Danylo Piliaiev [Mon, 30 Jan 2023 16:08:29 +0000 (17:08 +0100)]
freedreno, turnip: Clarify some RB_CCU_CNTL fields
There is no .gmem field, there is a ccu color cache size field
which tells the size as a fraction of depth cache used in direct
rendering.
There is also GMEM_FAST_CLEAR_DISABLE flag which is set on a608/a610.
Since these values will stop being the same between models,
make them configurable.
Credits to Connor Abbott for deciphering color cache size meaning.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20991>
Lionel Landwerlin [Mon, 10 Jul 2023 07:20:33 +0000 (10:20 +0300)]
anv: hide exec_flags selection inside the i915 backend
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Hyunjun Ko <zzoon@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24073>
Sil Vilerino [Wed, 12 Jul 2023 14:54:16 +0000 (10:54 -0400)]
util: Blake3 - Identify arm64ec as aarch64 instead of x64
ARM64EC is a new build target for Windows ARM64 devices for x64 support.
Currently that build flavor fails due to attempting to use x64 intrinsics.
This commit fixes it by changing the auto-detection to be aarch64
instead of x64 for arm64ec.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24119>
Filip Gawin [Sun, 2 Jul 2023 11:40:30 +0000 (13:40 +0200)]
crocus: Avoid fast-clear with incompatible view
Port of code from iris.
Original author: Nanley Chery
Helps with fast_color_clear@fcc-write-after-clear
Cc: mesa-stable
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24135>
Danylo Piliaiev [Mon, 10 Jul 2023 14:28:41 +0000 (16:28 +0200)]
freedreno/cffdec: Decode CP_DRAW_AUTO
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24074>
Pavel Ondračka [Thu, 13 Jul 2023 08:57:11 +0000 (10:57 +0200)]
r300: update RV370 failures
This was missed in
0bf6dcb785ce82006f9757217153735e39127834
There is a loop which iterates over a temp array. NIR optimization
moves the real work out of the loop and what remains are just ALU ops
with undefs. So after converting undefs to zero, the ALU ops are
optimized out and DCE kills the loop. This is a good thing in
general and we don't fail the linking due to the loop presence.
However than we hit the shader constants and ALU limits later :-(
So from dEQP POW we go from NotSupported to Fail.
Fixes:
0bf6dcb785ce82006f9757217153735e39127834
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24134>
Frank Binns [Mon, 10 Jul 2023 10:20:33 +0000 (11:20 +0100)]
pvr: clang-format fixes
Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Karmjit Mahil <Karmjit.Mahil@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24070>
Jordan Justen [Fri, 28 Apr 2023 01:35:27 +0000 (21:35 -0400)]
isl: Set MOCS to uncached for MTL stream-out
Without this change various OpenGL CTS tranform feedback tests were
failing.
Cc: 23.2 <mesa-stable>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23823>
Jordan Justen [Fri, 23 Jun 2023 01:05:41 +0000 (18:05 -0700)]
isl/dev: Add uncached MOCS value
Rework:
* Jordan: Add uncached for all platforms (Requested by Francisco)
* Jordan: Use gen7 & gen8 values suggested by Francisco
* Jordan: Fix IVB and CHV MOCS mistakes pointed out by Francisco
Cc: 23.2 <mesa-stable>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23823>
Jordan Justen [Mon, 10 Jul 2023 18:28:38 +0000 (11:28 -0700)]
genxml/chv: Add MEMORY_OBJECT_CONTROL_STATE_CHV to document compared to BDW
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23823>
Jordan Justen [Mon, 10 Jul 2023 18:17:57 +0000 (11:17 -0700)]
genxml/hsw: Add additional MOCS field enumerations
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23823>
Jordan Justen [Fri, 28 Apr 2023 01:35:27 +0000 (21:35 -0400)]
anv,iris,hasvk: Use ISL_SURF_USAGE_STREAM_OUT_BIT for setting stream-out MOCS
Cc: 23.2 <mesa-stable>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23823>
Jordan Justen [Fri, 23 Jun 2023 00:47:08 +0000 (17:47 -0700)]
isl: Add ISL_SURF_USAGE_STREAM_OUT_BIT
Cc: 23.2 <mesa-stable>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23823>
Helen Koike [Mon, 10 Jul 2023 14:23:31 +0000 (11:23 -0300)]
docs/ci: Add docs for EXTRA_LOCAL_PACKAGES
Add a section about reusing the CI scripts for other projects.
Signed-off-by: Helen Koike <helen.koike@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23894>
Helen Koike [Wed, 28 Jun 2023 00:01:12 +0000 (21:01 -0300)]
ci: add EXTRA_LOCAL_PACKAGES to apt-get install
This can make it more convenient for other projects to reuse these
scripts.
Signed-off-by: Helen Koike <helen.koike@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23894>
Helen Koike [Tue, 27 Jun 2023 23:43:25 +0000 (20:43 -0300)]
ci: re-add EXTRA_LOCAL_PACKAGES to rootfs
This variable was removed on commit
848f59deda3ae7bb99409a3d15ddafe96b763ea1 when file `create-rootfs.sh`
was splitted.
Re-add it.
This can make it more convenient for other projects to reuse these
scripts.
Signed-off-by: Helen Koike <helen.koike@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23894>
David Heidelberg [Thu, 13 Jul 2023 00:23:16 +0000 (02:23 +0200)]
ci/freedreno: update a530 flakes
Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24132>
Dylan Baker [Wed, 12 Jul 2023 23:03:53 +0000 (16:03 -0700)]
docs: Update release calendar for 23.2.0-rc1
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24131>
Faith Ekstrand [Wed, 12 Jul 2023 08:06:35 +0000 (03:06 -0500)]
nv50/ir: Convert to new-style NIR registers
Shader-db results on Turing:
total inst in shared programs :
11121531 ->
11121458 (-0.00%)
total gpr in shared programs : 1848287 -> 1848425 (0.01%)
total ugpr in shared programs : 0 -> 0 (0.00%)
total local in shared programs : 27200 -> 27200 (0.00%)
total shared in shared programs : 236476 -> 236476 (0.00%)
total bytes in shared programs :
177944496 ->
177943328 (-0.00%)
total cached in shared programs : 0 -> 0 (0.00%)
inst gpr ugpr local shared bytes cached
helped 470 50 0 0 0 470 0
hurt 327 197 0 0 0 327 0
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24110>
Yiwei Zhang [Wed, 12 Jul 2023 20:16:28 +0000 (20:16 +0000)]
venus: refactor query feedback cmd record
Now copy and reset are similar enough to unify.
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24130>
Yiwei Zhang [Wed, 12 Jul 2023 19:05:20 +0000 (19:05 +0000)]
venus: add a missing barrier before copying query feedback
The 1st sync scope of vkCmdCopyQueryPoolResults is not sufficient to
cover transfer writes against query feedback buffer. We must ensure
ordering against prior query reset cmd where the feedback buffer fill
gets injected.
Fixes:
de4593faa193 ("venus: add query pool feedback cmds")
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24130>
Mohamed Ahmed [Tue, 11 Jul 2023 17:23:59 +0000 (20:23 +0300)]
vulkan/util: Use ycbcr_info for multiplane helpers in vk_format.c
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Acked-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24096>
Mohamed Ahmed [Tue, 11 Jul 2023 17:12:35 +0000 (20:12 +0300)]
vulkan/util: Support VK_EXT_ycbcr_2plane_444_formats color formats in vk_format.c
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Acked-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24096>
Mohamed Ahmed [Tue, 11 Jul 2023 16:13:32 +0000 (19:13 +0300)]
vulkan/util: Support 10-bit and 12-bit color formats in ycbcr_info in vk_format.c
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Acked-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24096>
Yiwei Zhang [Wed, 12 Jul 2023 17:09:22 +0000 (17:09 +0000)]
venus: ensure consistency of query overflow behavior
Fixes:
e6cffa1f0e4e ("venus: use feedback for vkGetQueryPoolResults")
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24123>
Yiwei Zhang [Wed, 12 Jul 2023 16:39:30 +0000 (16:39 +0000)]
venus: handle query feedback creation failure
Fixes:
e6cffa1f0e4e ("venus: use feedback for vkGetQueryPoolResults")
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24123>
Friedrich Vock [Tue, 11 Jul 2023 17:31:51 +0000 (19:31 +0200)]
radv/ci: Set DRIVER_NAME in LAVA raven vkcts jobs
Some CTS tests work with RADV, but take a very long time, making
deqp-runner trigger timeout failures. These tests are supposed to be
skipped, so they're contained in radv-skips.txt. But without setting
DRIVER_NAME to "radv", deqp-runner.sh won't pick up that file.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24095>
Eric Engestrom [Wed, 12 Jul 2023 15:36:27 +0000 (16:36 +0100)]
ci: avoid running hardware jobs if lint fails - now on Windows too!
I missed this in !23774.
Fixes:
a1c1cce9dfc2d8400a67 ("ci: avoid running hardware jobs if there are already trivial issues")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24121>
Eric Engestrom [Wed, 12 Jul 2023 15:36:27 +0000 (16:36 +0100)]
ci: avoid running hardware jobs if lint fails - now on LAVA too!
I missed this in !23774.
Fixes:
a1c1cce9dfc2d8400a67 ("ci: avoid running hardware jobs if there are already trivial issues")
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24121>
Alyssa Rosenzweig [Mon, 12 Jun 2023 20:59:11 +0000 (16:59 -0400)]
zink: Switch to register intrinsics
SPIR-V does not have anything like nir_register natively, so we were already
inserting loads/stores for register sources/destinations. That means it's easy
to switch to register intrinsics, getting explicit load_reg/store_reg intrinsics
in the NIR and translating those to the SPIR-V load/stores, dropping the
handling for nir_register. There's no need to use any of the chasing helpers for
coalescing the load/stores, like a hardware backend would. (In
fact, the underlying Vulkan driver will probably turn this back into SSA.)
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24111>
Dylan Baker [Wed, 12 Jul 2023 17:32:49 +0000 (10:32 -0700)]
VERSION: bump to 23.3.0-devel
For further development
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24124>
Karol Herbst [Sat, 8 Jul 2023 16:53:38 +0000 (18:53 +0200)]
api/icd: drop static lifetime from `get_ref` return type
This was never correct as the object pointed to can be destroyed at any
moment.
Signed-off-by: Karol Herbst <git@karolherbst.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24061>
Karol Herbst [Sat, 8 Jul 2023 16:41:32 +0000 (18:41 +0200)]
rusticl/device: make it &'static
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24061>
Karol Herbst [Sat, 8 Jul 2023 15:38:02 +0000 (17:38 +0200)]
rusticl: Replace &Arc<Device> with &Device
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24061>
Karol Herbst [Mon, 10 Jul 2023 13:09:20 +0000 (15:09 +0200)]
rusticl/kernel: silence newer clippy warning
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24061>
Eric Engestrom [Tue, 20 Jun 2023 13:37:22 +0000 (14:37 +0100)]
ci: avoid running hardware jobs if there are already trivial issues
Suggested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Eric Engestrom <eric@igalia.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Acked-by: Jesse Natalie <jenatali@microsoft.com>
Acked-by: Christian Gmeiner <cgmeiner@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23774>
Danylo Piliaiev [Fri, 30 Jun 2023 13:30:06 +0000 (15:30 +0200)]
freedreno/regs: Document a7xx CP_BV_BR_COUNT_OPS
Fully tested on HW. Credits to Connor Abbott for finding out how
it works.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Thu, 29 Jun 2023 17:30:08 +0000 (19:30 +0200)]
freedreno/regs: Properly document a7xx CP_EVENT_WRITE, CP_WAIT_TIMESTAMP
Event write is changes so much in a7xx that it makes sense to
create a new event CP_EVENT_WRITE7.
All credits to Connor Abbott for finding out what different flags
in these commands are doing.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Thu, 20 Apr 2023 15:07:30 +0000 (17:07 +0200)]
tu: Use reg usage tables for stale reg dbg option
Defining regs to stomp as ranges in a separate header is a mistake
from maintenance standpoint. Now we have this information at the
point where reg is defined.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Tue, 11 Jul 2023 12:33:10 +0000 (14:33 +0200)]
tu: Allow reg stomping of compute related registers
We don't use draw states for dispatches, so the bound pipeline
could be overwritten by reg stomping in a renderpass or blit.
The solution is to re-emit pipeline's IB on every dispatch if
reg stomping is used.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Thu, 20 Apr 2023 15:06:21 +0000 (17:06 +0200)]
freedreno/regs: Define usage for all a6xx/a7xx regs
Could be used for knowing which regs to stomp and to verify that
only appropriate regs are emitted.
Each register that is actually being used by driver should have "usage"
defined, currently there are following usages:
- "cmd" - the register is used outside of renderpass and blits,
roughly corresponds to registers used in ib1 for Freedreno
- "rp_blit" - the register is used inside renderpass or blits
(ib2 for Freedreno)
It is expected that register with "cmd" usage may be written into only at
the start of the command buffer (ib1), while "rp_blit" usage indicates that
register is either overwritten by renderpass/blit (ib2) or not used if not
overwritten by a particular renderpass/blit.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Thu, 20 Apr 2023 15:03:29 +0000 (17:03 +0200)]
freedreno/regs: Generate per-gen reg usage tables
"reg" and "array" now could have `usage="a,b,c"` attribute, for each
usage a separate array is generated.
Would be used for register stomping debug option.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Tue, 27 Jun 2023 13:52:54 +0000 (15:52 +0200)]
freedreno/regs: Fix a7xx SP_FS_PREFETCH definition
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Mon, 3 Apr 2023 16:53:29 +0000 (18:53 +0200)]
freedreno/regs: Add more a7xx regs and reg fields
Deduced from a740 cmdtraces from running CTS on prop driver.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Wed, 5 Jul 2023 14:40:52 +0000 (16:40 +0200)]
freedreno/regs: Add some new a7xx events
There are many more a7xx events but they are left for later.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Wed, 5 Jul 2023 14:08:54 +0000 (16:08 +0200)]
freedreno/regs: Add 2 new a7xx modes to CP_COND_REG_EXEC
Also reworked how CP_COND_REG_EXEC is defined to print
less irrelevant fields.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Wed, 5 Jul 2023 14:04:49 +0000 (16:04 +0200)]
freedreno/regs: a7xx has a new source type CP_REG_TEST
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Wed, 5 Jul 2023 14:01:26 +0000 (16:01 +0200)]
freedreno/regs: Add a7xx pseudo-regs to CP_SET_PSEUDO_REG
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Wed, 5 Jul 2023 13:57:41 +0000 (15:57 +0200)]
freedreno/regs: Clarify polling on a7xx for CP_WAIT_REG_MEM/CP_COND_WRITE5
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Wed, 5 Jul 2023 13:52:47 +0000 (15:52 +0200)]
freedreno/regs: Document a7xx CP_MODIFY_TIMESTAMP
Clears, adds to local, or adds to global timestamp
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Wed, 5 Jul 2023 13:49:19 +0000 (15:49 +0200)]
freedreno/regs: Document CP_MEM_TO_SCRATCH_MEM
Best guess is that it is a faster way to fetch all the VSC_STATE registers
and keep them in a local scratch memory instead of fetching every time
when skipping IBs.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Connor Abbott [Wed, 28 Jun 2023 16:25:12 +0000 (17:25 +0100)]
freedreno/regs: Document a7xx CP_FIXED_STRIDE_DRAW_TABLE
Executes an array of fixed-size command buffers where each
buffer is assumed to have one draw call, skipping buffers with
non-visible draw calls.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Mon, 3 Apr 2023 11:25:12 +0000 (13:25 +0200)]
freedreno/regs: More CP commands are the same on a7xx as on a6xx
These ones are seen to be used by blob in CTS, the rest a6xx commands
were not seen beeing used.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Tue, 20 Jun 2023 14:04:55 +0000 (16:04 +0200)]
freedreno/regs: Change a7xx regs to have open range for generation
Until proven otherwise regs stay the same between gens.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Fri, 21 Apr 2023 10:36:38 +0000 (12:36 +0200)]
freedreno/rnn: Take into account array's variant for regs
Otherwise even if array only exists in one generation the code for
its registers is generated for all gens.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Thu, 11 May 2023 11:32:28 +0000 (13:32 +0200)]
freedreno/rnn: Make addvariant work for fields in the same reg
Previously if addvariant was processed after other fields in the reg
these fields would never get matched. Fix this by moving bitfields that
add variant to the beginning of the list.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Thu, 11 May 2023 11:28:58 +0000 (13:28 +0200)]
freedreno/rnn: Fix addvariant being set effectively once
Each time addvariant was added it was added to the end of ctx->vars
list, without previous variant being removed. While the check for
variant tests only the first one that has expected enum name.
Fix this by updating `variant` instead of appending new one if variant
with such enum already exists.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Danylo Piliaiev [Thu, 20 Apr 2023 14:59:02 +0000 (16:59 +0200)]
freedreno/regs: Print xml validation error if validation fails
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23881>
Eric Engestrom [Wed, 12 Jul 2023 10:31:40 +0000 (11:31 +0100)]
meson: clarify what "off-screen rendering" means
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24114>
Eric Engestrom [Wed, 12 Jul 2023 10:28:00 +0000 (11:28 +0100)]
meson: clarify description of `opengl` option
There was some confusion from users as to whether disabling this option
disables OpenGL ES as well, so let's remove the confusing "all versions"
note and specify this affects "desktop OpenGL" only.
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24113>
Marcin Ślusarz [Tue, 11 Jul 2023 12:59:32 +0000 (14:59 +0200)]
anv: limit stack usage for anv_surface_state
Each one is 136 bytes.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24109>
Marcin Ślusarz [Tue, 11 Jul 2023 12:52:44 +0000 (14:52 +0200)]
anv: pass anv_surface_state using a pointer
It's 136 bytes, so passing it by stack is wasteful.
CID: 1531860
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24109>
Marcin Ślusarz [Tue, 11 Jul 2023 08:12:49 +0000 (10:12 +0200)]
anv: fix how NULL buffer_view is handled in anv_descriptor_set_write_buffer_view
CID: 1531855
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24109>
Timur Kristóf [Wed, 12 Jul 2023 09:19:41 +0000 (11:19 +0200)]
aco: Add MESA_SHADER_KERNEL to instruction selection setup.
Treat it the same way as MESA_SHADER_COMPUTE.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24112>
Eric Engestrom [Wed, 12 Jul 2023 07:48:25 +0000 (08:48 +0100)]
broadcom/ci: add the renderonly folder to things that can affect v3d & vc4
Also, move the v3d/vc4 lines together so it's clearer.
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24105>
Christian Gmeiner [Sat, 8 Jul 2023 15:37:35 +0000 (17:37 +0200)]
nir/lower_tex: optimize offset lowering for has_texture_scaling
Generates much better code and even helps to beat a blob driver.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24054>
Christian Gmeiner [Mon, 10 Jul 2023 20:24:46 +0000 (22:24 +0200)]
nir: rename has_txs to has_texture_scaling
Convert it to an opt-in for backends to prefer and use nir_load_texture_scale
instead of txs for nir lowerings.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Suggested-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24054>
Christian Gmeiner [Sat, 8 Jul 2023 15:43:31 +0000 (17:43 +0200)]
nir: rename intrinsic to have a more generic nameing
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24054>
Eric Engestrom [Wed, 12 Jul 2023 07:51:43 +0000 (08:51 +0100)]
etnaviv/ci: drop duplicate line in etnaviv files list
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24106>
Karol Herbst [Thu, 22 Jun 2023 14:43:40 +0000 (16:43 +0200)]
nvc0: backport fp helper invocation fix to 2nd gen Maxwell+
Ben prefers that we use the firmware method where possible.
Cc: mesa-stable
Signed-off-by: Karol Herbst <git@karolherbst.de>
Acked-by: Dave Airlie <airlied@redhat.com>
Acked-by: M Henning <drawoc@darkrefraction.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23802>
Hyunjun Ko [Wed, 12 Jul 2023 05:07:23 +0000 (14:07 +0900)]
anv: Adds a workaround for HEVC decoding on some old platforms.
HEVC support on Gfx9 is only available on VCS0. So limit the number of video queues
to the first VCS engine instance.
We should be able to query HEVC support from the kernel using the engine query uAPI,
but this appears to be broken : https://gitlab.freedesktop.org/drm/intel/-/issues/8832
When this bug is fixed we should be able to check HEVC support to determine the
correct number of queues.
Closes: mesa/mesa#9172, mesa/mesa#9314
Signed-off-by: Hyunjun Ko <zzoon@igalia.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24065>
Dave Airlie [Tue, 11 Jul 2023 04:41:01 +0000 (14:41 +1000)]
ci: update fails for fixed tests due to llvmpipe linear changes.
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24066>
Dave Airlie [Mon, 10 Jul 2023 07:34:50 +0000 (17:34 +1000)]
llvmpipe/linear: add support for rgba color buffers.
This adds support to the linear rast for rgba outputs.
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24066>
Dave Airlie [Mon, 10 Jul 2023 07:32:53 +0000 (17:32 +1000)]
llvmpipe/linear: add support for sampling when cbuf order is different.
This rewrites bgra sampling when the output is rgba,
and vice-versa.
It allows to skip swaps if the sampling and cbuf match.
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24066>
Dave Airlie [Mon, 10 Jul 2023 07:21:59 +0000 (17:21 +1000)]
llvmpipe/linear: add sample routines for swapping r/b channels
This lets rgba textures be sampled in linear mode.
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24066>
Dave Airlie [Mon, 10 Jul 2023 05:39:35 +0000 (15:39 +1000)]
llvmpipe/linear/tgsi: calculate num_texs properly for nir.
This is a bit hacky, but it does the right thing and counts the number
of textures instructions so the linear path can work for multiple textures.
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24066>
Dave Airlie [Tue, 11 Jul 2023 03:56:23 +0000 (13:56 +1000)]
llvmpipe/linear: refactor linear samplers into templated code.
Before adding new copies of all of these for swapping start by
refactoring into macro templated code.
I avoided using inline functions because I want to test with
opts turned down, and this will kill perf.
Reviewed-by: Brian Paul <brianp@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24066>
Faith Ekstrand [Wed, 12 Jul 2023 00:48:02 +0000 (19:48 -0500)]
freedreno/ci: Update pixmark piano checksums
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 29 May 2023 17:05:56 +0000 (13:05 -0400)]
pan/mdg: Ingest new-style registers
Switch to register intrinsics, using the helpers. Since our backend copyprop
chokes on non-SSA moves, we get better coalescing with this approach, hence the
small improvements to instruction count / cycle count in shader-db. Changes to
register pressure seem to be noise from iteration order. I'm not too worried.
total instructions in shared programs: 1508444 -> 1508193 (-0.02%)
instructions in affected programs: 42581 -> 42330 (-0.59%)
helped: 482
HURT: 41
Inconclusive result (value mean confidence interval includes 0).
total bundles in shared programs: 643023 -> 643136 (0.02%)
bundles in affected programs: 16318 -> 16431 (0.69%)
helped: 230
HURT: 85
Inconclusive result (value mean confidence interval includes 0).
total quadwords in shared programs: 1125992 -> 1125600 (-0.03%)
quadwords in affected programs: 125366 -> 124974 (-0.31%)
helped: 507
HURT: 351
Quadwords are helped.
total registers in shared programs: 90632 -> 90554 (-0.09%)
registers in affected programs: 669 -> 591 (-11.66%)
helped: 114
HURT: 31
Registers are helped.
total threads in shared programs: 55607 -> 55600 (-0.01%)
threads in affected programs: 20 -> 13 (-35.00%)
helped: 1
HURT: 7
Inconclusive result (value mean confidence interval includes 0).
total spills in shared programs: 1371 -> 1437 (4.81%)
spills in affected programs: 44 -> 110 (150.00%)
helped: 0
HURT: 2
total fills in shared programs: 5133 -> 5273 (2.73%)
fills in affected programs: 84 -> 224 (166.67%)
helped: 0
HURT: 2
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Fri, 9 Jun 2023 15:03:09 +0000 (11:03 -0400)]
gallivm: Switch to reg intrinsics
This is pretty straightforward, since we don't try to "coalesce" register access
the way a GPU backend would. In the old path, we generated register load/store
instructions internally when hitting register sources/destinations. In the new
path, we just translate the register load/store intrinsics to the LLVM
loads/stores and we're back where we started. It's a bit more code, but it's
more straightforward.
Notably, although this continues to use registers, this does NOT use the chasing
helpers.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Tue, 23 May 2023 18:46:18 +0000 (14:46 -0400)]
nir: Remove nir_register-based unit tests
Non-SSA functionality will become obsolete after nir_register is removed, so
there's no need to keep the tests around, and they will interfere with the
nir_register de-clawing.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Fri, 26 May 2023 16:30:51 +0000 (12:30 -0400)]
nir: Remove nir_lower_regs_to_ssa
It is now unused, as all internal producers of registers have been switched over
to intrinsics and no drivers call it.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Thu, 25 May 2023 13:05:56 +0000 (09:05 -0400)]
nir/lower_shader_calls: Convert to register intrinsics
Yet another internal use of nir_register that gets lowered back to SSA after the
pass. Easy enough to replace with intrinsic-based registers instead.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 10 Jul 2023 21:15:42 +0000 (16:15 -0500)]
mesa: Convert PTN to register intrinsics
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 10 Jul 2023 21:06:50 +0000 (16:06 -0500)]
mesa: Return SSA defs from PTN ALU helpers
Mostly a big simplifcation. Some noise on Haswell shader-db:
total instructions in shared programs: 2978203 -> 2978161 (<.01%)
instructions in affected programs: 9812 -> 9770 (-0.43%)
helped: 61
HURT: 39
helped stats (abs) min: 1 max: 5 x̄: 1.44 x̃: 1
helped stats (rel) min: 0.27% max: 7.69% x̄: 1.76% x̃: 1.18%
HURT stats (abs) min: 1 max: 4 x̄: 1.18 x̃: 1
HURT stats (rel) min: 0.55% max: 16.67% x̄: 4.49% x̃: 3.45%
95% mean confidence interval for instructions value: -0.71 -0.13
95% mean confidence interval for instructions %-change: -0.11% 1.46%
Inconclusive result (%-change mean confidence interval includes 0).
total cycles in shared programs:
45346214 ->
45346684 (<.01%)
cycles in affected programs: 519970 -> 520440 (0.09%)
helped: 157
HURT: 157
helped stats (abs) min: 2 max: 2970 x̄: 166.80 x̃: 6
helped stats (rel) min: 0.05% max: 40.38% x̄: 5.01% x̃: 1.42%
HURT stats (abs) min: 2 max: 1922 x̄: 169.80 x̃: 10
HURT stats (rel) min: 0.04% max: 44.00% x̄: 6.28% x̃: 2.46%
95% mean confidence interval for cycles value: -49.93 52.92
95% mean confidence interval for cycles %-change: -0.49% 1.76%
Inconclusive result (value mean confidence interval includes 0).
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 10 Jul 2023 21:05:50 +0000 (16:05 -0500)]
mesa: Simplify ptn_log() a bit
Using fdiv lets us drop the fneg. nir_opt_algebraic will re-optimize
this if the driver implements fdiv using fmul and frcp.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 10 Jul 2023 21:00:16 +0000 (16:00 -0500)]
gallium: Convert TTN to register intrinsics
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Tue, 23 May 2023 22:13:05 +0000 (18:13 -0400)]
gallium: Return SSA values from TTN ALU helpers
This is a lot simpler!
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Fri, 9 Jun 2023 13:41:21 +0000 (09:41 -0400)]
nir: Add lower_vec_to_regs pass
This is a variant of nir_lower_vec_to_movs that produces register intrinsics
(store_reg with write masks) instead of masked moves.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Fri, 9 Jun 2023 14:20:43 +0000 (10:20 -0400)]
nir: Add intrinsics version of locals_to_regs
This isn't so bad. I still duplicated the pass because it makes a lot easier to
have them coexist, switch users over one by one, and then garbage collect the
old when we're done.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Faith Ekstrand [Fri, 9 Jun 2023 13:36:22 +0000 (09:36 -0400)]
nir/from_ssa: Support register intrinsics
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Faith Ekstrand [Wed, 31 May 2023 18:55:47 +0000 (13:55 -0500)]
nir/from_ssa: Make additional assumptions in coalescing
At this point, everything is SSA. Also, NIR no longer allows different
numbers of components on the two sides of a phi so we can just assert
rather than trying to gracefully handle mismatches.
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 22 May 2023 17:04:05 +0000 (13:04 -0400)]
nir: Produce intrinsics in lower_{phis,ssa_defs}_to_regs
A number of passes lower SSA partially to registers, do work that would be
invalid in SSA, and then go back into SSA with nir_lower_regs_to_ssa. As a step
towards replacing nir_register with intrinsics,
the nir_lower_{phis,ssa_defs}_to_regs passes are changed to produce intrinsics
instead of nir_registers, and their callers are updated to call
nir_lower_reg_intrinsics_to_ssa instead of nir_lower_regs_to_ssa to compensate.
Jointly authored with Faith.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Faith Ekstrand [Wed, 31 May 2023 18:26:53 +0000 (13:26 -0500)]
nir: Add a reg_intrinsics flag to nir_convert_from_ssa
It doesn't do anything yet. We leave that to the subsequent patches so we can
keep the tree-wide refactor as simple as possible.
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 22 May 2023 18:44:52 +0000 (14:44 -0400)]
nir: Add new version of lower_regs_to_ssa
in the sense of operating on register intrinsics instead of nir_registers.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Tue, 16 May 2023 21:07:20 +0000 (17:07 -0400)]
nir: Add legacy data structures & helpers
These are registerful versions of core nir_src/nir_dest which will become
SSA-only soon enough, and modifierful versions of nir_alu_src/nir_alu_dest.
The latter will let us remove modifiers from nir_alu_instr finally.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Thu, 18 May 2023 15:00:50 +0000 (11:00 -0400)]
nir: Add pass for trivializing register access
After running the pass, all register access intrinsics are guaranteed to be
"trivial" in the sense that the program is free of hazards preventing
propagating them away without inserting any copies.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Mon, 22 May 2023 18:39:52 +0000 (14:39 -0400)]
nir: Add helpers for walking register uses
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Alyssa Rosenzweig [Tue, 16 May 2023 15:19:49 +0000 (11:19 -0400)]
nir: Add intrinsics for register access
Note the writemask handling is chosen for consistency with the rest of NIR. In
every other instance, writemask=w requires a vec4 source. This is hardcoded into
nir_validate and nir_print as what it means to have a writemask.
More importantly, consistency with how register writemasks currently work.
nir_print hides it, but r0.w = fneg ssa_1.x is actually a vec4 instruction with
source ssa_1.xxxx. As a silly example nir_dest_num_components(that) = 4 in the
old model. I realize this is quite strange coming from a scalar ISA, but it's
perfectly natural for the class of vec4 hardware for which this was designed. In
that hardware, conceptually all instructions are vec4`, so the sequence "fneg
ssa_1 and write to channel w" is implemented as "fneg a vec4 with ssa_1.x in the
last component and write that vec4 out but mask to write only the w channel".
Isn't this inefficient? It can be. To save power, Midgard has scalar ALUs in
addition to vec4 ALUs. Those details are confined to the backend VLIW scheduler;
the instruction selection is still done as vec4. This mechanism has little in
common with AMD's SALUs. Midgard has a wave size of 1, with special hacks for
derivatives.
As a result, all backends consuming register writemasks are expecting this
pattern of code. Changing the store to take a vec1 instead of a vec4 would
require changing every backend to reswizzle the sources to resurrect the vec4. I
started typing a branch to do this yesterday, but it made a mess of both Midgard
and nir-to-tgsi. Without any good reason to think it'd actually help
performance, I abandoned the idea. Getting all 15 backends converted to the
helpers is enough of a challenge without forcing 10 backends to reswizzle their
sources too.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23089>
Eric Engestrom [Tue, 11 Jul 2023 15:18:21 +0000 (16:18 +0100)]
ci: split farm rules out of test-source-dep.yml
That file has become a bit of the new `.gitlab-ci.yml` with just about
everything in there, but a lot of its content doesn't need to be in the
same file anymore now that `!reference` exists.
Signed-off-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/24090>