mesa/src/amd/ci/gitlab-ci.yml
Martin Roukala (né Peres) 7f75ad546d radv/ci: reduce Renoir concurrency in vkcts
Deqp randomly dies at the beginning of VKCTS, which fails most
pipelines.

Reducing the job concurrency fixes the issue, so let's roll with that
until I go buy more RAM for this machine.

Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33486>
2025-02-12 16:51:19 +00:00

465 lines
12 KiB
YAML

include:
- local: 'src/amd/ci/gitlab-ci-inc.yml'
# Run five jobs in parallel each running 1/55th of the test suite
radv-stoney-vkcts:
extends:
- .lava-test-deqp:x86_64
- .radv-stoney-test:x86_64
parallel: 5
variables:
DEQP_SUITE: radv-stoney
DEQP_FRACTION: 18
radv-stoney-angle:
extends:
- .lava-test-deqp:x86_64
- .radv-stoney-test:x86_64
variables:
DEQP_SUITE: radv-stoney-angle
USE_ANGLE: 1
HWCI_START_XORG: 1
GPU_VERSION: angle-radv-stoney-aco
DEQP_FRACTION: 3
radv-stoney-angle-full:
extends:
- radv-stoney-angle
- .radv-collabora-manual-rules
parallel: 2
variables:
DEQP_SUITE: radv-stoney-angle-full
DEQP_FRACTION: 1
radeonsi-stoney-gl:
extends:
- .lava-test-deqp:x86_64
- .radeonsi-stoney-test:x86_64
parallel: 11
variables:
DEQP_SUITE: radeonsi-stoney
HWCI_START_WESTON: 1
DEQP_FRACTION: 2
radeonsi-stoney-gl-full:
extends:
- radeonsi-stoney-gl
- .radeonsi-manual-rules
variables:
DEQP_FRACTION: 1
radeonsi-stoney-traces:
extends:
- .lava-piglit-traces:x86_64
- .radeonsi-stoney-test:x86_64
variables:
EGL_PLATFORM: surfaceless
PIGLIT_TRACES_FILE: traces-amd.yml
PIGLIT_REPLAY_EXTRA_ARGS: --keep-image
radv-raven-vkcts:
extends:
- .lava-test-deqp:x86_64
- .radv-raven-test:x86_64
- .lava-asus-CM1400CXA-dalboz:x86_64
parallel: 9
variables:
DEQP_SUITE: radv-raven
DEQP_FRACTION: 6
amd-raven-skqp:
extends:
- .lava-test-deqp:x86_64
- .amd-raven-test:x86_64
- .lava-hp-x360-14a-cb0001xx-zork:x86_64
variables:
HWCI_START_WESTON: 1
DEQP_SUITE: amd-raven-skqp
parallel: 2
# Disabled as it was using an older version of Wine, with most of the traces
# being disabled due to being flaky.
.radv-raven-traces:
extends:
- .lava-piglit-traces:x86_64
- .radv-raven-test:x86_64
- .lava-asus-CM1400CXA-dalboz:x86_64
variables:
HWCI_START_XORG: 1
PIGLIT_PLATFORM: mixed_glx_egl # TODO, take wine/VK in account
PIGLIT_REPLAY_DEVICE_NAME: "vk-${GPU_VERSION}"
PIGLIT_TRACES_FILE: traces-amd.yml
radv-raven-traces-restricted:
extends:
- .lava-piglit-traces:x86_64
- .radv-raven-test:x86_64
- .lava-lenovo-TPad-C13-Yoga-zork:x86_64
- .radv-collabora-restricted-rules
variables:
HWCI_START_WESTON: 1
PIGLIT_REPLAY_ANGLE_TAG: "2023-02-10-1"
PIGLIT_TRACES_FILE: restricted-traces-amd.yml
PIGLIT_REPLAY_DEVICE_NAME: "vk-${GPU_VERSION}"
PIGLIT_REPLAY_EXTRA_ARGS: --keep-image --minio_bucket=${S3_TRACIE_PRIVATE_BUCKET} --jwt-file=${S3_JWT_FILE}
FDO_CI_CONCURRENT: 10
allow_failure: true
radeonsi-raven-piglit:
extends:
- .lava-test-deqp:x86_64
- .radeonsi-raven-test:x86_64
- .lava-lenovo-TPad-C13-Yoga-zork:x86_64
parallel: 2
variables:
DEQP_SUITE: radeonsi-raven
HWCI_START_WESTON: 1
# lower image opcodes to emulate CDNA compute devices
radeonsi-raven-cdna-lower-image:
extends:
- .lava-test-deqp:x86_64
- .radeonsi-raven-test:x86_64
- .lava-lenovo-TPad-C13-Yoga-zork:x86_64
variables:
DEQP_SUITE: radeonsi-raven-cdna
GPU_VERSION: radeonsi-raven-cdna
radeonsi-raven-va:
extends:
- .lava-test-deqp:x86_64
- .radeonsi-raven-test:x86_64
- .radeonsi-vaapi-rules
- .lava-hp-x360-14a-cb0001xx-zork:x86_64
variables:
HWCI_TEST_SCRIPT: "/install/gtest-runner.sh"
# While the tests don't need Xorg, it's an easy way to keep DRM open so we don't
# spend all our time in drm_lastclose()'s modesetting.
HWCI_START_WESTON: 1
GTEST: test_va_api
# The big 10-surface 10k allocations oomkill us at around 2.5G of GPU memory
# allocated if we run at the normal 4 threads.
# https://github.com/intel/libva-utils/issues/248
FDO_CI_CONCURRENT: 1
radeonsi-raven-vaapi-fluster:
extends:
- radeonsi-raven-va
- .radeonsi-raven-vaapi-fluster
timeout: 30m
parallel: 6
variables:
FLUSTER_CODECS: VP9 H.264 H.265
# FIXME: Downloading the vectors can take about 4-5 minutes
DEVICE_HANGING_TIMEOUT_SEC: 1200
############### VKCTS tests ###############
# VKCTS never finishes on gfx7 due to all the GPU resets and hangs.
# Hence, disable it for now.
.vkcts-kabini-valve:
extends:
- .vkcts-test-valve
- .kabini-test-valve
- .radv-valve-manual-rules # Make sure this job is *always* run manually
timeout: 12h
variables:
GPU_VERSION: radv-kabini-aco
B2C_TIMEOUT_BOOT_MINUTES: 710
B2C_TIMEOUT_OVERALL_MINUTES: 710
B2C_TIMEOUT_BOOT_RETRIES: 0
# Disabled for now due to Igalia farm issues.
.vkcts-stoney-valve:
extends:
- .vkcts-test-valve
- .radv-valve-manual-rules
parallel: 2
# The thin clients can't manage a full run in less than an hour
# currently. Interpolation suggests four more thin clients (6
# total) would get the runtime under 15 minutes.
timeout: 4h 10m
variables:
GPU_VERSION: radv-stoney-aco
# Note, it only has a wee 32g disk!
B2C_KERNEL_CMDLINE_EXTRAS: 'b2c.swap=8g'
B2C_TIMEOUT_BOOT_MINUTES: 240
B2C_TIMEOUT_OVERALL_MINUTES: 240
B2C_TIMEOUT_BOOT_RETRIES: 0
tags:
- farm:$RUNNER_FARM_LOCATION
- amdgpu:codename:STONEY
vkcts-tahiti-valve:
extends:
- .vkcts-test-valve
- .tahiti-test-valve
- .radv-valve-manual-rules # Make sure this job is *always* run manually
timeout: 6h
variables:
GPU_VERSION: radv-tahiti-aco
B2C_TIMEOUT_BOOT_MINUTES: 350
B2C_TIMEOUT_OVERALL_MINUTES: 350
B2C_TIMEOUT_BOOT_RETRIES: 0
vkcts-polaris10-valve:
extends:
- .vkcts-test-valve
- .polaris10-test-valve-kws
- .radv-valve-manual-rules
timeout: 1h 15m
variables:
GPU_VERSION: radv-polaris10-aco
B2C_TIMEOUT_BOOT_MINUTES: 70
B2C_TIMEOUT_OVERALL_MINUTES: 70
vkcts-vega10-valve:
extends:
- .vkcts-test-valve
- .vega10-test-valve
- .radv-valve-manual-rules
timeout: 1h 15m
variables:
GPU_VERSION: radv-vega10-aco
B2C_TIMEOUT_BOOT_MINUTES: 70
B2C_TIMEOUT_OVERALL_MINUTES: 70
RADV_PERFTEST: transfer_queue
vkcts-renoir-valve:
extends:
- .vkcts-test-valve
- .renoir-test-valve
- .radv-valve-manual-rules
timeout: 2h 15m
variables:
GPU_VERSION: radv-renoir-aco
B2C_TIMEOUT_BOOT_MINUTES: 130
B2C_TIMEOUT_OVERALL_MINUTES: 130
RADV_PERFTEST: transfer_queue
FDO_CI_CONCURRENT: 12 # HACK: Reduce the memory pressure
vkcts-navi10-valve:
parallel: 2
extends:
- .vkcts-test-valve
- .navi10-test-valve-kws
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-navi10-aco
RADV_PERFTEST: transfer_queue
vkcts-navi21-valve:
timeout: 35m
parallel: 3
extends:
- .vkcts-test-valve
- .navi21-test-valve
- .radv-valve-rules
variables:
GPU_VERSION: radv-navi21-aco
B2C_TIMEOUT_BOOT_MINUTES: 30
B2C_TIMEOUT_OVERALL_MINUTES: 30
RADV_PERFTEST: transfer_queue
# Disabled due to its extremelly-broken nature
.vkcts-navi21-llvm-valve:
extends:
- .vkcts-test-valve
- .navi21-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-navi21-llvm
RADV_DEBUG: llvm,checkir
vkcts-vangogh-valve:
parallel: 2
extends:
- .vkcts-test-valve
- .vangogh-test-valve
- .radv-valve-manual-rules
timeout: 2h 10m
variables:
GPU_VERSION: radv-vangogh-aco
RADV_PERFTEST: transfer_queue
.glcts-vangogh-valve-common:
extends:
- .b2c-x86_64-test-gl
- .b2c-deqp-test
- .vangogh-test-valve
variables:
GPU_VERSION: radeonsi-vangogh
DEQP_SUITE: radeonsi-vangogh
glcts-vangogh-valve:
extends:
- .glcts-vangogh-valve-common
- .radeonsi-valve-rules
timeout: 30m
variables:
DEQP_FRACTION: 4
glcts-vangogh-full-valve:
extends:
- .glcts-vangogh-valve-common
- .radeonsi-valve-manual-rules
timeout: 90m
vkcts-raphael-valve:
extends:
- .vkcts-test-valve
- .raphael-test-valve
- .radv-valve-manual-rules
timeout: 45m
variables:
GPU_VERSION: radv-raphael-aco
B2C_TIMEOUT_BOOT_MINUTES: 35
B2C_TIMEOUT_OVERALL_MINUTES: 40
RADV_PERFTEST: transfer_queue
vkcts-navi31-valve:
parallel: 2
extends:
- .vkcts-test-valve
- .navi31-test-valve
- .radv-valve-rules
variables:
GPU_VERSION: radv-navi31-aco
RADV_DEBUG: nomeshshader # Disable mesh shaders until task shaders stop hanging
RADV_PERFTEST: transfer_queue
############### Fossilize
radv-fossils:
extends:
- .radv-valve-rules
- .test-vk
- .test-radv
- .fossilize-test
script:
- export LD_PRELOAD=$PWD/install/lib/libamdgpu_noop_drm_shim.so
# Pitcairn (GFX6)
- AMDGPU_GPU_ID="PITCAIRN"
./install/fossilize-runner.sh
# Bonaire (GFX7)
- AMDGPU_GPU_ID="BONAIRE"
./install/fossilize-runner.sh
# Polaris10 (GFX8)
- AMDGPU_GPU_ID="POLARIS10"
./install/fossilize-runner.sh
# Vega10 (GFX9)
- AMDGPU_GPU_ID="VEGA10"
./install/fossilize-runner.sh
# Navi10 (GFX10)
- AMDGPU_GPU_ID="NAVI10"
./install/fossilize-runner.sh
# Navi21 (GFX10_3)
- AMDGPU_GPU_ID="NAVI21"
./install/fossilize-runner.sh
# RDNA3 (GFX11)
- AMDGPU_GPU_ID="NAVI31"
./install/fossilize-runner.sh
# RDNA3.5 (GFX1150)
- AMDGPU_GPU_ID="GFX1150"
./install/fossilize-runner.sh
############### vkd3d-proton
vkd3d-tahiti-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .tahiti-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-tahiti
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-polaris10-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .polaris10-test-valve-kws
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-polaris10
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-vega10-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .vega10-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-vega10
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-renoir-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .renoir-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-renoir
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-navi10-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .navi10-test-valve-mupuf
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-navi10
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-navi21-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .navi21-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-navi21
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-vangogh-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .vangogh-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-vangogh
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-raphael-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .raphael-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-raphael
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1
vkd3d-navi31-valve:
extends:
- .b2c-test-radv-vk
- .b2c-vkd3d-proton-test
- .navi31-test-valve
- .radv-valve-manual-rules
variables:
GPU_VERSION: radv-navi31
# Random tests fail/crash when trying to run them in parallel
FDO_CI_CONCURRENT: 1