
9 years agoutil: convert slab macros to inline functions
Brian Paul [Thu, 12 Mar 2015 21:50:20 +0000 (15:50 -0600)]
util: convert slab macros to inline functions

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
9 years agoegl: fix cast to silence compiler warning
Brian Paul [Thu, 12 Mar 2015 14:35:38 +0000 (08:35 -0600)]
egl: fix cast to silence compiler warning

eglcurrent.c: In function '_eglSetTSD':
eglcurrent.c:57:4: warning: passing argument 2 of 'tss_set' discards
'const' qualifier from pointer target type [enabled by default]
    tss_set(_egl_TSD, (const void *) t);
    ^
In file included from ../../../include/c11/threads.h:72:0,
                 from eglcurrent.c:32:
../../../include/c11/threads_posix.h:357:1: note: expected 'void *'
but argument is of type 'const void *'
 tss_set(tss_t key, void *val)
 ^

Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
9 years agogallivm: (trivial) Fix typo in comment introduced by 70dc8a
Alexandre Demers [Fri, 13 Mar 2015 00:50:08 +0000 (20:50 -0400)]
gallivm: (trivial) Fix typo in comment introduced by 70dc8a

Fix typo in comment introduced by 70dc8a

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Jose Fonseca <jfonseca@vmware.com>
9 years agomesa: improve ARB_copy_image internal format compat check
Seán de Búrca [Sat, 7 Mar 2015 09:23:53 +0000 (02:23 -0700)]
mesa: improve ARB_copy_image internal format compat check

The memory layout of compatible internal formats may differ in bytes per
block, so TexFormat is not a reliable measure of compatibility. For example,
GL_RGB8 and GL_RGB8UI are compatible formats, but GL_RGB8 may be laid out in
memory as B8G8R8X8. If GL_RGB8UI has a 3 byte-per-block memory layout, the
existing compatibility check will fail.

Additionally, the current check allows any two compressed textures which share
block size to be used, whereas the spec gives an explicit table of compatible
formats.

v2: Use a switch instead of array iteration for block class and show the
    correct GL error when internal formats are mismatched.
v3: Include spec citations for new compatibility checks, rearrange check
    order to ensure that compressed, view-compatible formats return the
    correct result, and make style fixes. Original commit message amended
    for clarity.
v4: Reformatted spec citations.

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agonir: Fix non-determinism in nir_lower_vars_to_ssa().
Kenneth Graunke [Tue, 10 Mar 2015 01:36:31 +0000 (18:36 -0700)]
nir: Fix non-determinism in nir_lower_vars_to_ssa().

Previously, we stored derefs in a hash table, using the malloc'd pointer
as the key.  Then, we walked through the hash table and generated code,
based on the order of the hash table's elements.

Memory addresses returned by malloc are pretty much random, which meant
that the hash was random, and the hash table's elements would be walked
in some random order.  This led to successive compiles of the same
shader using different variable names and slightly different orderings
of phi-nodes.  Code could not be diff'd, and the final assembly would
sometimes change slightly too.

It turns out the only point of the hash table was to avoid inserting
the same node multiple times for different dereferences.  We never
actually searched the hash table!  This patch uses an intrusive
linked list instead.  Since exec_list uses head and tail sentinels,
checking prev or next against NULL will tell us whether the node is
already in the list.

Pair programming with Jason Ekstrand.

Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
9 years agoutil: Fix foreach_list_typed_safe when exec_node is not at offset 0.
Jason Ekstrand [Tue, 10 Mar 2015 01:36:30 +0000 (18:36 -0700)]
util: Fix foreach_list_typed_safe when exec_node is not at offset 0.

__next and __prev are pointers to the structure containing the exec_node
link, not the embedded exec_node.  NULL checks would fail unless the
embedded exec_node happened to be at offset 0 in the parent struct.

v2: Jason Ekstrand <jason.ekstrand@intel.com>:
   Use "(__node)->__field.next != NULL" to check for the end of the list
   instead of the "&__next->__field != NULL".  The former is far more
   obviously correct as it matches what the non-safe versions do.  The
   original code tried to avoid any use of __next as the client code may
   delete it during its execution.  However, since the looping condition is
   checked after the iteration clause but before the client code is
   executed, we know that __node is valid during the looping condition.

Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
9 years agoi965: Use NIR for scalar VS when INTEL_USE_NIR is set.
Kenneth Graunke [Mon, 9 Mar 2015 08:58:59 +0000 (01:58 -0700)]
i965: Use NIR for scalar VS when INTEL_USE_NIR is set.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/fs: Add VS output support to nir_setup_outputs().
Kenneth Graunke [Mon, 9 Mar 2015 08:58:58 +0000 (01:58 -0700)]
i965/fs: Add VS output support to nir_setup_outputs().

Adapted from fs_visitor::visit(ir_variable *).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/fs: Handle VS inputs in the NIR backend.
Kenneth Graunke [Mon, 9 Mar 2015 08:58:57 +0000 (01:58 -0700)]
i965/fs: Handle VS inputs in the NIR backend.

(Jason noted that this is not a good long term solution, and we should
instead improve nir_lower_io so that this extra set of MOVs is
unnecessary.  I tend to agree, but decided we could do that as a
follow-up improvement.)

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/fs: Refactor fs_visitor::nir_setup_inputs().
Kenneth Graunke [Mon, 9 Mar 2015 08:58:56 +0000 (01:58 -0700)]
i965/fs: Refactor fs_visitor::nir_setup_inputs().

No functional change.  In preparation for supporting vertex shaders,
this adds a switch statement on shader stage (since vertex attributes
and fragment shader varyings will need different handling).  It also
renames "varying" to "input", to be more general.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965: Implement NIR intrinsics for loading VS system values.
Kenneth Graunke [Mon, 9 Mar 2015 08:58:55 +0000 (01:58 -0700)]
i965: Implement NIR intrinsics for loading VS system values.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agonir: Add intrinsics for SYSTEM_VALUE_BASE_VERTEX and VERTEX_ID_ZERO_BASE
Kenneth Graunke [Mon, 9 Mar 2015 08:58:54 +0000 (01:58 -0700)]
nir: Add intrinsics for SYSTEM_VALUE_BASE_VERTEX and VERTEX_ID_ZERO_BASE

Ian and I added these around the time Connor was developing NIR.  Now
that both exist, we should make them work together!

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/nir: Lower to registers a bit later.
Kenneth Graunke [Mon, 9 Mar 2015 08:58:53 +0000 (01:58 -0700)]
i965/nir: Lower to registers a bit later.

We can't safely call nir_optimize() with register present, since several
passes called in the loop can't handle registers, and will fail asserts.

Notably, nir_lower_vec_alus() and nir_opt_algebraic() really don't want
registers.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/nir: Optimize after nir_lower_var_copies().
Kenneth Graunke [Mon, 9 Mar 2015 08:58:52 +0000 (01:58 -0700)]
i965/nir: Optimize after nir_lower_var_copies().

Array variable copy splitting generates a bunch of stuff we want to
clean up before proceeding.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agoi965/fs: Store a pointer to brw_sampler_prog_key_data in the visitor.
Kenneth Graunke [Mon, 9 Mar 2015 08:58:51 +0000 (01:58 -0700)]
i965/fs: Store a pointer to brw_sampler_prog_key_data in the visitor.

The NIR backend hardcodes brw_wm_prog_key at the moment, which won't
work when we support scalar VS.  We could use get_tex(), but it's a
static method.  I was going to promote it to fs_visitor, but then
realized that both parameters (stage and key) are already members.

It then occured to me that we could just set up a pointer in the
constructor, and skip having a function altogether.

This patch also converts all existing users to use key_tex.

v2: Make key_tex a "const brw_sampler_prog_key_data *" instead of
    non-const; word-wrap some lines.  (Review comments from Topi.)

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
9 years agotnl: HAVE_LE32_VERTS is never defined, remove associated code
Brian Paul [Wed, 11 Mar 2015 23:10:53 +0000 (17:10 -0600)]
tnl: HAVE_LE32_VERTS is never defined, remove associated code

Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agomesa: move LONGSTRING into generated enums.c
Brian Paul [Wed, 11 Mar 2015 22:54:15 +0000 (16:54 -0600)]
mesa: move LONGSTRING into generated enums.c

enums.c is the only place this directive is needed.

Reviewed-by: Matt Turner <mattst88@gmail.com>
9 years agomesa: remove _ASMAPI, ASMAPIP
Brian Paul [Wed, 11 Mar 2015 14:38:09 +0000 (08:38 -0600)]
mesa: remove _ASMAPI, ASMAPIP

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
9 years agomesa: remove _XFORMAPI
Brian Paul [Wed, 11 Mar 2015 14:33:21 +0000 (08:33 -0600)]
mesa: remove _XFORMAPI

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