{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":208145128,"defaultBranch":"main","name":"iree","ownerLogin":"iree-org","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2019-09-12T20:57:39.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/107954215?v=4","public":true,"private":false,"isOrgOwned":true},"refInfo":{"name":"","listCacheKey":"v0:1716923167.0","currentOid":""},"activityList":{"items":[{"before":"74eb54e578b801f082da6957d38bb5a52cf9f215","after":"f485477a60fdbcd78e964c5d3d678ae1e2841ab0","ref":"refs/heads/gh-pages","pushedAt":"2024-05-28T22:49:53.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed aa0bc40 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed aa0bc40 with MkDocs version: 1.6.0"}},{"before":"29e70abcd352a4d037509e6ecf5e52507b738f4e","after":"aa0bc40235adafabbfc1d73cac54b97dba856841","ref":"refs/heads/main","pushedAt":"2024-05-28T22:46:27.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"qedawkins","name":"Quinn Dawkins","path":"/qedawkins","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/22101546?s=80&v=4"},"commit":{"message":"[Codegen][GPU] Add pass to fuse and hoist scf.forall ops (#17505)\n\nThis pass greedily fuses parallel loops together and tries to hoist them\r\nout of serial loops. It is left as TODO to include greedy fusion of\r\nuntiled consumers.","shortMessageHtmlLink":"[Codegen][GPU] Add pass to fuse and hoist scf.forall ops (#17505)"}},{"before":"a6a56a97ae070c2a1a290841e99ec1a396e99101","after":"29e70abcd352a4d037509e6ecf5e52507b738f4e","ref":"refs/heads/main","pushedAt":"2024-05-28T19:48:55.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"ScottTodd","name":"Scott Todd","path":"/ScottTodd","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/4010439?s=80&v=4"},"commit":{"message":"Update onnx package version minimum to 1.16.0. (#17504)\n\nAs discovered on\r\nhttps://github.com/iree-org/iree/pull/17476#discussion_r1617673191, this\r\nis needed to avoid an error running `iree-import-onnx` on certain test\r\nfiles:\r\n\r\n> onnx.onnx_cpp2py_export.checker.ValidationError: Your model ir_version\r\n10 is higher than the checker's (9).\r\n\r\nIf we care about supporting older onnx package versions, we could add\r\nsome fallback code to `iree-import-onnx`.\r\n\r\nWe may also want to update the version used in torch-mlir?\r\nhttps://github.com/llvm/torch-mlir/blob/e0a5adb1db38b0072c44b87570bc530eb3b324ad/setup.py#L264","shortMessageHtmlLink":"Update onnx package version minimum to 1.16.0. (#17504)"}},{"before":"8a25e9b1512aac14466c99d393ca419e92d8f2d4","after":"74eb54e578b801f082da6957d38bb5a52cf9f215","ref":"refs/heads/gh-pages","pushedAt":"2024-05-28T19:40:36.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed a6a56a9 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed a6a56a9 with MkDocs version: 1.6.0"}},{"before":"3d1364e1b184492165db2a1bf3556b281e8fcea3","after":"a6a56a97ae070c2a1a290841e99ec1a396e99101","ref":"refs/heads/main","pushedAt":"2024-05-28T19:35:46.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"IanWood1","name":"Ian Wood","path":"/IanWood1","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/75152913?s=80&v=4"},"commit":{"message":"Add `LinalgFusionInterface` to support fusion for linalg_ext ops (added `scatter` and `reverse`) (#17428)\n\n`LinalgFusionOpInterface` allows for fusion of both `Linalg` and\r\n`LinalgExt` operations. The new interface provides access to methods\r\nessential for performing fusion, allowing existing fusion logic to be\r\nused with `LinalgExt` operations.\r\n\r\nAs noted in #17392, it probably makes sense to move this into the\r\n`TilingInterface` + probably make it a bit more abstracted\r\n\r\n\r\n#### Changes\r\n- **`LinalgFusionOpInterface`**: Interface for fusion operations for\r\nboth `Linalg` and `LinalgExt` ops.\r\n - Implements methods to access indexing maps (or null\r\n- **Implementation for Linalg Ops**: The interface is implemented for\r\nstandard Linalg operations by forwarding to preexisting methods (e.g\r\n`getIndexingMaps()`). No changes to the ops themselves.\r\n- **Implementation for LinalgExt Ops**: The interface currently only\r\nimplemented for `iree_linalg_ext.scatter/reverse`.\r\n\r\n---------\r\n\r\nSigned-off-by: Ian Wood ","shortMessageHtmlLink":"Add LinalgFusionInterface to support fusion for linalg_ext ops (add…"}},{"before":"b44131d6a1c4171ad33ecaa426da8e1cf212b6b0","after":"8a25e9b1512aac14466c99d393ca419e92d8f2d4","ref":"refs/heads/gh-pages","pushedAt":"2024-05-28T19:26:04.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 3d1364e with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 3d1364e with MkDocs version: 1.6.0"}},{"before":"ab8f6680e20b2dbc3cb497656321302732ea0c19","after":"3d1364e1b184492165db2a1bf3556b281e8fcea3","ref":"refs/heads/main","pushedAt":"2024-05-28T19:22:45.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"qedawkins","name":"Quinn Dawkins","path":"/qedawkins","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/22101546?s=80&v=4"},"commit":{"message":"[Codegen][GPU] Add pattern to lower iree_gpu.multi_mma to intrinsics (#17457)\n\nThis takes multi_mma ops that have already been unrolled to a single\r\nintrinsic and simply lowers them to the intrinsic using the interfaced\r\nattribute carried by the op. Inserts shape casts at the boundary to\r\nmatch the expected type of the intrinsic.","shortMessageHtmlLink":"[Codegen][GPU] Add pattern to lower iree_gpu.multi_mma to intrinsics (#…"}},{"before":"5621283ad1df9adbb3ea047696926e13c0796b25","after":"b44131d6a1c4171ad33ecaa426da8e1cf212b6b0","ref":"refs/heads/gh-pages","pushedAt":"2024-05-28T19:08:34.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed e33ca89 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed e33ca89 with MkDocs version: 1.6.0"}},{"before":"b375644e638db9852eb54e64b77de9cfc46d180e","after":null,"ref":"refs/heads/revert-17446-transpose_rebase","pushedAt":"2024-05-28T19:06:07.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"bjacob","name":"Benoit Jacob","path":"/bjacob","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/79535?s=80&v=4"}},{"before":"e33ca8920e33f0108385a486884ff67b1aaacee5","after":"ab8f6680e20b2dbc3cb497656321302732ea0c19","ref":"refs/heads/main","pushedAt":"2024-05-28T19:06:06.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"bjacob","name":"Benoit Jacob","path":"/bjacob","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/79535?s=80&v=4"},"commit":{"message":"Revert \"Data tiling: transpose narrow-N into narrow-M\" (#17503)\n\nReverts iree-org/iree#17446\r\n\r\nReason: postsubmit failures on arm64,\r\nhttps://github.com/iree-org/iree/pull/17446#issuecomment-2135891362","shortMessageHtmlLink":"Revert \"Data tiling: transpose narrow-N into narrow-M\" (#17503)"}},{"before":"322d688348903c8e6c69586b15b7cbff298932b5","after":"e33ca8920e33f0108385a486884ff67b1aaacee5","ref":"refs/heads/main","pushedAt":"2024-05-28T19:05:03.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"Groverkss","name":"Kunwar Grover","path":"/Groverkss","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/51270680?s=80&v=4"},"commit":{"message":"[LinalgExt] Split TileAndDecomposeAttention (#17468)\n\nThis patch splits up TileAndDecomposeAttention into two separate passes.\r\nTileAttention will eventually be replaced once we have\r\nPartialOpTilingInterface implemented for it.","shortMessageHtmlLink":"[LinalgExt] Split TileAndDecomposeAttention (#17468)"}},{"before":"78291b69fd19170c0ca22d96769ca123f22b8a69","after":"5621283ad1df9adbb3ea047696926e13c0796b25","ref":"refs/heads/gh-pages","pushedAt":"2024-05-28T18:58:32.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 322d688 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 322d688 with MkDocs version: 1.6.0"}},{"before":"117cb43324f0b00d8c76fce539a985162838de6a","after":"322d688348903c8e6c69586b15b7cbff298932b5","ref":"refs/heads/main","pushedAt":"2024-05-28T18:54:28.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"qedawkins","name":"Quinn Dawkins","path":"/qedawkins","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/22101546?s=80&v=4"},"commit":{"message":"[Codegen][GPU] Add pattern to drop lead unit dims of multi_mma ops (#17456)\n\nDropping leading unit dims is a useful step that simplifies the process\r\nof lowering to intrinsics by removing the outer iteration space. The\r\ntypical lowering flow is to unroll the outer dims to 1 and then apply\r\nthis pattern to drop the unit outer dims.","shortMessageHtmlLink":"[Codegen][GPU] Add pattern to drop lead unit dims of multi_mma ops (#…"}},{"before":null,"after":"b375644e638db9852eb54e64b77de9cfc46d180e","ref":"refs/heads/revert-17446-transpose_rebase","pushedAt":"2024-05-28T18:42:43.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"bjacob","name":"Benoit Jacob","path":"/bjacob","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/79535?s=80&v=4"},"commit":{"message":"Revert \"Data tiling: transpose narrow-N into narrow-M (#17446)\"\n\nThis reverts commit 16bdaa90e2f02769db2a8949ab88e58c9443392d.","shortMessageHtmlLink":"Revert \"Data tiling: transpose narrow-N into narrow-M (#17446)\""}},{"before":"16bdaa90e2f02769db2a8949ab88e58c9443392d","after":"117cb43324f0b00d8c76fce539a985162838de6a","ref":"refs/heads/main","pushedAt":"2024-05-28T18:38:15.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"ScottTodd","name":"Scott Todd","path":"/ScottTodd","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/4010439?s=80&v=4"},"commit":{"message":"Test 'console' provider in 'tracing' job. (#16454)\n\nThis also refactors the job to use a GitHub-managed runner with less\r\nrandom GitHub Actions boilerplate for caching, env vars, etc. since the\r\n'tracing' job is only building the runtime (as of\r\nhttps://github.com/openxla/iree/pull/6709, from 2021).\r\n\r\nFixes https://github.com/openxla/iree/issues/16405\r\n\r\n---------\r\n\r\nSigned-off-by: Scott Todd ","shortMessageHtmlLink":"Test 'console' provider in 'tracing' job. (#16454)"}},{"before":"8fcfbdd30c0a4ca139ca2c87f66639165f4bd464","after":"d71f385ce84ce5e62c29bd7d92d13cdc3ac04e90","ref":"refs/heads/shared/tresleches-cpu","pushedAt":"2024-05-28T18:05:18.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"hanhanW","name":"Han-Chung Wang","path":"/hanhanW","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6949191?s=80&v=4"},"commit":{"message":"Revert \"Rank-reduce 3D->5D pack to 2D->4D pack to trigger ukernels\"\n\nThis reverts commit 57c3040bcccc2f0432a8ec934df38bab91eadeca.","shortMessageHtmlLink":"Revert \"Rank-reduce 3D->5D pack to 2D->4D pack to trigger ukernels\""}},{"before":"6c75aa1083d6f9a1fa7f2b1ddd032decc9e87aa7","after":"16bdaa90e2f02769db2a8949ab88e58c9443392d","ref":"refs/heads/main","pushedAt":"2024-05-28T17:31:48.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"ScottTodd","name":"Scott Todd","path":"/ScottTodd","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/4010439?s=80&v=4"},"commit":{"message":"Data tiling: transpose narrow-N into narrow-M (#17446)\n\n(This is a rebasing PR for https://github.com/iree-org/iree/pull/16890 )\r\n\r\nThis is a generic idea in the design of matrix multiplication\r\nimplementations: the M and N dimensions play symmetrical roles, so there\r\nis this opportunity to halve the problem space by transposition. The\r\nimmediate motivation is ukernels: we have chosen to implement narrow\r\nukernels only for the narrow-M cases, not narrow-N, in preparation for\r\nthis. This is the reason why this PR is a 5%-9% e2e speedup on multiple\r\nML models with ukernels (and a > 2x speedup on matvec microbenchmarks).\r\n\r\nThe idea should be beneficial beyond ukernels though:\r\n* With codegen (outside of ukernels), inner unit dimensions have often\r\ncaused codegen to fall off of good vectorization cases. This\r\ntransposition moves unit or generally smaller static dimensions to the\r\nouter dimensions, which will help with that.\r\n* When we get to serious distribution tiling (#16410), the reduction of\r\ngenerality will again greatly help.\r\n\r\nThis transposition is made easier by (and was all along part of the idea\r\nin) the RHS-transposition in `mmt4d` (the `t` in `mmt4d`), as generally\r\nwith matrix multiplication\r\n\r\n```\r\nB * Transpose(A) == Transpose( A * Transpose(B) )\r\n```\r\n\r\nso in `mmt4d` terms\r\n\r\n```\r\nmmt4d(B, A) == Transpose(mmt4d(A, B))\r\n```\r\n\r\nAs `pack` and `unpack` already have enough generality to perform these\r\ntranspositions, we just directly generate the right transposing `pack`\r\nand `unpack` ops. An earlier plan was to generate `linalg.transpose` and\r\nrely on a later folding pattern, but it turned out to just be simpler to\r\ndirectly generate the already-transposed `pack`, `unpack`.\r\n\r\nA legitimate question was: should this transposition be implemented at\r\n`SetEncoding` instead of at `MaterializeEncoding`? That would have been\r\nsimpler in some ways, but:\r\n* The benefit of the transposition depends on the backend, so it doesn't\r\nbelong in Flow.\r\n* SetEncoding must be reversible in case the back-end doesn't want to do\r\ndata-tiling. The transposition would be difficult to revert, and\r\ngenerally confusing in settings where it's not wanted.\r\n* The above mmt4d-specific trait simplifying the transposition only\r\nhelps since at MaterializeEncoding we know we are generating a mmt4d. We\r\ncouldn't so easily rely on that in SetEncoding.\r\n* Before MaterializeEncoding we would have to handle `linalg.generic`,\r\nnot just named matmul ops.\r\n\r\nCo-authored-by: Benoit Jacob \r\n\r\nbenchmark-extra: x86_64-dt-only, android-cpu-dt-only\r\n\r\nSigned-off-by: Alan Li \r\nCo-authored-by: Benoit Jacob ","shortMessageHtmlLink":"Data tiling: transpose narrow-N into narrow-M (#17446)"}},{"before":"2e752fa07ac3b549d9c4d92a1776ecb4683bb684","after":"78291b69fd19170c0ca22d96769ca123f22b8a69","ref":"refs/heads/gh-pages","pushedAt":"2024-05-28T13:31:53.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 46c6bf5 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 46c6bf5 with MkDocs version: 1.6.0"}},{"before":"a96f4233e1097329ee247eb7a82d8b57821396a3","after":"2e752fa07ac3b549d9c4d92a1776ecb4683bb684","ref":"refs/heads/gh-pages","pushedAt":"2024-05-27T22:03:29.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 6c75aa1 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 6c75aa1 with MkDocs version: 1.6.0"}},{"before":"1750e2bfd28f2dd369dd8c4424f87a2f3854ec15","after":"6c75aa1083d6f9a1fa7f2b1ddd032decc9e87aa7","ref":"refs/heads/main","pushedAt":"2024-05-27T21:58:06.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"qedawkins","name":"Quinn Dawkins","path":"/qedawkins","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/22101546?s=80&v=4"},"commit":{"message":"[Codegen][GPU] Allow iree_gpu.tensor_barrier to take vectors (#17479)\n\nThis allows synchronizing on vectors as well as tensors with similar\r\nsemantics. In a typical lowering flow, this will represent the\r\nread equivalent to a tensor barrier, in that a tensor barrier represents\r\na wait until all writes to a shared allocation has finished, while this\r\nrepresents a wait until all threads have read the value they need from\r\nthat shared allocation.\r\n\r\nRenames the operation to iree_gpu.value_barrier for clarity.","shortMessageHtmlLink":"[Codegen][GPU] Allow iree_gpu.tensor_barrier to take vectors (#17479)"}},{"before":"051c361f8b33094763a506f640344aed463343a9","after":"1750e2bfd28f2dd369dd8c4424f87a2f3854ec15","ref":"refs/heads/main","pushedAt":"2024-05-27T21:10:58.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"bjacob","name":"Benoit Jacob","path":"/bjacob","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/79535?s=80&v=4"},"commit":{"message":"Integrate LLVM at 855eef2abd81cb8c7543d4748353d5e378fdd4c2 (#17501)\n\nSigned-off-by: Benoit Jacob ","shortMessageHtmlLink":"Integrate LLVM at 855eef2abd81cb8c7543d4748353d5e378fdd4c2 (#17501)"}},{"before":"d89a83ff9b22643c02757023b31165986adc9034","after":"a96f4233e1097329ee247eb7a82d8b57821396a3","ref":"refs/heads/gh-pages","pushedAt":"2024-05-27T19:50:27.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 051c361 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 051c361 with MkDocs version: 1.6.0"}},{"before":"9e3d27ace8349b8875cae6d34450041f1f334122","after":"051c361f8b33094763a506f640344aed463343a9","ref":"refs/heads/main","pushedAt":"2024-05-27T19:46:12.000Z","pushType":"pr_merge","commitsCount":1,"pusher":{"login":"qedawkins","name":"Quinn Dawkins","path":"/qedawkins","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/22101546?s=80&v=4"},"commit":{"message":"NFC: Make a few loop transformations more accessible (#17489)\n\nThis moves `FuseForallOp` to the IREEGPU dialect because it generates\r\nIREEGPU operations, avoiding a global dependency on the dialect in\r\ncodegen. The `HoistForallFromFor` pattern is moved to\r\n`Codegen/Transforms` because it applies only to operations used across\r\nall codegen backends.","shortMessageHtmlLink":"NFC: Make a few loop transformations more accessible (#17489)"}},{"before":"e8bc68e5b13a3eb127213c1725b306f5914bf27d","after":"d89a83ff9b22643c02757023b31165986adc9034","ref":"refs/heads/gh-pages","pushedAt":"2024-05-27T13:34:15.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 46c6bf5 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 46c6bf5 with MkDocs version: 1.6.0"}},{"before":"055e47f13f39d75a29f55cabcfc75c9794cd0a46","after":"e8bc68e5b13a3eb127213c1725b306f5914bf27d","ref":"refs/heads/gh-pages","pushedAt":"2024-05-26T13:33:49.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 46c6bf5 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 46c6bf5 with MkDocs version: 1.6.0"}},{"before":"a076ad5dc0964b41c0bd2d7b660d041e5e13ba91","after":"055e47f13f39d75a29f55cabcfc75c9794cd0a46","ref":"refs/heads/gh-pages","pushedAt":"2024-05-25T13:31:47.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 46c6bf5 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 46c6bf5 with MkDocs version: 1.6.0"}},{"before":"c8144962204421e1b15e4cc89e4b4a5e92d191a4","after":"1b9a088dd9a90b576783e33546a4f6cd9ff84635","ref":"refs/heads/shared/tresleches-united","pushedAt":"2024-05-25T07:48:26.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"raikonenfnu","name":"Stanley Winata","path":"/raikonenfnu","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/68087699?s=80&v=4"},"commit":{"message":"Add gfx1150 support.","shortMessageHtmlLink":"Add gfx1150 support."}},{"before":"b3f859e162f95290027174fa60dfcd061c9ff883","after":"a076ad5dc0964b41c0bd2d7b660d041e5e13ba91","ref":"refs/heads/gh-pages","pushedAt":"2024-05-24T13:38:32.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"Deployed 46c6bf5 with MkDocs version: 1.6.0","shortMessageHtmlLink":"Deployed 46c6bf5 with MkDocs version: 1.6.0"}},{"before":"008add90d3de1333e9dcacc16eb9583989cfc492","after":"46c6bf5660022068cb1078bd7eeae527b13b30c2","ref":"refs/heads/latest-snapshot","pushedAt":"2024-05-24T13:33:20.000Z","pushType":"push","commitsCount":10,"pusher":{"login":"iree-github-actions-bot","name":null,"path":"/iree-github-actions-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/62120345?s=80&v=4"},"commit":{"message":"[CPU] Add support for pack ukernel preparation. (#17472)\n\nThe pack ukernel only works for 2D cases. Similar to what's happening in\r\nbatch_mmt4d, The revision adds the support to convert 3D pack into 2D\r\npack.\r\n\r\n---------\r\n\r\nSigned-off-by: hanhanW ","shortMessageHtmlLink":"[CPU] Add support for pack ukernel preparation. (#17472)"}},{"before":"2486565c4aa704f3b09b93e5d4d678a562892524","after":null,"ref":"refs/heads/nanobind20","pushedAt":"2024-05-24T02:13:31.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"stellaraccident","name":"Stella Laurenzo","path":"/stellaraccident","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/494208?s=80&v=4"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"djE6ks8AAAAEVlkkPgA","startCursor":null,"endCursor":null}},"title":"Activity · iree-org/iree"}