Commit Graph

  • 49b55af619 jit: simpler free_intermediates (#16099) nimlgen 2026-05-08 19:08:33 +03:00
  • 0f46c08582 div mixin cleanups (#16100) chenyu 2026-05-08 12:05:37 -04:00
  • 235044c9d8 Ops.IDIV -> Ops.CDIV, Ops.MOD -> Ops.CMOD (#16093) chenyu 2026-05-07 23:18:15 -04:00
  • faabe6aa42 nv: remaining firmware from /lib/firmware (#16088) Christopher Milan 2026-05-07 20:07:43 -07:00
  • 7ef901a81d llm: moe speedup (#16059) b1tg 2026-05-08 10:06:35 +08:00
  • 80da8a4b9c add spec to main tinygrad repo (#16092) George Hotz 2026-05-07 18:52:49 -07:00
  • 83eaefcd0f onnx: deduplicate simple proto parsers (#16085) June 2026-05-07 18:44:27 -07:00
  • c106c73e51 remove the gate from index (#16081) George Hotz 2026-05-07 18:42:00 -07:00
  • d11f4d0ec2 fix: don't copy on slice of DP weight (#16089) wozeparrot 2026-05-07 20:58:01 -04:00
  • 1d1b726cf6 hotfix: disable flaky framework pytest George Hotz 2026-05-07 17:05:06 -07:00
  • 9a6f7f7576 nv: look for fmc firmware in /lib/firmware (#16080) Christopher Milan 2026-05-07 15:08:27 -07:00
  • b796bbae87 fix valid in indexing tests (#16087) George Hotz 2026-05-07 14:11:28 -07:00
  • 4d1a9dca41 fix: don't copy precompiled custom kernel outputs (#16084) wozeparrot 2026-05-07 17:02:38 -04:00
  • f9083cf901 use subactions for benchmark.yml process replay [pr] (#13396) qazal 2026-05-07 21:46:25 +03:00
  • 2f0aa884d5 tinygpu: minimal is macos13 for resets (#16075) nimlgen 2026-05-07 21:25:56 +03:00
  • 072db9924c div to mixin (#16078) chenyu 2026-05-07 12:52:37 -04:00
  • 516b00e286 mod and fmod to mixin (#16077) chenyu 2026-05-07 12:13:39 -04:00
  • a9a87ad8fd viz/cli: less flags (#16076) qazal 2026-05-07 18:22:40 +03:00
  • f813a04b3f viz: pickle path in str (#16073) qazal 2026-05-07 12:49:21 +03:00
  • 730fa66bf3 llama speed 6 (#16071) wozeparrot 2026-05-06 23:51:03 -04:00
  • 7b91f7c90c nv: look for gsp firmware in /lib/firmware (#16068) Christopher Milan 2026-05-06 18:35:47 -07:00
  • 8e84317743 the renderer part of gate moving from index to load/store (#16064) George Hotz 2026-05-06 13:47:04 -07:00
  • ef085304bc stronger divmod_recombine (#16066) chenyu 2026-05-06 15:41:54 -04:00
  • d7d32d82ee viz/cli: print first uop with DEBUG=6 (#16065) qazal 2026-05-06 21:39:34 +03:00
  • af4140f3be fix divmod recombine for floordiv (#16062) chenyu 2026-05-06 14:22:42 -04:00
  • b9be9fbc77 Merge branch 'master' into move_gates_to_load_store move_gates_to_load_store George Hotz 2026-05-06 10:06:31 -07:00
  • 2ccefa11ec does this pass? George Hotz 2026-05-06 09:30:58 -07:00
  • c6ad3d3ac2 better divmod late rewrite (#16061) chenyu 2026-05-06 11:31:48 -04:00
  • aaabe42373 relax fold_divmod_general (#16058) chenyu 2026-05-05 21:37:56 -04:00
  • 1de14cf33a am: autogen soc (#16055) Christopher Milan 2026-05-05 17:39:43 -07:00
  • 869eae6b37 fix double div rewrites (#16054) chenyu 2026-05-05 19:34:35 -04:00
  • bd06ea9f97 am: simplify import_module (#16046) Christopher Milan 2026-05-05 16:25:53 -07:00
  • 95b0a651c2 fix decomp George Hotz 2026-05-05 15:46:42 -07:00
  • 795501e1da fix device in null graph events (#16053) qazal 2026-05-06 01:44:08 +03:00
  • ab6218bc92 llama mp fixes (#16050) wozeparrot 2026-05-05 18:35:32 -04:00
  • 76606eb386 push George Hotz 2026-05-05 15:34:05 -07:00
  • e74bf441f0 Revert "remove legacy stuff" George Hotz 2026-05-05 15:33:43 -07:00
  • 34fe37d64e use FLOORDIV and FLOORMOD (#16048) chenyu 2026-05-05 18:32:54 -04:00
  • 661eb76309 fix f2f George Hotz 2026-05-05 15:24:37 -07:00
  • a0c04a5e35 remove legacy stuff George Hotz 2026-05-05 15:14:20 -07:00
  • 13e0fbaba6 fix webgpu and some edge cases George Hotz 2026-05-05 15:07:28 -07:00
  • 58a09b22ac Merge branch 'master' into move_gates_to_load_store George Hotz 2026-05-05 14:58:32 -07:00
  • d09ea1d620 fix nir George Hotz 2026-05-05 14:56:34 -07:00
  • 7a00223bd3 Fix webgpu George Hotz 2026-05-05 14:50:53 -07:00
  • 5053148502 nir fix George Hotz 2026-05-05 14:48:49 -07:00
  • 76ff378007 autogen: fewer apt dependencies (#16049) Christopher Milan 2026-05-05 14:22:41 -07:00
  • ecf49474eb cleanups + fix nir George Hotz 2026-05-05 14:04:07 -07:00
  • 396d3f441a Merge branch 'master' into move_gates_to_load_store George Hotz 2026-05-05 13:55:08 -07:00
  • 6573c103f9 fix wrong load alt dtypes George Hotz 2026-05-05 13:53:38 -07:00
  • 5fa0016ffc supports_exec_item -> supports_uop (#16033) nimlgen 2026-05-05 22:41:13 +03:00
  • cee17e0d2f viz: fix diff color (#16045) qazal 2026-05-05 21:40:53 +03:00
  • 9c37a0c75d Ops.FLOORDIV and Ops.FLOORMOD (#16038) chenyu 2026-05-05 11:42:14 -04:00
  • d79bf356c2 viz: add CALL -> codegen link (#16044) qazal 2026-05-05 17:34:44 +03:00
  • fc2a289f61 fix nir George Hotz 2026-05-04 20:27:25 -07:00
  • 5736eee2f2 oops, inverted George Hotz 2026-05-04 20:22:30 -07:00
  • 651279c7ff Merge branch 'master' into move_gates_to_load_store George Hotz 2026-05-04 20:19:00 -07:00
  • 0821bef6b4 fix gated load George Hotz 2026-05-04 20:17:12 -07:00
  • 437205ae03 flip order, this is simpler George Hotz 2026-05-04 20:02:41 -07:00
  • cfefef479b add dtype George Hotz 2026-05-04 19:56:53 -07:00
  • 1c8cb0769a am: autogen asic_regs (#16004) Christopher Milan 2026-05-04 19:52:07 -07:00
  • 5d9431ecb9 fix ptx George Hotz 2026-05-04 19:44:01 -07:00
  • 0f3b12fcd8 fixes George Hotz 2026-05-04 19:30:25 -07:00
  • 60c8542320 work George Hotz 2026-05-04 19:21:22 -07:00
  • 4ec5487ad8 fix renderers George Hotz 2026-05-04 19:04:49 -07:00
  • 995a787d6c fix llvm crash George Hotz 2026-05-04 18:52:22 -07:00
  • 1b17762030 fix python George Hotz 2026-05-04 18:44:53 -07:00
  • c0f443cf47 Merge branch 'master' into move_gates_to_load_store George Hotz 2026-05-04 18:35:26 -07:00
  • 26406bed83 amd uses .valid, not index src valid (#16042) George Hotz 2026-05-04 18:35:15 -07:00
  • ff1258feef fix tests George Hotz 2026-05-04 17:33:01 -07:00
  • 51b13466dd fix amd George Hotz 2026-05-04 17:23:43 -07:00
  • 416878db9e i hate ai George Hotz 2026-05-04 17:14:53 -07:00
  • e00b3b4065 fix George Hotz 2026-05-04 17:09:59 -07:00
  • d810bd2b41 Merge branch 'master' into move_gates_to_load_store George Hotz 2026-05-04 17:05:21 -07:00
  • 09ec34437d fix oob validation George Hotz 2026-05-04 16:55:32 -07:00
  • a357a0449a Tensor.div cleanup (#16041) chenyu 2026-05-04 19:27:36 -04:00
  • 36383298be move gates to load/store George Hotz 2026-05-04 14:56:37 -07:00
  • 8f397f5c7c move load gates George Hotz 2026-05-04 14:45:42 -07:00
  • 5b4f62519d cache buffer_views as well (#16039) nimlgen 2026-05-05 00:00:09 +03:00
  • 8e99c4f097 fetch checks sha256 (#16037) Christopher Milan 2026-05-04 13:08:38 -07:00
  • 1884f67a39 simplify full_rewrite_to_sink spec (#16035) George Hotz 2026-05-04 11:44:13 -07:00
  • a4fccd23b2 remove kwargs in UOp.vectorize [pr] (#16034) chenyu 2026-05-04 12:46:38 -04:00
  • b1d88ebf02 viz/cli: aggregate flops in -t (#16031) qazal 2026-05-04 17:35:02 +03:00
  • c02e390c2b viz: encode flops, mem and metadata in json (#16032) qazal 2026-05-04 17:06:18 +03:00
  • 4024d8438f runtime/graph: avoid core_id runtimevar merge conflicts (#16026) bigyoshi 2026-05-03 12:16:02 -04:00
  • 9684334dfe viz: fix flops in graph, add null graph tracing (#16024) qazal 2026-05-03 16:32:44 +03:00
  • 419d525553 feat: handle multioutput kernel grads (#16028) wozeparrot 2026-05-03 01:31:45 -04:00
  • 9717d3a3a2 hotfix: prepend LD_LIBRARY_PATH to DLL posix search dirs (#16023) mefengl 2026-05-03 01:45:19 +08:00
  • 7daf4b7d52 viz: split cli test (#16015) qazal 2026-05-02 19:47:11 +03:00
  • d65b8ca25f jit: remove *input_list from the graph sources (#16021) nimlgen 2026-05-02 14:42:47 +03:00
  • 7dae9e6f7f viz: keep VIZ.value = 0 during python shutdown, cleanup launch (#16022) qazal 2026-05-02 14:35:53 +03:00
  • 637bdd5530 am: only support CDNA3/4 and RDNA3/4 (#16017) Christopher Milan 2026-05-01 21:02:14 -07:00
  • c0c120bf58 cleanups gate-on-load-store George Hotz 2026-05-01 23:52:45 +00:00
  • 9596d13550 move gate to load/store George Hotz 2026-05-01 23:37:31 +00:00
  • 4a2e1f1076 STORE doesn't have ranges anymore (#16019) George Hotz 2026-05-01 15:00:27 -07:00
  • 0bffbc5f8a onnx fmod uses fmod (#16018) chenyu 2026-05-01 16:47:11 -04:00
  • 782d1ff80f Tensor.fmod (#16014) chenyu 2026-05-01 16:02:18 -04:00
  • 35e13e08d5 improve shapes to make them behave like dtype.count, try 2 dtype_shape_2 George Hotz 2026-05-01 10:27:54 -07:00
  • 1079441332 revoke bus master (#16007) nimlgen 2026-05-01 18:00:01 +03:00
  • 8b147a9ed5 minimal repro for llama copies 2 (#16011) qazal 2026-05-01 16:23:47 +03:00
  • a29dd7b19b Revert "cleanup: untrack wait Metal buffers (#15954)" (#16010) qazal 2026-05-01 15:18:19 +03:00