{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":17872428,"defaultBranch":"master","name":"libflame","ownerLogin":"flame","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2014-03-18T16:07:22.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/6494486?v=4","public":true,"private":false,"isOrgOwned":true},"refInfo":{"name":"","listCacheKey":"v0:1656022918.989851","currentOid":""},"activityList":{"items":[{"before":"646ed9075cc45aeffb53632a3ec88defb423fefa","after":"f38dd83e30136b4e25eb2343813ee4fbd7c16681","ref":"refs/heads/master","pushedAt":"2024-05-20T18:37:39.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Support --with-fc (#106)\n\nDetails:\r\n* Adding --with-fc as a configure flag\r\n* FC and FCFLAGS can be adjusted at configure time\r\n\r\nCo-authored-by: ItsPitt ","shortMessageHtmlLink":"Support --with-fc (#106)"}},{"before":"70c19e770ead0ae846c59b59216deb16d236b40c","after":"646ed9075cc45aeffb53632a3ec88defb423fefa","ref":"refs/heads/master","pushedAt":"2024-02-27T00:05:17.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Updates for netlib LAPACK 3.12.0 and ROCm 6.0 (#103)\n\nDetails:\r\n- Upgrade non-legacy LAPACK version to netlib 3.12.0\r\n- Change call signature of ROCm trmm for ROCm 6.0","shortMessageHtmlLink":"Updates for netlib LAPACK 3.12.0 and ROCm 6.0 (#103)"}},{"before":"7a5801f65c3f6f465e4067f5814d42e081aa1832","after":"70c19e770ead0ae846c59b59216deb16d236b40c","ref":"refs/heads/master","pushedAt":"2023-07-13T19:33:42.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Improve HIP locking in FLASH (#99)\n\nDetails:\r\n- instead of tracking of to be flushed blocks, kick off async transfers and wait on them afterwards\r\n- similarly, remove the victim block and instead transfer that single block instantly\r\n- add write locking for memory allocations\r\n- add read locking for final flush\r\n- while there, enable managed memory by default (reduces latencies as it eschews cache allocations)","shortMessageHtmlLink":"Improve HIP locking in FLASH (#99)"}},{"before":"a19fb8d7144e729353c5498d110016d4910b6413","after":"7a5801f65c3f6f465e4067f5814d42e081aa1832","ref":"refs/heads/master","pushedAt":"2023-07-13T19:32:08.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Optimize locks (#98)\n\nDetails:\r\n- target the single thread, single queue, single cache case (present in typical HIP offloading)\r\n- do not allocate or initialize the locks - set them to NULL instead\r\n- add a short circuit in the lock wrappers to return on NULL'd lock pointers\r\n- add a warning if a NULL'd lock pointer is supposed to be initialized\r\n- turn the all_lock into a heap allocated pointer","shortMessageHtmlLink":"Optimize locks (#98)"}},{"before":"9eae1d98f37c30774e53d0cefb31188102ab742b","after":"a19fb8d7144e729353c5498d110016d4910b6413","ref":"refs/heads/master","pushedAt":"2023-07-13T19:25:00.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Add LTO (link-time optimization) option. (#97)\n\nDetails:\r\n- add a configure-time option to --enable-lto (off by default)\r\n- adds -flto and -fno-semantic-interposition to CFLAGS and LDFLAGS\r\n- add a FLA_ENABLE_LTO macro","shortMessageHtmlLink":"Add LTO (link-time optimization) option. (#97)"}},{"before":"c7dad20d92df24de194afcd7ab7d46feea37a1ec","after":"9eae1d98f37c30774e53d0cefb31188102ab742b","ref":"refs/heads/master","pushedAt":"2023-06-24T01:23:06.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Conditionally enable HIP FLASH queue and managed memory. (#96)\n\nDetails:\r\n- at init of the HIP FLASH queue, check HSA agents present on the\r\n system for their chip ID.\r\n- find the highest chip ID (since CPU is present and we can assume\r\n strictly homogeneous accelerator setups for this purpose).\r\n- if only CPU agents are found, do NOT enable FLASH queue.\r\n- if the highest found ID matches a specific APU family, enable\r\n HIP queue and managed memory by default to improve performance.","shortMessageHtmlLink":"Conditionally enable HIP FLASH queue and managed memory. (#96)"}},{"before":"52f172267cfdbfd0b4a02177f57166e410b33e30","after":"c7dad20d92df24de194afcd7ab7d46feea37a1ec","ref":"refs/heads/master","pushedAt":"2023-04-28T01:57:24.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Auto-disable HIP for small problems in lapack2flash wrappers (#95)\n\nDetails:\r\n- instead of using the HIP queue independent of problem size, check if blocked size is\r\n above or below a crossover point. If it is below, disable the HIP queue prior to running\r\n the problem and re-enable it afterwards.\r\n- while there, fix a few small build issues and add the lapack2flash header to FLAME.h\r\n- Co-authored-by: Michael Pittard ","shortMessageHtmlLink":"Auto-disable HIP for small problems in lapack2flash wrappers (#95)"}},{"before":"450c1094972439d76d9eb0632e386ca3fb9d991f","after":"52f172267cfdbfd0b4a02177f57166e410b33e30","ref":"refs/heads/master","pushedAt":"2023-04-28T01:14:54.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Implement hegst through FLASH (#94)\n\nDetails:\r\n- implement hegst through FLASH_Eig_gest\r\n- while here, fix a min() in trsyl\r\n- Authored-by: Michael Pittard ","shortMessageHtmlLink":"Implement hegst through FLASH (#94)"}},{"before":"7e94bbd0cb203d7fbd43d30c1890614ceac3b02a","after":"450c1094972439d76d9eb0632e386ca3fb9d991f","ref":"refs/heads/master","pushedAt":"2023-04-28T01:13:54.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Prepare APIs for non-default streams (#93)\n\nDetails:\r\n- instead of using the default null stream in HIP queue functions, get\r\n the stream from the rocblas_handle associated with the thread\r\n- add a stream sync function\r\n- rename the HIP write API to write_async to correctly signal intent\r\n- change the device allocation to be asynchronous (and rename API)\r\n- write something unique into the buffer_hip pointer in the managed memory\r\n case to ensure comparisons are correct","shortMessageHtmlLink":"Prepare APIs for non-default streams (#93)"}},{"before":"d48aa5d42525b0b8d077336763e70e669365b727","after":"7e94bbd0cb203d7fbd43d30c1890614ceac3b02a","ref":"refs/heads/master","pushedAt":"2023-03-11T01:03:28.077Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"fgvanzee","name":"Field G. Van Zee","path":"/fgvanzee","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5487570?s=80&v=4"},"commit":{"message":"Update Fortran LAPACK and tests to 3.11 (#92)\n\nDetails:\r\n* update Fortran LAPACK and netlib tests to version 3.11\r\n* add a README how to build and use\r\n* consolidate build steps and account for Fortran suffices\r\n\r\nAuthored-by: Michael Pittard \r\nCo-authored-by: ItsPitt ","shortMessageHtmlLink":"Update Fortran LAPACK and tests to 3.11 (#92)"}}],"hasNextPage":false,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"djE6ks8AAAAETxZnKgA","startCursor":null,"endCursor":null}},"title":"Activity ยท flame/libflame"}