)]}' { "log": [ { "commit": "56d941a8cede7c0d6aa4dc19e8f0b95de6f97e1b", "tree": "bbfb98617dc70f376ea7a6067eb7d5fe3e465599", "parents": [ "cb6f3c46a841e47718564ca40f49af8bcfc57c3a" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Tue Oct 15 17:15:26 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Tue Oct 15 17:15:26 2019 +0000" }, "message": "[libomptarget][nfc] Make interface.h target independent\n\nSummary:\n[libomptarget][nfc] Make interface.h target independent\n\nMove interface.h under a top level include directory.\nRemove #includes to avoid the interface depending on the implementation.\n\nReviewers: ABataev, jdoerfert, grokos, ronlieb, RaviNarayanaswamy\n\nReviewed By: jdoerfert\n\nSubscribers: mgorny, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D68615\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@374919 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "cb6f3c46a841e47718564ca40f49af8bcfc57c3a", "tree": "5b7d1c38316e0ff0b44ab49d4d5c9c22c0f1474d", "parents": [ "ace8505db1e11227e9007a2ac55ffaeccf88a65c" ], "author": { "name": "David Carlier", "email": "devnexen@gmail.com", "time": "Tue Oct 08 21:25:30 2019 +0000" }, "committer": { "name": "David Carlier", "email": "devnexen@gmail.com", "time": "Tue Oct 08 21:25:30 2019 +0000" }, "message": "[OpenMP] Enable thread affinity on FreeBSD\n\nReviewers: chandlerc, jlpeyton, jdoerfert, dim\n\nReviewed-By: dim\n\nDifferential Revision: https://reviews.llvm.org/D68580\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@374118 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ace8505db1e11227e9007a2ac55ffaeccf88a65c", "tree": "e3739778d3d034b2e280687b975f5520321245a5", "parents": [ "9dfddd6912d2af7f5a4e27a9b7782214bf2045e9" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Oct 08 12:39:04 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Oct 08 12:39:04 2019 +0000" }, "message": "Don\u0027t assume Type from `readelf -d` has parentheses\n\nPatch by jbeich (Jan Beich)\n\nDifferential Revision: https://reviews.llvm.org/D68053\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@374038 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "9dfddd6912d2af7f5a4e27a9b7782214bf2045e9", "tree": "fb373e6b9870c835defd1d5186e8f202cf8d79a6", "parents": [ "731769fe89c8e17850bc9df82d2ed248632a59ad" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Oct 08 12:23:25 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Oct 08 12:23:25 2019 +0000" }, "message": "Don\u0027t link libm with -Wl,--as-needed on FreeBSD\n\nPatch by jbeich (Jan Beich)\n\nDifferential Revision: https://reviews.llvm.org/D68051\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@374037 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "731769fe89c8e17850bc9df82d2ed248632a59ad", "tree": "928beb2e1f028b9bbcdc347c79baa2aa99182fea", "parents": [ "3bdcaec89abb1affe3113894bd96e64f275395c4" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Fri Oct 04 22:30:28 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Fri Oct 04 22:30:28 2019 +0000" }, "message": "[libomptarget][nfc] Update remaining uint32 to use lanemask_t\n\nSummary:\n[libomptarget][nfc] Update remaining uint32 to use lanemask_t\n\nUpdate a few functions in the API to use lanemask_t instead of i32. NFC for\nnvptx. Also update the ActiveThreads type in DataSharingStateTy.\nThis removes a lot of #ifdef from the downsteam amdgcn implementation.\n\nReviewers: ABataev, jdoerfert, grokos, ronlieb, RaviNarayanaswamy\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D68513\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@373806 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "3bdcaec89abb1affe3113894bd96e64f275395c4", "tree": "fa8b7770ed7b217daa32f7ae60b60048ec26cf4e", "parents": [ "85e24591e4ebdbae12cf85f55ade8c80653219cb" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Fri Oct 04 21:39:22 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Fri Oct 04 21:39:22 2019 +0000" }, "message": "Use named constant to indicate all lanes, to handle 32 and 64 wide architectures\n\nSummary: Use named constant to indicate all lanes, to handle 32 and 64 wide architectures\n\nReviewers: ABataev, jdoerfert, grokos, ronlieb\n\nReviewed By: grokos\n\nSubscribers: ronlieb, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D68369\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@373793 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "85e24591e4ebdbae12cf85f55ade8c80653219cb", "tree": "a2fc5f4834ecb36ea477c56aab73d2c835654af2", "parents": [ "dbb024cf94e6a7603ac0484be3c44d2f29eeceb6" ], "author": { "name": "David Carlier", "email": "devnexen@gmail.com", "time": "Sat Sep 28 19:01:59 2019 +0000" }, "committer": { "name": "David Carlier", "email": "devnexen@gmail.com", "time": "Sat Sep 28 19:01:59 2019 +0000" }, "message": "[OpenMP] FreeBSD address check if mapped more native\n\n/proc unless Linux layer compatibility is activated for CentOS is activated is not present\nthus relying on a more native for checking the address.\n\nReviewers: Hahnfeld, kongyl, jdoerfert, jlpeyton, AndreyChurbanov, emaster, dim\n\nReviewed By: Hahnfeld\n\nDifferential Revision: https://reviews.llvm.org/D67326\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@373152 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "dbb024cf94e6a7603ac0484be3c44d2f29eeceb6", "tree": "0997c6cc14be0d337ed58eeef97879e62104bbee", "parents": [ "2cb5d49f712a6c4c502ab4f155acf72d86bc604f" ], "author": { "name": "Sergey Dmitriev", "email": "serguei.n.dmitriev@intel.com", "time": "Fri Sep 27 20:00:51 2019 +0000" }, "committer": { "name": "Sergey Dmitriev", "email": "serguei.n.dmitriev@intel.com", "time": "Fri Sep 27 20:00:51 2019 +0000" }, "message": "[Clang][OpenMP Offload] Create start/end symbols for the offloading entry table with a help of a linker\n\nLinker automatically provides __start_\u003csection name\u003e and __stop_\u003csection name\u003e symbols to satisfy unresolved references if \u003csection name\u003e is representable as a C identifier (see https://sourceware.org/binutils/docs/ld/Input-Section-Example.html for details). These symbols indicate the start address and end address of the output section respectively. Therefore, renaming OpenMP offload entries section name from \".omp.offloading_entries\" to \"omp_offloading_entries\" to use this feature.\n\nThis is the first part of the patch for eliminating OpenMP linker script (please see https://reviews.llvm.org/D64943).\n\nDifferential Revision: https://reviews.llvm.org/D68070\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@373118 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "2cb5d49f712a6c4c502ab4f155acf72d86bc604f", "tree": "723dfccd79fb4ba9ded0e23a495b6b254803b478", "parents": [ "66ddc2110ae411ee6187b7f674e3698e8cc11872" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Sep 25 15:25:52 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Sep 25 15:25:52 2019 +0000" }, "message": "fixed test: eliminated race condition which might cause deadlock\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@372887 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "66ddc2110ae411ee6187b7f674e3698e8cc11872", "tree": "66531a4ae061d6eb18ce557789e30e29e5a2b245", "parents": [ "2b135c5afc8ffa96cebe7a9ce9ae3e7251b1fadc" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Sep 25 14:40:19 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Sep 25 14:40:19 2019 +0000" }, "message": "Enable tasks dependencies hashmaps resizing.\n\nPatch by viroulep (Philippe Virouleau)\n\nDifferential Revision: https://reviews.llvm.org/D67447\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@372879 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "2b135c5afc8ffa96cebe7a9ce9ae3e7251b1fadc", "tree": "045369d50015c7884206c76ef085f1badac9212f", "parents": [ "b93e8ad8939f319cbc53f13ba0f6f724504acfe2" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Wed Sep 04 17:47:37 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Wed Sep 04 17:47:37 2019 +0000" }, "message": "[OpenMP] Change initialization of __kmp_global\n\nThere\u0027s no need to initialize variables with static storage duration\nbecause they\u0027re implicitly initialized to zero. See\nhttps://en.cppreference.com/w/c/language/initialization#Implicit_initialization\n\nI think that\u0027s already relied upon because the supplied 0 only sets\n\u0027kmp_time_global_t g_time;\u0027 in \u0027struct kmp_base_global\u0027. The other fields\nare not set in the code, but implicitly initialized by the compiler.\n\nDifferential Revision: https://reviews.llvm.org/D66292\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370943 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "b93e8ad8939f319cbc53f13ba0f6f724504acfe2", "tree": "a331d6e47a7ad6c9b716922de7c8890222ba400f", "parents": [ "537de38cfe7387cbeecb51e03fe87fd38933a363" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Sep 03 18:11:50 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Sep 03 18:11:50 2019 +0000" }, "message": "[OPENMP][NVPTX]Fix parallel level counter in non-SPMD mode.\n\nSummary:\nIn non-SPMD mode we may end up with the divergent threads when trying to\nincrement/decrement parallel level counter. It may lead to incorrect\ncalculations of the parallel level and wrong results when threads are\ndivergent. We need to reconverge the threads before trying to modify the\nparallel level counter.\n\nReviewers: grokos, jdoerfert\n\nSubscribers: guansong, openmp-commits, caomhin, kkwli0\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66802\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370803 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "537de38cfe7387cbeecb51e03fe87fd38933a363", "tree": "834892cb2a35f14a39bb2db3b0d6fcc084296bb3", "parents": [ "1fe4d6217edb8c431631aa39a1f75fd2519bb7eb" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Tue Sep 03 16:31:30 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Tue Sep 03 16:31:30 2019 +0000" }, "message": "[libomptarget] Refactor activemask macro to inline function\n\nSummary:\n[libomptarget] Refactor activemask macro to inline function\nSee also abandoned D66846, split into this diff and others.\n\nReviewers: jdoerfert, ABataev, grokos, ronlieb, gregrodgers\n\nReviewed By: jdoerfert, ABataev\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66851\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370781 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "1fe4d6217edb8c431631aa39a1f75fd2519bb7eb", "tree": "f31720090e4a2f28ae2a846c63221f38ccaa456a", "parents": [ "ee031606128fec949b4f90152e25c37fdab13a8c" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 15:04:06 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 15:04:06 2019 +0000" }, "message": "Use target_impl functions to replace more inline asm\n\nSummary:\nUse target_impl functions to replace more inline asm\nFollow on from D65836. Removes remaining asm shuffles and lanemask accessors\nAlso changes the types of target_impl bitwise functions to unsigned.\n\nReviewers: jdoerfert, ABataev, grokos, Hahnfeld, gregrodgers, ronlieb, hfinkel\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66809\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370216 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ee031606128fec949b4f90152e25c37fdab13a8c", "tree": "e16724f50b1edfb4e8de86c97abdad879ceba304", "parents": [ "7ed5372fe0c8ae8328b2c8274d88cf3762b3b776" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 14:22:35 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 14:22:35 2019 +0000" }, "message": "[libomptarget] Refactor syncthreads macro to inline function\n\nSummary:\n[libomptarget] Refactor syncthreads macro to inline function\nSee also abandoned D66846, split into this diff and others.\n\nRev 2 of D66855\n\nReviewers: jdoerfert, ABataev, grokos, ronlieb, gregrodgers\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66861\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370210 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "7ed5372fe0c8ae8328b2c8274d88cf3762b3b776", "tree": "59ce30cfd397b5ba2493bbc3e44fdbb14f382ef0", "parents": [ "bae5e5989e667b8d0db15815dc9345a9776273e4" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 02:02:53 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 02:02:53 2019 +0000" }, "message": "[libomptarget] Refactor syncwarp macro to inline function\n\nSummary:\n[libomptarget] Refactor syncwarp macro to inline function\nSee also abandoned D66846, split into this diff and others.\n\nReviewers: jdoerfert, ABataev, grokos, ronlieb, gregrodgers\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66857\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370149 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "bae5e5989e667b8d0db15815dc9345a9776273e4", "tree": "b1531cb1485009a0a0d6f32dd4a14966d963c68e", "parents": [ "77d31c11e83957e751f9d9854e58110e6b532c0c" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 01:56:26 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 01:56:26 2019 +0000" }, "message": "Fix build break due to close brace lost in merge\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370148 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "77d31c11e83957e751f9d9854e58110e6b532c0c", "tree": "56b9cad81677305847b2c1ed1ad485aa98bf98d2", "parents": [ "bfca7f70c5f1756d7e74e92a1de5630dc41f54fc" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 01:47:41 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 01:47:41 2019 +0000" }, "message": "[libomptarget] Refactor shfl_down_sync macro to inline function\n\nSummary:\n[libomptarget] Refactor shfl_down_sync macro to inline function\nSee also abandoned D66846, split into this diff and others.\n\nReviewers: jdoerfert, ABataev, grokos, ronlieb, gregrodgers\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66853\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370146 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "bfca7f70c5f1756d7e74e92a1de5630dc41f54fc", "tree": "ee0b027c830074f8d87ce232f9778fba7c280db3", "parents": [ "ad72f5e57bb95e2f765d693821562b4c79406fa9" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 01:31:04 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 28 01:31:04 2019 +0000" }, "message": "[libomptarget] Refactor shfl_sync macro to inline function\n\nSummary:\n[libomptarget] Refactor shfl_sync macro to inline function\nSee also abandoned D66846, split into this diff and others.\n\nReviewers: jdoerfert, ABataev, grokos, ronlieb, gregrodgers\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66852\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@370144 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ad72f5e57bb95e2f765d693821562b4c79406fa9", "tree": "fe28f8f43d8d8311ebb1e9c3c1b09370ef4393e8", "parents": [ "cd9405f3b99b1c2d92d43b32cd227fd79ff11bfd" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon Aug 26 17:32:45 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon Aug 26 17:32:45 2019 +0000" }, "message": "[OPENMP][NVPTX]Add __kmpc_syncwarp(int32_t) function.\n\nSummary:\nAdded function void __kmpc_syncwarp(int32_t) to expose it to the\ncompiler. It is required to fix the problem with the critical regions in\nCuda9.0+. We cannot use barrier in the critical region, but still need\nto reconverge the threads in the warp after. This function allows to do\nthis.\n\nReviewers: grokos, jdoerfert\n\nSubscribers: guansong, openmp-commits, kkwli0, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D66672\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@369933 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "cd9405f3b99b1c2d92d43b32cd227fd79ff11bfd", "tree": "195da40dd31585a3bc4fdb34dbe7866f8cc519fa", "parents": [ "3f4c51a258bcfbf588265dd6e80726c67305f8eb" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri Aug 23 18:34:48 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri Aug 23 18:34:48 2019 +0000" }, "message": "[OPENMP][NVPTX]Use __syncwarp() to reconverge the threads.\n\nSummary:\nIn Cuda 9.0 it is not guaranteed that threads in the warps are\nconvergent. We need to use __syncwarp() function to reconverge\nthe threads and to guarantee the memory ordering among threads in the\nwarps.\nThis is the first patch to fix the problem with the test\nlibomptarget/deviceRTLs/nvptx/src/sync.cu on Cuda9+.\nThis patch just replaces calls to __shfl_sync() function with the call\nof __syncwarp() function where we need to reconverge the threads when we\ntry to modify the value of the parallel level counter.\n\nReviewers: grokos\n\nSubscribers: guansong, jfb, jdoerfert, caomhin, kkwli0, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D65013\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@369796 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "3f4c51a258bcfbf588265dd6e80726c67305f8eb", "tree": "148fd9fecac7f7e503ba80a4c3b256e9afd02ce4", "parents": [ "3eae253b1ac2da31d8928950dcb50956b64e81d1" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue Aug 20 19:39:17 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue Aug 20 19:39:17 2019 +0000" }, "message": "Force honoring nthreads-var and thread-limit-var inside teams construct on host\n\nThis patch fixes https://bugs.llvm.org/show_bug.cgi?id\u003d42906, via adding\nadjustment of number of threads on enter to the teams construct on host\naccording to user settings. This allows to pass checks and avoid assertions\nat time of team of threads creation.\n\nPatch by Andrey Churbanov\n\nDifferential Revision: https://reviews.llvm.org/D66351\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@369430 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "3eae253b1ac2da31d8928950dcb50956b64e81d1", "tree": "dbfa8de26d4b10cbd2dd1f1637132d97907007fa", "parents": [ "7875131e8d4ab26991ccdb0c26904b902d0ee49e" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:26:55 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:26:55 2019 +0000" }, "message": "[OpenMP] Enable warning about \"implicit fallthrough\"\n\nFix last warned location in ittnotify_static.cpp using the defined\nmacro KMP_FALLTHROUGH().\n\nDifferential Revision: https://reviews.llvm.org/D65871\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@369003 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "7875131e8d4ab26991ccdb0c26904b902d0ee49e", "tree": "e8574aafd56c184df41eae2873e40ed5a3aa5c8a", "parents": [ "b0cb65c5088c8ae7f0fee1c5f879994cd0317be2" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:26:41 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:26:41 2019 +0000" }, "message": "[OpenMP] Remove \u0027unnecessary parentheses\u0027\n\nThe variables in kmp_lock.cpp are really arrays of function pointers\nthat return void or int, not pointers to functions that return void*\nor int*. The other changes are only cosmetic.\n\nDifferential Revision: https://reviews.llvm.org/D65870\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@369002 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "b0cb65c5088c8ae7f0fee1c5f879994cd0317be2", "tree": "d44c73505f4891070f87193b5c476f2c1cdf99c7", "parents": [ "04b98c18752c76ede5eb20a1b108d542c7d46b12" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:26:29 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:26:29 2019 +0000" }, "message": "[OMPT] Resolve warnings because of ints in if conditions\n\nThe implementation status can only be one of\nompt_event_UNIMPLEMENTED \u003d ompt_set_never \u003d 1\nompt_event_MAY_ALWAYS \u003d ompt_set_always \u003d 5\n\nIn both cases, the condition was already true, so just remove\nthe check.\n\nDifferential Revision: https://reviews.llvm.org/D65869\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@369001 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "04b98c18752c76ede5eb20a1b108d542c7d46b12", "tree": "49ea3eeb278b113620769a23c6503bd90a15e0bb", "parents": [ "923dcc5544f82b1ec999b591dcc648fbdd448def" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:11:50 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Aug 15 13:11:50 2019 +0000" }, "message": "[OpenMP] Turn on -Wall compiler warnings by default\n\nInstead, maintain a list of disabled options to still build libomp and\nlibomptarget without warnings. This includes -Wno-error and -Wno-pedantic\nto silence warnings that LLVM enables when building in-tree.\n\nI tested the following compilers:\n * Clang 6.0, 7.0, 8.0\n * GCC 4.8.5 (CentOS 7), GCC 6, 7, 8, 9\n * Intel Compiler 16, 17, 18, 19\n\nRFC thread on openmp-dev mailing list:\nhttp://lists.llvm.org/pipermail/openmp-dev/2019-August/002668.html\n\nDifferential Revision: https://reviews.llvm.org/D65867\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368999 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "923dcc5544f82b1ec999b591dcc648fbdd448def", "tree": "6e6d629effc59fef88a7129c00ea36f0cb11bd68", "parents": [ "ee9c0c5007ed135b73b9590f37cd83cbf584de54" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Tue Aug 13 21:41:47 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Tue Aug 13 21:41:47 2019 +0000" }, "message": "Factor architecture dependent code out of loop.cu\n\nSummary:\n[libomptarget] Factor architecture dependent code out of loop.cu\n\nRelated to the patch series starting D64217. Added subscribers to said series as reviewers. This effort is smaller in scope.\n\nThis patch factors out just enough architecture dependent code from loop.cu to allow the same source to be used with amdgcn, given a different target_impl.h. Testing is that the same bitcode (modulo variable names) is generated for libomptarget before and after the refactor, for nvptx and the out of tree amdgcn.\n\nReviewers: jdoerfert, ABataev, bollu, jfb, tra, grokos, Hahnfeld, guansong, xtian, gregrodgers, ronlieb, hfinkel, gtbercea, guraypp, arpith-jacob\n\nReviewed By: jdoerfert, ABataev\n\nSubscribers: dexonsmith, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D65836\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368751 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ee9c0c5007ed135b73b9590f37cd83cbf584de54", "tree": "3d3276c24b6819725bc51ff3894d46eaeaad93ad", "parents": [ "dabebaf3ab75720dc9786a0a2c851eabf2a6b4e1" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon Aug 12 12:37:30 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon Aug 12 12:37:30 2019 +0000" }, "message": "Cleanup unused variable.\n\nThis patch fixes problem raised in post-review comments of the\nhttps://reviews.llvm.org/D65285. Developers of ittnotify confirmed\nthat dll_path_ptr field of the __itt_global structure is never used\nby ittnotify library, so it is safe to remove the dll_path array.\n\nDifferential Revision: https://reviews.llvm.org/D65885\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368559 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "dabebaf3ab75720dc9786a0a2c851eabf2a6b4e1", "tree": "38f126af2a77ba1697bd395960958cef00a65237", "parents": [ "9ea8071852391148611c7ac3da94a2af3e447c51" ], "author": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Fri Aug 09 21:32:57 2019 +0000" }, "committer": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Fri Aug 09 21:32:57 2019 +0000" }, "message": "[OpenMP][libomptarget] Add support for close map modifier\n\nSummary:\nThis patch adds support for the close map modifier.\n\nThe close map modifier will overwrite the unified shared memory requirement and create a device copy of the data.\n\nReviewers: ABataev, Hahnfeld, caomhin, grokos, jdoerfert, AlexEichenberger\n\nReviewed By: Hahnfeld, AlexEichenberger\n\nSubscribers: guansong, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D65340\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368488 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "9ea8071852391148611c7ac3da94a2af3e447c51", "tree": "197e67a16f19a8b6f16c12e8424d553f788a9bec", "parents": [ "3654abe1c43bb77b33278463bf892a7b1effac1f" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Fri Aug 09 19:20:39 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Fri Aug 09 19:20:39 2019 +0000" }, "message": "[libomptarget] Remove duplicate RTLRequiresFlags per device\n\nWe have one global RTLs.RequiresFlags, I don\u0027t see a need to make a\ncopy per device that the runtime manages. This was problematic anyway\nbecause the copy happened during the first __tgt_register_lib(). This\nmade it impossible to call __tgt_register_requires() from normal user\nfuntions for testing.\nHence, this change also fixes unified_shared_memory/shared_update.c for\nolder versions of Clang that don\u0027t call __tgt_register_requires() before\n__tgt_register_lib().\n\nDifferential Revision: https://reviews.llvm.org/D66019\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368465 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "3654abe1c43bb77b33278463bf892a7b1effac1f", "tree": "f98fbccb5871e0e129a9ddd606f80dcdb125f1f2", "parents": [ "2535f2b25498d11c34b4c3affa4985ae977fc0ac" ], "author": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Wed Aug 07 17:29:45 2019 +0000" }, "committer": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Wed Aug 07 17:29:45 2019 +0000" }, "message": "[OpenMP][libomptarget] Add support for unified memory for regular maps\n\nSummary:\nThis patch adds support for using unified memory in the case of regular maps that happen when a target region is offloaded to the device.\n\nFor cases where only a single version of the data is required then the host address can be used. When variables need to be privatized in any way or globalized, then the copy to the device is still required for correctness.\n\nReviewers: ABataev, jdoerfert, Hahnfeld, AlexEichenberger, caomhin, grokos\n\nReviewed By: Hahnfeld\n\nSubscribers: mgorny, guansong, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D65001\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368192 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "2535f2b25498d11c34b4c3affa4985ae977fc0ac", "tree": "8b0d8e2a8cf1031fa1d685b22b56d8611774632f", "parents": [ "c8294f02f99a6db8c0f3b5b0cf67c594d843d5cf" ], "author": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 07 15:24:12 2019 +0000" }, "committer": { "name": "Jon Chesterfield", "email": "jonathanchesterfield@gmail.com", "time": "Wed Aug 07 15:24:12 2019 +0000" }, "message": "Use forceinline. Necessary for nvcc to inline small functions within the bitcode library\n\nSummary:\n[libomptarget] Use forceinline. Necessary for nvcc to inline small functions within the bitcode library\nSuggested in D65836\n\nReviewers: ABataev, jdoerfert, grokos, gregrodgers\n\nSubscribers: openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D65876\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368177 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "c8294f02f99a6db8c0f3b5b0cf67c594d843d5cf", "tree": "129add6264b132e3e25fac0b30eb03dc4326e262", "parents": [ "3c7e9228cad7da4c5f148b788c59f2548df1b119" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Aug 06 18:13:39 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Aug 06 18:13:39 2019 +0000" }, "message": "[OPENMP][OFFLOADING]Fix the test, NFC.\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@368068 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "3c7e9228cad7da4c5f148b788c59f2548df1b119", "tree": "0bf0506e597e0638c2a2d63487346b1a61e42396", "parents": [ "6a2cf6907c3f80527e4fb04d5d4cc34c9c78f931" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Mon Aug 05 23:23:52 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Mon Aug 05 23:23:52 2019 +0000" }, "message": "[OpenMP] Add support for GOMP_*_nonmonotonic_* functions\n\nPatch by Isuru Fernando\n\nDifferential Revision: https://reviews.llvm.org/D65714\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367949 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "6a2cf6907c3f80527e4fb04d5d4cc34c9c78f931", "tree": "db30005e2f664fd82cfd9c97cb3595cb304f252e", "parents": [ "64bfa8497f298e773f9145fe7e8294eacdcfd7b0" ], "author": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Mon Aug 05 21:46:13 2019 +0000" }, "committer": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Mon Aug 05 21:46:13 2019 +0000" }, "message": "[OpenMP] Fix broken build due to new OMPT tests\n\nNew OMPT tests with teams construct should be disabled for GCC as it\nemits code with a GOMP entry not supported in the LLVM runtime.\n\nDifferential Revision: https://reviews.llvm.org/D65757\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367939 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "64bfa8497f298e773f9145fe7e8294eacdcfd7b0", "tree": "daa56f8d721257d3fdd01ec374ab2ddf9e61d59d", "parents": [ "e8bd614a37e220abc0cbc5f24cfea266526eecea" ], "author": { "name": "Michael Kruse", "email": "llvm@meinersbur.de", "time": "Mon Aug 05 19:12:10 2019 +0000" }, "committer": { "name": "Michael Kruse", "email": "llvm@meinersbur.de", "time": "Mon Aug 05 19:12:10 2019 +0000" }, "message": "[libomptarget] Harmonize emitting CUDA errors and general debug messages.\n\nEnsures that CUDA fail reasons (such as \"No CUDA-capable device detected\")\nare printed together with libomptarget\u0027s debug message\n(e.g. \"Error when setting CUDA context\"). Previously, the former was\nprinted only in CMAKE_BUILD_TYPE\u003dDebug builds while the latter was\nenabled by LIBOMPTARGET_ENABLE_DEBUG.\n\nWith this change, also only call cuGetErrorString when the error will be\nprinted.\n\nSuggested-by: Ye Luo \u003cxw111luoye@gmail.com\u003e\n\nDifferential Revision: https://reviews.llvm.org/D65687\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367910 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "e8bd614a37e220abc0cbc5f24cfea266526eecea", "tree": "61208ebcf970506f608c29d11d8724c2ea806a44", "parents": [ "716f6ba727f8f606255a39c84cd974f50d51c386" ], "author": { "name": "Michael Kruse", "email": "llvm@meinersbur.de", "time": "Sun Aug 04 04:18:28 2019 +0000" }, "committer": { "name": "Michael Kruse", "email": "llvm@meinersbur.de", "time": "Sun Aug 04 04:18:28 2019 +0000" }, "message": "[OpenMP 5.0] libomptarget interface for declare mapper functions.\n\nThis patch implements the libomptarget runtime interface for OpenMP 5.0\ndeclare mapper functions. The declare mapper functions generated by\nClang will call them to complete the mapping of members.\nkmpc_mapper_num_components gets the current number of components for a\nuser-defined mapper; kmpc_push_mapper_component pushes back one\ncomponent for a user-defined mapper.\n\nThe design slides can be found at\nhttps://github.com/lingda-li/public-sharing/blob/master/mapper_runtime_design.pptx\n\nPatch by Lingda Li \u003clildmh@gmail.com\u003e\n\nDifferential Revision: https://reviews.llvm.org/D60972\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367772 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "716f6ba727f8f606255a39c84cd974f50d51c386", "tree": "d0b033defcb2396167edc5282b33ce7b7ee7e2ec", "parents": [ "6cbafacdcf6066b4a4073d817a66ac13cb19fddd" ], "author": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Sat Aug 03 02:38:53 2019 +0000" }, "committer": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Sat Aug 03 02:38:53 2019 +0000" }, "message": "Add OMPT support for teams construct\n\nThis change adds OMPT support for events from teams construct.\n\nDifferential Revision: https://reviews.llvm.org/D64025\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367746 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "6cbafacdcf6066b4a4073d817a66ac13cb19fddd", "tree": "e50ef2c80406bfb7df9b5d7817b111f733ae1c75", "parents": [ "34b2830c8305adcd0710076127e8a817ba4ba4db" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Tue Jul 30 18:37:28 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Tue Jul 30 18:37:28 2019 +0000" }, "message": "[OpenMP] Rename last file to cpp and remove LIBOMP_CFLAGS\n\nAll other files are already C++ and the build system has always\npassed \u0027-x c++\u0027 for C files, effectively compiling them as C++.\n\nTo stay warning free we need one fix in ittnotify_static.{c,cpp}:\nThe variable dll_path can be written to, so it must not be const.\nGCC complained with -Wcast-qual and I think it\u0027s right.\n\nDifferential Revision: https://reviews.llvm.org/D65285\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367343 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "34b2830c8305adcd0710076127e8a817ba4ba4db", "tree": "b52997bd1c3636049ce1439450fd067111814003", "parents": [ "ca7d371ddfa08d131688282b20a2f678a4339b9f" ], "author": { "name": "Yi Kong", "email": "yikong@google.com", "time": "Thu Jul 25 22:29:55 2019 +0000" }, "committer": { "name": "Yi Kong", "email": "yikong@google.com", "time": "Thu Jul 25 22:29:55 2019 +0000" }, "message": "[openmp] Workaround bug in old Android pthread_attr_setstacksize\n\nRound the stack size to a multiple of the page size. Older versions of\nAndroid (until KitKat) would fail pthread_attr_setstacksize with\nEINVAL if the stack size was not a multiple of the page size.\n\nPatch by Dan Albert \u003cdanalbert@google.com\u003e.\n\nTest: Build, copied into the NDK, passed openmp test on ICS.\nBug: https://github.com/android-ndk/ndk/issues/9\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367070 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ca7d371ddfa08d131688282b20a2f678a4339b9f", "tree": "23953088110867a0a00dd76b53fa7e3c3037afae", "parents": [ "b9b98019a08fabd76e4d5b3f161702c3696ccd33" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Jul 25 17:51:24 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Jul 25 17:51:24 2019 +0000" }, "message": "[OpenMP] Fix build of stubs library, NFC.\n\nBoth Clang and GCC complained that they cannot initialize a return\nobject of type \u0027kmp_proc_bind_t\u0027 with an \u0027int\u0027. While at it, also\nfix a warning about missing parentheses thrown by Clang.\n\nDifferential Revision: https://reviews.llvm.org/D65284\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367041 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "b9b98019a08fabd76e4d5b3f161702c3696ccd33", "tree": "f99123c27841072831cf76c7ec34d912fb3c8b3b", "parents": [ "57b2fafae6ea3c35e4acf5a520a0e5f14f71ba13" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu Jul 25 15:02:28 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu Jul 25 15:02:28 2019 +0000" }, "message": "[OPENMP][NVPTX]Perform memory flush if number of threads to sync is 1 or less.\n\nSummary:\nAccording to the OpenMP standard, barrier operation must perform\nimplicit flush operation. Currently, if there is only one thread in the\nteam, barrier does not flush the memory. Patch fixes this problem.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jdoerfert, openmp-commits, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D62398\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367024 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "57b2fafae6ea3c35e4acf5a520a0e5f14f71ba13", "tree": "ecf941df08d8b50b06016013fad4490feb3ec47f", "parents": [ "b840a856c2cc27c1aa6db219a095a11d278344c5" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Jul 25 14:36:20 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Jul 25 14:36:20 2019 +0000" }, "message": "[OpenMP] RISCV64 port\n\nThis is a port of libomp for the RISC-V 64-bit Linux target.\n\nWe have tested this port on a HiFive Unleashed development board\nusing a downstream LLVM that has support for the missing bits in\nupstream. As of now, all tests are passing, including OMPT.\n\nPatch by Ferran Pallarès!\n\nDifferential Revision: https://reviews.llvm.org/D59880\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@367021 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "b840a856c2cc27c1aa6db219a095a11d278344c5", "tree": "d4bf009f854f517ca552ec3c0f63a6268bd0943a", "parents": [ "9ad60c0ccd9fe0b733850895d259881124748797" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Tue Jul 23 14:20:48 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Tue Jul 23 14:20:48 2019 +0000" }, "message": "[libomptarget] Handle offload policy in push_tripcount\n\nIf the first target region in a program calls the push_tripcount\nfunction, libomptarget didn\u0027t handle the offload policy correctly.\nThis could lead to unexpected error messages as seen in\nhttp://lists.llvm.org/pipermail/openmp-dev/2019-June/002561.html\n\nTo solve this, add a check calling IsOffloadDisabled() as all other\nentry points already do. If this method returns false, libomptarget\nis effectively disabled.\n\nDifferential Revision: https://reviews.llvm.org/D64626\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366810 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "9ad60c0ccd9fe0b733850895d259881124748797", "tree": "f2b706b01b95fb4adff0190a80bc11f16a726e14", "parents": [ "8dabfbaf5111356cdcb7a2dc4486eafbab6cfe4c" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Mon Jul 22 18:46:02 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Mon Jul 22 18:46:02 2019 +0000" }, "message": "[OMPT] Cleanup reset of exit_frame pointer\n\nThis is done at call-site and does not need to be handled in\n__kmp_invoke_microtask. It was already absent from the x86\nand x86_64 assembly, this patch removes it from the generic\nimplementation in z_Linux_util.cpp and adds documentation for\nAArch64 and PPC64 that it\u0027s actually not needed. I can\u0027t test\non these architectures, so I don\u0027t want to change the code just\nbecause it looks right :)\n\nWhile at it, rename some variables for consistency and add a\ncheck in test/ompt/parallel/normal.c that the pointer was reset\nbefore entering the barrier.\n\nDifferential Revision: https://reviews.llvm.org/D64442\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366721 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "8dabfbaf5111356cdcb7a2dc4486eafbab6cfe4c", "tree": "395317ff4dae4778e7258ca45169589a3a4c1454", "parents": [ "b0376454faefb83111b7e9058985e154b1725413" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Mon Jul 22 18:11:06 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Mon Jul 22 18:11:06 2019 +0000" }, "message": "Delete empty file\n\nThis is a left-over from r356288 which was reviewed in D58989.\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366716 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "b0376454faefb83111b7e9058985e154b1725413", "tree": "0e2e8417047e70ebc5df6da5148815fb182354a4", "parents": [ "c1b1ef1c009f285690b40a96f0853ec2b1573fbb" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu Jul 18 13:54:01 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu Jul 18 13:54:01 2019 +0000" }, "message": "[OpenMP][libomptarget] Suppress C++ 11 related warnings when building libomptarget-nvptx bitcode library, by Doru Bercea.\n\nSummary: Pass -std\u003dc++11 flag to compiler to suppress C++ 11 related warnings when building NVPTX bitcode library.\n\nReviewers: ABataev, caomhin, Hahnfeld\n\nReviewed By: ABataev, Hahnfeld\n\nSubscribers: jdoerfert, Hahnfeld, jholewinski, mgorny, guansong, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D55772\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366438 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "c1b1ef1c009f285690b40a96f0853ec2b1573fbb", "tree": "f3cdcd90d34ce959dde11e010742da176697bb86", "parents": [ "07f4283e71aa62bc8c2c5755985c7b850276b90f" ], "author": { "name": "Ron Lieberman", "email": "ronlieb.g@gmail.com", "time": "Wed Jul 17 17:07:52 2019 +0000" }, "committer": { "name": "Ron Lieberman", "email": "ronlieb.g@gmail.com", "time": "Wed Jul 17 17:07:52 2019 +0000" }, "message": "[OPENMP] Resolve lost LoopTripCnt for subsequent loops in same thread.\n\nRemove loopTripCnt from threaded device stack after consuming it.\nAdded a libomptarget DP message to aid in future debugging and to\nvalidate the added testcase, which only runs in Debug build.\n\nDifferential Revision: https://reviews.llvm.org/D64808\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366349 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "07f4283e71aa62bc8c2c5755985c7b850276b90f", "tree": "ea6adc29c94fc03849d336782b9fbc26461e94d0", "parents": [ "cff3557e55737a09e4b4adfc1dc41205bb12dba3" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Wed Jul 17 15:41:00 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Wed Jul 17 15:41:00 2019 +0000" }, "message": "Remove REQUIRES OMP spec version within lit tests\n\nThis is a follow up patch to D64534 (r365963) which removed all OMP\nspec versioning within the OpenMP runtime codebase. This patch removes\nREQUIRES: openmp-x.y lines from lit tests.\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366341 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "cff3557e55737a09e4b4adfc1dc41205bb12dba3", "tree": "74262d732cfdc74a453f82eee43d5a49ddfc6df6", "parents": [ "0519a2c6b21e1cec91c1d2fe1434054f718a3137" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Tue Jul 16 17:16:43 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Tue Jul 16 17:16:43 2019 +0000" }, "message": "[OpenMP] Move header inclusion out of \u0027extern \"C\"\u0027\n\nThis leads to problems when compiling C++ code with libc++ for Nvidia GPUs\nbecause Clang now uses wrappers for math functions that might include\nC++ templates not allowed in \u0027extern \"C\"\u0027.\n\nDifferentiel Revision: https://reviews.llvm.org/D64625\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366229 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "0519a2c6b21e1cec91c1d2fe1434054f718a3137", "tree": "8026f1e32f68e8d7f4f08dcf21261c4f7a9401bd", "parents": [ "946967528b804d3de059b9ad6fbe7e22ebbdd9bb" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Jul 16 16:07:10 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Jul 16 16:07:10 2019 +0000" }, "message": "[OPENMP][NVPTX]Fixed checks for cuda versions.\n\nSummary:\nWe used CUDART_VERSION macro to check for the installed cuda version\nbut this macro is defined in cuda_runtime_api.h, which is not used by\nproject. Better to use CUDA_VERSION macro, which is defined in cuda.h.\nAlso, added the check if this macro is defined. If macro is undefined,\nthere is something wrong with the cuda configuration and we should not\ncontinue the compilation.\nThis also fixes problems with runtime building in cuda 10+.\n\nReviewers: grokos\n\nSubscribers: guansong, jdoerfert, caomhin, kkwli0, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D64648\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366224 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "946967528b804d3de059b9ad6fbe7e22ebbdd9bb", "tree": "23a2141e26b3b950b8d72c87320b39fb49992db1", "parents": [ "d4b114904a3bb0ee08a7833c52450b793d6c7781" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Jul 16 15:51:32 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Tue Jul 16 15:51:32 2019 +0000" }, "message": "[OPENMP]Fix threadid in __kmpc_omp_taskwait call for dependent target calls.\n\nSummary:\nWe used to call __kmpc_omp_taskwait function with global threadid set to\n0. It may crash the application at the runtime if the thread executing\n target region is not a master thread.\n\nReviewers: grokos, kkwli0\n\nSubscribers: guansong, jdoerfert, caomhin, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D64571\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@366220 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "d4b114904a3bb0ee08a7833c52450b793d6c7781", "tree": "0caa40955997edd9c35d09db5a24defee3732836", "parents": [ "bb2634a97ed88cd542993339e10bbc98ffa37e41" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Fri Jul 12 21:45:36 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Fri Jul 12 21:45:36 2019 +0000" }, "message": "[OpenMP] Remove OMP spec versioning\n\nRemove all older OMP spec versioning from the runtime and build system.\n\nPatch by Terry Wilmarth\n\nDifferential Revision: https://reviews.llvm.org/D64534\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@365963 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "bb2634a97ed88cd542993339e10bbc98ffa37e41", "tree": "bb7045635447b49a4591134d8bc0eab76a460511", "parents": [ "8e626ba79d528b118f922b8a40634a36c1f06438" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Fri Jul 12 10:21:55 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Fri Jul 12 10:21:55 2019 +0000" }, "message": "[libomptarget] Fix typos and grammar in error messages, NFC.\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@365890 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "8e626ba79d528b118f922b8a40634a36c1f06438", "tree": "ed260273383f78717d38c38631891846e406ce9e", "parents": [ "201ba9ed4cda7c2407a525b95eaea115cbc06981" ], "author": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Jul 11 20:12:51 2019 +0000" }, "committer": { "name": "Jonas Hahnfeld", "email": "hahnjo@hahnjo.de", "time": "Thu Jul 11 20:12:51 2019 +0000" }, "message": "[libomptarget-nvptx] Remove dead functions\n\nThese entry points are never called by Clang trunk nor clang-ykt. If\nXL doesn\u0027t use them either, they can finally go away.\n\nDifferential Revision: https://reviews.llvm.org/D52700\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@365817 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "201ba9ed4cda7c2407a525b95eaea115cbc06981", "tree": "d00cc4741f4ad4c02a620c0e86ebc3c78580f1e9", "parents": [ "3a22f66a9bd7435d012f259cf434901ae5f2faba" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jul 10 15:09:37 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jul 10 15:09:37 2019 +0000" }, "message": "NFC: fixed typo #ifdef --\u003e #if to allow macro set to 0 work correctly\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@365642 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "3a22f66a9bd7435d012f259cf434901ae5f2faba", "tree": "5edc6b088a688eba9156dc273e38fbae5395a72c", "parents": [ "0b1328156a6e3127fe0964426bb3e1cd5ec6bcdd" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon Jul 08 18:45:48 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon Jul 08 18:45:48 2019 +0000" }, "message": "[OPENMP]Rename loopTripCnt member data to LoopTripCnt, NFC.\n\nRename variable to follow LLVM coding standard.\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@365368 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "0b1328156a6e3127fe0964426bb3e1cd5ec6bcdd", "tree": "12139df3a0b81cbaf3ff1f5caaa508ae8db8d7c0", "parents": [ "bbbee1dd60a97871f4b6cafd381d44bb32c9aa0e" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon Jul 08 15:30:23 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon Jul 08 15:30:23 2019 +0000" }, "message": "[OPENMP]Make __kmpc_push_tripcount thread safe.\n\nSummary:\n__kmpc_push_tripcount function is not thread safe and may lead to data\nrace when the target regions are executed in parallel threads. The patch\nmakes loopTripCnt counter thread aware and stores the tripcount value\nper thread in the map. Access to map is guarded by mutex to prevent\ndata race in the map itself.\nTest is for NVPTX target because it does not work correctly on the\nhost. Seems to me, there is a problem in libomp with target regions in\nthe parallel threads.\n\nReviewers: grokos\n\nSubscribers: guansong, jfb, jdoerfert, openmp-commits, kkwli0, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D64080\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@365332 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "bbbee1dd60a97871f4b6cafd381d44bb32c9aa0e", "tree": "71891fae12db8eea083115813ddbb129409142b5", "parents": [ "f9ad1a04db542c46a3650d0090321a4e06fa8be3" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Jul 02 15:10:20 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Jul 02 15:10:20 2019 +0000" }, "message": "Create a runtime option to disable task throttling.\n\nPatch by viroulep (Philippe Virouleau)\n\nDifferential Revision: https://reviews.llvm.org/D63196\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@364934 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "f9ad1a04db542c46a3650d0090321a4e06fa8be3", "tree": "35f64df6b29dba7a086463e9bd0325e07e7d4bd6", "parents": [ "a31ae2e44a8f91ca20aa03319603829d994a1635" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Jul 02 13:45:40 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Tue Jul 02 13:45:40 2019 +0000" }, "message": "Cleanup of unused code\n\nPatch by Terry Wilmarth\n\nDifferential Revision: https://reviews.llvm.org/D63891\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@364925 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "a31ae2e44a8f91ca20aa03319603829d994a1635", "tree": "0e6fc882cb8358b6a15396206b3e415e679aed68", "parents": [ "0afc673b1708437b5d9ff8ce0d50b41ab3b776b8" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu Jun 27 18:33:09 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu Jun 27 18:33:09 2019 +0000" }, "message": "[OPENMP][NVPTX]Relax flush directive.\n\nSummary:\nAccording to the OpenMP standard, flush makes a thread’s temporary view of memory consistent with memory and enforces an order on the memory operations of the variables explicitly specified or implied.\n\nAccording to the Cuda toolkit documentation (https://docs.nvidia.com/cuda/archive/8.0/cuda-c-programming-guide/index.html#memory-fence-functions), __threadfence() functions provides required functionality.\n\n__threadfence_system() also provides required functionality, but it also\nincludes some extra functionality, like synchronization of page-locked\nhost memory, synchronization for the host, etc. It is not required per\nthe standard and we can use more relaxed version of memory fence\noperation.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jfb, jdoerfert, openmp-commits, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D62397\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@364572 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "0afc673b1708437b5d9ff8ce0d50b41ab3b776b8", "tree": "9632b0f0ef712f0388f72aebd1c5d31cfc1cc307", "parents": [ "4df92eda05cd22bfd4a97f4ce821b97cad57f4f7" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jun 26 18:11:26 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jun 26 18:11:26 2019 +0000" }, "message": "Fixed memory use-after-free problem.\n\nBug reported in https://bugs.llvm.org/show_bug.cgi?id\u003d42269.\nFreeing of the contention group (CG) stucture by master thread looks wrong,\nbecause workers can leave the CG later on. Intead the freeing\nis now done by the last thread leaving the CG.\n\nDifferential Revision: https://reviews.llvm.org/D63599\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@364456 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "4df92eda05cd22bfd4a97f4ce821b97cad57f4f7", "tree": "d5fed73832681f872073c8e494a4e5443be2080b", "parents": [ "79d5d6d37a486e8670aa89932df971d4cf5123c7" ], "author": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Wed Jun 19 15:48:10 2019 +0000" }, "committer": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Wed Jun 19 15:48:10 2019 +0000" }, "message": "[OpenMP][libomptarget] Add support for declare target to clause under unified memory\n\nSummary:\nThis patch adds support for handling variables under the:\n\n```\n#pragma omp declare target to()\n```\n\nclause when the \n\n```\n#pragma omp requires unified_shared_memory\n```\n\nis used.\n\nThe address of the host variable is copied into the device pointer just like for the declare target link case.\n\nReviewers: ABataev, caomhin, grokos, AlexEichenberger\n\nReviewed By: grokos\n\nSubscribers: jcownie, guansong, jdoerfert, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D63106\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@363825 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "79d5d6d37a486e8670aa89932df971d4cf5123c7", "tree": "6074a627e16d15be4743b9a42dcce405521e3bfb", "parents": [ "3f381f546ec9b065f9133d1fcd5d2711affb646a" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Wed Jun 19 14:20:34 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Wed Jun 19 14:20:34 2019 +0000" }, "message": "[OPENMP][CUDA]Use __syncthreads when compiled by nvcc and clang \u003e\u003d 9.0.\n\nSummary:\nThe problems with __syncthreads() were fixed in clang \u003e\u003d 9.0 and the\noriginal __syncthreads() can be used instead of the ptx instruction.\n\nReviewers: grokos\n\nSubscribers: guansong, jdoerfert, openmp-commits, kkwli0, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D63515\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@363807 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "3f381f546ec9b065f9133d1fcd5d2711affb646a", "tree": "40673bbd358f75971676016138225056a87d331c", "parents": [ "e9791d7f3768e71fbae69ac21317e3edd0382b0d" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jun 19 13:23:28 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jun 19 13:23:28 2019 +0000" }, "message": "New implementation of OpenMP 5.0 detached tasks.\n\nPatch by Alex Duran\n\nDifferential Revision: https://reviews.llvm.org/D62485\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@363799 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "e9791d7f3768e71fbae69ac21317e3edd0382b0d", "tree": "a3fea99e3d44618baa3808a6cf3153ea66254451", "parents": [ "402c7439d7ba1cfa49e205554dfa8087a5a64313" ], "author": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Fri Jun 14 20:15:15 2019 +0000" }, "committer": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Fri Jun 14 20:15:15 2019 +0000" }, "message": "[OpenMP] Add task alloc function\n\nSummary: Add the target task allocation function to the interface.\n\nReviewers: ABataev, AlexEichenberger, caomhin, jlpeyton, AndreyChurbanov, RaviNarayanaswamy, hbae\n\nReviewed By: AlexEichenberger, hbae\n\nSubscribers: hbae, RaviNarayanaswamy, cfe-commits, Hahnfeld, guansong, jdoerfert, openmp-commits\n\nTags: #openmp, #clang\n\nDifferential Revision: https://reviews.llvm.org/D63010\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@363449 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "402c7439d7ba1cfa49e205554dfa8087a5a64313", "tree": "0e71aaa42cca5894240f1e02d1097614979ce8cf", "parents": [ "dba741d61d3b460f090810bf14ec2f6cd5ecc32d" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jun 05 16:14:47 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed Jun 05 16:14:47 2019 +0000" }, "message": "Added propagation of not big initial stack size of master thread to workers.\n\nCurrently implemented only for non-Windows 64-bit platforms.\n\nDifferential Revision: https://reviews.llvm.org/D62488\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@362618 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "dba741d61d3b460f090810bf14ec2f6cd5ecc32d", "tree": "e56155421cec60ea8dcc12a8d4ea426b07d37ad2", "parents": [ "6973da5c32719dead93113061a40fccbc7e8f662" ], "author": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Tue Jun 04 15:05:53 2019 +0000" }, "committer": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Tue Jun 04 15:05:53 2019 +0000" }, "message": "[OpenMP][libomptarget] Enable usage of unified memory for declare target link variables\n\nSummary: This patch enables the usage of a host variable on the device for declare target link variables when unified memory is available.\n\nReviewers: ABataev, caomhin, grokos\n\nReviewed By: grokos\n\nSubscribers: Hahnfeld, guansong, jdoerfert, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D60884\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@362505 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "6973da5c32719dead93113061a40fccbc7e8f662", "tree": "117cabf62e7f795efd1e882d4548de9e396483ac", "parents": [ "74db2a4213dd1f4d7b0056ebecbce6cfca3d3e30" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon Jun 03 14:21:59 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon Jun 03 14:21:59 2019 +0000" }, "message": "Fixed build warning with -DLIBOMP_USE_HWLOC\u003d1\n\nMade type of depth of hwloc object to correapond with\nchange from unsigned in hwloc 1,x to int in hwloc 2.x.\nThis eliminates the warning on signed-unsigned comparison.\n\nDifferential Revision: https://reviews.llvm.org/D62332\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@362401 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "74db2a4213dd1f4d7b0056ebecbce6cfca3d3e30", "tree": "00ae7822eef2c07552d4614911e5920400f19919", "parents": [ "48eeef98d8147c59f9db18c8d7e646741c28f738" ], "author": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Thu May 30 18:35:07 2019 +0000" }, "committer": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Thu May 30 18:35:07 2019 +0000" }, "message": "Fix OMP_TARGET_OFFLOAD parsing\n\nCurrent parsing allows trailing string after the permitted value,\nMANDATORY|DISABLED|DEFAULT -- e.g., \"mandatorynot\" is also recognized\nas \"MANDATORY\". Such cases should be recognized as incorrect/unknown\nvalue.\n\nDifferential Revision: https://reviews.llvm.org/D62431\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@362125 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "48eeef98d8147c59f9db18c8d7e646741c28f738", "tree": "52c0b48d00681a133c59445c3a524080fb368e23", "parents": [ "93fe5cefbde9bfced3c6577219a98791ff7e3481" ], "author": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Thu May 30 16:32:20 2019 +0000" }, "committer": { "name": "Hansang Bae", "email": "hansang.bae@intel.com", "time": "Thu May 30 16:32:20 2019 +0000" }, "message": "Add checks before pointer dereferencing\n\nThis change adds checks before dereferencing a pointer returned from a\nfunction.\n\nDifferential Revision: https://reviews.llvm.org/D62224\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@362111 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "93fe5cefbde9bfced3c6577219a98791ff7e3481", "tree": "44ce235f8a57f650cd1ebd6b8e1bce38da428997", "parents": [ "34def82866d4eb9c52a0ae65aa2abefbc4a923c5" ], "author": { "name": "Michal Gorny", "email": "mgorny@gentoo.org", "time": "Tue May 28 14:10:47 2019 +0000" }, "committer": { "name": "Michal Gorny", "email": "mgorny@gentoo.org", "time": "Tue May 28 14:10:47 2019 +0000" }, "message": "[openmp] [test] Skip kernel-breaking tests on NetBSD\n\nThe omp_taskloop_num_tasks and omp_taskwait have deadlooped\non the NetBSD buildbot previously, practically hanging the host running\nit. Disable them until we can find a good solution, or make the kernel\nless fragile.\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361825 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "34def82866d4eb9c52a0ae65aa2abefbc4a923c5", "tree": "a349541443c5aaab055383ae1eaf93f8bef567a4", "parents": [ "926938a039dbca89d50348b06cf7cda4924c60f8" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 24 14:06:47 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 24 14:06:47 2019 +0000" }, "message": "Revert \"[OPENMP][NVPTX]Fix barriers and parallel level counters, NFC.\"\n\nThis reverts commit r361421 to split the patch into 3 parts.\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361638 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "926938a039dbca89d50348b06cf7cda4924c60f8", "tree": "598c3dd691ace052d265ea91935f19505dd54894", "parents": [ "f8e8d598bf8b2725baf87892119e6952819fce34" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Wed May 22 19:50:32 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Wed May 22 19:50:32 2019 +0000" }, "message": "[OPENMP][NVPTX]Fix barriers and parallel level counters, NFC.\n\nSummary:\nParallel level counter should be volatile to prevent some dangerous\noptimiations by the ptxas. Otherwise, ptxas optimizations lead to\nundefined behaviour in some cases.\nAlso, use __threadfence() for #pragma omp flush and if the barrier\nshould not be used (we have only one thread in the team), still perform\nflush operation since the standard requires implicit flush when\nexecuting barriers.\n\nReviewers: gtbercea, kkwli0, grokos\n\nSubscribers: guansong, jfb, jdoerfert, openmp-commits, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D62199\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361421 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "f8e8d598bf8b2725baf87892119e6952819fce34", "tree": "a349541443c5aaab055383ae1eaf93f8bef567a4", "parents": [ "74813230782d697dab344b00b0db3788593237ee" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed May 22 16:48:05 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed May 22 16:48:05 2019 +0000" }, "message": "Fixed third issue reported in https://bugs.llvm.org/show_bug.cgi?id\u003d41584.\n\nRemoved wrong debug assertion.\n\nDifferential Revision: https://reviews.llvm.org/D62251\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361408 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "74813230782d697dab344b00b0db3788593237ee", "tree": "814f71988b52b4fc5b7743cd4763d9196d370905", "parents": [ "9994ae3c1c21dd4bbc57bb8f491f0a3cbc198336" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue May 21 20:51:05 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue May 21 20:51:05 2019 +0000" }, "message": "[OpenMP] Add implementation to two OMPT API routines\n\nThis change adds implementation to ompt_finalize_tool() and\nompt_get_task_memory().\n\nPatch by Hansang Bae\n\nDifferential Revision: https://reviews.llvm.org/D61657\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361309 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "9994ae3c1c21dd4bbc57bb8f491f0a3cbc198336", "tree": "6f5cb694097ccc5631d83d9fedffa9455ae53545", "parents": [ "665d5c5112ab6ee1cba56733ec3a1b37e86aa55b" ], "author": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Tue May 21 19:35:02 2019 +0000" }, "committer": { "name": "Gheorghe-Teodor Bercea", "email": "gheorghe-teod.bercea@ibm.com", "time": "Tue May 21 19:35:02 2019 +0000" }, "message": "[OpenMP][libomptarget] Enable requires flags for target libraries.\n\nSummary:\nTarget link variables are currently implemented by creating a copy of the variables on the device side and unified memory never gets exploited.\n\nWhen the prgram uses the:\n\n```\n#pragma omp requires unified_shared_memory\n```\n\ndirective in conjunction with a declare target link, the linked variable is no longer allocated on the device and the host version is used instead.\n\nThis behavior is overridden by performing an explicit mapping.\n\nA Clang side patch is required.\n\nReviewers: ABataev, AlexEichenberger, grokos, Hahnfeld\n\nReviewed By: AlexEichenberger, grokos, Hahnfeld\n\nSubscribers: Hahnfeld, jfb, guansong, jdoerfert, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D60223\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361294 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "665d5c5112ab6ee1cba56733ec3a1b37e86aa55b", "tree": "5748ec44f2f0eb2dab358c757e72b16a0f7dd194", "parents": [ "dac3432cdb45db48112a0a63478ae94a184b5d45" ], "author": { "name": "Joachim Protze", "email": "protze@itc.rwth-aachen.de", "time": "Mon May 20 14:21:42 2019 +0000" }, "committer": { "name": "Joachim Protze", "email": "protze@itc.rwth-aachen.de", "time": "Mon May 20 14:21:42 2019 +0000" }, "message": "[OpenMP][OMPT] Fix locking testcases for 32 bit architectures\n\nhttps://reviews.llvm.org/D58454 did not fix the problem for a typical use\ncase of building LLVM with gcc or icc and then testing with the newly built\nclang compiler.\nThe compilers do not agree on how to extend a 32-bit pointer to uint64, so\nmake the pointer unsigned first, before adjusting the size.\n\nPatch by Joachim Protze\n\nDifferential Revision: https://reviews.llvm.org/D58506\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361158 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "dac3432cdb45db48112a0a63478ae94a184b5d45", "tree": "3bece2569825eb13b59c28341ef0ed3919052edd", "parents": [ "ba56714719294ad7aa15a1351c201e642445e2ab" ], "author": { "name": "Joachim Protze", "email": "protze@itc.rwth-aachen.de", "time": "Mon May 20 14:21:36 2019 +0000" }, "committer": { "name": "Joachim Protze", "email": "protze@itc.rwth-aachen.de", "time": "Mon May 20 14:21:36 2019 +0000" }, "message": "[OMPT] Handling of the events of initial-task-begin and initial-task-end\n\nOpenMP 5.0 says that the callback for the events initial-task-begin and\ninitial-task-end has to be ompt_callback_implicit_task.\n\nPatch by Tim Cramer\n\nDifferential Revision: https://reviews.llvm.org/D58776\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@361157 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ba56714719294ad7aa15a1351c201e642445e2ab", "tree": "f41aa105ca07416199aa3f533efccb1792023ec0", "parents": [ "9a0d0f00e90c4f9e0cbd281ab5578db28fafdfcb" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Thu May 16 17:52:53 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Thu May 16 17:52:53 2019 +0000" }, "message": "Fixed second issue reported in https://bugs.llvm.org/show_bug.cgi?id\u003d41584.\n\nAdded synchronization for possible concurrent initialization of mutexes\nby multiple threads. The need of synchronization caused by commit r357927\nwhich added the use of mutexes at threads movement to/from common pool\n(earlier the mutexes were used only at suspend/resume).\n\nPatch by Johnny Peyton.\n\nDifferential Revision: https://reviews.llvm.org/D61995\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360919 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "9a0d0f00e90c4f9e0cbd281ab5578db28fafdfcb", "tree": "4ff648af7974c22d23bcbe88361cbbb3a2e6abbd", "parents": [ "a86b5e4738990f8eeaff4ad1ee05b1af42bc34ea" ], "author": { "name": "Paul Osmialowski", "email": "pawel.osmialowski@arm.com", "time": "Thu May 16 13:16:24 2019 +0000" }, "committer": { "name": "Paul Osmialowski", "email": "pawel.osmialowski@arm.com", "time": "Thu May 16 13:16:24 2019 +0000" }, "message": "Fix hwloc topology traversal code unable to handle situation where L2 cache is common for the packages\n\nCurrently cores within package that share the same L2 cache are grouped together.\nThe current logic behind this assumes that the L2 cache is always at deeper\n(or the same) level than the package itself. In case when L2 cache is common\nfor all packages (and the packages are at deeper level than L2 cache) the whole of\nthe further topology discovery fails to find any computational units resulting in\nfollowing assertion:\n\nAssertion failure at kmp_affinity.cpp(715): nActiveThreads \u003d\u003d __kmp_avail_proc.\nOMP: Error #13: Assertion failure at kmp_affinity.cpp(715).\n\nThis patch adds a bit of a logic that prevents such situation from occurring.\n\nDifferential Revision: https://reviews.llvm.org/D61796\n\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360890 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "a86b5e4738990f8eeaff4ad1ee05b1af42bc34ea", "tree": "6b49996237ec173bbfe05c71f1042955b192311d", "parents": [ "0a3b7b60f61f1dbfd56aa5fe1bd613388788417c" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed May 15 16:53:45 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed May 15 16:53:45 2019 +0000" }, "message": "Fixed https://bugs.llvm.org/show_bug.cgi?id\u003d41584.\n\nRemoved unconditional and unsafe decrement of counter \nof active threads in pool at shutdown time.\n\nDifferential Revision: https://reviews.llvm.org/D61944\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360784 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "0a3b7b60f61f1dbfd56aa5fe1bd613388788417c", "tree": "1264797f93f4dc3e27137777b2727a4e7a6c38e5", "parents": [ "c1985baed25b1a7dd45412cb6cc62c32ac5743d0" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed May 15 13:45:36 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Wed May 15 13:45:36 2019 +0000" }, "message": "Introduce new OpenMP 5.0 depend object type.\n\nThe implementation should be done by compiler, user can only declare\nobjects of this type and use them in OpenMP directives.\n\nDifferential Revision: https://reviews.llvm.org/D61860\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360774 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "c1985baed25b1a7dd45412cb6cc62c32ac5743d0", "tree": "a636138d603fb2fef093dc98a5bd8e5846fe7d9c", "parents": [ "8db7163070b885e918ab79314943bc3c1df12a7e" ], "author": { "name": "Eli Friedman", "email": "efriedma@quicinc.com", "time": "Tue May 14 21:44:54 2019 +0000" }, "committer": { "name": "Eli Friedman", "email": "efriedma@quicinc.com", "time": "Tue May 14 21:44:54 2019 +0000" }, "message": "[OpenMP][AArch64] Fix compile with LLVM trunk.\n\nThe code is currently using the ambiguous instruction\n\"sub sp, sp, w9, lsl #4\". The ARM reference manual says this isn\u0027t\nvalid, and it\u0027s not clear whether it\u0027s supposed to mean uxtw or uxtx.\n\nIt doesn\u0027t matter which instruction we use here, since the high\nbits of the operand are zero anyway, so I arbitrarily choose uxtw, to\npreserve the register name.\n\nSee https://reviews.llvm.org/D60840 for the LLVM patch.\n\nDifferential Revision: https://reviews.llvm.org/D61770\n\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360711 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "8db7163070b885e918ab79314943bc3c1df12a7e", "tree": "c0c42ce77f916332ddeaa00f827b2082fd7e6e91", "parents": [ "fe501ad8d1dd02a9f6ab7bbe6b62beaf695097cf" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon May 13 17:04:32 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon May 13 17:04:32 2019 +0000" }, "message": "fixed typo made by commit r360595\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360602 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "fe501ad8d1dd02a9f6ab7bbe6b62beaf695097cf", "tree": "d9d1480f0b9836bd7484da039ad7a42d2bc35f2c", "parents": [ "d5bb0a8d360f1c053b1397618fbf9f8877365cec" ], "author": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon May 13 16:07:37 2019 +0000" }, "committer": { "name": "Andrey Churbanov", "email": "Andrey.Churbanov@intel.com", "time": "Mon May 13 16:07:37 2019 +0000" }, "message": "Fixed creation of aliases in Windows build.\n\nChanged file extension of the destination of the copy of libomp.lib\n(it was mistakely .dll, now it is .lib) in installation on Windows.\n\nDifferential Revision: https://reviews.llvm.org/D61673\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360595 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "d5bb0a8d360f1c053b1397618fbf9f8877365cec", "tree": "11691459bc764d975dbb3645eabeccca38270253", "parents": [ "d140b83ae9d0526ccc8339a2b845287db728cfc7" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon May 13 14:21:46 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Mon May 13 14:21:46 2019 +0000" }, "message": "[OPENMP][NVPTX]Simplify handling of thread limit, NFC.\n\nSummary:\nPatch improves performance of the full runtime mode by moving\nthreads limit counter to the shared memory. It also allows to save\nglobal memory.\n\nReviewers: grokos, kkwli0, gtbercea\n\nSubscribers: guansong, jdoerfert, openmp-commits, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D61801\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360584 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "d140b83ae9d0526ccc8339a2b845287db728cfc7", "tree": "0f72a13c61a4c154237fb61245c40be8133ce375", "parents": [ "e55ab9929284e9fa908126748b79385dc56be764" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 10 18:56:05 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 10 18:56:05 2019 +0000" }, "message": "[OPENMP][NVPTX]Improve number of threads counter, NFC.\n\nSummary:\nPatch improves performance of the full runtime mode by moving\nnumber-of-threads counter to the shared memory. It also allows to save\nglobal memory.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jfb, jdoerfert, openmp-commits, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D61785\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360457 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "e55ab9929284e9fa908126748b79385dc56be764", "tree": "de86609e4eac30e4d2744e78da4e852ad1caea74", "parents": [ "aebe831622403337fabe44206f5d497e066c761e" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Wed May 08 23:12:31 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Wed May 08 23:12:31 2019 +0000" }, "message": "[OpenMP] Workaround gfortran bugzilla build bug 41755\n\nThis patch provides workaround to allow gfortran to compile the\nOpenMP Fortran modules.\n\nFrom the gfortran manual:\nhttps://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BOZ-literal-constants.html\n\n\"Note that initializing an INTEGER variable with a statement such as\nDATA i/Z\u0027FFFFFFFF\u0027/ will give an integer overflow error rather than the desired\nresult of -1 when i is a 32-bit integer on a system that supports 64-bit\nintegers. The -fno-range-check option can be used as a workaround for legacy\ncode that initializes integers in this manner.\"\n\nBug filed: https://bugs.llvm.org/show_bug.cgi?id\u003d41755\n\nDifferential Revision: https://reviews.llvm.org/D61603\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360299 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "aebe831622403337fabe44206f5d497e066c761e", "tree": "a860bd2a91b1a6217770f9b96fcead514c24f3ae", "parents": [ "ae9830cdb86118db1283a1d00a6dde82ab2b201e" ], "author": { "name": "Dimitry Andric", "email": "dimitry@andric.com", "time": "Mon May 06 17:58:03 2019 +0000" }, "committer": { "name": "Dimitry Andric", "email": "dimitry@andric.com", "time": "Mon May 06 17:58:03 2019 +0000" }, "message": "Add non-SSE wrapper for __kmp_{load,store}_mxcsr\n\nSummary:\nTo be able to successfully build OpenMP on FreeBSD/i386, which still\nuses i486 as its default processor, I had to provide wrappers for the\n`__kmp_load_mxcsr` and `__kmp_store_mxcsr` functions.\n\nIf the compiler signals that SSE is not available, loading and storing\nmxcsr does not make sense anway, so in that case the inline functions\nare empty. This gives the minimum amount of code churn.\n\nSee also https://svnweb.freebsd.org/changeset/base/345283\n\nReviewers: emaste, jlpeyton, Hahnfeld\n\nReviewed By: jlpeyton\n\nSubscribers: hfinkel, krytarowski, jdoerfert, openmp-commits, llvm-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D60916\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@360062 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ae9830cdb86118db1283a1d00a6dde82ab2b201e", "tree": "22d64581e03fd5a9ec452c1f182851e54beb2f63", "parents": [ "49325fc6abe74ff4a67ef19409999d128d17dca6" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 03 20:00:38 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 03 20:00:38 2019 +0000" }, "message": "[OPENMP][NVPTX]Improve thread limit counter, NFC.\n\nSummary:\nPatch improves performance of the full runtime mode by moving\nthread-limit counter to the shared memory. It also allows to save\nglobal memory.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jdoerfert, caomhin, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D61526\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359922 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "49325fc6abe74ff4a67ef19409999d128d17dca6", "tree": "a609edaa7dcd9d697ef9be6ee217cb8efa19ee78", "parents": [ "2d4c5f915a723961181c1725daa66a299ab65e81" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 03 14:47:20 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri May 03 14:47:20 2019 +0000" }, "message": "[OPENMP][NVPTX]Improved several standard OpenMP functions, NFC.\n\nSummary:\nUsed parallelLevel[] counter to simplify and improve implementation of\nthe existing standard OpenMP functions. Functions are tested already in\nseveral tests, the patch is NFC.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jdoerfert, caomhin, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D61459\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359892 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "2d4c5f915a723961181c1725daa66a299ab65e81", "tree": "5f32eaf163ca0b3e99adac63e7653142bfa1856b", "parents": [ "db75b85d572b5526020f54055f37fc6bba22e08b" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu May 02 20:05:01 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu May 02 20:05:01 2019 +0000" }, "message": "[OPENMP][NVPTX]Improve code by using parallel level counter.\n\nSummary:\nPreviously for the different purposes we need to get the active/common\nparallel level and with full runtime we iterated over all the records to\ncalculate this level. Instead, we can used the warp-based parallel level\ncounters used in no-runtime mode.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jfb, jdoerfert, caomhin, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D61395\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359822 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "db75b85d572b5526020f54055f37fc6bba22e08b", "tree": "33b8d952586c002a6684cf8fa9b553968d212617", "parents": [ "573fae727555eddbc20088d0b7e5abe547bfcf39" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu May 02 14:52:52 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu May 02 14:52:52 2019 +0000" }, "message": "[OPENMP][NVPTX]Improve omp_get_max_threads() function.\n\nSummary:\nFunction omp_get_max_threads() can always return 1 if current execution\nmode is SPMD.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jdoerfert, caomhin, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D61379\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359792 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "573fae727555eddbc20088d0b7e5abe547bfcf39", "tree": "bac0ef5f69ec1edb1a5dc2bca47450328da43781", "parents": [ "ca840d66bb83070331bc6ed0e21ac01d1b90ba43" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu May 02 14:46:32 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Thu May 02 14:46:32 2019 +0000" }, "message": "[OPENMP][NVPTX]Improved omp_get_thread_limit() function.\n\nSummary:\nFunction omp_get_thread_limit() in SPMD mode can return the maximum\navailable number of threads as a result.\n\nReviewers: grokos, gtbercea, kkwli0\n\nSubscribers: guansong, jdoerfert, openmp-commits, caomhin\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D61378\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359790 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "ca840d66bb83070331bc6ed0e21ac01d1b90ba43", "tree": "8abb0409a07ff15860149d9f061a8766e11ce032", "parents": [ "9377fa96534671f478c956e38fa258a58457d543" ], "author": { "name": "Dimitry Andric", "email": "dimitry@andric.com", "time": "Wed May 01 19:32:58 2019 +0000" }, "committer": { "name": "Dimitry Andric", "email": "dimitry@andric.com", "time": "Wed May 01 19:32:58 2019 +0000" }, "message": "Enable OpenMP build for 32-bit FreeBSD\n\nSummary:\nTo be able to successfully build OpenMP on 32-bit FreeBSD, such as\nFreeBSD/i386, I first had to provide a few wrappers (see D60916), and\nthen add `KMP_OS_FREEBSD` to the list of defines checked for 32-bit\narchitectures in `kmp_runtime.cpp`.\n\nI have successfully built libomp.so and ran a bunch of test programs on\nFreeBSD/i386 with this.\n\nSee also https://svnweb.freebsd.org/changeset/base/345283\n\nReviewers: emaste, jlpeyton, Hahnfeld\n\nReviewed By: jlpeyton\n\nSubscribers: krytarowski, guansong, jdoerfert, openmp-commits, llvm-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D60917\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359716 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "9377fa96534671f478c956e38fa258a58457d543", "tree": "e550a9c4dfa84ccc2ce0af74440a04a921c4776e", "parents": [ "67ef47a8a17b8d44181a0651d684c45b828a4653" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Wed May 01 17:54:01 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Wed May 01 17:54:01 2019 +0000" }, "message": "[OpenMP] Implement task modifier for reduction clause\n\nImplemented task modifier in two versions - one without taking into account\nomp_orig variable (the omp_orig still can be processed by compiler without help\nof the library, but each reduction object will need separate initializer with\nglobal access to omp_orig), another with omp_orig variable included into\ninterface (single initializer can be used for multiple reduction objects of\nthe same type). Second version can be used when the omp_orig is not globally\naccessible, or to optimize code in case of multiple reduction objects\nof the same type.\n\nPatch by Andrey Churbanov\n\nDifferential Revision: https://reviews.llvm.org/D60976\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359710 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "67ef47a8a17b8d44181a0651d684c45b828a4653", "tree": "2b7b619e29a01da2b52670d227156515aec890a7", "parents": [ "03f8ac57f0aa1a4b1a18ebf03a9c9ba33fec0222" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue Apr 30 19:20:35 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue Apr 30 19:20:35 2019 +0000" }, "message": "[OpenMP] Add OpenMP 5.0 nonmonotonic code\n\nThis patch adds:\n* New omp_sched_monotonic flag to omp_sched_t which is handled within the runtime\n* Parsing of monotonic/nonmonotonic in OMP_SCHEDULE\n* Tests for the monotonic flag and envirable parsing\n* Logic to force monotonic when hierarchical scheduling is used\n\nDifferential Revision: https://reviews.llvm.org/D60979\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359601 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "03f8ac57f0aa1a4b1a18ebf03a9c9ba33fec0222", "tree": "e8ab1a17cc455078402c1982ad8a112601eb4934", "parents": [ "867a57389a1512800a2b2bdff6967f0978d398d6" ], "author": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue Apr 30 19:13:37 2019 +0000" }, "committer": { "name": "Jonathan Peyton", "email": "jonathan.l.peyton@intel.com", "time": "Tue Apr 30 19:13:37 2019 +0000" }, "message": "[OpenMP] Eliminate some compiler warnings\n\n* Remove accidental \u003d\u003d for \u003d\n* Assign values to variables to appease compiler\n* Surround debug code with KMP_DEBUG\n* Remove unused local typedefs\n\nDifferential Revision: https://reviews.llvm.org/D60983\n\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359599 91177308-0d34-0410-b5e6-96231b3b80d8\n" }, { "commit": "867a57389a1512800a2b2bdff6967f0978d398d6", "tree": "f0cf09e44d2bbe8662498803026995d59754f5c9", "parents": [ "777f60f39b80b52f0b3ce3f2aa1af5ffb7aaf7bc" ], "author": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri Apr 26 19:30:34 2019 +0000" }, "committer": { "name": "Alexey Bataev", "email": "a.bataev@hotmail.com", "time": "Fri Apr 26 19:30:34 2019 +0000" }, "message": "[OPENMP][NVPTX]Correctly handle L2 parallelism in SPMD mode.\n\nSummary:\nThe parallelLevel counter must be on per-thread basis to fully support\nL2+ parallelism, otherwise we may end up with undefined behavior.\nIntroduce the parallelLevel on per-warp basis using shared memory. It\nallows to avoid the problems with the synchronization and allows fully\nsupport L2+ parallelism in SPMD mode with no runtime.\n\nReviewers: gtbercea, grokos\n\nSubscribers: guansong, jdoerfert, caomhin, kkwli0, openmp-commits\n\nTags: #openmp\n\nDifferential Revision: https://reviews.llvm.org/D60918\n\ngit-svn-id: https://llvm.org/svn/llvm-project/openmp/trunk@359341 91177308-0d34-0410-b5e6-96231b3b80d8\n" } ], "next": "777f60f39b80b52f0b3ce3f2aa1af5ffb7aaf7bc" }