)]}' { "log": [ { "commit": "a7b82bbc588847235b52f3597d874d0862cbaa3d", "tree": "be34fb9dfe0659652614f9ecaeb39da433683a8f", "parents": [ "c0c3d8e30a8329c48a7e81f40d18a710f4eb4074" ], "author": { "name": "Lurie97", "email": "109333972+Lurie97@users.noreply.github.com", "time": "Wed Oct 14 09:40:02 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Oct 13 18:50:12 2026 -0700" }, "message": "[libclc] Produce subnormal results for exp/exp2 on FP32 (#212696)\n\nThe FP32 __clc_exp/__clc_exp2 implementations scale the reduced result\nby 2^p using integer bit manipulation (as_int(y) + (p \u003c\u003c 23)). That\nrepresentation cannot encode a subnormal, so inputs whose result falls\nbelow the smallest normal were unconditionally flushed to zero via \"x \u003c\nllim ? 0.0f\". On devices that report CL_FP_DENORM this loses the correct\nsubnormal result (e.g. erfc(), which is built on exp(), failed CTS\nmath_brute_force with the reference subnormal vs a returned 0).\n\nRecompute the subnormal range with __clc_ldexp, which does produce\nsubnormals, gated on __clc_fp32_subnormals_supported() so flush-to-zero\ntargets keep the previous fast path. The ldexp path is restricted to\n[subnorm_llim, llim); inputs below subnorm_llim (including -inf and\nlarge-magnitude negatives) still flush to zero, since their reduced\nexponent argument would be ill-formed for ldexp.\n\nGitOrigin-RevId: 7cec6725695788dc9a82aa4464611a9dacde2549\n" }, { "commit": "c0c3d8e30a8329c48a7e81f40d18a710f4eb4074", "tree": "da1bf59722fd6eb85d0dec82a23f59904cf1bc7e", "parents": [ "02fb88c692a4f806dacb3dc6e1253675f46eed69" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Fri Oct 09 07:58:00 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Oct 09 06:02:24 2026 -0700" }, "message": "[flang-rt][libclc] Install multilib variants in their own directory (#230251)\n\nA multilib runtimes build sets \u003cPROJECT\u003e_LIBDIR_SUBDIR. flang-rt\nlibraries and libclc bitcode were installed with the same name as the\nbase library. This doesn\u0027t really work for `libclc` because the clang\ndriver doesn\u0027t look through multilibs currently, but it should still go\nsomewhere else so it doesn\u0027t clobber.\n\nGitOrigin-RevId: ef05fab1df3fe07794c2e32fea294334c3100289\n" }, { "commit": "02fb88c692a4f806dacb3dc6e1253675f46eed69", "tree": "3331f5c27cb8f2897410d9e6d7903851acbf8f56", "parents": [ "5932501d8a1543d8c4796a04e65596fc5f81f621" ], "author": { "name": "Jared Wyles", "email": "jared.wyles@gmail.com", "time": "Sun Oct 04 14:37:04 2026 +1100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sat Oct 03 20:56:07 2026 -0700" }, "message": "Wyles/update libclctests (#228681)\n\nUsing libclc with different versions of clang emit different output.\ncomputeConstantRange now pushes ranges through zext and sext so we now\ncan prove noundef. This test returns AMDGPU workgroup ID which is zero\nextended, so we can now prove it needs noundef.\n\nAt some point we may need to update the CI targets to look at libclc?\n\nGitOrigin-RevId: 72b39c854fd7f9fd008bf14df4ce2c98b0ea364d\n" }, { "commit": "5932501d8a1543d8c4796a04e65596fc5f81f621", "tree": "478e3651a46109bff1b1df160d5c375935d9501f", "parents": [ "12e4086b6dc0aa77e121467b58a5a02b2f3a532d" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Thu Sep 10 09:56:16 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Sep 10 01:03:18 2026 -0700" }, "message": "libclc: Update fmod implementations (#222369)\n\nThis was originally ported from rocm device libs in\n93af966747b59d37c57312a0c0242151076c072b. Merge in more\nrecent changes. This should also approximately match the default\nexpansion in ExpandIRInsts\n\nCo-authored-by: Claude \u003cnoreply@anthropic.com\u003e\nGitOrigin-RevId: ab7245d75cc97db490d73037b5f021a561381d0d\n" }, { "commit": "12e4086b6dc0aa77e121467b58a5a02b2f3a532d", "tree": "84247af408b58038bd3cc453513ca3547165187d", "parents": [ "0f522c63e191c1e0afd47f1d0c1270f12c7a0e93" ], "author": { "name": "Romaric Jodin", "email": "rjodin@google.com", "time": "Wed Sep 02 23:42:33 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Sep 02 14:50:41 2026 -0700" }, "message": "[libclc] Fix remainder calculation in clc_remquo for subnormals (#217925)\n\nThe remainder t was previously computed using:\n\n __CLC_GENTYPE t \u003d __clc_mad(y, -__CLC_CONVERT_GENTYPE(qsgn), x);\n\nMultiplying y by -qsgn (+-1.0) introduces an unnecessary intermediate\nmultiplication step. On platforms or execution modes where subnormals\nare flushed to zero computing `y * -qsgn` can prematurely flush a\nsubnormal `y` to zero, resulting in `0.0 + x \u003d x` instead of computing\nthe subtraction `x - y` (or `x + y`).\n\nReplace `__clc_mad` with a direct addition/subtraction based on the sign\nof the quotient:\n\n __CLC_GENTYPE t \u003d qsgn \u003e 0 ? (x - y) : (x + y);\n\nThis avoids multiplication by +-1.0, eliminates unwanted subnormal\nflushing on intermediate products in FTZ modes, and computes the exact\nremainder.\n\nGitOrigin-RevId: 71a85ef89fc1997205104b72447e0dc2c4ad83da\n" }, { "commit": "0f522c63e191c1e0afd47f1d0c1270f12c7a0e93", "tree": "69f5645dbadbb1094a1168f4d21bac2a22c4de29", "parents": [ "529203c0e5e35ba4c144fdfbc4253087863fad7b" ], "author": { "name": "Romaric Jodin", "email": "rjodin@google.com", "time": "Wed Aug 19 12:15:50 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Aug 19 03:22:06 2026 -0700" }, "message": "[libclc] Use generic subnormal_config for spirv-vulkan (#217240)\n\nSince clspv properly handles denorm with\nhttps://github.com/google/clspv/pull/1616\nwe should rely on the generic implementation of subnormal_config.\n\nGitOrigin-RevId: d4c19355521fe0d7af54dd80c766fa00d1c6a278\n" }, { "commit": "529203c0e5e35ba4c144fdfbc4253087863fad7b", "tree": "b9aa6476d69f498afe06cd2a28841ca4339b5e85", "parents": [ "b2c23617a51268c1cde8a535f7cb887e6caad9a8" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Fri Aug 07 21:08:18 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Aug 07 19:10:30 2026 -0700" }, "message": "[libclc] Add math smoke test for cos special values (#214786)\n\nSummary:\nBasic math version of unit tests used to give reference for how these\ncan test the known special values. SImilar to LLVM libc\u0027s smoke math\ntests on the GPU. More can be added later,\nGitOrigin-RevId: 756e26ab79917987f678bb9a81a470c4c61516f4\n" }, { "commit": "b2c23617a51268c1cde8a535f7cb887e6caad9a8", "tree": "84dd8ff89d1b4f5f7de01e2d4f6c50a5f4db9c9b", "parents": [ "84e72f4635dbaffa63f52ef1ba710249ebd94546" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat Aug 08 03:37:31 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Aug 07 18:40:13 2026 -0700" }, "message": "[libclc][test] Update cos.cl checks after fneg FP-options fix (#214906)\n\nceaff22ed2c3 ([Clang][CodeGen] Respect FP pragma options for fneg and\ncalls) scopes fneg\u0027s fast-math flags to the expression\u0027s own FPOptions\ninstead of the ambient IRBuilder state, so one fneg in cos.cl\u0027s\ngenerated IR no longer carries contract. Regenerate check.\n\nCo-authored-by: Claude Sonnet 5 \u003cnoreply@anthropic.com\u003e\nGitOrigin-RevId: db116a9b319818a63d9e5acf27920fea063b0f20\n" }, { "commit": "84e72f4635dbaffa63f52ef1ba710249ebd94546", "tree": "da5ab2e353c3264a81801c34a005e677358f9cad", "parents": [ "daf08853bcb6125509cfe7cfc424ccf3752c86c8" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat Aug 08 02:39:26 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Aug 07 17:40:27 2026 -0700" }, "message": "[libclc][CMake] Add cmake cache files (#201480)\n\nThis simplifies cmake configuration without need to pass verbose\nvariables in cmake command line. Settings in cache file can still be\noverriden by passing `-D` flags.\n\nUpdate `libclc/README.md` to document the usage.\n\nGitOrigin-RevId: d34b0ab0434285bad750b5da1af3d7c9c2492f3a\n" }, { "commit": "daf08853bcb6125509cfe7cfc424ccf3752c86c8", "tree": "9baefd42ea7c37d89ce71821925a674a619ed02c", "parents": [ "b8e291301acb2cdc610e3e3775c17812d803ca58" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Thu Aug 06 09:00:25 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Aug 06 07:05:45 2026 -0700" }, "message": "[libclc] Improve libclc device tests with feature checks (#214217)\n\nSummary:\nMake sure the tests can conditionally run on the correct features.\n\nI think that it would be nice if we could print from these tests, but\nthat would require linking against `libc` for GPUs and I\u0027m unsure if\nthat\u0027s worth it, so for now I\u0027m just sticking with traps as the failure\nmechanism.\n\nGitOrigin-RevId: c90a5cbb0d3366526a5d44032b06b3e624a5249b\n" }, { "commit": "b8e291301acb2cdc610e3e3775c17812d803ca58", "tree": "1f7923e027605d74581f546db2ff3e51beb526b4", "parents": [ "fb3b88a5c1cd941b7790e0789734bbe099f66747" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue Aug 04 22:47:00 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Aug 04 20:50:43 2026 -0700" }, "message": "[libclc] Add initial support for libclc execution conformance tests (#214072)\n\nSummary:\nThis uses the `llvm-gpu-loader` tool to invoke OpenCL test kernels. This\nroutes OpenCL through the standard compute runtimes, which aren\u0027t\nexactly 1-to-1 with OpenCL, but it\u0027s close enough. Obviously, the best\nsolution would be to have a real OpenCL host library that can launch\nthese, but this is the best we can get in-tree and is very sipmle, write\nkernel, test kernel.\n\nThe goal is to have a lot more tests run here, but this just gets the\nbasic infrastructure in place. We won\u0027t be able to do full conformance\nlike external suites, like exhaustive math, but I think its\u0027 a step in\nthe right direction. It looks like this on my side:\n\n```console\n$ ninja check-libclc-amdgpu-amd-amdhsa-llvm\n```\n\nThis requires building with the LLVM `offload/` runtime enabled for the\nsource compiler.\n\nGitOrigin-RevId: d87c685d5d88fe00b0594fc238b9afb16ba6eb64\n" }, { "commit": "fb3b88a5c1cd941b7790e0789734bbe099f66747", "tree": "e1d1a2543072ccc8483f1d4714516bbd106a4c53", "parents": [ "a07884a652cfc69626b7b3f17adddfdcfa311267" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Thu Jul 23 06:24:44 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jul 22 21:30:36 2026 -0700" }, "message": "libclc: Fix signed 0 handling in sincospi (#211260)\n\nGitOrigin-RevId: 55df5985398653fcff135c6e93345d2b1af98f55\n" }, { "commit": "a07884a652cfc69626b7b3f17adddfdcfa311267", "tree": "c2d97aa301fc4c7eaeea90fbe1861ce8111fd961", "parents": [ "af4ebee220fe19fdb9ac6522575a70a10947e6b5" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Jul 22 06:41:39 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Jul 21 21:45:34 2026 -0700" }, "message": "libclc: Fix UB in trig functions with nan inputs (#211047)\n\nGitOrigin-RevId: 6793bb48ac7517ea1e965de540c75359e465c355\n" }, { "commit": "af4ebee220fe19fdb9ac6522575a70a10947e6b5", "tree": "e11c756653ea8e1712c7c6eac747fa1757dee57d", "parents": [ "08886a2db52bf93e74d72780722278dbe69a324b" ], "author": { "name": "Petr Hosek", "email": "phosek@google.com", "time": "Mon Jun 29 11:16:27 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 29 11:21:06 2026 -0700" }, "message": "[CMake] Warn when the version of CMake is older than 3.31.0 (#204203)\n\nThis is a preparation to raise the minimum required CMake version to\n3.31.0 in LLVM 24.\n\nSee\nhttps://discourse.llvm.org/t/rfc-raising-minimum-required-cmake-version-to-3-31/91086.\n\nGitOrigin-RevId: e192ec42b426f6ffbc25419238a90f3c59088587\n" }, { "commit": "08886a2db52bf93e74d72780722278dbe69a324b", "tree": "cd5e90d1fd8b8726b7ea2969a6a5d42bc2ee2e0e", "parents": [ "0dcd4a4abc076d5f37005606dd4d576bc49d1561" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Thu Jun 25 02:47:23 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 24 17:50:45 2026 -0700" }, "message": "[libclc] Delete wrong implementation nvptx clc_isinf (#205699)\n\nThe file calls __nv_isinf which return 1 for true on vector input, while\nthe generic clc_isinf which return -1 for true on vector input. Using\nnvptx clc_isinf in OpenCL isinf violates OpenCL spec.\n\nFound the issue in https://github.com/intel/llvm/pull/22413\n\nGitOrigin-RevId: 89ac672f5b9242e040598eedb7d240c97a8da44b\n" }, { "commit": "0dcd4a4abc076d5f37005606dd4d576bc49d1561", "tree": "aace25fe0f0919fa9cafd813f8f800e61fd4850a", "parents": [ "dfbdd4e9505247d637185d682472a2f059942487" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Mon Jun 15 07:26:32 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Jun 15 05:30:36 2026 -0700" }, "message": "[libclc] Use unchecked division for f64 AMDGPU reciprocal (#203809)\n\nSummary:\nThis matches what the AMD device libraries does. We can avoid extra\nsteps by only performing the two steps of the Newton-Raphson\napproximation of 1 / x. The exceptional cases should not appear in these\nmath functions, this is local to AMDGPU, and I verified they are bitwise\nidentical to the AMD math functions with parity in performance now.\nGitOrigin-RevId: cac31111d37cd27b222aa2ddb11e2238db27017f\n" }, { "commit": "dfbdd4e9505247d637185d682472a2f059942487", "tree": "f450e75146b93d5428b5650dbd5ff8d62927d838", "parents": [ "65691fdf8c6721a19cc74d5c7795ea823bccfaf1" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Sun Jun 14 20:33:36 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Jun 14 18:36:08 2026 -0700" }, "message": "[libclc] Use __CLC_SCALAR instead of nonexistant __CLC_SCALAR1 for sin (#203807)\n\nSummary:\nThis seems to be a typo? Every other case is guarded by `__CLC_SCALAR`\nbut this case had a `1` after it. Removing this improved performance on\nsin/cos/tan to match the ROCm version.\nGitOrigin-RevId: 303400f0809a808d0f478c05eeaf0073d6fa149a\n" }, { "commit": "65691fdf8c6721a19cc74d5c7795ea823bccfaf1", "tree": "10ec9b06b4b72bd0e68eff916d7e78d41668602d", "parents": [ "a77f1f6a2481a41f575d45354589f6f7d3d064a8" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Sun Jun 14 20:32:57 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Jun 14 18:36:01 2026 -0700" }, "message": "[libclc] Use FMA for the pi reconstruction in acos / atan (#203804)\n\nSummary:\nThis should recombine the split constant for this case. The performance\nshould be negligible for such large math functions, we get an extra add,\nbut in exchange the results should improve 1 ULP.\n\nThis was primarily done to match what AMD\u0027s math libraries do, with this\nchange we are byte-for-byte identical in output.\n\nGitOrigin-RevId: 8b6551e49395be0dc878e7acff1edd31f4204143\n" }, { "commit": "a77f1f6a2481a41f575d45354589f6f7d3d064a8", "tree": "10940d19f794ef67e062f6001397a04475eb3667", "parents": [ "a877d636b9415acec863647b0414620247848fca" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Sun Jun 14 20:32:00 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Jun 14 18:35:53 2026 -0700" }, "message": "[libclc] Improve performance and precision of reciprocal functions (#203805)\n\nSummary:\nSmall change to improve the performance and output of functions using\nthe reciprocol. This makes these functions *byte-for-byte* identical\nwith their OCML counterparts in ROCm.\n\nAdditionally ensure tanpi is correctly rounded to match ROCm.\n\nGitOrigin-RevId: 375b70bafedafa71fabfb0e8b1e759daeef0f703\n" }, { "commit": "a877d636b9415acec863647b0414620247848fca", "tree": "3eb14c31490315dc57a88b698a7225be43ca7973", "parents": [ "801280349cd7b513ba48923be33fc4d07dcf0224" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat Jun 06 07:18:19 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 16:20:45 2026 -0700" }, "message": "[libclc][Test] Don\u0027t hard code `clang` path in update_libclc_tests.py (#201806)\n\nIn #201773, build folder is `redhat-linux-build` rather than `build`.\nDetect clang from PATH. Also add --clang-binary option to specify clang path.\n\nGitOrigin-RevId: 985e49989fa6ac24a1318969f6865bb5407574c6\n" }, { "commit": "801280349cd7b513ba48923be33fc4d07dcf0224", "tree": "822166a34a87fdef903bf97a57a2d6c6c1679176", "parents": [ "7028afc2a7e477dde73745d8e593492ffdff6234" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat Jun 06 07:17:55 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 16:20:38 2026 -0700" }, "message": "[libclc][Test] Add -fno-discard-value-names to RUN line (#201808)\n\nEntry block label could be missing. Add -fno-discard-value-names to make\nsure it exists.\n\nResolves #201773\n\nGitOrigin-RevId: e8753fdde70b8a2bd0d3219000228704587935bd\n" }, { "commit": "7028afc2a7e477dde73745d8e593492ffdff6234", "tree": "58ceff87cb0fce5eaa6a02c6d6fc65566a745fd7", "parents": [ "d0b74930d5f9a6b92d515a0e9131a50c4ad96fe2" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat Jun 06 07:17:42 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Jun 05 16:20:31 2026 -0700" }, "message": "[libclc] Fix LIBCLC_OUTPUT_LIBRARY_DIR when find_package(LLVM) fails (#201816)\n\nIf we configure llvm from llvm-project/build folder, find_package(LLVM)\nsucceeds, `LLVM_LIBRARY_OUTPUT_INTDIR` is `build/./lib` and\n`LLVM_BINARY_DIR` is `build`.\n\nHowever, in #201773, find_package(LLVM) fails due to\nLLVM_LIBDIR_SUFFIX\u003d64, LLVM_LIBRARY_OUTPUT_INTDIR is\n`redhat-linux-build/runtimes/runtimes-nvptx64-nvidia-cuda-bins/lib64`\nand LLVM_BINARY_DIR is `redhat-linux-build`.\n\nUse LLVM_BINARY_DIR, which is stable for both cases, for libclc output dir.\n\nFixes #201773\n\nGitOrigin-RevId: 517a0755f5d128d950fc8a08b2bd45e629e40a95\n" }, { "commit": "d0b74930d5f9a6b92d515a0e9131a50c4ad96fe2", "tree": "5bffc24a328edc7f4d16269eddcec3a75efcf2e9", "parents": [ "3934cf88dc2b825fec7c272628f38b47254b11ad" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Wed Jun 03 16:01:00 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Jun 03 01:05:54 2026 -0700" }, "message": "[libclc] Use generic spirv*-unknown-unknown clang triple for SPIR-V targets (#199618)\n\nspirv-diff shows only numbering change to\nspirv64-unknown-unknown/libclc.spv. No change in `llvm-spirv -to-text`\noutputs. llvm-diff shows no change on reverse-translated bitcode files.\n\nAlso fixes a bug that spirv32-unknown-unknown was incorrectly using\n64-bit triple.\n\nUpdate README.md to use the generic target triple for SPIR-V targets.\n\nGitOrigin-RevId: 6a7b8876153815cf3847d952614440e79e82fc86\n" }, { "commit": "3934cf88dc2b825fec7c272628f38b47254b11ad", "tree": "ebcc4996bcc20fdc9fa970b38e477704ab8d7eeb", "parents": [ "e741a08a2ee085c2e011833fa0d54cb3c4e670c6" ], "author": { "name": "michaelselehov", "email": "michael.selehov@amd.com", "time": "Fri May 29 20:16:27 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 29 11:20:39 2026 -0700" }, "message": "[libclc][test] Update math/cos.cl check lines after #199981 revert (#200353)\n\nThe libclc lit test \u0027math/cos.cl\u0027 (introduced by #197151 with\nauto-generated FileCheck assertions) started failing on the staging\nbuildbot after the revert in #199981 of the AMDGPU ABI coercion change\n(#185083). The revert restores the older codegen path, which produces a\nslightly different IR shape for the select instruction inside the cos()\nimplementation:\n\n - the select\u0027s \u0027contract\u0027 fast-math flag is no longer present, and\n - the result is named %.v.i.i instead of a numbered SSA temporary.\n\nRegenerate the AMDGCN CHECK lines with\nlibclc/test/update_libclc_tests.py to reflect the post-revert IR. After\nthis change, all 10 libclc tests in check-libclc-amdgcn-amd-amdhsa-llvm\npass.\n\nNote: running update_libclc_tests.py also produces a purely cosmetic\ndiff in math/rsqrt.cl (it renames a few FileCheck capture names like\nMETA13 to META12 to match the actual metadata indices); those capture\nrenames are not required for the test to pass and are not included here\nto keep the change minimal.\n\nCo-authored-by: mselehov \u003cmselehov@amd.com\u003e\nGitOrigin-RevId: 5945eacb934590c2a7091d73f53289e5dbcf2c40\n" }, { "commit": "e741a08a2ee085c2e011833fa0d54cb3c4e670c6", "tree": "1914ef9742bf8eadd6b283b473faba7366ef9b02", "parents": [ "d7c35a601181f1cc00e554a8c96bbc1dc7205a5a" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Fri May 29 13:28:17 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu May 28 22:30:26 2026 -0700" }, "message": "[libclc] Use float sw_fma for spirv*-mesa-mesa3d targets (#199626)\n\nThis partially reverts 39f3d72b72fd, which changes almost all targets to\nuse __builtin_elementwise_fma.\nThe same revert was done for clspv in 1d719edd9935.\n\nRestore __clc_runtime_has_hw_fma32 so that Mesa can dynamically change\nits behavior to enable native fma for specific hardware.\n\nFixes #198542\n\nGitOrigin-RevId: 767f01e0910d9211486d4022c5c5e7d71c85dff2\n" }, { "commit": "d7c35a601181f1cc00e554a8c96bbc1dc7205a5a", "tree": "7ba55923c5848f0370a5201e3d9e9871f31fc776", "parents": [ "daafba5c870187d972f603e8f5f0bab1dc8f9e5c" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Thu May 28 08:06:47 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed May 27 17:11:36 2026 -0700" }, "message": "[libclc] Optimize and vectorize signbit (#199497)\n\nReplace element-wise scalarizing implementation with bitcast and shift.\nFor example,\ndefine hidden range(i32 -1, 1) \u003c2 x i32\u003e @_Z7signbitDv2_f(\u003c2 x float\u003e\nnoundef %0) #0 {\n %2 \u003d bitcast \u003c2 x float\u003e %0 to \u003c2 x i32\u003e\n %3 \u003d extractelement \u003c2 x i32\u003e %2, i64 0\n %4 \u003d lshr i32 %3, 31\n %5 \u003d insertelement \u003c2 x i32\u003e poison, i32 %4, i64 0\n %6 \u003d extractelement \u003c2 x i32\u003e %2, i64 1\n %7 \u003d lshr i32 %6, 31\n %8 \u003d insertelement \u003c2 x i32\u003e %5, i32 %7, i64 1\n %9 \u003d icmp ne \u003c2 x i32\u003e %8, zeroinitializer\n %10 \u003d sext \u003c2 x i1\u003e %9 to \u003c2 x i32\u003e\n ret \u003c2 x i32\u003e %10\n}\nis changed to:\ndefine hidden noundef range(i32 -1, 1) \u003c2 x i32\u003e @_Z7signbitDv2_f(\u003c2 x\nfloat\u003e noundef %0) #0 {\n %2 \u003d bitcast \u003c2 x float\u003e %0 to \u003c2 x i32\u003e\n %3 \u003d ashr \u003c2 x i32\u003e %2, splat (i32 31)\n ret \u003c2 x i32\u003e %3\n}\n\nGitOrigin-RevId: 9cc6c9375faa11722435a150c9829df2452b48ca\n" }, { "commit": "daafba5c870187d972f603e8f5f0bab1dc8f9e5c", "tree": "d49619f349fc0b4972cbc7eb0f07b9158e78351f", "parents": [ "8f171418830410024eef856956b38156e04688b7" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat May 23 07:30:59 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 22 16:36:22 2026 -0700" }, "message": "[libclc] Fix .cl tests in standalone build (#199164)\n\nlibclc standalone build puts libclc.bc in ${CMAKE_CURRENT_BINARY_DIR}/\n${TARGET_TRIPLE} dir. check-libclc fails because .cl test is looking for\nlibclc in clang resource dir.\nFix them by adding `--libclc-lib\u003d:{path}` flag for standalone build,\nwhere `path` is path to libclc.bc.\nNote: this flag is not used in in-tree build.\nGitOrigin-RevId: 0de4a8cd41efaac4d63ce056e8d4177397fdc1a7\n" }, { "commit": "8f171418830410024eef856956b38156e04688b7", "tree": "fabeacf15cac2fc6ed83b111c0eca2160711b54d", "parents": [ "b8da9daece744fdc788fa4eab60fda531e188736" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Wed May 20 09:27:39 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue May 19 18:30:47 2026 -0700" }, "message": "[libclc][NFC] Use clc/shared/ternary_def.inc for bitselect (#198495)\n\nDelete opencl/lib/generic/relational/bitselect.inc\n\nGitOrigin-RevId: 05c1a266b84d811f7869d8326674da4f5aee7a89\n" }, { "commit": "b8da9daece744fdc788fa4eab60fda531e188736", "tree": "9c0dc9c8a5f701d4427e00ebd7d43756d06f09d7", "parents": [ "2aaefe60c3add9954e67a429acde79db6926bb6c" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Fri May 15 17:46:44 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 15 02:50:33 2026 -0700" }, "message": "[libclc][CMake] Use set instead of APPEND for LIBCLC_ARCHS_ALL initialization (#197866)\n\nGitOrigin-RevId: bf9916a5b046cd0612cb314fea1fc88f94aed9aa\n" }, { "commit": "2aaefe60c3add9954e67a429acde79db6926bb6c", "tree": "2aef4438b436e4472e0d6b85e5e5827cd43ecded", "parents": [ "28b0b2a4414ea3ddc5ce3a0630875646ede46d07" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Fri May 15 10:17:35 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu May 14 19:21:12 2026 -0700" }, "message": "[libclc] Move sources collection into libclc_add_builtin_library, merge opencl library creation into libclc_add_library (#197654)\n\nMerging opencl library creation into libclc_add_library undos part of\nchange in 14bf8e7.\nThis simplifies creating different libraries by invoking\nlibclc_add_library multiple times, without the need of splitting\nintermediate library creation and its finalization.\n\nadd_subdirectory(clc) and add_subdirectory(opencl) only updates sources\nlist of existing library target, so it is safe to move inside\nlibclc_add_builtin_library.\n\nGitOrigin-RevId: 3d7c5e0bbdf5107ed5e18afcbc83a0d84d9685bb\n" }, { "commit": "28b0b2a4414ea3ddc5ce3a0630875646ede46d07", "tree": "6f6b896090aa46efb79b2f72cba66f49c70f4528", "parents": [ "9bedef2a158b62bac48540c67c00a1aa97e10978" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Wed May 13 15:26:30 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed May 13 13:30:37 2026 -0700" }, "message": "[libclc] Add optimized DPP scan functions for AMDGPU (#197543)\n\nSummary:\nThis uses the update_dpp function to efficiently provide scans. DPP\nallows for swizzling within a group of 16, so we do four of those,\nfollowed by a permlane on GFX10, or a native DPP shift on GFX9.\n\n Right now these are not actually used because we do not compile for\n multiple targets yet, but I verified them by setting\n `CMAKE_CLC_FLAGS\u003dmcpu\u003dgfx1030` or similar. The intention is to have\n these waiting as a motivational implementation once we start doing\n variable builds.\n\n The output is optimal as far as I am aware. Here is the example for\n gfx1030 generation:\n\n ```asm\n_Z28sub_group_scan_inclusive_addi: ; @_Z28sub_group_scan_inclusive_addi\n; %bb.0:\n\ts_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)\n\tv_mov_b32_e32 v1, 0\n\ts_mov_b32 s4, s33\n\ts_mov_b32 s33, s32\n\ts_mov_b32 s33, s4\nv_add_nc_u32_dpp v0, v0, v0 row_shr:1 row_mask:0xf bank_mask:0xf\nbound_ctrl:1\nv_add_nc_u32_dpp v0, v0, v0 row_shr:2 row_mask:0xf bank_mask:0xf\nbound_ctrl:1\nv_add_nc_u32_dpp v0, v0, v0 row_shr:4 row_mask:0xf bank_mask:0xf\nbound_ctrl:1\nv_add_nc_u32_dpp v0, v0, v0 row_shr:8 row_mask:0xf bank_mask:0xf\nbound_ctrl:1\n\tv_permlanex16_b32 v1, v0, -1, -1\nv_add_nc_u32_dpp v0, v1, v0 quad_perm:[0,1,2,3] row_mask:0xa\nbank_mask:0xf\n\ts_setpc_b64 s[30:31]\n ```\nGitOrigin-RevId: f230e44ab888a3688d99ac5f28a2e3edbf3aa596\n" }, { "commit": "9bedef2a158b62bac48540c67c00a1aa97e10978", "tree": "d63a47841638079ce277cd58e032f765045f304a", "parents": [ "13f44f73f0ce31a10facd3aa3546f8c71827beff" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Wed May 13 18:21:36 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed May 13 03:25:35 2026 -0700" }, "message": "[libclc] Move external-funcs.test to static file and use IR checks for .cl tests (#197151)\n\nThis PR supercedes #87989.\n\nMoving external-funcs.test to static file simplifies\ntest/CMakeLists.txt. Static files follows llvm standard lit pattern and\nenables fine-grained check of missing symbols in specific libraries.\n\n.cl test files uses %target, %cpu and %check_prefix, which are replaced\nwith specific values during `ninja check-libclc` or `llvm-lit\nbuild/runtimes/runtimes-${triple}-llvm-bins/libclc/test`. This allows\nchecking outputs of multiple triples in the same test file.\n\nAdd script libclc/test/update_libclc_tests.py, which wraps\nutils/update_cc_test_checks.py to update CHECK lines in libclc .cl tests\nfor a given arch. Example usage:\n`libclc/test/update_libclc_tests.py amdgpu`\n\nAssisted-by: Claude Sonnet 4.6\n\n---------\n\nCo-authored-by: Matt Arsenault \u003carsenm2@gmail.com\u003e\nGitOrigin-RevId: f6007cf3ada595da3e003e0bc7f8dc1a86bff27b\n" }, { "commit": "13f44f73f0ce31a10facd3aa3546f8c71827beff", "tree": "b995752ad8eededa54038e8731b469be9d72f863", "parents": [ "60cc1979e08525744f60683d0616bc0a38e950a2" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue May 12 20:29:25 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue May 12 18:30:45 2026 -0700" }, "message": "[libclc] Apply hidden visibility to amdgpu / nvptx builds (#197235)\n\nSummary:\n This is not currently used because we force `--internalize` for\nllvm-link, but if you ever want to link this normally we\u0027d need hidden\nvisibility. SPIR-V does not currently handle hidden visibility, as it is\n an extension still under discussion, so it is omitted for now.\nGitOrigin-RevId: 9b6e77dfb688ff56149f9a03ba0186b8eeb07810\n" }, { "commit": "60cc1979e08525744f60683d0616bc0a38e950a2", "tree": "daa8242a23a9293be51d683aba736554252f2a7e", "parents": [ "c834cbf18cf842e1147a5f6ea99c2cc21092afcf" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue May 12 20:29:06 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue May 12 18:30:39 2026 -0700" }, "message": "[libclc] Create a static `.a` file in addition to the `.bc` file (#197247)\n\nSummary:\nThis changes the libraries to be object libraries instead of static\nlibraries. We can then just link these into final static libraries that\ncontain everything we need.\n\nThe desire here is that we\u0027d need static libraries if we wanted to move\naway from `-mlink-builtin-bitcode` appraoches.\n\nEffectively we\u0027ll have `libclc.a` next to `libclc.bc` and the idea is\nthat we could alternatively link it and let the target linker handle it.\n\nGitOrigin-RevId: 4106333fe05be0c2154aedfc4d060c4e835e6c8e\n" }, { "commit": "c834cbf18cf842e1147a5f6ea99c2cc21092afcf", "tree": "0d38837c93648895336b28907ccfafbe48cf2239", "parents": [ "147a184f560f3ebec921136dc16884a479bc6439" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue May 12 12:20:50 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue May 12 10:25:55 2026 -0700" }, "message": "[libclc] Consolidate `amdgpu` and `amdgcn` architectures consistently (#197233)\n\nSummary:\nCurrently we did not pass all checks with amdgpu triple as we did with\namdgcn. SPIR-V set this pattern so let\u0027s make it consistent.\nGitOrigin-RevId: 9a39c60406638ee8545d432e7f22a5ecd253c24d\n" }, { "commit": "147a184f560f3ebec921136dc16884a479bc6439", "tree": "b3e17a343ac1e68b72b628371b34ff0326241254", "parents": [ "6f69bfa764abadc6c09680b2789add9df869b977" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Tue May 12 07:04:11 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue May 12 05:05:57 2026 -0700" }, "message": "[libclc] Base the build around `add_sources` instead of source list (#197034)\n\nSummary:\nThe current build uses a curated + deduplicated source list. This PR\nseeks to simplify this a little bit and canonicalize the behavior.\n\nNow we create the target up-front, `clc` and `opencl`. We add the\ndirectories which add sources to this target. We normalize the\narchitecture to the variants. We always add target specific versions\nfirst. When we add sources we check if the file already exists and defer\nto the architecture specific one.\n\nThis normalized the behavior, the directories are now laid out like this\n`clc/\u003carch\u003e/\u003cos\u003e`. We normalize these to `amdgpu`, `nvptx`, and `spirv`\nrespectively. We use the OS for the newly created vulkan target. We now\ncontrol variants via checking if the directory for that exists, so it\u0027s\nnested more naturally.\n\nHopefully this makes more sense, the goal is to exercise the fact that\nwe have individual builds now. Previously this did not work because you\ncould not add_subdirectory more than once.\n\nGitOrigin-RevId: 14bf8e76c021c200b6e37b9df996dbfc0b5dbd4e\n" }, { "commit": "6f69bfa764abadc6c09680b2789add9df869b977", "tree": "91e97a35365947464588ffaa03fcf29a2f9655a0", "parents": [ "e82f74931543f91cedca387bfa7515e94482f39f" ], "author": { "name": "Konrad Kleine", "email": "kkleine@redhat.com", "time": "Mon May 11 12:33:44 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon May 11 03:35:24 2026 -0700" }, "message": "libclc: group spirv archs in LIBCLC_ARCHS_SPIRV (#196911)\n\nThis was done do remove repetitive comparisons.\n\nGitOrigin-RevId: 363c871c8b5a50d10b58f409ff7125f2f41c08ca\n" }, { "commit": "e82f74931543f91cedca387bfa7515e94482f39f", "tree": "3d53167061f12cb33becba6aa7e946f32442f3f1", "parents": [ "73959697c351eb303776dd88f5e838a321ca9ac7" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Fri May 08 08:41:58 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 08 06:45:11 2026 -0700" }, "message": "[libclc] Canonicalize \u0027clspv\u0027 to the \u0027spirv-unknown-vulkan\u0027 triple (#196351)\n\nSummary:\nThe libclc project has clspv support for exporting OpenCL standard\nlibrary utilities to Vulkan consumers. This was previously exposed as a\nhack into the build system that renamed the triple and relied on macro\ndefines. Recent changes allowed us to use `vulkan` as an OS for the\nspir-V target. This should make the intention more clear and allow the\nsystem to inherit the same triple handling the other targets use.\n\nTested the build, but I will need @rjodinchr and @alan-baker to verify.\n\nGitOrigin-RevId: 6ef96710068e2fe1d1c4ebaa957ec215ff42c2c2\n" }, { "commit": "73959697c351eb303776dd88f5e838a321ca9ac7", "tree": "3554e7609f1427d4fb55911cb8cd289b419f0c6b", "parents": [ "e9ea3b800961c38fd1f5975e7b7f980fb45d46ac" ], "author": { "name": "Konrad Kleine", "email": "kkleine@redhat.com", "time": "Fri May 08 11:18:45 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri May 08 02:20:42 2026 -0700" }, "message": "[libclc] Use spirv[64]-mesa-mesa3d triple in README.md (#196483)\n\nNow that #194607 landed we use a normalized triple in the README for the\nSPIRV targets. Before `spirv-mesa3d-` and `spirv64-mesa3d-` were being\nused and those will be normalized to `spirv-unknown-mesa3d` and\n`spirv64-unknown-mesa3d` by the following command in\n`runtimes/CMakeLists.txt` with this command:\n\n```console\n$ clang --target\u003dspirv-mesa3d- -print-target-triple\nspirv-unknown-mesa3d\n```\n\nThis is because in `llvm/lib/TargetParser/Triple.cpp` the term `mesa3d`\nis recognized as an OS and placed in third position. The install path\nfor `libclc.spv` there ends up in `spirv-unknown-mesa3d/libclc.spv`.\n\nWith this change we suggest to use triples that \"survive\" the\nnormalization:\n\n```console\n$ clang --target\u003dspirv-mesa-mesa3d -print-target-triple\nspirv-mesa-mesa3d\n```\n\nSee also this discussion:\nhttps://github.com/llvm/llvm-project/pull/194607#issuecomment-4378126607\n\nGitOrigin-RevId: 016894861f0dc650b4fc22d57034c2f41b7f8481\n" }, { "commit": "e9ea3b800961c38fd1f5975e7b7f980fb45d46ac", "tree": "369792049c278663c91a551b1cb9da1001f46ee5", "parents": [ "f5a74a2046c7e8a2dbd31aced1c2ae00406a7402" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Thu May 07 10:40:56 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu May 07 08:44:37 2026 -0700" }, "message": "[libclc] Fix using normalized triple for the directory name (#194607)\n\nSummary:\nlibclc needs to normalize triples, mostly on account of the fact that\nthings like spv and clspv aren\u0027t real triples but are used anyway. That\nis, we convert the user-value to whatever is passed to `--target\u003d`. The\nproblem is that we were not using the normal triple for the installation\ndirectory, so something like `spirv64-mesa32-unknown` would be installed\nin `spirv64--`.\n\nThis *might* have the side effect of putting these in\n`spirv64-unknown-unknown`. I actually do not know if that\u0027s a problem\nwith the clang handling, I\u0027ll double check.\n\nGitOrigin-RevId: 87d42c13cd6b119240781f31e5869981d500a186\n" }, { "commit": "f5a74a2046c7e8a2dbd31aced1c2ae00406a7402", "tree": "25649787079829f1d5ffe20dbd20629a43e950da", "parents": [ "df72e3bb2d7de5a4af7370a6b1e417a39ce42ba5" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Sat Apr 25 20:52:20 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sat Apr 25 18:55:39 2026 -0700" }, "message": "[libclc] Only check the triple architecture for libclc (#194149)\n\nSummary:\nPreviously, `nvptx64--` would reject `nvptx64-unknown-unknown`. Two\noptions, either normalize all the triples in CMake, or just check the\narchitecture. I went with the former because it makes it easier for\npeople to pass different values.\nGitOrigin-RevId: e7164d42243b8cac55c6c1f91a507608c414c361\n" }, { "commit": "df72e3bb2d7de5a4af7370a6b1e417a39ce42ba5", "tree": "f00ec1157c317d4c2ad061d65ca99423170f1c12", "parents": [ "b871fbf9010fef242b59f096e04896d304e461de" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Fri Apr 24 10:41:28 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Apr 24 08:45:48 2026 -0700" }, "message": "[libclc] Make sure PACKAGE_VERSION is set for libclc (#193966)\n\nSummary:\nThis can be unset because CMake does not expose this as a raw variable\nwhen you use the find_package interface. If it is not set as in the case\nof standalone builds the clang resource directory won\u0027t be found\nGitOrigin-RevId: eb17a2e19c854b0f4f3018f3e17d70340a3eb45f\n" }, { "commit": "b871fbf9010fef242b59f096e04896d304e461de", "tree": "fe6414db7b052ef447974647c67cff41d856fc6a", "parents": [ "0222ef03d3067360f59de31039cadc5183d36135" ], "author": { "name": "Joseph Huber", "email": "huberjn@outlook.com", "time": "Fri Apr 24 08:54:09 2026 -0500" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Apr 24 06:57:54 2026 -0700" }, "message": "[libclc] Use \u0027LLVM_DEFAULT_TARGET_TRIPLE\u0027 instead of \u0027LLVM_RUNTIMES_TARGET\u0027 (#193969)\n\nSummary:\nThe \u0027LLVM_RUNTIMES_TARGET\u0027 variable is the raw value used by the LLVM\nCMake. It can contain multilib arguments which will not compile when\nused as a triple. The more canonical value is\n`LLVM_DEFAULT_TARGET_TRIPLE`, which is used by flang-rt, libc, openmp,\netc.\nGitOrigin-RevId: 6c7d16c0bfd38afc8c4131510740737113451f22\n" }, { "commit": "0222ef03d3067360f59de31039cadc5183d36135", "tree": "c1884ba0f5dd897ab4314bf39a1056558bccdd5a", "parents": [ "e931074a49001c5ee4da1b44c80280b4cf66fde7" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Fri Apr 24 08:46:06 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 23 17:50:23 2026 -0700" }, "message": "[libclc] Allow testing unresolved symbols on multiple libraries (#193647)\n\nOur downstream generates multiple libraries for a single target. This\nchange allows testing multiple libraries.\n\nGitOrigin-RevId: 969247cc47a35ad73155f97f7d66c3f558ea96c5\n" }, { "commit": "e931074a49001c5ee4da1b44c80280b4cf66fde7", "tree": "2adb168654790768a7923139e9bcc87131ad9936", "parents": [ "f2d83982d865a4fd8145590c8a7302467f9061f5" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Fri Apr 24 08:45:51 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 23 17:50:14 2026 -0700" }, "message": "[libclc][CMake] Remove CMAKE_C_COMPILER_ID check (#186717)\n\nWhen LLVM_TARGETS_TO_BUILD does not contain the host target, runtime\nbuild can not identify a compatible target triple for the host compiler.\nCMAKE_C_COMPILER is set to clang, and CMAKE_C_COMPILER_ID is empty\nalthough the compiler is functional.\n\nRemove CMAKE_C_COMPILER_ID check. CMakeTestCLCCompiler.cmake already\nprovides a real functional guard.\n\nThis change ensures valid configurations can proceed, e.g.:\nLLVM_TARGETS_TO_BUILD\u003dAMDGPU;\nRUNTIMES_amdgcn-amd-amdhsa-llvm_LLVM_ENABLE_RUNTIMES\u003dlibclc;\nLLVM_RUNTIME_TARGETS\u003d\"amdgcn-amd-amdhsa-llvm\".\n\nNote CMAKE_C_COMPILER_WORKS is always true after `project(Runtimes C CXX\nASM)` in runtimes/CMakeLists.txt.\n\nGitOrigin-RevId: f3192382c33691e4d3cff07ab645aad3343de81b\n" }, { "commit": "f2d83982d865a4fd8145590c8a7302467f9061f5", "tree": "40c008f1dbd45fe789b1677b452fa1317c632bb7", "parents": [ "301c793bd04242fd403cf199f3d25de0a3653534" ], "author": { "name": "Victor Mustya", "email": "victor.mustya@intel.com", "time": "Thu Apr 16 16:41:51 2026 -0700" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 16 16:45:57 2026 -0700" }, "message": "[libclc] Fix atomic_fetch_add/sub overloads for uintptr_t (#192570)\n\nThe overloads taking the memory order and/or scope parameters should\nhave the `_explicit` suffix, according to the OpenCL C specification.\n\nGitOrigin-RevId: daf814c4319b4cbf183a55df0030d0706630999e\n" }, { "commit": "301c793bd04242fd403cf199f3d25de0a3653534", "tree": "8d3aee252d5f059ea787c1db8faa642d61dc6614", "parents": [ "22290b167015b0cad1b55ae0237648cf56228926" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Wed Apr 15 08:45:49 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 14 17:47:04 2026 -0700" }, "message": "[libclc] Only add test folder when LLVM_INCLUDE_TESTS is ON (#191948)\n\nGitOrigin-RevId: c1fc739b5d6b731bf0fb4477d4605102407a933b\n" }, { "commit": "22290b167015b0cad1b55ae0237648cf56228926", "tree": "85d8fd4345b6671eb225b8fcb13b14ba95d2d1cf", "parents": [ "308250d103bb30a71716c7fedc41883c88cfdd0a" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Wed Apr 15 08:45:29 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 14 17:46:57 2026 -0700" }, "message": "[libclc][CMake][NFC] Delete dead code LLVM_PACKAGE_VERSION (#191943)\n\nUse of LLVM_PACKAGE_VERSION in AddLibclc.cmake was dropped by e20ae16ce672.\n\nGitOrigin-RevId: 19d1b348f02f21fba5ca17bef5001d2a063ee430\n" }, { "commit": "308250d103bb30a71716c7fedc41883c88cfdd0a", "tree": "df83d25419880d31b7dbd3d2a7779ef97b19a44d", "parents": [ "0ccfb3748fc42fbf171fdd21486a394743881535" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Wed Apr 15 08:45:04 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Apr 14 17:46:47 2026 -0700" }, "message": "[runtimes] Aggregate per-target runtime checks in top-level check-${runtime_name} (#191743)\n\nWhen a per-target runtime build exports a\ncheck-${runtime_name}-${target} proxy, make the top-level\ncheck-${runtime_name} target depend on it, creating\ncheck-${runtime_name} on demand (it may not exist).\n\nThis applies regardless of whether the runtime comes from the default\nLLVM_ENABLE_RUNTIMES set or from a target-specific\nRUNTIMES_\u003ctarget\u003e_LLVM_ENABLE_RUNTIMES override.\n\nThis allows a single `check-${runtime_name}` command to trigger all\nper-target tests for that runtime.\n\nGitOrigin-RevId: 18f63d1375d00fa8cac9951340e2ca0f68b4b6fa\n" }, { "commit": "0ccfb3748fc42fbf171fdd21486a394743881535", "tree": "80775d098b53c30242553dd611fcda33ec39e7aa", "parents": [ "d309d31a8dd222d83bd0242cc4d3846d7e1b88a3" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Tue Apr 14 10:40:51 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Apr 13 19:45:16 2026 -0700" }, "message": "Revert \"[libclc][CMake] Use clang/llvm-ar on Windows (#186726)\" (#191745)\n\nThis reverts commit 4abb927bacf37f18f6359a41639a6d1b3bffffb5.\n\nThe code is not needed since 121f5a96ff38 because the C compiler is now\nalways just-built clang in in-tree build. In addition, CMAKE_AR is\nllvm-ar and CMAKE_RANLIB is llvm-ranlib.\n\nGitOrigin-RevId: 12f636d4425ae2c89479ae65a581b8bc5ef53571\n" }, { "commit": "d309d31a8dd222d83bd0242cc4d3846d7e1b88a3", "tree": "fbb422a7cfa04f03f02aebd61ef81b85287b0403", "parents": [ "13d7a2f99f64aadbe4c0f2db600c08d174ea0d8a" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Mon Apr 13 08:19:08 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Apr 12 17:20:45 2026 -0700" }, "message": "[libclc] Enable LLVM_RUNTIME_TARGETS in build system (#189892)\n\nlibclc target is now passed in from LLVM_RUNTIME_TARGETS.\n\nThe old configure flow based on `-DLLVM_ENABLE_RUNTIMES\u003dlibclc` is\ndeprecated because libclc no longer has a default target.\n`-DLLVM_ENABLE_RUNTIMES\u003dlibclc -DLLVM_RUNTIME_TARGETS\u003d\"\u003ctarget-triple\u003e\"`\nstill works but it is considered legacy.\n\nThe new standard build requires:\nEach target must now be selected explicitly on the CMake command line\nthrough the runtimes target-specific cache entry and\nLLVM_RUNTIME_TARGETS.\nFor example:\n-DRUNTIMES_amdgcn-amd-amdhsa-llvm_LLVM_ENABLE_RUNTIMES\u003dlibclc\n-DLLVM_RUNTIME_TARGETS\u003d\"amdgcn-amd-amdhsa-llvm\"\n-DRUNTIMES_nvptx64-nvidia-cuda_LLVM_ENABLE_RUNTIMES\u003dlibclc\n-DLLVM_RUNTIME_TARGETS\u003d\"nvptx64-nvidia-cuda\"\n-DRUNTIMES_clspv--_LLVM_ENABLE_RUNTIMES\u003dlibclc\n-DLLVM_RUNTIME_TARGETS\u003d\"clspv--\"\n-DRUNTIMES_clspv64--_LLVM_ENABLE_RUNTIMES\u003dlibclc\n-DLLVM_RUNTIME_TARGETS\u003d\"clspv64--\"\n-DRUNTIMES_spirv-mesa3d-_LLVM_ENABLE_RUNTIMES\u003dlibclc\n-DLLVM_RUNTIME_TARGETS\u003d\"spirv-mesa3d-\"\n-DRUNTIMES_spirv64-mesa3d-_LLVM_ENABLE_RUNTIMES\u003dlibclc\n-DLLVM_RUNTIME_TARGETS\u003d\"spirv64-mesa3d-\"\n\nTo build multiple targets, pass them as a semicolon-separated list in\n`LLVM_RUNTIME_TARGETS` and provide a matching\n`RUNTIMES_\u003ctarget-triple\u003e_LLVM_ENABLE_RUNTIMES\u003dlibclc` entry for each\ntarget.\n\nUpdated README.md to document the new build flow.\n\n---------\n\nCo-authored-by: Copilot \u003c175728472+Copilot@users.noreply.github.com\u003e\nGitOrigin-RevId: 121f5a96ff38ec0c5d5f7274b1b0ca0df26a1cee\n" }, { "commit": "13d7a2f99f64aadbe4c0f2db600c08d174ea0d8a", "tree": "32d246f8fa3f3c2d07cfefc44ffc38d7956c51d6", "parents": [ "68674c3a68931d18ae280e54170120a608accec4" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Mon Apr 13 08:16:06 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Sun Apr 12 17:20:37 2026 -0700" }, "message": "[libclc] Refine generic __clc_get_sub_group_size with fast full sub-group path (#188895)\n\nAdd a fast path for the common case that total work-group size is\nmultiple of max sub-group size.\n\nThe fallback path is ported from amdgpu/workitem/clc_get_sub_group_size.cl.\n\nCompiler can generate predicated instructions for the fallback path to\navoid branches.\n\nGitOrigin-RevId: 7b94b9ae13c69075a85d3af8b987a38610fbce5a\n" }, { "commit": "68674c3a68931d18ae280e54170120a608accec4", "tree": "cb50a4f02d102a0897e31c3404e75aad91de5a12", "parents": [ "24422dcfd7eb5f240db34cf7ffddc7da9069fcfb" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Thu Apr 02 22:18:17 2026 +0200" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Apr 02 13:20:39 2026 -0700" }, "message": "libclc: Use special division for atan2 for DAZ (#190248)\n\nThe AMDGPU DAZ fdiv works fine in this case, so there\u0027s\nmaybe something better we could do here.\n\nGitOrigin-RevId: eed1f2749d0defc815344531e9e234e620323eae\n" }, { "commit": "24422dcfd7eb5f240db34cf7ffddc7da9069fcfb", "tree": "473508c785ac17e6f7f358ed9f99827bf6687b4c", "parents": [ "b36ae2c5f06eaa2353130f6f2bf455f99a412009" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Sat Mar 28 00:16:07 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 16:19:26 2026 -0700" }, "message": "libclc: Simplify fract implementation (#189080)\n\nGitOrigin-RevId: 15bc5b0267ad4b8206b108ecd596a842020b9ba9\n" }, { "commit": "b36ae2c5f06eaa2353130f6f2bf455f99a412009", "tree": "751e01f37dfb01cb095bc5caea3c912b3a7fd05d", "parents": [ "e3620a2ade18ccfbdc9e56ad6797e8ea79747013" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat Mar 28 07:06:23 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 16:08:01 2026 -0700" }, "message": "[libclc] Fix llvm-spirv dependency when llvm-spirv is built in-tree (#188896)\n\nWhen SPIRV-LLVM-Translator is built in-tree (i.e., placed in\nllvm/projects folder), llvm-spirv target exists.\n\nDrop legacy llvm-spirv_target dependency (was for non-runtime build) and\nadd llvm-spirv to runtimes dependencies.\n\nGitOrigin-RevId: 7be9972cb2b11bdf70459459cb7f5dcdb0839ea7\n" }, { "commit": "e3620a2ade18ccfbdc9e56ad6797e8ea79747013", "tree": "19122c02ae5eb188cd037691f935d87eaa107e1b", "parents": [ "6b7f3bcbe308ad52ff60ddf242151f013c605b61" ], "author": { "name": "Romaric Jodin", "email": "rjodin@google.com", "time": "Fri Mar 27 11:36:31 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 03:40:58 2026 -0700" }, "message": "libclc: do not use int64 in sincos helpers (#188056)\n\nint64 is optional, thus we do not want to force its usage for clspv.\n\nGitOrigin-RevId: e0a1e78738b2ac2644db4b0a7d27dffc621aacdb\n" }, { "commit": "6b7f3bcbe308ad52ff60ddf242151f013c605b61", "tree": "4cc4efafbd702db19eb90e34dbdb3381aa4634c9", "parents": [ "7d0fbe6cc1b8cb03ef8e32e35a33ea2a6e460860" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 27 10:47:44 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 02:48:57 2026 -0700" }, "message": "libclc: Partially implement nonuniform subgroup reduce functions (#188929)\n\nFor AMDGPU these are identical to the uniform case. Stub out the missing\ncases with traps to avoid test failures from undefined symbols while\nkeeping the structure consistent.\n\nGitOrigin-RevId: 35781a7d43fbd61cf18fd63923df95cb1d2a861c\n" }, { "commit": "7d0fbe6cc1b8cb03ef8e32e35a33ea2a6e460860", "tree": "9405d34cccaebf17e666e435bc970f0777e38acd", "parents": [ "b1f2d028935f02bee29ef8f47965f4700061dfde" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 27 09:53:35 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 01:55:15 2026 -0700" }, "message": "libclc: Add work group scan functions (#188829)\n\nGitOrigin-RevId: 56e1510d21f0782af95b32ee41fe799fa63b93d1\n" }, { "commit": "b1f2d028935f02bee29ef8f47965f4700061dfde", "tree": "90281cc48135a8081d0c12e08c268ecbe074f3cd", "parents": [ "e5316bf9c5fdab84d88afcbff6e8df231b9a155a" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 27 09:37:27 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 01:40:03 2026 -0700" }, "message": "libclc: Add subgroup scan functions (#188828)\n\nAdd the base implementation using ds_swizzle which should work\non all subtargets. There are at least 2 more paths available for\nnewer targets.\n\nGitOrigin-RevId: 1a32a4185b84bba30474305cd915d1fbcfa1a352\n" }, { "commit": "e5316bf9c5fdab84d88afcbff6e8df231b9a155a", "tree": "f3b0d96fbf3328a3e6d7f0f2c1eb74633af170a1", "parents": [ "a2482fd0579eab295bc48bc27b67378a434e7670" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 27 08:59:00 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 01:00:24 2026 -0700" }, "message": "libclc: Update atan2pi (#188707)\n\nThis was originally ported from rocm device libs in\n37406a209c75a09f850cd5e5498568d34a6f05d1. Merge in more\nrecent changes.\n\nGitOrigin-RevId: 8430e9e8f043d2ea28925e299157f760711cbedb\n" }, { "commit": "a2482fd0579eab295bc48bc27b67378a434e7670", "tree": "821134e398871fe9d92da3b512b5a3343fbed763", "parents": [ "7f0e5bc626c7fcad87f599527ad02ce20c64f147" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 27 08:27:36 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 27 00:30:50 2026 -0700" }, "message": "libclc: Fix missing overloads for atomic_fetch_add/sub (#188478)\n\nGitOrigin-RevId: 8e9e9228d45463ae52a9dde006b26fb921cd23ea\n" }, { "commit": "7f0e5bc626c7fcad87f599527ad02ce20c64f147", "tree": "4907667eba0f913c9847d85cf09a55ba31330bed", "parents": [ "e1965d9283ec2131e8880290ddaca84e31b96b8f" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Thu Mar 26 11:20:45 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 26 03:25:31 2026 -0700" }, "message": "libclc: Implement get_sub_group_id with get_local_linear_id (#188713)\n\nGitOrigin-RevId: 9a51b1eebe6b0d20b9a62a411f036aa50268c1d1\n" }, { "commit": "e1965d9283ec2131e8880290ddaca84e31b96b8f", "tree": "4ac76cd259f7e1b40899fb8b369a97f18148dd30", "parents": [ "02886eca15060554e2c00f04fae4acac2e0e421b" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Thu Mar 26 11:00:17 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 26 03:02:11 2026 -0700" }, "message": "libclc: Update atan2 (#188706)\n\nThis was originally ported from rocm device libs in\nf9caca8b9dd26a9e7a13e5ca8be57100578e3432. Update for more\nrecent changes.\n\nGitOrigin-RevId: 158282b17652acd3a5ab282dbcc54967fd9837d7\n" }, { "commit": "02886eca15060554e2c00f04fae4acac2e0e421b", "tree": "c722433123e6bc76f8ee1ffaec98334f28e8551b", "parents": [ "63da043b209e32995de1b2daee48c17d395229eb" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Thu Mar 26 09:31:06 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 26 01:35:22 2026 -0700" }, "message": "libclc: Update erfc (#188570)\n\nThis was originally ported from rocm device libs in\n2cf4d5f31204c873d76953bfe3c8b5602b29e789. Merge in more\nrecent changes.\n\nGitOrigin-RevId: 3aeea1037144d5f825b879147908ee3cdc498579\n" }, { "commit": "63da043b209e32995de1b2daee48c17d395229eb", "tree": "addf35aabf0f7c2227efdc69330ee5c8daaf0ab7", "parents": [ "003c547d72d8559c73e2015a3c48ba5bf922e51f" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Thu Mar 26 09:21:49 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Thu Mar 26 01:25:54 2026 -0700" }, "message": "libclc: Update erf (#188569)\n\nThis was originally ported from rocm device libs in\nc374cb76f467f01a3f60740703f995a0e1f7a89a. Merge in more\nrecent changes. Also enables vectorization.\n\nGitOrigin-RevId: 2f11484baa53b0cd95d2368b03d9cac5e8dae8a1\n" }, { "commit": "003c547d72d8559c73e2015a3c48ba5bf922e51f", "tree": "21b08ebf79cf6e79bf77ee49f923b6d9daabc2a7", "parents": [ "ef5306df79b84b2c95d71e3283db94db02751585" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 23:59:21 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 16:00:55 2026 -0700" }, "message": "libclc: Use prefetch builtin to implement default prefetch (#188491)\n\nGitOrigin-RevId: 180ae2f2b5674947f302ff73f14a47cc81ef69e9\n" }, { "commit": "ef5306df79b84b2c95d71e3283db94db02751585", "tree": "11545351c79d844df9274b103c398648b8d2ca8e", "parents": [ "08f93a583b0c37d7bbce48f9236497b85a327064" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 23:59:06 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 16:00:48 2026 -0700" }, "message": "libclc: Fix directly adding vector booleans (#188540)\n\nVector compares return -1 instead of 1, so explicitly select a 0 or 1\ninstead of directly adding the result.\n\nGitOrigin-RevId: fce6d53850714ba521951f586cdfe97438b52130\n" }, { "commit": "08f93a583b0c37d7bbce48f9236497b85a327064", "tree": "6a552e3d294f2c97c99a3f91dbd510782323933c", "parents": [ "fcf3ad5f0ae1d8a43acdc4ecf3a16fd7dfa763e1" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 23:58:40 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 16:00:41 2026 -0700" }, "message": "libclc: Fix amdgpu sub_group_broadcast for double (#188594)\n\nGitOrigin-RevId: e46a3299ac455f6c445c48db0674e464b93e69c5\n" }, { "commit": "fcf3ad5f0ae1d8a43acdc4ecf3a16fd7dfa763e1", "tree": "67cf4405652663048341a10aea233c9fc536462c", "parents": [ "b9b2f4199d21dd5a350034baa61a2df2848e965c" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 22:02:56 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 14:06:23 2026 -0700" }, "message": "libclc: Fix amdgpu subgroup reduce for max u64 (#188598)\n\nGitOrigin-RevId: dd57b4552202f3ef816c6d49781d6f1550e98484\n" }, { "commit": "b9b2f4199d21dd5a350034baa61a2df2848e965c", "tree": "d4c836ee910b4d7c13750a9981a208b744c51f62", "parents": [ "e2670ae4d6410c8a774e4f6f96916cce1cd6c04f" ], "author": { "name": "Romaric Jodin", "email": "rjodin@google.com", "time": "Wed Mar 25 19:23:18 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 11:25:22 2026 -0700" }, "message": "libclc: clspv does not need workaround for flush_if_daz (#188555)\n\nGitOrigin-RevId: b492b55b8cad286629b78a8274a98bd8dfcd6486\n" }, { "commit": "e2670ae4d6410c8a774e4f6f96916cce1cd6c04f", "tree": "d44379c76ffefd4ebf6a7f2718f37e83e147021e", "parents": [ "214f3e96687ad91705609800ec1674cf5e74534b" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 13:51:15 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 05:56:11 2026 -0700" }, "message": "libclc: Update acospi (#188455)\n\nThis was originally ported from rocm device libs in\n084124a8fab6fd71d49ac4928d17c3ef8b350ead. Merge in more\nrecent changes.\n\nGitOrigin-RevId: d642f598ba421d3d3de791f4854a1131154c3e16\n" }, { "commit": "214f3e96687ad91705609800ec1674cf5e74534b", "tree": "a6999c39ced3b125d28c12d3364e89b2b6d52598", "parents": [ "96e65feaae17169dc6cc74705c76cbf8ec8b9781" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 11:27:09 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 03:30:39 2026 -0700" }, "message": "libclc: Update asinpi (#188454)\n\nThis was originally ported from rocm device libs in\neea0997566cad3be13df897a06dfda74cbd684b9. Update for more recent\nchanges.\n\nGitOrigin-RevId: 0effa7caf1263b4ca7fccb2b6b1953b363923375\n" }, { "commit": "96e65feaae17169dc6cc74705c76cbf8ec8b9781", "tree": "eb5a26d897f544a373f3bc9bdf1e5c37d711add2", "parents": [ "c7cce36f44d88088d6ada394d7a6144c1114be4b" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 09:49:01 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 01:50:19 2026 -0700" }, "message": "libclc: Update atanpi (#188424)\n\nThis was originally ported from rocm device libs in\n03dc366e79cd01afe0bbfad2a7ede3087d6c9356. Merge in more\nrecent changes.\n\nGitOrigin-RevId: ebe2454dc518d8edcd352fcb2038797eccc13d42\n" }, { "commit": "c7cce36f44d88088d6ada394d7a6144c1114be4b", "tree": "4093c5e8506d99c5a215043c0dc0c95fc7c3deaa", "parents": [ "89b508568ad7d72012a700f1463101ea7aa911cc" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 09:39:45 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 01:40:35 2026 -0700" }, "message": "libclc: Force assuming fast float fma for AMDGPU (#188245)\n\nCurrently the build uses the default dummy target, which assumes\nFMA is slow. Force this to assume fast fma, which is the case on\nany remotely new hardware. In the future if we want better support\nfor older targets, there should be a separate build of the math\nfunctions for the slow fma case.\n\nGitOrigin-RevId: 0e49be0d985e30592b400961e77e2582082ed196\n" }, { "commit": "89b508568ad7d72012a700f1463101ea7aa911cc", "tree": "3129a42d1055e1b8582e1ee2564934fc320f42a4", "parents": [ "98d97c5ca46162f4a63154e9ff257fdded47af3e" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Wed Mar 25 09:24:23 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Wed Mar 25 01:25:35 2026 -0700" }, "message": "libclc: Unify fast FMA controls (#188244)\n\nThis was defined in multiple places with different names. Consolidate\non one, with a gentype wrapper for it. Also set the value based on the\nstandard FP_FAST_FMA* macros.\n\nGitOrigin-RevId: c1cd0d57c319c28cb0ded983eb7f1fb07163faea\n" }, { "commit": "98d97c5ca46162f4a63154e9ff257fdded47af3e", "tree": "35d17770899b0565b3188b8c068bc09b85cdc4f5", "parents": [ "bd34e424d7e2eb9b546ceeeaa7c05ca9a850a362" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 13:07:33 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 05:10:52 2026 -0700" }, "message": "libclc: Update atanh (#188225)\n\nGitOrigin-RevId: 46205ff9b4cbe2bc18cc6600f0acf76bb4810053\n" }, { "commit": "bd34e424d7e2eb9b546ceeeaa7c05ca9a850a362", "tree": "a8d78947d73662c47f3b8c76c3c486c37e474e46", "parents": [ "d9cedeb02eba1660ae5d921548dad7949229ff0c" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 13:07:00 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 05:10:42 2026 -0700" }, "message": "libclc: Update acosh (#188224)\n\nlibclc: Update acosh\n\nThis was originally ported from rocm device libs\nin ca4d382e119e1389c83dbb07d9ca0085e88b2944. Merge in\nmore recent changes.\n\nRemove unused ep_log.\n\nGitOrigin-RevId: 8b224162fa6aa6ebb6dd963c89b45571b1b0c8e5\n" }, { "commit": "d9cedeb02eba1660ae5d921548dad7949229ff0c", "tree": "b7e2de5efa9f42ea184fa4927525c94edebae3ae", "parents": [ "5f0d93f6ae7173569e3e88ffc0adc6c910a5c5d7" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 12:20:34 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 04:25:32 2026 -0700" }, "message": "libclc: Update asinh (#188219)\n\nThis was originally ported from rocm device libs in\n2b4ef39b2f46acf29294f1fbb223ea5a243c2567. Merge in more\nrecent changes.\n\nGitOrigin-RevId: 80245f3b1471c264cdac87f1d8ceb7169f5f68d6\n" }, { "commit": "5f0d93f6ae7173569e3e88ffc0adc6c910a5c5d7", "tree": "aac7c31fd8e447262a6ae4925a471d515a410b99", "parents": [ "bd5434ebc5f0253ddf02293eb6de96c7add59c7d" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 12:01:30 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 04:05:41 2026 -0700" }, "message": "libclc: Update tanh (#188215)\n\nThis was originally ported from rocm device libs in\nf51df5ba8c4512dbeb7828ac0c34f89177b551d6. Merge in more\nrecent changes.\n\nGitOrigin-RevId: f046f4518dc7c707289b2fc8b748d3034cfa4c15\n" }, { "commit": "bd5434ebc5f0253ddf02293eb6de96c7add59c7d", "tree": "206aefc4c59c795ebde754964ec5ca3516f10e19", "parents": [ "d92480081a648fac5c69f75548f6f89e780dfe38" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 11:58:59 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 04:01:38 2026 -0700" }, "message": "libclc: Update cosh (#188214)\n\nThis was originally ported from rocm device libs in\n9cb070f96a8a9af5f513ffba0a8eed362623f216. Merge in more\nrecent changes.\n\nGitOrigin-RevId: aeba5d62e5352aa3ed2f0d5a6a563ac873dde716\n" }, { "commit": "d92480081a648fac5c69f75548f6f89e780dfe38", "tree": "5d08d19b31732dc6a48b74dbd3eb7ccdccf9aab9", "parents": [ "d8e4f9607bcacb198ee24e41540cd59fa0abed92" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 11:56:05 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 04:01:27 2026 -0700" }, "message": "libclc: Update sinh (#188213)\n\nThis was originally ported from rocm device libs in\n9f7172965c650627c020264e9dbdb32d005ce69e. Merge in more\nrecent changes.\n\nGitOrigin-RevId: d491f702062d5ca14a7f633baf4b307a900fc02f\n" }, { "commit": "d8e4f9607bcacb198ee24e41540cd59fa0abed92", "tree": "e1793d76f59f83f5df05a8928012200df98429ca", "parents": [ "3f8ae83dc44ae555720aec06903f1fb53ddc088c" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 11:28:34 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 03:30:21 2026 -0700" }, "message": "libclc: Update expm1 (#188209)\n\nThis was originally ported from rocm device libs in\n900bd7eb7f5426ad13f624cbf29716afe376c878. Merge in more\nrecent changes.\n\nGitOrigin-RevId: 1b255ef5acdc5ebb14320efb135377f3b417c958\n" }, { "commit": "3f8ae83dc44ae555720aec06903f1fb53ddc088c", "tree": "b59ac2d6a0e77c7defd34d1b4c474e83b88961a4", "parents": [ "f5193bfe84b7749c35a23baf032d9ff24022d755" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 10:45:07 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 02:50:30 2026 -0700" }, "message": "libclc: Improve tgamma handling (#188066)\n\nGitOrigin-RevId: 641569de4e07f3a38af6720cabd84a91c73d8143\n" }, { "commit": "f5193bfe84b7749c35a23baf032d9ff24022d755", "tree": "4c0f2d4c2db45104eab771f60c706c45e6546524", "parents": [ "a76a5c9588d3cce713471bebd0201a0595248674" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 10:43:00 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 02:45:26 2026 -0700" }, "message": "libclc: Update lgamma_r (#188065)\n\nThis was originally ported from rocm device libs in\n0ab07e1bde7d002f1a4c30babb6241c0cc366320. Merge\nin more recent changes.\n\nGitOrigin-RevId: ff3632cdf37c94a4eb830c6cd501b3067a79163d\n" }, { "commit": "a76a5c9588d3cce713471bebd0201a0595248674", "tree": "6336d36153fcd20d90b8e9e6ab1124d077d021b9", "parents": [ "a5d146933a44c5e2da00d4217a789ceac88def72" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Tue Mar 24 15:55:10 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 01:00:04 2026 -0700" }, "message": "[libclc][NFC] Remove __CLC_BIT_INTN macro (#188023)\n\nThis macro was originally introduced in 64735ad63975 for relational\nbuilt-ins. It is functionally identical to __CLC_S_GENTYPE. Replacing it\nsimplifies gentype.inc, which is widely used in the library.\n\nGitOrigin-RevId: b929f2f0189ef3698d7b65168ff51976bb6f86e2\n" }, { "commit": "a5d146933a44c5e2da00d4217a789ceac88def72", "tree": "2ce9f85dbe19aa076ab291d67b65dc9264ade6a1", "parents": [ "da0cc9fbbdb700ed5e690cbf635f4afad69b128b" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 08:41:30 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 00:45:24 2026 -0700" }, "message": "libclc: Update log1p (#188186)\n\nGitOrigin-RevId: 1a5e52168c0bef0c5abecb89e9f18253f8a43dda\n" }, { "commit": "da0cc9fbbdb700ed5e690cbf635f4afad69b128b", "tree": "b4a475b5447f514c6efbb88a28d51aaa904afc1d", "parents": [ "39960c5385a3b152747bfdc44e82e4a13dd1125c" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 08:27:11 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 00:30:18 2026 -0700" }, "message": "libclc: Fix -cl-denorms-are-zero for rtp and rtn conversions (#188148)\n\nGitOrigin-RevId: 32c6a53b767850183208aff8b783e6bc08f21515\n" }, { "commit": "39960c5385a3b152747bfdc44e82e4a13dd1125c", "tree": "1a8825c7e3e3c785798411f4a19a895c5fd78b80", "parents": [ "5704a67e30bb81be3d5c08d207056724988e4646" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 08:11:41 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 00:15:40 2026 -0700" }, "message": "libclc: Update atan (#188095)\n\nThis was originally ported from rocm device libs in\n47882923c7b48c00d6c0ea9960b5457e957093c4. Merge in more\nrecent changes.\n\nGitOrigin-RevId: cba948b2b93cc5d51f70f0fa0d66bc5cacef9d76\n" }, { "commit": "5704a67e30bb81be3d5c08d207056724988e4646", "tree": "772fa0126b1be173e4e7c735e50b1ee18142d8e6", "parents": [ "9aa7d957aaabe20afe5e333f961362155b91e88e" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 08:11:24 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 00:15:31 2026 -0700" }, "message": "libclc: Update asin (#188094)\n\nThis was originally ported from rocm device libs in\n64a8e1b83e14836f97dab4d28dae498e897804e6. Update for more\nrecent changes.\n\nGitOrigin-RevId: ff68154480eccdfeaa18d87fd6d83906e93a85c1\n" }, { "commit": "9aa7d957aaabe20afe5e333f961362155b91e88e", "tree": "270baa4b067f4ed456ef8781c0f44e12b522ffbb", "parents": [ "606abce308d2add384481b26f9a293baa32e87c8" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 08:06:01 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 00:11:02 2026 -0700" }, "message": "libclc: Use nextup and nextdown in place of nextafter (#188141)\n\nUnfortunately it seems the optimizer isn\u0027t able to clean this\nup, so this is a code quality improvement.\n\nGitOrigin-RevId: 31aa52086fb2208f451be9e22197ae1a63cdc4af\n" }, { "commit": "606abce308d2add384481b26f9a293baa32e87c8", "tree": "7a414cf80984c7f5532b6ea8dc2d4fed9bce99e3", "parents": [ "c5cb7c524a9ae7ba155dda3cc9463ffa91bdd91a" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Tue Mar 24 08:05:38 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Tue Mar 24 00:10:47 2026 -0700" }, "message": "libclc: Simplify rtz conversion (#188137)\n\nGitOrigin-RevId: ef5658a289e24fa145663a1c22e0751d6c12991c\n" }, { "commit": "c5cb7c524a9ae7ba155dda3cc9463ffa91bdd91a", "tree": "447b159f34f6800cd4a96841354e25e1f321aafd", "parents": [ "fb945a37fa6d40e7d7b5e362df44083638170d9c" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Tue Mar 24 08:32:07 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Mar 23 17:35:19 2026 -0700" }, "message": "[libclc][NFC] Simplify __CLC_GENTYPE/__CLC_U_GENTYPE/__CLC_S_GENTYPE define in gentype.inc (#188027)\n\nReduce macro re-definition overhead.\n\nGitOrigin-RevId: 99a4d78bef28fc0a32e8be24a560ee26f459925d\n" }, { "commit": "fb945a37fa6d40e7d7b5e362df44083638170d9c", "tree": "e4c5816cdbf0056b9d6dfb8d437eab4e11bc9b72", "parents": [ "78e756ef6574965c603dbc7977a07fa724339998" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Mon Mar 23 11:08:13 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Mar 23 03:10:57 2026 -0700" }, "message": "libclc: Implement remainder with remquo\n (#187999)\n\nThis fixes conformance failures for double and\nwithout -cl-denorms-are-zero. Optimizations are\nable to eliminate the unusued quo handling without\nduplicating most of the code.\n\nGitOrigin-RevId: befad798a91ef8c74f605bdb15c352528f8d9807\n" }, { "commit": "78e756ef6574965c603dbc7977a07fa724339998", "tree": "06fba1556f6877efd91e72d1eb17e3d8748add8b", "parents": [ "7b7874f3dfba43dce02753b4a7d55061b96bfd4e" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Mon Mar 23 11:06:59 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Mon Mar 23 03:10:52 2026 -0700" }, "message": "libclc: Update remquo (#187998)\n\nThis was failing in the float case without -cl-denorms-are-zero\nand failing for double. This now passes in all cases.\n\nThis was originally ported from rocm device libs in\n8db45e4cf170cc6044a0afe7a0ed8876dcd9a863. This is mostly a port\nin of more recent changes with a few changes.\n\n- Templatification, which almost but doesn\u0027t quite enable\n vectorization yet due to the outer branch and loop.\n\n- Merging of the 3 types into one shared code path, instead of\n duplicating per type with 3 different functions implemented together.\n There are only some slight differences for the half case, which mostly\n evaluates as float.\n\n- Splitting out of the is_odd tracking, instead of deriving it from the\n accumulated quotient. This costs an extra register, but saves several\ninstructions. This also enables automatic elimination of all of the quo\n output handling when this code is reused for remainder. I\u0027m guessing\n this would be unnecessary if SimplifyDemandedBits handled phis.\n\n- Removal of the slow FMA path. I don\u0027t see how this would ever be\n faster with the number of instructions replacing it. This is really a\n problem for the compiler to solve anyway.\n\nGitOrigin-RevId: 1a9fe1769a0f9684dc9ecfcb7a2cec2d66077cc3\n" }, { "commit": "7b7874f3dfba43dce02753b4a7d55061b96bfd4e", "tree": "0ae29d3061910519f512f9e4689e775b99879ab0", "parents": [ "fb2b55a8092e733e152394bea8bd19c661e3b6f3" ], "author": { "name": "Wenju He", "email": "wenju.he@intel.com", "time": "Sat Mar 21 09:01:44 2026 +0800" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 20 18:05:17 2026 -0700" }, "message": "[libclc] Replace llvm-dis with llvm-nm in check-external-funcs.test (#187190)\n\nllvm-nm is covered by extra_deps in runtime build when\nLLVM_INCLUDE_TESTS is true.\n\nGitOrigin-RevId: 8eccc21e47fbd2a81b10d3f5041c685d273d0e97\n" }, { "commit": "fb2b55a8092e733e152394bea8bd19c661e3b6f3", "tree": "332e1c39e961b7c34ea95c4e2eb863d8adc159c8", "parents": [ "6c4ae5a527832fed31e89fbe6a1de7a90a37e8ab" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 20 12:43:44 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 20 04:45:41 2026 -0700" }, "message": "libclc: Update acos (#187666)\n\nThis was originally ported from rocm device libs in\nefeafa1bdaa715733fc100bcd9d21f93c7272368, merge in more\nrecent changes.\n\nGitOrigin-RevId: 22f5b8db125c073af66b40e966ed77d70603013b\n" }, { "commit": "6c4ae5a527832fed31e89fbe6a1de7a90a37e8ab", "tree": "01ce279278dcad64decf99ecd1689251ca885d42", "parents": [ "8edf6dd9ceec3d053d5f4032991e307254605a4a" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 20 08:34:17 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 20 00:36:59 2026 -0700" }, "message": "libclc: Override cbrt for AMDGPU (#187560)\n\nGitOrigin-RevId: c8dd82916bc3867c8e08e3605622ef23c99d671e\n" }, { "commit": "8edf6dd9ceec3d053d5f4032991e307254605a4a", "tree": "68d8fc4c636dbe3b5ceeb00b19213538b1d00e68", "parents": [ "d8f9df7b4a1eef5bd2d0480f0f2eed40f9924ab9" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 20 08:33:31 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 20 00:36:53 2026 -0700" }, "message": "libclc: Use log intrinsic for half and float cases for amdgpu (#187538)\n\nThis is pretty verbose and ugly. We\u0027re pulling the base implementation\nin for the double cases, and scalarizing it. Also fully defining the\nhalf and float cases to directly use the intrinsic, for all vector\ntypes. It would be much more convenient if we had linker based overrides\nfor the generic implementations, rather than per source file.\n\nGitOrigin-RevId: edbe8277c10433ded3b8b28375dd8b0e85d108f1\n" }, { "commit": "d8f9df7b4a1eef5bd2d0480f0f2eed40f9924ab9", "tree": "58b27dca22f5ecb1b650418cf962adf8a7359499", "parents": [ "031721d75e1856e87c9de362845c3db88e64ac62" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 20 08:33:16 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 20 00:36:47 2026 -0700" }, "message": "libclc: Rewrite log implementation as gentype inc file (#187537)\n\nFollow the ordinary gentype conventions for the log implementation,\ninstead of using a plain header. This doesn\u0027t quite yet enable\nvectorization, due to how the table is currently indexed. This should\nmake it easier for targets to selectively overload the function for\na subset of types.\n\nGitOrigin-RevId: a5de509e4ecd6ac0e2ab34bc5b96a7f76f52ccc0\n" }, { "commit": "031721d75e1856e87c9de362845c3db88e64ac62", "tree": "dd8fc2a4e2bc8d08dcfea4d2e49e7a5abfa499e5", "parents": [ "e30b0a0377b0cb03c2a2157ee653bcc2eb3cb40f" ], "author": { "name": "Matt Arsenault", "email": "Matthew.Arsenault@amd.com", "time": "Fri Mar 20 08:24:23 2026 +0100" }, "committer": { "name": "Copybara-Service", "email": "copybara-worker@google.com", "time": "Fri Mar 20 00:26:10 2026 -0700" }, "message": "libclc: Update trigpi functions (#187579)\n\nThese were originally ported from rocm device\nlibs in bc81ebefb7d9d9d71d20bfee2ce4cccb09701e9b.\nMerge in more recent changes.\n\nGitOrigin-RevId: 421bf13e4bf1c19a0cc48fbb0af63b902d3118b1\n" } ], "next": "e30b0a0377b0cb03c2a2157ee653bcc2eb3cb40f" }