Before rebasing on top of Ken's split-SEND optimization (see !17018),
this commit just caused some scheduling changes in various tessellation
and geometry shaders. These changes were caused by the addition of real
latency information for the URB messages.
With the addition of the split-SEND optimization, the changes
are... staggering. All of the shaders helped for spills and fills are
vertex shaders from Batman Arkham Origins. What surprises me is that
these shaders account for such a high percentage of the spills and fills
in fossil-db. 85%?!?
v2: Use FIXED_GRF instead of BRW_GENERAL_REGISTER_FILE in an assertion.
Suggested by Ken.
Tiger Lake, Ice Lake, and Skylake had similar results. (Ice Lake shown)
total instructions in shared programs: 20013625 -> 19954020 (-0.30%)
instructions in affected programs: 4007157 -> 3947552 (-1.49%)
helped: 31161
HURT: 0
helped stats (abs) min: 1 max: 400 x̄: 1.91 x̃: 2
helped stats (rel) min: 0.08% max: 59.70% x̄: 2.20% x̃: 1.83%
95% mean confidence interval for instructions value: -1.97 -1.86
95% mean confidence interval for instructions %-change: -2.22% -2.18%
Instructions are helped.
total cycles in shared programs: 859337569 -> 858636788 (-0.08%)
cycles in affected programs: 74168298 -> 73467517 (-0.94%)
helped: 13812
HURT: 16846
helped stats (abs) min: 1 max: 291078 x̄: 82.83 x̃: 4
helped stats (rel) min: <.01% max: 37.09% x̄: 3.47% x̃: 2.02%
HURT stats (abs) min: 1 max: 1543 x̄: 26.31 x̃: 14
HURT stats (rel) min: <.01% max: 77.97% x̄: 4.11% x̃: 2.58%
95% mean confidence interval for cycles value: -55.10 9.39
95% mean confidence interval for cycles %-change: 0.62% 0.77%
Inconclusive result (value mean confidence interval includes 0).
Broadwell
total cycles in shared programs: 904844939 -> 904832320 (<.01%)
cycles in affected programs: 525360 -> 512741 (-2.40%)
helped: 215
HURT: 4
helped stats (abs) min: 4 max: 1018 x̄: 60.16 x̃: 39
helped stats (rel) min: 0.14% max: 15.85% x̄: 2.16% x̃: 2.04%
HURT stats (abs) min: 79 max: 79 x̄: 79.00 x̃: 79
HURT stats (rel) min: 1.31% max: 1.57% x̄: 1.43% x̃: 1.43%
95% mean confidence interval for cycles value: -75.02 -40.22
95% mean confidence interval for cycles %-change: -2.37% -1.81%
Cycles are helped.
No shader-db changes on any older Intel platforms.
Tiger Lake, Ice Lake, and Skylake had similar results. (Ice Lake shown)
Instructions in all programs: 142622800 -> 141461114 (-0.8%)
Instructions helped: 197186
Cycles in all programs: 9101223846 -> 9099440025 (-0.0%)
Cycles helped: 37963
Cycles hurt: 151233
Spills in all programs: 98829 -> 13695 (-86.1%)
Spills helped: 2159
Fills in all programs: 128142 -> 18400 (-85.6%)
Fills helped: 2159
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17379>
The lowering is currently fake. It just changes the opcode from the
_LOGICAL version to the non-_LOGICAL version.
v2: Remove some rebase cruft. 's/gfx8_//;s/simd8_/' in
brw_instruction_name. Both suggested by Ken.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17379>
If these checks had been in place previously, some bugs
that... eh-hem... practically took down the Intel CI would have been
caught earlier. *blush*
v2: Update to account for split sends.
v3: Add some more Gfx version checks. Remove the redundant "src0 is a
GRF" check. Both suggested by Ken.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17379>
An argument could be made that all stage-specific opcodes for vec4
stages should be prefixed with VEC4_ like the stage-agnostic opcodes.
I'll leave those additional sed jobs for another day.
egrep -lr '(VS|GS|TCS)_OPCODE_URB_WRITE' src |\
while read f; do
sed --in-place 's/\(VS\|GS\|TCS\)_OPCODE_URB_WRITE/VEC4_\1_OPCODE_URB_WRITE/g' $f
done
egrep -lr 'T.S_OPCODE[_A-Z]*URB_OFFSETS' src |\
while read f; do
sed --in-place 's/\(T.S_OPCODE[_A-Z]*URB_OFFSETS\)/VEC4_\1/g' $f
done
Suggested-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17379>
According to LLVM, image_bvh64_intersect_ray does not write results in
order with other VMEM instructions.
fossil-db (navi21):
Totals from 7 (0.00% of 162293) affected shaders:
Instrs: 39978 -> 39985 (+0.02%)
CodeSize: 219356 -> 219384 (+0.01%)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17079>
When the field was first introduces, the numbers were reporting the number of
vec4 instead of the number of float. Do not propagate them if they are wrong.
Fixes: d92c1ca01b
Signed-off-by: Corentin Noël <corentin.noel@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17415>
We should be explicit that we are cancelling jobs once the script finds
some log messages that are linked with known issues. That means the
script preemptively retried the job without giving chances to recover.
Adds magenta color to cancelled jobs.
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17389>
Mark test_full_yaml_log with this new marker to be easily run by the
developers.
Make `debian-testing` skip this test with `not slow` marker hint.
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17389>
Implement a log-based retry hint for R8152 issue described in #6681,
which is based on detecting these two consecutive lines:
```
r8152 <USB> eth0: Tx status -71
nfs: server <IP> not responding, still trying
```
Where <IP> and <USB> could be any IP and USB addresses, respectfully.
This commit is a temporary fix since it requires a section-aware log
follower, implemented in !16323. When the cited MR is merged, one will
make a proper fix on top of that.
Closes: #6681
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17389>
if a texture barrier occurs while clears are pending, these clears should
show up if the fb attachments are read in shaders, so trigger a renderpass
to flush out the clears
cc: mesa-stable
fixes#6766
fixes (radv):
dEQP-GLES3.functional.draw_buffers_indexed.overwrite_common.common_advanced_blend_eq_buffer_advanced_blend_eq
dEQP-GLES3.functional.draw_buffers_indexed.overwrite_common.common_blend_eq_buffer_advanced_blend_eq
dEQP-GLES3.functional.draw_buffers_indexed.overwrite_common.common_separate_blend_eq_buffer_advanced_blend_eq
dEQP-GLES3.functional.draw_buffers_indexed.overwrite_indexed.common_advanced_blend_eq_buffer_advanced_blend_eq
dEQP-GLES3.functional.draw_buffers_indexed.overwrite_indexed.common_advanced_blend_eq_buffer_blend_eq
dEQP-GLES3.functional.draw_buffers_indexed.overwrite_indexed.common_advanced_blend_eq_buffer_separate_blend_eq
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17363>
This got removed by mistake and broke
RADV_DEBUG=shaders,nocache,prologs.
Fixes: 9fe2b6b748 ("aco/radv: provide a vs prolog callback from aco to radv.")
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/17413>
The Vulkan specification states:
> Query commands, for the same query and submitted to the same queue,
> execute in their entirety in submission order, relative to each other. In
> effect there is an implicit execution dependency from each such query
> command to all query commands previously submitted to the same queue.
Fixes dEQP-VK.query_pool.statistics_query.reset_after_copy.*
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17400>
This seems to be a relic from before aco added per generation opcodes.
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17405>
PS is the only shader type where unpacked 16-bit outputs can be used,
so use ac_shader_abi::is_16bit to pass the proper type to LLVMBuildLoad2.
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17361>
Outputs are always f32 except for FS that may use unpacked f16.
Store this information here to make it available to later processing.
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17361>
This commit replaces LLVMBuildLoad usage by LLVMBuildLoad2
where possible (= where the pointee type is known).
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17361>
Disabling opaque pointers in LLVM doesn't fix all the issues but
it makes pointers non-opaque by default (eg LLVMPointerType()
returns a typed pointer).
Reviewed-by: Mihai Preda <mhpreda@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17361>
Fixed:
- VK_QUERY_TYPE_PRIMITIVES_GENERATED_EXT was able to stop prim counter
when pipeline stats query is running.
- This may have happened when prim gen query was in secondary cmdbuf.
- VK_QUERY_TYPE_PRIMITIVES_GENERATED_EXT counting geometry in each tile.
- VK_QUERY_TYPE_PRIMITIVES_GENERATED_EXT counting geometry in each tile
when pipeline stats query is started inside prim gen query and inside
a renderpass.
The matter of VK_QUERY_TYPE_PRIMITIVES_GENERATED_EXT and pipeline stats
interaction is solved by tracking whether pipeline stats query is
running both on CPU (for non secondary cmdbuf case) and on GPU (for
secondary cmdbuf).
Note, prim gen query is not allowed with secondary command buffers, so
only pipeline stats query is tracked on gpu.
See https://gitlab.khronos.org/vulkan/vulkan/-/issues/3142
Counting geometry per each tile is solved by:
- Conditionally executing START/STOP_PRIMITIVE_CTRS to not run in tiling
pass. Solves the case when prim gen query is inside a renderpass.
- Stop prim counters before executing `draw_cs` and restarting them
afterwards. Solves prim gen query being outside a renderpass.
Fixes GL CTS tests with Zink + `TU_DEBUG=gmem`:
GTF-GL46.gtf30.GL3Tests.transform_feedback.transform_feedback_max_separate
GTF-GL46.gtf40.GL3Tests.transform_feedback2.transform_feedback2_basic
GTF-GL46.gtf40.GL3Tests.transform_feedback2.transform_feedback2_framebuffer
GTF-GL46.gtf40.GL3Tests.transform_feedback3.transform_feedback3_streams_overflow
GTF-GL46.gtf40.GL3Tests.transform_feedback3.transform_feedback3_streams_queried
GTF-GL46.gtf40.GL3Tests.transform_feedback2.transform_feedback2_states
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6602
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17163>
For a5xx+ renamed:
- RST_PIX_CNT -> START_FRAGMENT_CTRS
- RST_VTX_CNT -> STOP_FRAGMENT_CTRS
- TILE_FLUSH -> START_COMPUTE_CTRS
- STAT_EVENT -> STOP_COMPUTE_CTRS
I'm not sure about a5xx itself but I'll take a chance of it being
similar to a6xx in this regard.
Knowing this emit_begin_stat_query/emit_end_stat_query can now emit
only events that are needed for the pool's flags.
Also primitive generated query clearly doesn't need fragment and
compute counters.
Passes tests:
dEQP-VK.query_pool.statistics_query.*
dEQP-VK.transform_feedback.primitives_generated_query.*
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17163>
Rob Clark
Ian Romanick
Jesse Natalie
pal1000
David Heidelberg
Samuel Pitoiset
Rhys Perry
Corentin Noël
Guilherme Gallo
Mike Blumenkrantz
Tatsuyuki Ishi
Georg Lehmann
Pierre-Eric Pelloux-Prayer
Danylo Piliaiev