ydb/build/scripts/compile_cuda.py, branch oidc-1.2.8.12-devMirror of YDB github reposhttps://code.mastervirt.ru/ydb/atom?h=oidc-1.2.8.12-dev2026-02-16T17:42:44Zcreate sanitizer toolchains with CUDA2026-02-16T17:42:44Zabityugovabityugov@yandex-team.com2026-02-16T17:00:38Zurn:sha1:a9ecb4c26f5e3244bc5d3e5d3b04e7dfbd77a8c2
Добавляю тулчейны asan и msan, собранные с CUDA
commit_hash:3731067c580fc1343521a3a08cccbfb363f12ce2
[build] cuda: Introduce CUDA_SRCS macro utilizing parallelized device code compilation2026-02-01T17:51:53Zdeshevoydeshevoy@yandex-team.com2026-02-01T17:25:47Zurn:sha1:2726a4c04b0e9ddc111f2d1ea56390624cb67228
Instead of a single graph node launching NVCC to compile .cu for both host and all device architectures
CUDA_SRCS generates multiple nodes:
- node per each device architecture producing PTX and CUBIN
- node merging all PTX and CUBIN files into a single FATBIN blob
- node producing .cpp with host code
- node compiling host .cpp with embedded FATBIN blob
CUDA_ARCHITECTURES variable is used to determine the list of architectures to compile device code for.
ISSUE:
commit_hash:0a4c2797dd238ae062482af30694df6978301278
Bump CUDA_HOST_TOOLCHAIN2025-10-07T07:38:03Zmikhnenkomikhnenko@yandex-team.com2025-10-07T07:25:21Zurn:sha1:e07309516bf008045b92028d09d482923b65baea
commit_hash:076fe96026b57c22a3f65ea7b482b7d880430570
[build/scripts/compile_cuda] Support mcdc coverage option2025-08-01T04:20:55Zprettyboyprettyboy@yandex-team.com2025-08-01T04:09:09Zurn:sha1:a2a86b907d9b84fc8028dc3afea623af23226b81
commit_hash:52fc45f6924fc03f61c4bd37c9feaf8da3fb4e95
futurize build/scripts2024-09-26T07:44:46Zv-korovinv-korovin@yandex-team.com2024-09-26T07:35:27Zurn:sha1:c9843510b39357d5510030691beecfab56e3bd17
commit_hash:41fb885eb1e03094e65521671349e66f4225321b
clean some blacklist2024-09-02T21:04:55Zarkhip0v0arkhip0v0@yandex-team.com2024-09-02T20:50:23Zurn:sha1:f521542f2363de26174dc2fa8b2eb1a8806c3bd5
occupancy_grid_utils
maintenance_tools
incident_detectors
topic_commutator
sdcbag_player
launcherd
f4ef1453a86ab809b9ceefb57dc6a5fb45a90ec9
mock getpid func2024-08-28T08:03:37Zzhukoff-pavelzhukoff-pavel@yandex-team.com2024-08-28T07:40:30Zurn:sha1:9f71049204d63d8b302cf8d2608843e70ee18379
171c556f349e96400867de4937fe0c703849c0c8
Fix commands on Windows2024-06-12T09:20:00Zspreisspreis@yandex-team.com2024-06-12T03:54:26Zurn:sha1:f5553951374bace336d651d286a1132405c42225
In order to get stable command regardless of platform we shall use linux way of representing commands everywhere. Happily this almost always works on Windows except direct file presence checks like in compile_cuda.py
1412fdabc2fcf1732cd2578e0c6ffe91b220aa2e
`build/scripts` ya style --py2023-11-30T09:56:46Ziaz1607iaz1607@yandex-team.com2023-11-30T09:16:39Zurn:sha1:8951ddf780e02616cdb2ec54a02bc354e8507c0fadd ymake export to ydb2023-06-13T08:05:01Zalexv-smirnovalex@ydb.tech2023-06-13T08:05:01Zurn:sha1:bf0f13dd39ee3e65092ba3572bb5b1fcd125dcd0