ed04848770
r600/sfn: update readports before trying to schedule group instrutions
...
We only do minimal checks to ensure that copy propagation doesn't break
the readport setup, but we don't update the groups readport setup. So
before scheduling the group do this update. Also check the readport
constellation when scheduling a group is finished.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36572 >
2025-08-11 22:10:08 +00:00
070f56237f
r600/sfn: Fix update readports method
...
- If a readport reservation is not successful then we have to reset the
readport reservation.
- Since the scheduler can add instructions in any order, we have to
update the readports in the same order the slots were filled when
re-evaluating.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36572 >
2025-08-11 22:10:08 +00:00
316d3a3f95
r600/sfn: reuse readport for already loaded registers
...
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36572 >
2025-08-11 22:10:07 +00:00
68d5f1c456
r600/sfn: unify and fix naming of group readport reserver
...
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36572 >
2025-08-11 22:10:07 +00:00
9ce6d9ff1f
r600/sfn: Simplify test code when scheduling a vec instr into trans
...
Now that we track the free slots right away, we can make use of this
information when testing whether a vec instruction can be scheduled into
a trans slot.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36584 >
2025-08-11 21:54:38 +00:00
a6d97070a7
r600/sfn: rename free_slots and improve updating it
...
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36584 >
2025-08-11 21:54:38 +00:00
67d0f51029
r600/sfn: check number of fsat64 source uses properly
...
Fixes: 255eee10acgert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36585 >
2025-08-11 21:40:10 +00:00
206ec1ff77
r600/sfn: lower u2f64 and i2f64 in nir
...
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36587 >
2025-08-11 21:26:20 +00:00
f7552429ef
r600/sfn: remove first call to r600_split_64bit_alu_and_phi
...
The pass is run later a second time, but there is no need to do it twice.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36587 >
2025-08-11 21:26:20 +00:00
6fea840c2c
r600/sfn: Allow f2f64 to use vec2
...
Lowering u2f64 and i2f64 will create such instructions and with that
the ALU groups are filled without the need to do scheduler trickery with
two-slot ops that have also two dest registers.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36587 >
2025-08-11 21:26:20 +00:00
fdaf105178
r600/sfn: lower b2f64 in nir
...
Signed-off-by: Gert Wollny <gert.wollny@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36587 >
2025-08-11 21:26:20 +00:00
f12ad5da70
microsoft/clc: Improve clc_compiler_test.cpp to use defined expect value
...
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36723 >
2025-08-11 21:11:18 +00:00
c2f6efbc8a
microsoft/clc: {} for struct initialize to avoid warning
...
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36723 >
2025-08-11 21:11:18 +00:00
6ece4f3fa0
vulkan: Add a vk_video_session_finish() helper
...
It's always better if init/finish come in pairs.
Reviewed-by: Dave Airlie <airlied@redhat.com >
Reviewed-by: Hyunjun Ko <zzoon@igalia.com >
Reviewed-by: David Rosca <david.rosca@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36646 >
2025-08-11 20:34:56 +00:00
22a1f34728
radv: Delete radv_video_session_params
...
It was just a dummy wrapper around the runtime struct. We do, however,
have to keep at least the Create/Update entrypoints because RADV has to
do some patching for video encode. Since we're keeping Create, we keep
Destroy as well.
Reviewed-by: Dave Airlie <airlied@redhat.com >
Reviewed-by: David Rosca <david.rosca@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36646 >
2025-08-11 20:34:56 +00:00
e0b25e78ff
anv: Delete anv_video_session_params
...
It's just a dummy wrapper around the common struct.
Reviewed-by: Dave Airlie <airlied@redhat.com >
Reviewed-by: Hyunjun Ko <zzoon@igalia.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36646 >
2025-08-11 20:34:56 +00:00
8f03f9b443
vulkan: Add common VideoSessionParametersKHR entrypoints
...
Reviewed-by: Dave Airlie <airlied@redhat.com >
Reviewed-by: Hyunjun Ko <zzoon@igalia.com >
Reviewed-by: David Rosca <david.rosca@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36646 >
2025-08-11 20:34:56 +00:00
9d6f65db9a
vulkan: Add handle casts for vk_video_session[_parameters]
...
This also allows us to simplify the interface to
vk_video_session_parameters_create().
Reviewed-by: Dave Airlie <airlied@redhat.com >
Reviewed-by: Hyunjun Ko <zzoon@igalia.com >
Reviewed-by: David Rosca <david.rosca@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36646 >
2025-08-11 20:34:56 +00:00
f8086c6472
vulkan/video: Switch vk_video_session_parameters to create/destroy
...
These are never created on the stack or deep inside other objects so it
makes sense to use create/destroy instead of init/finish.
Reviewed-by: Dave Airlie <airlied@redhat.com >
Reviewed-by: Hyunjun Ko <zzoon@igalia.com >
Reviewed-by: David Rosca <david.rosca@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36646 >
2025-08-11 20:34:56 +00:00
d93156c2a2
ac: merge AC_ARG_INT & AC_ARG_FLOAT into single AC_ARG_VALUE
...
nothing uses the type anymore
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:30 -04:00
bb8d2e55d1
ac/llvm: make AC_ARG_FLOAT equal to AC_ARG_INT
...
Nothing cares about the type anymore.
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:29 -04:00
9cede6e8b0
radeonsi: bitcast shader args to float in LLVM IR manually
...
This will allow us to merge AC_ARG_FLOAT/INT because nothing else uses
the type.
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:27 -04:00
324a7f7e24
ac/llvm: make ac_get_arg non-inline
...
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:25 -04:00
54fe9aa664
ac: simplify AC_ARG_CONST_*PTR enums
...
The pointer type doesn't matter anymore.
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:23 -04:00
d5d5726908
ac/nir: remove unused ac_get_ptr_arg & ac_arg_type_to_pointee_type
...
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:19 -04:00
dbefe0c26d
ac/nir: inline ac_get_ptr_arg
...
so that we can get rid of ac_get_ptr_args.
RADV uses AC_ARG_CONST_PTR for num_work_groups, which maps to i8, which
seems wrong.
No functional change.
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:17 -04:00
4edcd8a87f
ac/llvm: inline ac_array_in_const*_addr_space
...
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36696 >
2025-08-11 15:32:15 -04:00
a1278095d3
ir3: align alias sequences to work around hardware bug
...
There seems to be a hardware bug that sometimes causes a GPU hang when
an alias...sam sequence crosses an instruction cache line boundary. This
commit adds a workaround pass that inserts padding nops to ensure no
such sequence cross a cache line. Until an alternative solution is
found, this is the best we can do.
While the number of nops we have to insert is fixed at this point, we
can try to minimize the number of nops executed at runtime by replacing
nops encoded in instructions by standalone nops. That is, if this pass
has to insert one nop, it will try to make one of the following
replacements:
- (rptN)nop -> (rptN-1)nop; nop
- (nopN)foo -> (nopN-1)foo; nop
It does so by keeping track of "insert points". Each insert point keeps
track of the instruction and the maximum number of nops that can be
inserted there without pushing any subsequent alias sequences over the
next cache line. Whenever we need to insert nops, we first try it at the
encountered insert points and only if that doesn't work, we insert them
right before the first alias. The pass makes sure the insert points are
only visited a bounded number of times in total to keep the whole pass
O(n).
Totals:
Instrs: 48207402 -> 48278230 (+0.15%)
CodeSize: 101907026 -> 102294524 (+0.38%)
NOPs: 8386320 -> 8457148 (+0.84%)
(ss)-stall: 4013046 -> 4012931 (-0.00%)
(sy)-stall: 16741190 -> 16741033 (-0.00%)
Preamble Instrs: 11506988 -> 11520671 (+0.12%)
Last helper: 11686328 -> 11701615 (+0.13%)
Cat0: 9241457 -> 9312285 (+0.77%)
Totals from 25237 (15.32% of 164705) affected shaders:
Instrs: 22172360 -> 22243188 (+0.32%)
CodeSize: 44372164 -> 44759662 (+0.87%)
NOPs: 4201698 -> 4272526 (+1.69%)
(ss)-stall: 1982473 -> 1982358 (-0.01%)
(sy)-stall: 7379552 -> 7379395 (-0.00%)
Preamble Instrs: 4552074 -> 4565757 (+0.30%)
Last helper: 6260280 -> 6275567 (+0.24%)
Cat0: 4616677 -> 4687505 (+1.53%)
Signed-off-by: Job Noorman <jnoorman@igalia.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36639 >
2025-08-11 17:07:10 +00:00
aef01d73d8
ci: uprev vkd3d
...
Fixes all the workgraph failures, but also introduces two new failures
for lavapipe.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36712 >
2025-08-11 15:57:35 +00:00
657080fe59
nvk: drop some needless definitions and deps
...
We already have these two:
- dep_android comes from vulkan_lite_runtime_deps, through
idep_vulkan_runtime.
- VK_USE_PLATFORM_ANDROID_KHR comes from idep_vulkan_wsi_defines,
through idep_vulkan_runtime.
So let's remove these two, as they don't really add anything new.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36716 >
2025-08-11 15:35:24 +00:00
3bd0badd3a
util: SWAP macro implementation for older MSVC versions
...
Reviewed-by: Jesse Natalie <jenatali@microsoft.com >
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36682 >
2025-08-11 13:50:02 +00:00
fe42a3d0eb
meson/util: properly detect sysconf
...
There's two functions that use sysconf(), and they don't seem to agree
on what combination of platforms supports the function. Let's perform
proper function detection instead.
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: Eric Engestrom <eric@igalia.com >
Reviewed-by: Marek Olšák <marek.olsak@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36621 >
2025-08-11 12:30:23 +00:00
e636991c78
prefer _SC_PAGESIZE over _SC_PAGE_SIZE
...
The POSIX spec says that_SC_PAGE_SIZE is a synonym for _SC_PAGESIZE, and
both will have the same value. Let's be consistent in which one we use,
and let's use the one that points directly to the authoritative
documentation.
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: Eric Engestrom <eric@igalia.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36621 >
2025-08-11 12:30:23 +00:00
590ad7fd42
winsys/radeon: use util_get_cpu_caps()-helper
...
Reviewed-by: Marek Olšák <marek.olsak@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36621 >
2025-08-11 12:30:22 +00:00
91232524bf
winsys/radeon: use os_get_page_size and error-check
...
Reviewed-by: Marek Olšák <marek.olsak@amd.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36621 >
2025-08-11 12:30:22 +00:00
ad3445b6df
panfrost: use os_get_page_size()
...
We have an abstraction for this, let's use it to try and keep the code
as portable as possible.
Reviewed-by: John Anthony <john.anthony@arm.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36621 >
2025-08-11 12:30:22 +00:00
44bc8557ae
iris: use os_get_total_physical_memory instead of open-coding
...
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36621 >
2025-08-11 12:30:22 +00:00
2213a76c0a
crocus: use os_get_total_physical_memory instead of open-coding
...
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36621 >
2025-08-11 12:30:22 +00:00
c5c196e6d5
pan/bi: Revamp bi_optimize_nir
...
This reorder things a bit, ensure we attempt more agressive vectorisation,
attempt to optimize cf and more.
Inspiration from NAK's optimize_nir function.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com >
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com >
Acked-by: Eric R. Smith <eric.smith@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36629 >
2025-08-11 11:38:57 +00:00
ef7095c85b
pan/bi: Handle needless conversions in nir_lower_bool_to_bitsize
...
We can end up with conversion instructions to the same type of integer
with nir_lower_bool_to_bitsize so let's make
bifrost_nir_lower_algebraic_late handle those cases.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com >
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com >
Reviewed-by: Eric R. Smith <eric.smith@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36629 >
2025-08-11 11:38:57 +00:00
9ec1bb0111
pan/bi: Vectorize UBOs load/store
...
We can benifit from it, let's allow it when we aren't forced to follow
robustness2.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com >
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com >
Reviewed-by: Eric R. Smith <eric.smith@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36629 >
2025-08-11 11:38:57 +00:00
b4ce8998d7
pan/bi: Switch to nir_lower_alu_width
...
Embrace modernity and consistency between alu and vectorization passes.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com >
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com >
Reviewed-by: Eric R. Smith <eric.smith@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36629 >
2025-08-11 11:38:57 +00:00
0c7aaa7aad
panvk: Call nir_opt_access
...
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com >
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com >
Reviewed-by: Eric R. Smith <eric.smith@collabora.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36629 >
2025-08-11 11:38:56 +00:00
3d85580779
anv: subclass vk_pipeline
...
We'll have to handle RT pipelines inside of Anv because the runtime
doesn't have support for that.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: Ivan Briano <ivan.briano@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36711 >
2025-08-11 11:14:53 +00:00
44aaf88425
anv: rework gfx state emission (again)
...
Previously we had 2 stages :
runtime -> precomputed values -> packing/emission
With this change we use 3 stages :
runtime -> precomputed values -> packing -> emission
Now blorp & other changes to the pipeline should not retrigger
repacking of instructions.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: Ivan Briano <ivan.briano@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36711 >
2025-08-11 11:14:53 +00:00
5a2fb0da32
anv: actually use the COMPUTE_WALKER_BODY prepacked field
...
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: José Roberto de Souza <jose.souza@intel.com >
Reviewed-by: Ivan Briano <ivan.briano@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36711 >
2025-08-11 11:14:52 +00:00
9c8571794a
anv: use a local variable for batch
...
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: Ivan Briano <ivan.briano@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36711 >
2025-08-11 11:14:52 +00:00
f2eea4ec6d
anv: remove pipeline_stage unused field
...
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: Ivan Briano <ivan.briano@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36711 >
2025-08-11 11:14:52 +00:00
e7aeed1f09
anv: pass active stages to push descriptor flushing
...
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com >
Reviewed-by: Ivan Briano <ivan.briano@intel.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36711 >
2025-08-11 11:14:51 +00:00
efab2205fa
panvk: Add utrace tracepoints in queue_submit
...
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com >
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36519 >
2025-08-11 10:12:30 +00:00
Gert Wollny