Marek Olšák [Thu, 23 Nov 2017 19:22:25 +0000 (20:22 +0100)]
radeonsi: set COMPUTE_RESOURCE_LIMITS.FORCE_SIMD_DIST when profitable
ported from Vulkan
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Dave Airlie [Tue, 14 Nov 2017 05:11:39 +0000 (15:11 +1000)]
ac/nir: don't write tcs outputs to LDS that aren't read back.
If the TCS doesn't read back the outputs, no need to store them
to LDS in the first place. (except for tess factors).
This seems to give about 50fps (3290->3330) with tessellation demo.
I haven't tested if it impacts DoW3 at all.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Dave Airlie [Tue, 14 Nov 2017 05:10:44 +0000 (15:10 +1000)]
nir: fill outputs_read field and add patch outputs read (v2)
This is to be used for TCS optimisations on radv.
v2: don't set written on reads (nha)
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Dave Airlie [Thu, 25 May 2017 21:57:52 +0000 (07:57 +1000)]
r600/eg: dump event type in dumps
This just makes it easier to debug some things.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Tobias Klausmann [Sun, 12 Nov 2017 01:51:55 +0000 (02:51 +0100)]
nouveau/compiler: Allow to omit line numbers when printing instructions
This comes in handy when checking "NV50_PROG_DEBUG=1" outputs with diff!
V2:
- Use environmental variable (Karol Herbst)
V3:
- Use the already populated nv50_ir_prog_info to forward information to the
print pass (Pierre Moreau)
V4:
- get rid of default value in PrintPass constructor
Signed-off-by: Tobias Klausmann <tobias.johannes.klausmann@mni.thm.de>
Reviewed-by: Pierre Moreau <pierre.morrow@free.fr>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Nicolai Hähnle [Wed, 22 Nov 2017 16:52:43 +0000 (17:52 +0100)]
radeonsi: try flushing unflushed fences in si_fence_finish even when timeout == 0
Under certain conditions, waiting on a GL sync objects should act like
a flush, regardless of the timeout.
Portal 2, CS:GO, and presumably other Source engine games rely on this
behavior and hang during loading without this fix.
Fixes: bc65dcab3bc4 ("radeonsi: avoid syncing the driver thread in si_fence_finish")
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103902
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103904
Ilia Mirkin [Thu, 16 Nov 2017 06:48:20 +0000 (01:48 -0500)]
nv50/ir: move LateAlgebraicOpt to the very end
Memory loads can take offsets, but the SHLADD will often attempt to
consume the offsets too. As there may be multiple memory loads with the
same base but different offsets, those would end up in a SHLADD instead
of the offset of the memory operation.
This moves the pass after we've had a chance to attempt to propagate
immediate adds into the indirect offset.
total instructions in shared programs :
6580681 ->
6567716 (-0.20%)
total gprs used in shared programs : 944261 -> 943375 (-0.09%)
total shared used in shared programs : 0 -> 0 (0.00%)
total local used in shared programs : 15328 -> 15328 (0.00%)
total bytes used in shared programs :
60339896 ->
60221504 (-0.20%)
local shared gpr inst bytes
helped 0 0 555 2698 2698
hurt 0 0 138 336 336
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Ilia Mirkin [Thu, 16 Nov 2017 04:32:16 +0000 (23:32 -0500)]
nv50/ir: when merging immediates/consts, load directly
When a MERGE operation gets its constraint moves added, it
susbstantially extends live ranges to be reusing an immediate from
earlier in the program (not to mention the silliness of loading an
immediate into a register, and then moving into another register).
We detect these scenarios and insert moves that take the immediate or
constbuf load directly into the register. If it's the last use, then we
can just move that operation to the closer location.
With SM35 (255 regs) we get these results:
total instructions in shared programs :
6583670 ->
6580681 (-0.05%)
total gprs used in shared programs : 950818 -> 944261 (-0.69%)
total shared used in shared programs : 0 -> 0 (0.00%)
total local used in shared programs : 15328 -> 15328 (0.00%)
total bytes used in shared programs :
60367456 ->
60339896 (-0.05%)
local shared gpr inst bytes
helped 0 0 4584 3186 3186
hurt 0 0 55 968 968
I suspect they will be better for SM20 and SM30.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Ilia Mirkin [Sat, 11 Nov 2017 03:10:46 +0000 (22:10 -0500)]
nv50/ir: add optimization for modulo by a non-power-of-2 value
We can still use the optimized division methods which make use of
multiplication with overflow.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tobias Klausmann <tobias.johannes.klausmann@mni.thm.de>
Ilia Mirkin [Sat, 11 Nov 2017 02:47:59 +0000 (21:47 -0500)]
nv50/ir: optimize signed integer modulo by pow-of-2
It's common to use signed int modulo in GLSL. As it happens, the GLSL
specs allow the result to be undefined, but that seems fairly
surprising. It's not that much more effort to get it right, at least for
positive modulo operators.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Matt Turner [Sun, 26 Nov 2017 00:45:27 +0000 (16:45 -0800)]
util: Just give up and define PIPE_ARCH_LITTLE_ENDIAN on MSVC
MSVC doesn't support #warning?! Getting really tired of this.
Andres Gomez [Sun, 26 Nov 2017 00:15:43 +0000 (02:15 +0200)]
docs: remove bug 103626 from fix list as per 17.2.6
Bug https://bugs.freedesktop.org/show_bug.cgi?id=103626 was
incorrectly listed as fixed.
Signed-off-by: Andres Gomez <agomez@igalia.com>
(cherry picked from commit
b9b60dbf55a1307a60a333c70c3add3643243c36)
Matt Turner [Sat, 25 Nov 2017 23:56:43 +0000 (15:56 -0800)]
util: Use preprocessor correctly
Fixes: 6a353479a757 ("util: Assume little endian in the absence of
platform-specific handling")
Andres Gomez [Sat, 25 Nov 2017 23:46:25 +0000 (01:46 +0200)]
docs: update calendar, add news item and link release notes for 17.2.6
Signed-off-by: Andres Gomez <agomez@igalia.com>
Andres Gomez [Sat, 25 Nov 2017 23:40:36 +0000 (01:40 +0200)]
docs: add sha256 checksums for 17.2.6
Signed-off-by: Andres Gomez <agomez@igalia.com>
(cherry picked from commit
93c2beafc0a7fa2f210b006d22aba61caa71f773)
Andres Gomez [Sat, 25 Nov 2017 23:32:53 +0000 (01:32 +0200)]
docs: add release notes for 17.2.6
Signed-off-by: Andres Gomez <agomez@igalia.com>
(cherry picked from commit
00b52f8e99653316a090826914509a138a1c78f7)
Ilia Mirkin [Sun, 19 Nov 2017 21:36:08 +0000 (16:36 -0500)]
freedreno/a4xx: add ARB_framebuffer_no_attachments support
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Ilia Mirkin [Sun, 19 Nov 2017 21:32:12 +0000 (16:32 -0500)]
freedreno/a4xx: add indirect draw support
This is a copy of the a5xx logic. Fails a few tests, but basic
functionality is there.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Ilia Mirkin [Sun, 19 Nov 2017 21:31:02 +0000 (16:31 -0500)]
freedreno: regenerate pm4 header, adjust code for new names
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Ilia Mirkin [Sun, 19 Nov 2017 20:13:41 +0000 (15:13 -0500)]
freedreno/a4xx: add stencil texturing support
Copied from a5xx, should be identical.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Ilia Mirkin [Sun, 19 Nov 2017 17:28:53 +0000 (12:28 -0500)]
freedreno/ir3: add a pass to lower tg4 to txl, enable gather on a4xx
Unfortunately Adreno A4xx hardware returns incorrect results with the
GATHER4 opcodes. As a result, we have to lower to 4 individual texture
calls (txl since we have to force lod to 0). We achieve this using
offsets, including on cube maps which normally never have offsets.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Ilia Mirkin [Sun, 19 Nov 2017 17:27:12 +0000 (12:27 -0500)]
nir: allow texture offsets with cube maps
GL doesn't have this, but some hardware supports it. This is convenient
for lowering tg4 to plain texture calls, which is necessary on Adreno
A4xx hardware.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Matt Turner [Thu, 23 Nov 2017 18:41:34 +0000 (10:41 -0800)]
util: Fix disk_cache index calculation on big endian
The cache-test test program attempts to create a collision (using key_a
and key_a_collide) by making the first two bytes identical. The idea is
fine -- the shader cache wants to use the first four characters of a
SHA1 hex digest as the index.
The following program
unsigned char array[4] = {1, 2, 3, 4};
int *ptr = (int *)array;
for (int i = 0; i < 4; i++) {
printf("%02x", array[i]);
}
printf("\n");
printf("%08x\n", *ptr);
prints
01020304
04030201
on little endian, and
01020304
01020304
on big endian.
On big endian platforms reading the character array back as an int (as
is done in disk_cache.c) does not yield the same results as reading the
byte array.
To get the first four characters of the SHA1 hex digest when we mask
with CACHE_INDEX_KEY_MASK, we need to byte swap the int on big endian
platforms.
Bugzilla: https://bugs.freedesktop.org/103668
Bugzilla: https://bugs.gentoo.org/637060
Bugzilla: https://bugs.gentoo.org/636326
Fixes: 87ab26b2ab35 ("glsl: Add initial functions to implement an
on-disk cache")
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Matt Turner [Wed, 22 Nov 2017 23:10:47 +0000 (15:10 -0800)]
util: Add a SHA1 unit test program
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Matt Turner [Thu, 23 Nov 2017 06:39:51 +0000 (22:39 -0800)]
util: Fix SHA1 implementation on big endian
The code defines a macro blk0(i) based on the preprocessor condition
BYTE_ORDER == LITTLE_ENDIAN. If true, blk0(i) is defined as a byte swap
operation. Unfortunately, if the preprocessor macros used in the test
are no defined, then the comparison becomes 0 == 0 and it evaluates as
true.
Fixes: d1efa09d342b ("util: import sha1 implementation from OpenBSD")
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Matt Turner [Sat, 25 Nov 2017 04:25:04 +0000 (20:25 -0800)]
util: Assume little endian in the absence of platform-specific handling
Marek Olšák [Wed, 15 Nov 2017 22:53:04 +0000 (23:53 +0100)]
mesa: shrink VERT_ATTRIB bitfields to 32 bits
There are only 32 vertex attribs now.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Marek Olšák [Wed, 15 Nov 2017 22:24:56 +0000 (23:24 +0100)]
mesa: remove unused vertex attrib WEIGHT
We don't support ARB_vertex_blend.
Note that the attribute aliasing check for ARB_vertex_program had to be
rewritten.
vbo_context: 20344 -> 20008 bytes
gl_context: 74672 -> 74616 bytes
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Marek Olšák [Wed, 15 Nov 2017 21:58:58 +0000 (22:58 +0100)]
mesa: don't assign numbers to vertex attrib enums manually
I plan to remove one of them.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Marek Olšák [Sun, 19 Nov 2017 20:29:46 +0000 (21:29 +0100)]
gallium/hud: add HUD sharing within a context share group
This is needed for profiling multi-context applications like Chrome.
One context can record queries and another context can draw the HUD.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sun, 19 Nov 2017 20:04:07 +0000 (21:04 +0100)]
gallium/hud: update the HUD interface for multiple contexts
This is the boring subset of the following commit.
All new parameters are optional.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sun, 19 Nov 2017 03:36:38 +0000 (04:36 +0100)]
gallium/hud: prevent a crash if the recording context is inactive
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 17:07:40 +0000 (18:07 +0100)]
gallium/hud: separate code for record context init/release
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 17:07:40 +0000 (18:07 +0100)]
gallium/hud: separate code for draw context init/release
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 16:53:34 +0000 (17:53 +0100)]
gallium/hud: don't use hud->pipe in hud_parse_env_var
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 16:46:51 +0000 (17:46 +0100)]
gallium/hud: use cso_get_pipe_context
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 16:43:42 +0000 (17:43 +0100)]
cso: add cso_get_pipe_context
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 15:25:52 +0000 (16:25 +0100)]
gallium/hud: pass pipe_context explicitly to most functions
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 14:23:23 +0000 (15:23 +0100)]
gallium/hud: split hud_draw into 3 separate functions
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 23:24:40 +0000 (00:24 +0100)]
st/dri: remove dead code and incorrect comment around make_current
Core Mesa already handles flushing based on ContextReleaseBehavior,
so the comment is wrong.
Also, old_st is always NULL, because unbind_context always precedes
make_current.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Sat, 18 Nov 2017 23:19:19 +0000 (00:19 +0100)]
st/dri: clean up dri_unbind_context
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Tue, 21 Nov 2017 00:47:30 +0000 (01:47 +0100)]
radeonsi: expose all CB performance counters on Stoney
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Mon, 20 Nov 2017 00:51:50 +0000 (01:51 +0100)]
radeonsi: handle imported textures with DCC robustly
now you can hack the driver to enable DCC for displayable textures and
Glamor that doesn't enable that by default won't crash anymore.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Fri, 17 Nov 2017 16:52:09 +0000 (17:52 +0100)]
radeonsi: fix a typo in creating monolithic ES-GS
This has no effect because both occupy the same memory in a union.
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Fri, 17 Nov 2017 03:56:13 +0000 (04:56 +0100)]
radeonsi: don't write undefined output channels to LDS in LS
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Thu, 9 Nov 2017 22:34:26 +0000 (23:34 +0100)]
radeonsi: use ac.lds for shared memory
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Marek Olšák [Thu, 9 Nov 2017 22:25:34 +0000 (23:25 +0100)]
radeonsi: do 64-bit LDS loads recursively
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Jon Turney [Sat, 11 Nov 2017 14:48:10 +0000 (14:48 +0000)]
mapi: Teach es{1,2}api/ABI-check shared library names on Cygwin
Ideally we'd be able to get the library filename from libtool, but that
doesn't seem to be a feature...
Use of ${uname} is presumably ok here as we won't be running 'make check' if
we are cross-compiling
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Samuel Pitoiset [Wed, 22 Nov 2017 15:13:28 +0000 (16:13 +0100)]
Revert "radv: remove unnecessary memset() in radv_AllocateCommandBuffers()"
This fixes two CTS regressions:
- dEQP-VK.api.object_management.alloc_callback_fail_multiple.command_buffer_primary
- dEQP-VK.api.object_management.alloc_callback_fail_multiple.command_buffer_secondary
These two tests are part the mustpass lists, so presumably they
are correct and my change was wrong.
This reverts commit
0f68208f1d1d3b7b2963dab40e84c60212518692.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 22 Nov 2017 19:13:26 +0000 (20:13 +0100)]
radv/winsys: improve error messages when the buffer list creation failed
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Samuel Pitoiset [Wed, 22 Nov 2017 19:13:25 +0000 (20:13 +0100)]
radv/winsys: do not try to create a BO list with 0 buffers
This happens when all BOs have the RADEON_FLAG_NO_INTERPROCESS_SHARING
(DRM version >= 3.23) flag set. This flag is mainly used for reducing
overhead on the userspace side because we don't have to put those BOs
inside the list.
Though, if the driver tries to create a list with 0 buffers inside it,
libdrm returns -EINVAL and the app just crashes.
This fixes a bunch of CTS dEQP-VK.sparse_resources.* fails (~100).
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Iago Toral Quiroga [Tue, 21 Nov 2017 10:33:53 +0000 (11:33 +0100)]
i965/vec4: fix splitting of interleaved attributes
When we split an instruction that reads an uniform value
(vstride 0) we need to respect the vstride on the second
half of the instruction (that is, the second half should
read the same region as the first).
We were doing this already, but we didn't account for
stages that have interleaved input attributes which also
have a vstride of 0 and need the same treatment.
Fixes the following on Haswell:
KHR-GL45.enhanced_layouts.varying_locations
KHR-GL45.enhanced_layouts.varying_array_locations
KHR-GL45.enhanced_layouts.varying_structure_locations
Reviewed-by: Matt Turner <mattst88@gmail.com>
Acked-by: Andres Gomez <agomez@igalia.com>
Wladimir J. van der Laan [Thu, 23 Nov 2017 09:08:34 +0000 (10:08 +0100)]
etnaviv: Emit vertex buffers consecutively
Vertex buffer legacy state is no longer picked up with new drawing
commands. Change to use different cases depending on the number of
vertex streams in the GPU specs.
This results in slightly more compact state emission as well, on all
vivantes.
Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Eric Engestrom [Thu, 9 Nov 2017 17:38:25 +0000 (17:38 +0000)]
REVIEWERS: add Alexander von Gluck IV as a reviewer for Haiku
There's been some Haiku-related activity lately, so let's document who
to cc on these patches.
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Acked-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Eric Engestrom [Wed, 22 Nov 2017 10:11:13 +0000 (10:11 +0000)]
genxml: fix assert guards
This removes a few hundred warnings on debug builds with asserts off.
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Eric Engestrom [Tue, 21 Nov 2017 15:07:50 +0000 (15:07 +0000)]
meson: add variable for mapi_abi.py instead of going back up the tree
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Eric Engestrom [Tue, 21 Nov 2017 15:07:11 +0000 (15:07 +0000)]
meson: reorder subdirs to avoid directly including more than one level
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Eric Engestrom [Tue, 21 Nov 2017 14:24:01 +0000 (14:24 +0000)]
meson: fix strtof locale support check
Fixes: d1992255bb29054fa5176 "meson: Add build Intel "anv" vulkan driver"
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Roland Scheidegger [Wed, 22 Nov 2017 02:11:33 +0000 (03:11 +0100)]
r600: set DX10_CLAMP for compute shader too
I really intended to set this for all shader stages by
3835009796166968750ff46cf209f6d4208cda86 but missed it for compute shaders
(because it's in a different source file...).
Reviewed-by: Dave Airlie <airlied@redhat.com>
Lionel Landwerlin [Fri, 17 Nov 2017 17:29:26 +0000 (17:29 +0000)]
anv: flag batch & instruction BOs for capture
When the kernel support flagging our BO, let's mark batch &
instruction BOs for capture so then can be included in the error
state.
v2: Only add EXEC_CAPTURE if supported (Kristian)
v3: Fix operator precedence issue (Lionel)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Lionel Landwerlin [Fri, 17 Nov 2017 17:26:59 +0000 (17:26 +0000)]
anv: setup BO flags at state_pool/block_pool creation
This will allow to set the flags on any anv_bo created/filled from a
state pool or block pool later.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Gert Wollny [Wed, 15 Nov 2017 09:29:11 +0000 (10:29 +0100)]
r600/shader: Fix all warnings issed with "-Wall -Wextra"
- fix a number of -Wsign-compare warnings
- fix two warnings for -Woverride-init because TGSI_OPCODE_CEIL == 83, and
the according field was defined two times.
[airlied: don't use -1 with unsigned type,
fix whitespace]
Signed-off-by: Gert Wollny <gw.fossdev@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Gert Wollny [Fri, 17 Nov 2017 11:13:40 +0000 (12:13 +0100)]
r600: Emit EOP for more CF instruction types
So far on pre-cayman chipsets the CF instructions CF_OP_LOOP_END,
CF_OP_CALL_FS, CF_OP_POP, and CF_OP_GDS an extra CF_NOP instruction
was added to add the EOP flag, even though this is not actually
needed, because all these instrutions support the EOP flag.
This patch removes the fixup code, adds setting the EOP flag for the
according instructions as well as others like CF_OP_TEX and CF_OP_VTX,
and adds writing out EOP for this type of instruction in the disassembler.
This also fixes a bug where shaders were created that didn't actually have
the EOP flag set in the last CF instruction, which might have resulted
in GPU lockups.
[airlied: cleaned up a little]
Signed-off-by: Gert Wollny <gw.fossdev@gmail.com>
Cc: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Dylan Baker [Tue, 21 Nov 2017 00:34:28 +0000 (16:34 -0800)]
meson: replace with_*dri with with_dri_platform
This fixes the windows and macos stubs to be consistent with the *nix
path.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Sat, 28 Oct 2017 00:20:52 +0000 (17:20 -0700)]
meson: add logic to select apple and windows dri
This is still not fully correct (haiku and BSD is notably probably not
correct), but Linux is not regressed and this should be correct for
macOS and Windows.
v2: - set the dri_platform to windows on Cygwin as well (Jon)
v3: - Add a better todo for Hurd (Eric)
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Tue, 21 Nov 2017 00:26:06 +0000 (16:26 -0800)]
meson: Fix LLVM requires for radeonsi
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Sat, 18 Nov 2017 00:37:50 +0000 (16:37 -0800)]
meson: convert llvm option to tristate
This option has been acting as a strange sort of half-tri state anyway.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Thu, 16 Nov 2017 01:31:32 +0000 (17:31 -0800)]
meson: Convert platform to auto
This is necessary to support operating systems other than the *nix
family (excluding macOS). For Linux nothing has changed, the defaults
are still the same.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Thu, 16 Nov 2017 01:30:52 +0000 (17:30 -0800)]
meson: Remove duplicate _GNU_SOURCE
There is one provided unconditionally, and one guarded by platform ==
linux. Remove the unconditional one.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Thu, 16 Nov 2017 01:09:33 +0000 (17:09 -0800)]
meson: Remove completed or irrelevant TODO comments
These are all either done already, or are autotools specific. The
misspelled gallium G3DVL is the autotools specific bit, meson is
handling that via build_by_default.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Thu, 16 Nov 2017 01:07:37 +0000 (17:07 -0800)]
meson: Fix TODO for missing dl_iterate_phdr function
This function is required for both the Intel "Anvil" vulkan driver and
the i965 GL driver. Error out if either of those is enabled but this
function isn't found.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Thu, 16 Nov 2017 00:53:40 +0000 (16:53 -0800)]
meson: disable x86 asm in fewer cases.
This patch allows building asm for x86 on x86_64 platforms, when the
operating system is the same. Previously cross compile always turned off
assembly. This allows using a cross file to cross compile x86 binaries
on x86_64 with asm.
This could probably be relaxed further thanks to meson's "exe_wrapper",
which is way to specify an emulator or compatibility layer (wine) that
can run the foreign binaries on the build system. Since the meson build
at this point only supports building on Linux I can't test this and I
don't want to write/enable code that cannot even be build tested.
v4: - set condition to build == x86_64 and host == x86 and
build.system == host.system
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Dylan Baker [Thu, 16 Nov 2017 00:09:22 +0000 (16:09 -0800)]
meson: Enable SSE4.1 optimizations
This patch checks for an and then enables sse4.1 optimizations if the
host machine will be x86/x86_64.
v2: - Don't compile code, it's unnecessary since we require a compiler
which always has SSE4.1 (Matt)
v3: - x64 -> x86_64 (Matt)
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Eric Anholt [Wed, 22 Nov 2017 00:32:33 +0000 (16:32 -0800)]
broadcom/vc5: Fix BASE_LEVEL handling with txl.
The HW doesn't add the base level anywhere (the min/max lod clamping is
what does base level), so we need to add it manually in this case.
Fixes piglit tex-miplevel-selection *Lod 2D.
Eric Anholt [Wed, 22 Nov 2017 00:05:49 +0000 (16:05 -0800)]
broadcom/vc5: Fix array texture layer count setup.
Fixes piglit array-texture.
Eric Anholt [Tue, 21 Nov 2017 23:27:20 +0000 (15:27 -0800)]
broadcom/vc5: Don't increment primitive queries while they're paused.
Fixes ext_transform_feedback-generatemipmap prims_generated
Eric Anholt [Tue, 21 Nov 2017 23:20:31 +0000 (15:20 -0800)]
broadcom/vc5: Fix incorrect padding of TF outputs.
After the first output, we were padding by an extra size of the previous
output. Fixes piglit ext_transform_feedback-output-type mat4x3[2] and
friends.
Eric Anholt [Tue, 21 Nov 2017 23:00:36 +0000 (15:00 -0800)]
broadcom/vc5: Fix UIF surface size setup for ARB_fbo's mismatched sizes.
The HW was computing an implicit height for the surface based on the image
size, but that may be smaller than the surface with ARB_fbo mismatched
sizes. In that case, we need to tell it about the pad, either with the
little 4-bit field in the RT config, or the extended field in
CLEAR_COLORS_PART3.
Fixes piglit arb_framebuffer_object-mixed-buffer-sizes.
Wladimir J. van der Laan [Sat, 18 Nov 2017 09:44:25 +0000 (10:44 +0100)]
etnaviv: Put HALTI level in specs
The HALTI level is an indication of the gross architecture of the GPU.
It determines for significant part what feature level the GPU has, what
state (especially frontend state) is there, and where it is located.
Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Wladimir J. van der Laan [Sat, 18 Nov 2017 09:44:24 +0000 (10:44 +0100)]
etnaviv: Const-correctness etnaviv_emit.h
The relocation structure is never changed by submitting it.
Signed-off-by: Wladimir J. van der Laan <laanwj@gmail.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Juan A. Suarez Romero [Tue, 21 Nov 2017 11:38:27 +0000 (12:38 +0100)]
meson: add si_driinfo.h in libgallium_dri
v2: generate target conditionally (Dylan)
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Iago Toral Quiroga [Thu, 16 Nov 2017 07:53:07 +0000 (08:53 +0100)]
nir/gather_info: recognize load_patch_vertices_in as a system value
This intrinsic is produced to load SYSTEM_VALUE_VERTICES_IN, which is
generated to load gl_PatchVerticesIn in the SPIR-V path for both
Vulkan and OpenGL.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Jordan Justen [Wed, 15 Nov 2017 00:27:34 +0000 (16:27 -0800)]
i965: Support decoding INTERFACE_DESCRIPTOR_DATA with INTEL_DEBUG=bat
This will dump the INTERFACE_DESCRIPTOR_DATA along with the associated
samplers & surfaces.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Scott D Phillips <scott.d.phillips@intel.com>
Kristian H. Kristensen [Wed, 30 Nov 2016 05:07:57 +0000 (21:07 -0800)]
intel/genxml: Add helpers for determining field type
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Matt Turner [Mon, 20 Nov 2017 22:21:43 +0000 (14:21 -0800)]
i965/fs: Check ADD/MAD with immediates in satprop unit test
The gen had to be changed from 4 to 6 so that we could test MAD, which
is new on Gen6.
mad_imm_float_neg_mov_sat tests the case fixed by the previous commit.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Matt Turner [Mon, 20 Nov 2017 22:24:57 +0000 (14:24 -0800)]
i965/fs: Handle negating immediates on MADs when propagating saturates
MADs don't take immediate sources, but we allow them in the IR since it
simplifies a lot of things. I neglected to consider that case.
Fixes: 4009a9ead490 ("i965/fs: Allow saturate propagation to propagate
negations into MADs.")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103616
Reported-and-Tested-by: Ruslan Kabatsayev <b7.10110111@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Juan A. Suarez Romero [Wed, 15 Nov 2017 16:49:21 +0000 (16:49 +0000)]
mesa/teximage: add TEXTURE_CUBE_MAP_ARRAY target for CompressedTexImage3D
From section 8.7, page 179 of OpenGL ES 3.2 spec:
An INVALID_OPERATION error is generated by CompressedTexImage3D
if internalformat is one of the the formats in table 8.17 and target
is not TEXTURE_2D_ARRAY, TEXTURE_CUBE_MAP_ARRAY or TEXTURE_3D.
An INVALID_OPERATION error is generated by CompressedTexImage3D if
internalformat is TEXTURE_CUBE_MAP_ARRAY and the “Cube Map Array”
column of table 8.17 is not checked, or if internalformat is
TEXTURE_3D and the “3D Tex.” column of table 8.17 is not checked.
So far it was only considering TEXTURE_2D_ARRAY as valid target. But as
"Cube Map Array" column is checked for all the cases, in practice we can
consider also TEXTURE_CUBE_MAP_ARRAY.
This fixes KHR-GLES32.core.texture_cube_map_array.etc2_texture
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>
Tapani Pälli [Mon, 20 Nov 2017 08:57:17 +0000 (10:57 +0200)]
intel: fix disasm_info memory leaks
Fixes: 4f82b1728719 ("i965: Rewrite disassembly annotation code")
Cc: Matt Turner <mattst88@gmail.com>
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Timothy Arceri [Thu, 16 Nov 2017 00:16:10 +0000 (11:16 +1100)]
st/glsl_to_nir: don't generate nir twice for gs
This was left out of
c980a3aa3133
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Roland Scheidegger [Sat, 18 Nov 2017 05:23:35 +0000 (06:23 +0100)]
llvmpipe: fix snorm blending
The blend math gets a bit funky due to inverse blend factors being
in range [0,2] rather than [-1,1], our normalized math can't really
cover this.
src_alpha_saturate blend factor has a similar problem too.
(Note that piglit fbo-blending-formats test is mostly useless for
anything but unorm formats, since not just all src/dst values are
between [0,1], but the tests are crafted in a way that the results
are between [0,1] too.)
v2: some formatting fixes, and fix a fairly obscure (to debug)
issue with alpha-only formats (not related to snorm at all), where
blend optimization would think it could simplify the blend equation
if the blend factors were complementary, however was using the
completely unrelated rgb blend factors instead of the alpha ones...
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Dave Airlie [Fri, 13 May 2016 04:35:33 +0000 (14:35 +1000)]
r600: add cull distance support
This passes all the tests in piglit.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Aravindan Muthukumar [Thu, 9 Nov 2017 05:45:28 +0000 (11:15 +0530)]
i965: Optimize bucket index calculation
Reducing Bucket index calculation to O(1).
This algorithm calculates the index using matrix method. Assuming
PAGE_SIZE is 4096, matrix arrangement is as below:
1*4096 2*4096 3*4096 4*4096
5*4096 6*4096 7*4096 8*4096
10*4096 12*4096 14*4096 16*4096
20*4096 24*4096 28*4096 32*4096
... ... ... ...
... ... ... ...
... ... ... max_cache_size
From this matrix its clearly seen that every row follows the below way:
... ... ... n
n+(1/4)n n+(1/2)n n+(3/4)n 2n
Row is calculated as log2(size/PAGE_SIZE) Column is calculated as
converting the difference between the elements to fit into power size of
two and indexing it.
Final Index is (row*4)+(col-1)
Tested with Intel Mesa CI.
Improves performance of 3DMark on BXT by 0.705966% +/- 0.229767% (n=20)
v4: Review comments on style and code comments implemented (Ian).
v3: Review comments implemented (Ian).
v2: Review comments implemented (Jason).
Signed-off-by: Aravindan Muthukumar <aravindan.muthukumar@intel.com>
Signed-off-by: Kedar Karanje <kedar.j.karanje@intel.com>
Reviewed-by: Yogesh Marathe <yogesh.marathe@intel.com>
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Dylan Baker [Wed, 15 Nov 2017 01:04:27 +0000 (17:04 -0800)]
meson: Guard the gallium dri componenet
Currently the target has a redundant guard, and the state tracker isn't
properly guarded.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Dylan Baker [Wed, 15 Nov 2017 01:03:39 +0000 (17:03 -0800)]
meson: don't build gallium subdir unless we're building gallium
This will allow us to simplify some guards within the gallium directory.
Signed-off-by: Dylan Baker <dylanx.c.baker@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Eric Anholt [Mon, 20 Nov 2017 18:14:38 +0000 (10:14 -0800)]
broadcom/vc5: Align 1D texture miplevels to 64b.
Fixes tex-miplevel-selection GL2:texture() 1D
Eric Anholt [Mon, 20 Nov 2017 18:07:24 +0000 (10:07 -0800)]
broadcom/vc5: Clamp min lod to the last level.
Otherwise, the simulator would complain in tex-miplevel-selection that the
min/max clamp was out of order. The actual HW seems to have clamped to
the max anyway.
Eric Anholt [Mon, 20 Nov 2017 20:26:49 +0000 (12:26 -0800)]
broadcom/vc5: Increase simulator memory for tex-miplevel-selection.
We were overflowing, because of all the little 4k allocations for CLs that
were getting expanded to 128kb in the simulator due to the GMP alignment.
Tim Rowley [Fri, 10 Nov 2017 22:45:38 +0000 (16:45 -0600)]
swr/rast: Repair simd8 frontend code rot
Keep non-default simd8 frontend code running for comparison purposes.
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
Tim Rowley [Thu, 9 Nov 2017 01:17:24 +0000 (19:17 -0600)]
swr/rast: Implement AVX-512 GATHERPS in SIMD16 fetch shader
Disabled for now.
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
Tim Rowley [Wed, 8 Nov 2017 20:07:33 +0000 (14:07 -0600)]
swr/rast: Simplify GATHER* jit builder api
General cleanup, and prep work for possibly moving to llvm masked
gather intrinsic.
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>