mesa.git
13 months agoaco: skip unnecessary compiler pass for the trap handler program
Samuel Pitoiset [Fri, 21 Aug 2020 09:33:22 +0000 (11:33 +0200)]
aco: skip unnecessary compiler pass for the trap handler program

The shader is written by hands with assigned registers, so most of
the pass are unnecessary.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6384>

13 months agoaco: add a helper for building a trap handler shader
Samuel Pitoiset [Tue, 18 Aug 2020 16:39:20 +0000 (18:39 +0200)]
aco: add a helper for building a trap handler shader

It's way easier to write a trap handler shader using ACO IR
instead of writing disassembly by hand + clrxasm + copy&paste.

This trap handler is quite simple for now, it just loads a
buffer descriptor from the TMA BO, it saves ttmp0-1 which
contain various info about the faulty instruction, and it
stores some hw registers about the wave/trap status.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6384>

13 months agoaco: validate that SMEM operands can use fixed registers
Samuel Pitoiset [Wed, 19 Aug 2020 07:41:42 +0000 (09:41 +0200)]
aco: validate that SMEM operands can use fixed registers

To fix a validation error when loading the scalar tma buffer
descriptor because it's not a temp but a fixed reg (tma_lo/tma_hi).

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6384>

13 months agoaco: add TBA/TMA/TTMP0-11 physical registers definitions
Samuel Pitoiset [Tue, 18 Aug 2020 07:26:48 +0000 (09:26 +0200)]
aco: add TBA/TMA/TTMP0-11 physical registers definitions

The TBA/TMA scalar registers are only available on GFX6-GFX8.

On GFX9+, TBA/TMA addr are stored in hardware registers and
the number of TTMP scalar registers is thus increased by 4.
Just keep in mind that tba_lo is actually ttmp0. Best would
be to support ttmp registers in RA but that's more complicated.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6384>

13 months agoamd/registers: add some SQ_WAVE_* register definitions
Samuel Pitoiset [Tue, 18 Aug 2020 12:54:51 +0000 (14:54 +0200)]
amd/registers: add some SQ_WAVE_* register definitions

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6384>

13 months agoamd/registers: add missing TBA registers on GFX6-GFX8
Samuel Pitoiset [Tue, 18 Aug 2020 06:34:33 +0000 (08:34 +0200)]
amd/registers: add missing TBA registers on GFX6-GFX8

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6384>

13 months agonir: add iabs-lowering code
Erik Faye-Lund [Fri, 10 Jan 2020 21:59:54 +0000 (22:59 +0100)]
nir: add iabs-lowering code

Microsoft's DXIL is based on LLVM, which doesn't have an integer ABS
opcode, but instead needs it lowered to NEG + MAX. We need to do this
with an option, to prevent an already existing optimization rule from
undoing this.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5211>

13 months agoegl: drop invalid shebang
Eric Engestrom [Mon, 3 Aug 2020 23:31:52 +0000 (01:31 +0200)]
egl: drop invalid shebang

This file exports a variable that is then used in a python script,
but it can never be executed by itself, so having a shebang here
makes no sense.

Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6166>

13 months agoclover: Call clang with -O0 for the SPIR-V path
Jason Ekstrand [Fri, 14 Aug 2020 20:48:59 +0000 (15:48 -0500)]
clover: Call clang with -O0 for the SPIR-V path

SPIRV-LLVM-Translator isn't really built for handling optimized LLVM IR.
It tends to fall over when, for instance, an optimization generates an
i96 type which isn't legal in SPIR-V.  Pass -O0 to avoid these cases.
See https://github.com/KhronosGroup/SPIRV-LLVM-Translator/issues/203.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6432>

13 months agoclover/spirv: Don't call llvm::regularizeLlvmForSpirv
Jason Ekstrand [Thu, 13 Aug 2020 20:12:15 +0000 (15:12 -0500)]
clover/spirv: Don't call llvm::regularizeLlvmForSpirv

writeSpirv() already takes care of that, and calling it twice seems to
duplicate functions and cause problems when processing execution modes.

Fixes: 2043c5f37cf "clover/llvm: Add functions for compiling from..."
Acked-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6306>

13 months agointel/nir: Allow splitting a single load into up to 32 loads
Jason Ekstrand [Fri, 21 Aug 2020 04:59:54 +0000 (23:59 -0500)]
intel/nir: Allow splitting a single load into up to 32 loads

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6405>

13 months agointel/fs: Fix an assert in load_scratch
Jason Ekstrand [Wed, 19 Aug 2020 23:53:32 +0000 (18:53 -0500)]
intel/fs: Fix an assert in load_scratch

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6405>

13 months agoiris: Always re-upload sysvals when we have kernel inputs
Jason Ekstrand [Fri, 21 Aug 2020 04:47:51 +0000 (23:47 -0500)]
iris: Always re-upload sysvals when we have kernel inputs

They can change on every dispatch and clover never gives us a heads up.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6405>

13 months agoiris: Normalize all compute shaders to MESA_SHADER_COMPUTE
Jason Ekstrand [Fri, 21 Aug 2020 16:31:18 +0000 (11:31 -0500)]
iris: Normalize all compute shaders to MESA_SHADER_COMPUTE

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6405>

13 months agoiris: ref/unref the GLSL type singleton in screen_create/destroy
Jason Ekstrand [Sat, 15 Aug 2020 05:26:05 +0000 (00:26 -0500)]
iris: ref/unref the GLSL type singleton in screen_create/destroy

Otherwise, we can run into trouble if the driver is dynamically loaded
via pipe-loader.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6405>

13 months agoiris: Stop advertising clover-only caps
Jason Ekstrand [Wed, 19 Aug 2020 23:56:01 +0000 (18:56 -0500)]
iris: Stop advertising clover-only caps

Both of these are clover-only caps.  We don't really support clover and,
even if we did, the number of address bits is wrong and we definitely
don't support the CL path for images.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6405>

13 months agoiris: Stop advertising PIPE_SHADER_IR_NIR_SERIALIZED
Jason Ekstrand [Thu, 20 Aug 2020 15:30:23 +0000 (10:30 -0500)]
iris: Stop advertising PIPE_SHADER_IR_NIR_SERIALIZED

This prevents clover from trying to initialize and crashing.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6405>

13 months agoclover/nir: add support for global invocation id offsets
Karol Herbst [Wed, 19 Aug 2020 14:02:55 +0000 (16:02 +0200)]
clover/nir: add support for global invocation id offsets

v2: create variables only once

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Jesse Natalie <jenatali@microsoft.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agonir: fix nir_variable_create for kernels
Karol Herbst [Fri, 21 Aug 2020 16:46:33 +0000 (18:46 +0200)]
nir: fix nir_variable_create for kernels

We ended up with INTERP_MODE_SMOOTH but we really just want to have NONE.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agospirv: Use new global invocation offset system value
Jesse Natalie [Tue, 28 Jul 2020 17:37:09 +0000 (10:37 -0700)]
spirv: Use new global invocation offset system value

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agonir: Add options to nir_lower_compute_system_values to control compute ID base lowering
Jesse Natalie [Fri, 21 Aug 2020 17:40:45 +0000 (10:40 -0700)]
nir: Add options to nir_lower_compute_system_values to control compute ID base lowering

If no options are provided, existing intrinsics are used.
If the lowering pass indicates there should be offsets used for global
invocation ID or work group ID, then those instructions are lowered to
include the offset.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agonir: Move compute system value lowering to a separate pass
Jesse Natalie [Fri, 21 Aug 2020 17:18:14 +0000 (10:18 -0700)]
nir: Move compute system value lowering to a separate pass

The actual variable -> intrinsic lowering stays where it is, but
ops which convert one intrinsic to be implemented in terms of
another have moved.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agocompiler/nir: make lowering global-id to local-id optional
Erik Faye-Lund [Mon, 13 Jan 2020 09:35:40 +0000 (10:35 +0100)]
compiler/nir: make lowering global-id to local-id optional

For D3D12, we don't want to lower this, as there's a dedicated global-id
system-value that might be faster to use, depending on the hardware.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agonir: Add new system values and intrinsics for dealing with CL work offsets
Jesse Natalie [Mon, 27 Jul 2020 23:56:21 +0000 (16:56 -0700)]
nir: Add new system values and intrinsics for dealing with CL work offsets

New intrinsics are added for global invocation IDs and work group IDs to
deal with offsets in both. The only one of these that needs a system value
is global invocation offset, for CL's get_global_offset().

Note that CL requires very large work group sizes, so these intrinsics
are modified to be able to use 64bit values, for 64bit SPIR-V.

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agonir: Populate some places where existing system values were missing
Jesse Natalie [Fri, 29 May 2020 18:09:37 +0000 (11:09 -0700)]
nir: Populate some places where existing system values were missing

Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>

13 months agoci/bare-metal: Try rebooting chezas again if they get stuck during tftp.
Eric Anholt [Wed, 19 Aug 2020 18:41:51 +0000 (11:41 -0700)]
ci/bare-metal: Try rebooting chezas again if they get stuck during tftp.

Occasionally something goes weird in the network and a group of chezas
will produce streams of these errors during the tftp process, eventually
timing out after 60 minutes in the job.  By the time we notice, the next
jobs seem to go through fine, so watch for them and try rebooting the
cheza to see if that gets our jobs to pass again.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6398>

13 months agoci/bare-metal: Retry booting chezas instead of failing when !POWER_GOOD
Eric Anholt [Wed, 19 Aug 2020 23:28:27 +0000 (16:28 -0700)]
ci/bare-metal: Retry booting chezas instead of failing when !POWER_GOOD

If we get this error, we can just try rebooting again and see if it comes
up then.  The POWER_GOOD failures are clustered in time, but it's better
to retry a few times in a row in one job (which has its own 60min timeout)
than to spuriously fail someone's pipeline.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6398>

13 months agoci/bare-metal: Convert the main cros-servo boot code to python
Eric Anholt [Wed, 19 Aug 2020 22:23:24 +0000 (15:23 -0700)]
ci/bare-metal: Convert the main cros-servo boot code to python

Switching this part to python makes the code clearer and cleans up our
logs as well.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6398>

13 months agoci/bare-metal: Use a new serial buffer tool.
Eric Anholt [Wed, 19 Aug 2020 20:00:25 +0000 (13:00 -0700)]
ci/bare-metal: Use a new serial buffer tool.

This one uses python threads to move some of our logic from shell
pipelines to python, and opens the door to doing better serial output
tracking in the future (the SerialBuffer.lines() method)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6398>

13 months agovirgl: update protocol headers
Chia-I Wu [Thu, 2 Jul 2020 18:40:43 +0000 (11:40 -0700)]
virgl: update protocol headers

Copied from virglrenderer.  Some in-development features are guarded by
VIRGL_RENDERER_UNSTABLE_APIS and they should not be used without knowing
the consequences.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6235>

13 months agovirgl: move protocol headers to a common place
Chia-I Wu [Thu, 2 Jul 2020 18:06:15 +0000 (11:06 -0700)]
virgl: move protocol headers to a common place

Tested with meson.  Android.mk changes are not tested.

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6235>

13 months agoutil/ralloc: fix ralloc alignment.
Lepton Wu [Fri, 14 Aug 2020 02:14:13 +0000 (19:14 -0700)]
util/ralloc: fix ralloc alignment.

On some malloc implementation, malloc doesn't always align to 16
bytes even on 64 bits system. To make sure ralloc_header always
starts at the wanted alignment, just force the size to be aligned at
the alignment of ralloc_header. This fixes crashed on instruction
like "movaps %xmm0,0x10(%rax)" which requires aligned memory access.

Signed-off-by: Lepton Wu <lepton@chromium.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6314>

13 months agosvga: fix draw elements with 8-bits indices
Charmaine Lee [Thu, 20 Aug 2020 22:45:47 +0000 (15:45 -0700)]
svga: fix draw elements with 8-bits indices

With commit 2122b902b80, u_index_translator can return U_TRANSLATE_MEMCPY
for 8-bits indices, and in this case we need to call the translation function
instead of a simple passthrough to the device.

Fixes piglit spec@nv_primitive_restart tests.

Fixes: 2122b902b8 "gallium/indices: don't expand prim-type for 8-bit indices"
Reviewed-by: Neha Bhende <bhenden@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6414>

13 months agowinsys/svga: fix display corruption after surface_init
Charmaine Lee [Thu, 20 Aug 2020 23:43:00 +0000 (16:43 -0700)]
winsys/svga: fix display corruption after surface_init

When we initialize the buffer surface, do not map the existing storage
with DONTBLOCK, leave it as a synchronized map.
This patch also sets the surface rebind flag after it is bound to a
new buffer and sets the surface buffer pointer accordingly.

This fixes display corruption issue seen with running steam.

Reviewed-by: Neha Bhende <bhenden@vmware.com>
Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6415>

13 months agost/mesa: increase size of gl_register_file bitfields
Charmaine Lee [Fri, 21 Aug 2020 00:11:22 +0000 (17:11 -0700)]
st/mesa: increase size of gl_register_file bitfields

Bump up the size of the bitfields for gl_register_file type for MSVC.
Also add ASSERT_BITFIELD_SIZE check where this bitfield is used.

Fixes spec@arb_shader_atomic_counter_ops tests in MSVC.

Reviewed-by: Neha Bhende <bhenden@vmware.com>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6417>

13 months agonir: rename nir_op_fne to nir_op_fneu
Karol Herbst [Tue, 18 Aug 2020 17:51:57 +0000 (19:51 +0200)]
nir: rename nir_op_fne to nir_op_fneu

It was always fneu but naming it fne causes confusion from time to time. So
lets rename it. Later we also want to add other unordered and fne, this is
a smaller preparation for that.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6377>

13 months agobifrost: use nir_intrinsic_has_type
Rhys Perry [Thu, 20 Aug 2020 13:01:58 +0000 (14:01 +0100)]
bifrost: use nir_intrinsic_has_type

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6402>

13 months agoaco: use nir_intrinsic_has_access
Rhys Perry [Thu, 20 Aug 2020 13:01:36 +0000 (14:01 +0100)]
aco: use nir_intrinsic_has_access

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6402>

13 months agonir: add and use nir_intrinsic_has_ helpers
Rhys Perry [Thu, 20 Aug 2020 13:01:23 +0000 (14:01 +0100)]
nir: add and use nir_intrinsic_has_ helpers

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6402>

13 months agoaco: fix non-rtz pack_half_2x16
Rhys Perry [Mon, 17 Aug 2020 10:28:07 +0000 (11:28 +0100)]
aco: fix non-rtz pack_half_2x16

We were using the wrong conversion opcode. The high bits are also not
zero'd on GFX10, which can cause v_cvt_pk_u16_u32 to clamp.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: df645fa369d ('aco: implement VK_KHR_shader_float_controls')
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6346>

13 months agoglsl: Eliminate out-of-bounds triop_vector_insert
Danylo Piliaiev [Tue, 18 Aug 2020 07:41:31 +0000 (10:41 +0300)]
glsl: Eliminate out-of-bounds triop_vector_insert

Section 5.11 (Out-of-Bounds Accesses) of the GLSL 4.60 spec says:

 "In the subsections described above for array, vector, matrix and
  structure accesses, any out-of-bounds access produced undefined
  behavior.... Out-of-bounds writes may be discarded or overwrite
  other variables of the active program."

Fixes crashes when dereferencing gl_ClipDistance and gl_TessLevel*, e.g:

  int index = -1;
  gl_ClipDistance[index] = -1;

When LowerCombinedClipCullDistance is true.

CC: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6363>

13 months agoglsl: Eliminate assigments to out-of-bounds elements of vector
Danylo Piliaiev [Mon, 17 Aug 2020 15:22:47 +0000 (18:22 +0300)]
glsl: Eliminate assigments to out-of-bounds elements of vector

Several optimization paths, including constant folding, can lead to
indexing vector with an out of bounds index.

Out-of-bounds writes could be eliminated per spec:

Section 5.11 (Out-of-Bounds Accesses) of the GLSL 4.60 spec says:

"In the subsections described above for array, vector, matrix and
 structure accesses, any out-of-bounds access produced undefined
 behavior.... Out-of-bounds writes may be discarded or overwrite
 other variables of the active program."

Fixes piglit tests:
spec@glsl-1.20@execution@vector-out-of-bounds-access@fs-vec4-out-of-bounds-1
spec@glsl-1.20@execution@vector-out-of-bounds-access@fs-vec4-out-of-bounds-6

CC: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6363>

13 months agoir_constant: Return zero on out-of-bounds vector accesses
Danylo Piliaiev [Mon, 17 Aug 2020 15:13:24 +0000 (18:13 +0300)]
ir_constant: Return zero on out-of-bounds vector accesses

Several optimization paths, including constant folding, can lead to
accessing an ir_constant vector with an out of bounds index.

Return 0 since GL_ARB_robustness and GL_KHR_robustness encourage
us to do so.

Fixes piglit tests:
spec@glsl-1.20@execution@vector-out-of-bounds-access@fs-vec4-out-of-bounds-2
spec@glsl-1.20@execution@vector-out-of-bounds-access@fs-vec4-out-of-bounds-4
spec@glsl-1.20@execution@vector-out-of-bounds-access@fs-vec4-out-of-bounds-5

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/2604
CC: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6363>

13 months agoanv: support fd==-1 in ImportSemaphoreFdKHR
Ricardo Quesada [Thu, 2 Jul 2020 13:46:11 +0000 (06:46 -0700)]
anv: support fd==-1 in ImportSemaphoreFdKHR

If fd==-1 is passed in ImportSemaphoreFdKHR, instead of importing the
fd, in creates an already signaled syncobj.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6427>

13 months agoaco: add ACO_DEBUG=force-waitcnt to emit wait-states
Samuel Pitoiset [Wed, 19 Aug 2020 08:40:35 +0000 (10:40 +0200)]
aco: add ACO_DEBUG=force-waitcnt to emit wait-states

Sounds useful for debugging missing wait-states and for improving
detection of the faulty instruction in case of memory violations.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6386>

13 months agowgl: Switch to Win10 version defines to enable usage of Win10 WGL callbacks
Jesse Natalie [Tue, 17 Mar 2020 17:53:33 +0000 (10:53 -0700)]
wgl: Switch to Win10 version defines to enable usage of Win10 WGL callbacks

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6228>

13 months agoradv: ignore BB labels when splitting the disassembly string
Samuel Pitoiset [Wed, 19 Aug 2020 11:42:57 +0000 (13:42 +0200)]
radv: ignore BB labels when splitting the disassembly string

Ignore everything that is not an instruction. This should fix
the annotated UMR disassembly.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6387>

14 months agoanv: take depth in to account in anv_GetImageSubresourceLayout
Tapani Pälli [Thu, 20 Aug 2020 09:49:46 +0000 (12:49 +0300)]
anv: take depth in to account in anv_GetImageSubresourceLayout

Fixes dEQP-VK.image.subresource_layout.3d* tests on gen9.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6400>

14 months agoiris: Add missing newline to debug log message
Caleb Callaway [Thu, 20 Aug 2020 00:12:10 +0000 (17:12 -0700)]
iris: Add missing newline to debug log message

Without this newline, log messages get lumped together on a single line,
making the debug log difficult to read.

Reviewed-by: Mark Janes <mark.a.janes@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6413>

14 months agoci/freedreno: List more common flakes reported recently.
Eric Anholt [Wed, 19 Aug 2020 17:51:50 +0000 (10:51 -0700)]
ci/freedreno: List more common flakes reported recently.

Now that adding to our flakes list won't disturb the test order
(potentially just moving the flake to a different result), we can be more
aggressive in listing new flakes.

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6392>

14 months agoci/freedreno: Move our skips lists over to being known-flakes lists.
Eric Anholt [Wed, 19 Aug 2020 17:43:54 +0000 (10:43 -0700)]
ci/freedreno: Move our skips lists over to being known-flakes lists.

This makes sure that we keep executing the tests so that we can get our
alerts in IRC and know whether the tests are still flaking.  It also keeps
us from having adjustments to the skip list causing failures/flakes to
move to different tests (as seen with a530 having to move some xfails
around after changing the skip list)

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6392>

14 months agoci/deqp-runner: Add a post-deqp-run filter list for known flakes.
Eric Anholt [Wed, 19 Aug 2020 17:30:12 +0000 (10:30 -0700)]
ci/deqp-runner: Add a post-deqp-run filter list for known flakes.

So far, we've been putting our known flakes that intermittently fail CI
into the skips list.  This has two downsides:

1) You don't know when the flakes stop happening and when to delist them
   from skips, unless you go do a bunch of manual runs with the skips list
   cleared.

2) If the flake was because the previous test left some broken state in
   the HW, you may just move your intermittent to a new test.

With this new path, you can list your flakes in the flakes file to keep
them from erroring out people's pipelines.  They still get run and
reported as is.

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6392>

14 months agoci/deqp-runner: Drop unused "count" variable
Eric Anholt [Wed, 19 Aug 2020 17:17:17 +0000 (10:17 -0700)]
ci/deqp-runner: Drop unused "count" variable

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6392>

14 months agoci/deqp-runner: Drop stale comment from deqp-runner.sh.
Eric Anholt [Wed, 19 Aug 2020 17:15:12 +0000 (10:15 -0700)]
ci/deqp-runner: Drop stale comment from deqp-runner.sh.

We don't rerun to detect flakes ourselves any more since the runner does
it internally.

Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6392>

14 months agonir: Use 'unsigned' instead of enum types in nir_variable::data
Jesse Natalie [Wed, 19 Aug 2020 21:34:11 +0000 (14:34 -0700)]
nir: Use 'unsigned' instead of enum types in nir_variable::data

MSVC treats enums as signed, so storing values that use the topmost
bit of the explicitly sized field loads as a negative value instead.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6393>

14 months agonir: Add an LOD parameter to image_*_size
Jason Ekstrand [Wed, 19 Aug 2020 23:21:33 +0000 (18:21 -0500)]
nir: Add an LOD parameter to image_*_size

The OpenCL image_width/height/depth functions have variants which can
take an LOD parameter.  More importantly, LLVM-SPIRV-Translator always
generates OpImageQuerySizeLod even if the LOD is guaranteed to be zero.
Given that over half the hardware out there has an LOD field for image
size queries (based on a rudimentary scan through their NIR -> whatever
code), we may as well just add the source to the NIR intrinsic.  If this
is ever a problem for anyone, the lowering is pretty trivial.

I've also added asserts to everyone's drivers that should alert them if
they ever see an LOD other than zero.  This will never happen with GL or
Vulkan so there's no need for panic.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6396>

14 months agobroadcom/compiler: Enable PER_QUAD for UBO and SSBO loads.
Arcady Goldmints-Orlov [Mon, 10 Aug 2020 03:17:01 +0000 (22:17 -0500)]
broadcom/compiler: Enable PER_QUAD for UBO and SSBO loads.

Helper invocations need to be able to read from UBOs since those values
can be used for flow control, but writes from helper invocations need to
be dropped.

Fixes CTS tests:
  dEQP-VK.glsl.derivate.*.uniform_loop.*

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6356>

14 months agobroadcom/compiler: Add a constant folding pass after nir_lower_io
Arcady Goldmints-Orlov [Sat, 25 Jul 2020 15:39:31 +0000 (10:39 -0500)]
broadcom/compiler: Add a constant folding pass after nir_lower_io

The nir_lower_io pass produces a bunch of constant arithmetic, and
assumes that constant folding will simplify it away.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6356>

14 months agobroadcom/compiler: support nir_intrinsic_load_sample_id
Arcady Goldmints-Orlov [Sat, 25 Jul 2020 20:20:02 +0000 (15:20 -0500)]
broadcom/compiler: support nir_intrinsic_load_sample_id

This adds support for the intrinsic as well as the vir_SAMPID
instruction that corresponds to it in vir.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6356>

14 months agoclover/spirv: pass list of supported extensions to the translator
Karol Herbst [Sat, 15 Aug 2020 19:49:05 +0000 (21:49 +0200)]
clover/spirv: pass list of supported extensions to the translator

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5038>

14 months agoclover/nir: Register callback for translation messages (v2)
Pierre Moreau [Sun, 10 May 2020 21:21:56 +0000 (23:21 +0200)]
clover/nir: Register callback for translation messages (v2)

This allows us to add SPIR-V to NIR translation failure messages to the
program’s compilation log, which can then be queried by the user.

v2: Replace the if-statement in `debug_function()` with an assert.

Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Serge Martin <edb@sigluy.net>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5038>

14 months agoclover/llvm: Use the highest supported SPIR-V version (v4)
Pierre Moreau [Mon, 18 May 2020 16:08:37 +0000 (18:08 +0200)]
clover/llvm: Use the highest supported SPIR-V version (v4)

v2:
   a) Move `supported_spirv_verssions()` to `spirv::` (Francisco Jerez)
   b) Introduce a `SPV_MAKE_VERSION` macro to avoid making mistakes and
     making it more readable than its uint representation.
v3: Replaced an if-statement with a `std::min()` in
    `spirv::get_spirv_translator_options()` (Karol Herbst)
v4: Turn `SPV_MAKE_VERSION()` into a function (Francisco Jerez)

Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5038>

14 months agomeson: Raise minimum version for SPIR-V OpenCL deps (v4)
Pierre Moreau [Tue, 5 May 2020 11:44:12 +0000 (13:44 +0200)]
meson: Raise minimum version for SPIR-V OpenCL deps (v4)

SPIRV-LLVM-Translator had API-breaking changes during version 0.2.1.
Since the new API was being used, a higher version has to be requested.

While at it and since SPIRV-LLVM-Translator added a new API to request a
specific SPIR-V version as well as allowed SPIR-V extensions, increase
the version number to the first including those new features.

v2: Require an LLVM version which is compatible with the
    SPIRV-LLVM-Translator that was found; that requirement was
    previously implicit and would be ensured when that library was
    built. Making it explicit will help in cases where multiple versions
    of LLVM might be installed.
v3: fix use of undefined _minimum_llvmspirvlib_version_array
v4 (Karol): fix creating the minimum requested version
            Simplifing the code

Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Serge Martin <edb@sigluy.net>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5038>

14 months agoclover/spirv: Print linked SPIR-V module if asked
Pierre Moreau [Tue, 5 May 2020 11:14:26 +0000 (13:14 +0200)]
clover/spirv: Print linked SPIR-V module if asked

Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Serge Martin <edb@sigluy.net>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5038>

14 months agoclover/spirv: Remove unused tuple header
Pierre Moreau [Tue, 5 May 2020 11:16:55 +0000 (13:16 +0200)]
clover/spirv: Remove unused tuple header

Fixes: 21473865050 ("clover/spirv: Add functions for parsing arguments, linking programs, etc.")
Signed-off-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Serge Martin <edb@sigluy.net>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5038>

14 months agoclover/spirv: rework handling of spirv extensions
Karol Herbst [Sat, 15 Aug 2020 11:19:52 +0000 (13:19 +0200)]
clover/spirv: rework handling of spirv extensions

What extensions we support depends on spirv_to_nir but it doesn't give us a
list. So hardcode one and add extensions we know we support and hit in the
wild.

v2: move into spirv lib

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Pierre Moreau <dev@pmoreau.org>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5038>

14 months agopanfrost: Fix alignment on Bifrost
Alyssa Rosenzweig [Thu, 20 Aug 2020 17:36:46 +0000 (13:36 -0400)]
panfrost: Fix alignment on Bifrost

This was real "fun" to debug...

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Fixes: 373a204bdd7 ("panfrost: Pass alignments explicitly")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6410>

14 months agotu: Enable VK_KHR_multiview
Connor Abbott [Thu, 2 Jul 2020 09:34:54 +0000 (11:34 +0200)]
tu: Enable VK_KHR_multiview

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agotu: Implement multiview pipeline state
Connor Abbott [Thu, 2 Jul 2020 09:33:42 +0000 (11:33 +0200)]
tu: Implement multiview pipeline state

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agotu: Add multiview lowering pass
Connor Abbott [Mon, 13 Jul 2020 17:27:53 +0000 (19:27 +0200)]
tu: Add multiview lowering pass

For now this only handles an a630 quirk where PC_MULTIVIEW_MASK doesn't
exist. However in the future it will also handle multi-position output.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agotu: Implement multiview query interactions
Connor Abbott [Thu, 2 Jul 2020 09:12:13 +0000 (11:12 +0200)]
tu: Implement multiview query interactions

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agotu: Improve timestamp queries
Connor Abbott [Thu, 2 Jul 2020 09:07:34 +0000 (11:07 +0200)]
tu: Improve timestamp queries

As the original comment says, we can't really give the user what they
want if there's a timestamp inside a GMEM renderpass, but we can give
them a better approximation of it. At least sysmem renderpasses will now
have an accurate timestamp.

Also, don't emit the WFI if it's not necessary, based on the stage
flags.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agotu: Implement multiview clear/resolve interactions
Connor Abbott [Thu, 2 Jul 2020 09:25:54 +0000 (11:25 +0200)]
tu: Implement multiview clear/resolve interactions

Loads, stores, clears, and resolves now happen per-view. Since we only
support multiview with sysmem rendering, we only implement this for
sysmem clears and resolves.

There aren't any tests that mix multiview and MSAA, so no coverage of
the resolve path.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agotu: Parse multiview render pass info
Connor Abbott [Thu, 2 Jul 2020 09:16:50 +0000 (11:16 +0200)]
tu: Parse multiview render pass info

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agotu: Translate VkRenderPassMultiviewCreateInfo to VkRenderPassCreateInfo2
Connor Abbott [Thu, 2 Jul 2020 08:46:27 +0000 (10:46 +0200)]
tu: Translate VkRenderPassMultiviewCreateInfo to VkRenderPassCreateInfo2

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agoir3: Add support for gl_ViewIndex in VS & FS
Connor Abbott [Wed, 1 Jul 2020 14:46:22 +0000 (16:46 +0200)]
ir3: Add support for gl_ViewIndex in VS & FS

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agofreedreno/a6xx: Add multiview registers
Connor Abbott [Tue, 30 Jun 2020 11:04:50 +0000 (13:04 +0200)]
freedreno/a6xx: Add multiview registers

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5720>

14 months agofreedreno/a6xx: disable LRZ when color channels are masked
Rob Clark [Thu, 20 Aug 2020 18:14:36 +0000 (11:14 -0700)]
freedreno/a6xx: disable LRZ when color channels are masked

From the PoV of early-z tests, having masked color channels is basically
like blend, ie. we do actually care about the fragments (or at least
parts of them) from previous draws.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6409>

14 months agofreedreno/cffdump: add arg to filter by process name
Rob Clark [Thu, 20 Aug 2020 18:13:32 +0000 (11:13 -0700)]
freedreno/cffdump: add arg to filter by process name

Usueful when you have a cmdstream trace which consists of multiple
different processes.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6409>

14 months agonir: Shrink store intrinsic num_components to the size used by the writemask.
Eric Anholt [Thu, 23 Jul 2020 21:32:13 +0000 (14:32 -0700)]
nir: Shrink store intrinsic num_components to the size used by the writemask.

This cuts a bunch of vector setup for undef components in the i965 vec4
backend.  Noticed while looking into codegen regressions in nir-to-tgsi.

brw results:
total instructions in shared programs: 3893221 -> 3881461 (-0.30%)
total cycles in shared programs: 113792154 -> 113810288 (0.02%)

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6054>

14 months agonir: Expand opt_undef to handle undef channels in a store intrinsic.
Eric Anholt [Thu, 23 Jul 2020 21:24:15 +0000 (14:24 -0700)]
nir: Expand opt_undef to handle undef channels in a store intrinsic.

Instead of only handling all-undef stores, shrink the writemask to just
the defined channels.  Cleans up a bunch of writemasks on prog_to_nir
output in particular.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6054>

14 months agonir/opt_undef: Handle a couple more normal store intrinsics.
Eric Anholt [Thu, 23 Jul 2020 20:39:56 +0000 (13:39 -0700)]
nir/opt_undef: Handle a couple more normal store intrinsics.

They've got the value in the same slot as the others, and the same undef
behavior should be fine.

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6054>

14 months agoradv: Move nir_opt_shrink_vectors() into the opt loop.
Eric Anholt [Wed, 19 Aug 2020 16:16:15 +0000 (09:16 -0700)]
radv: Move nir_opt_shrink_vectors() into the opt loop.

Upcoming changes to opt_undef will result in this pass doing more work and
generating vector MOVs that need re-scalarizing (which is inside of the
main opt loop).

Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6054>

14 months agopanfrost: Reduce attribute buffer allocations
Alyssa Rosenzweig [Mon, 17 Aug 2020 18:49:52 +0000 (14:49 -0400)]
panfrost: Reduce attribute buffer allocations

With vertex ID / instance ID in use, vs->attribute_count will be
expanded appropriately. Without them in use, everything is 1:1 (or 1:2
in the worst case of NPOT everything).

Down to 3 slabs + 64184 bytes.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Don't reserve for NPOT w/o instancing
Alyssa Rosenzweig [Mon, 17 Aug 2020 18:46:56 +0000 (14:46 -0400)]
panfrost: Don't reserve for NPOT w/o instancing

It's impossible. Down to 5 slabs + 45368.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Don't overallocate attributes
Alyssa Rosenzweig [Mon, 17 Aug 2020 18:42:41 +0000 (14:42 -0400)]
panfrost: Don't overallocate attributes

We only need vertex ID / instance ID if the shader actually requests it.
Cuts -bideas's beginning from (7 slab + 63352 bytes) to (7 slabs +
32568).

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Fix attribute buffer underallocation
Alyssa Rosenzweig [Mon, 17 Aug 2020 18:37:44 +0000 (14:37 -0400)]
panfrost: Fix attribute buffer underallocation

Breaks on the pathological where everything is NPOT.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Fixes: e646c861fc6 ("panfrost: Use packs for vertex attribute buffers")
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Pass alignments explicitly
Alyssa Rosenzweig [Mon, 17 Aug 2020 18:27:57 +0000 (14:27 -0400)]
panfrost: Pass alignments explicitly

In most cases, GPU data structures need only be self-aligned; the
worst-case 128 byte alignment is wasteful. By passing explicit
alignments, we can reduce memory usage, avoid extra allocations, and
improve descriptor cache locality.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Free batch->dependencies
Alyssa Rosenzweig [Tue, 18 Aug 2020 12:41:37 +0000 (08:41 -0400)]
panfrost: Free batch->dependencies

On glmark2-es2 -bterrain:

594.05KB leaked over 9282 calls from:
    panfrost_batch_update_bo_access
      at ../src/gallium/drivers/panfrost/pan_job.c:462
      in /home/alyssa/rockchip_dri.so
    panfrost_batch_add_bo
      at ../src/gallium/drivers/panfrost/pan_job.c:560
    panfrost_batch_add_bo
      at ../src/gallium/drivers/panfrost/pan_job.c:519
      in /home/alyssa/rockchip_dri.so
    panfrost_batch_add_resource_bos
      at ../src/gallium/drivers/panfrost/pan_job.c:569
    panfrost_batch_add_fbo_bos
      at ../src/gallium/drivers/panfrost/pan_job.c:588
      in /home/alyssa/rockchip_dri.so
    panfrost_create_batch
      at ../src/gallium/drivers/panfrost/pan_job.c:126

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Cc: mesa-stable
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Use memctx for sysvals
Alyssa Rosenzweig [Tue, 18 Aug 2020 12:31:42 +0000 (08:31 -0400)]
panfrost: Use memctx for sysvals

8.74KB leaked over 52 calls from:
    0xffffbb5b9fc3
      in ??
    _mesa_hash_table_init
      at ../src/util/hash_table.c:163
      in /home/alyssa/rockchip_dri.so
    _mesa_hash_table_create
      at ../src/util/hash_table.c:186
    _mesa_hash_table_u64_create
      at ../src/util/hash_table.c:701
      in /home/alyssa/rockchip_dri.so
    panfrost_nir_assign_sysvals
      at ../src/panfrost/util/pan_sysval.c:130
      in /home/alyssa/rockchip_dri.so
    midgard_compile_shader_nir
      at ../src/panfrost/midgard/midgard_compile.c:2905
      in /home/alyssa/rockchip_dri.so

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Cc: mesa-stable
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopan/mdg: Free previous liveness
Alyssa Rosenzweig [Tue, 18 Aug 2020 12:26:40 +0000 (08:26 -0400)]
pan/mdg: Free previous liveness

Before we drop the reference.

160 calls with 0B peak consumption from:
    0xffffbd9d2fc3
      in ??
    pan_compute_liveness
      at ../src/panfrost/util/pan_liveness.c:127
      in /home/alyssa/rockchip_dri.so
    mir_compute_liveness
      at ../src/panfrost/midgard/midgard_liveness.c:55
      in /home/alyssa/rockchip_dri.so
    midgard_opt_dead_code_eliminate
      at ../src/panfrost/midgard/midgard_opt_dce.c:118
      in /home/alyssa/rockchip_dri.so

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Cc: mesa-stable
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Free hash_to_temp map
Alyssa Rosenzweig [Tue, 18 Aug 2020 12:23:13 +0000 (08:23 -0400)]
panfrost: Free hash_to_temp map

No need to put it on the context, we can keep it local in mir_squeeze
and drop when we're done.

15.77KB leaked over 85 calls from:
    0xffffaed3bfc3
      in ??
    _mesa_hash_table_rehash
      at ../src/util/hash_table.c:368
      in /home/alyssa/rockchip_dri.so
    hash_table_insert
      at ../src/util/hash_table.c:403
      in /home/alyssa/rockchip_dri.so
    find_or_allocate_temp
      at ../src/panfrost/midgard/mir_squeeze.c:48
      in /home/alyssa/rockchip_dri.so
    find_or_allocate_temp
      at ../src/panfrost/midgard/mir_squeeze.c:35
      in /home/alyssa/rockchip_dri.so
    mir_squeeze_index
      at ../src/panfrost/midgard/mir_squeeze.c:76

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Cc: mesa-stable
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Free NIR of blit shaders
Alyssa Rosenzweig [Tue, 18 Aug 2020 12:10:25 +0000 (08:10 -0400)]
panfrost: Free NIR of blit shaders

After we compile from NIR to a native binary, we can throw away the NIR.

17.47KB leaked over 104 calls from:
    0xffff87dcafc3
      in ??
    _mesa_hash_table_init
      at ../src/util/hash_table.c:163
      in /home/alyssa/rockchip_dri.so
    _mesa_hash_table_create
      at ../src/util/hash_table.c:186
    nir_lower_vars_to_ssa_impl
      at ../src/compiler/nir/nir_lower_vars_to_ssa.c:717
      in /home/alyssa/rockchip_dri.so
    nir_lower_vars_to_ssa
      at ../src/compiler/nir/nir_lower_vars_to_ssa.c:817
    optimise_nir
      at ../src/panfrost/midgard/midgard_compile.c:504
      in /home/alyssa/rockchip_dri.so
    midgard_compile_shader_nir
      at ../src/panfrost/midgard/midgard_compile.c:2895
      in /home/alyssa/rockchip_dri.so
    panfrost_build_blit_shader
      at ../src/panfrost/lib/pan_blit.c:103
      in /home/alyssa/rockchip_dri.so

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Cc: mesa-stable
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Free cloned NIR shader
Alyssa Rosenzweig [Tue, 18 Aug 2020 12:07:15 +0000 (08:07 -0400)]
panfrost: Free cloned NIR shader

Fixes heaptrack leak:

19.37KB leaked over 63 calls from:
    0xffff92bbefc3
      in ??
    nir_alu_instr_create
      at ../src/compiler/nir/nir.c:442
      in /home/alyssa/rockchip_dri.so
    clone_alu
      at ../src/compiler/nir/nir_clone.c:277
      in /home/alyssa/rockchip_dri.so
    clone_instr
      at ../src/compiler/nir/nir_clone.c:495
      in /home/alyssa/rockchip_dri.so
    clone_block
      at ../src/compiler/nir/nir_clone.c:544
    clone_cf_list
      at ../src/compiler/nir/nir_clone.c:594
    clone_function_impl
      at ../src/compiler/nir/nir_clone.c:672
      in /home/alyssa/rockchip_dri.so
    nir_shader_clone
      at ../src/compiler/nir/nir_clone.c:744
      in /home/alyssa/rockchip_dri.so
    panfrost_shader_compile
      at ../src/gallium/drivers/panfrost/pan_assemble.c:154
      in /home/alyssa/rockchip_dri.so

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Cc: mesa-stable
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Fix blend leak for render targets 5-8
Alyssa Rosenzweig [Mon, 17 Aug 2020 23:57:54 +0000 (19:57 -0400)]
panfrost: Fix blend leak for render targets 5-8

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Fixes: 375d4c2c74d ("panfrost: Extend blending to MRT")
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Keep finalized blend state constant
Alyssa Rosenzweig [Mon, 17 Aug 2020 23:52:20 +0000 (19:52 -0400)]
panfrost: Keep finalized blend state constant

It's probably fine, but the writes to an uninitialized struct make me
nervous. Let's do the obvious thing instead.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Drop depth-only case in blend finalize
Alyssa Rosenzweig [Mon, 17 Aug 2020 23:43:24 +0000 (19:43 -0400)]
panfrost: Drop depth-only case in blend finalize

Since last commit, we don't call this function at all if there's no
matching colour buffer (it wouldn't make sense to!)

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Explicitly handle nr_cbufs=0 case
Alyssa Rosenzweig [Mon, 17 Aug 2020 23:41:48 +0000 (19:41 -0400)]
panfrost: Explicitly handle nr_cbufs=0 case

Avoids some indirection around colour buffer count.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Drop implicit blend pooling
Alyssa Rosenzweig [Mon, 17 Aug 2020 23:32:59 +0000 (19:32 -0400)]
panfrost: Drop implicit blend pooling

We dropped blend shader pools a long time ago for various reasons, but
the indirection required remained. Oops.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>

14 months agopanfrost: Share tiler_heap across batches/contexts
Alyssa Rosenzweig [Mon, 17 Aug 2020 17:14:54 +0000 (13:14 -0400)]
panfrost: Share tiler_heap across batches/contexts

There's only one tiler, so this is safe. (The blob does the same
optimization.) This avoids allocating multiple heaps for multiple
batches, which wastes memory and CPU time.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6373>