Commit Graph

  • 44d84228ff move comgr_3 logic back to the old place (#13266) George Hotz 2025-11-13 20:05:54 -08:00
  • 09f3aae169 In-tree autogen: all C libraries (#13220) Christopher Milan 2025-11-13 21:57:44 -05:00
  • 777cbec5b3 tk: rename rt tile dims to base (#13265) wozeparrot 2025-11-13 18:43:02 -08:00
  • 7eb0d8e744 feat: mixins on tiles (#13246) wozeparrot 2025-11-13 16:52:52 -08:00
  • ba84d415fe work from benchmarking tinybox red v2 (#13264) George Hotz 2025-11-13 16:38:40 -08:00
  • 547304c471 tk: group cleanup (#13262) wozeparrot 2025-11-13 14:19:51 -08:00
  • f9010fdfc9 works George Hotz 2025-11-13 14:01:37 -08:00
  • bf116deb5a imagenet labels George Hotz 2025-11-13 13:44:55 -08:00
  • 4ada51618f tk: don't flatten in clear (#13249) wozeparrot 2025-11-13 13:38:01 -08:00
  • 6b1bae6614 ruff format mixin (#13261) George Hotz 2025-11-13 10:10:38 -08:00
  • 3049f3edda support _rebuild_tensor method interception (#13253) Faizaan Gagan 2025-11-13 23:11:21 +05:30
  • 3af231904e openpilot compile tests: assert pre-rangify speeds (#12775) Harald Schäfer 2025-11-13 09:39:06 -08:00
  • 8179a07477 Merge branch 'master' into more_apps George Hotz 2025-11-13 09:33:27 -08:00
  • 86d1e42ed8 Merge branch 'master' into algebraic_upat algebraic_upat George Hotz 2025-11-13 09:21:05 -08:00
  • faf68c03a8 more mi350x matmul work (#13138) George Hotz 2025-11-13 09:09:28 -08:00
  • 5c13504bc1 Merge branch 'master' into algebraic_upat George Hotz 2025-11-13 09:07:54 -08:00
  • 256f81bb02 Fix tracemeta 0 (#13049) Ayman Jabr 2025-11-13 20:07:11 +03:00
  • 7e0aaadecd feat: add repro command to summary (#10930) alpharush 2025-11-13 10:52:27 -06:00
  • 6be86dde17 nv: add timeout when repsonding to rpc (#13260) nimlgen 2025-11-14 00:42:21 +08:00
  • f9b7586e08 roc: fix blob gc (#13256) nimlgen 2025-11-13 23:38:35 +08:00
  • 263b724143 one cache and bump it (#13258) George Hotz 2025-11-13 07:33:31 -08:00
  • 5efa727b83 move _pool to MovementMixins (#13257) George Hotz 2025-11-13 07:28:52 -08:00
  • bcdfc109b5 hotfix: disable flaky test George Hotz 2025-11-13 06:19:28 -08:00
  • 006dea4c3e roc: only save instruction execs (#13254) qazal 2025-11-13 21:28:40 +08:00
  • f9586b38ba system: pci mask and val (#13251) nimlgen 2025-11-13 20:44:58 +08:00
  • 7316da3253 new readme (#13250) George Hotz 2025-11-13 00:48:28 -08:00
  • 17aa3379e9 hotfix: improve self_tokenize George Hotz 2025-11-13 00:18:57 -08:00
  • 4e5a9132e7 JIT_BATCH_SIZE=0 in compile3 (#13245) chenyu 2025-11-12 20:12:45 -08:00
  • 759557f633 feat: move tk tests to testextra (#13242) wozeparrot 2025-11-12 17:06:53 -08:00
  • 3f939f3d3c update pm_simplify_valid (#13241) chenyu 2025-11-12 16:40:02 -08:00
  • f9851a852f minor update to uop_given_valid [pr] (#13243) chenyu 2025-11-12 16:03:18 -08:00
  • fe2876a6d8 hotfix: second GB/s in viz (#13240) qazal 2025-11-13 07:14:27 +08:00
  • a23dea202b actually make AMD_LLVM not default (#13238) George Hotz 2025-11-12 15:07:23 -08:00
  • ab9fa964d8 DISABLE_COMPILER_CACHE -> CCACHE (#13234) George Hotz 2025-11-12 15:07:09 -08:00
  • be2e24cb25 roc: requires sudo to install (#13237) qazal 2025-11-13 05:59:22 +08:00
  • 8f1f195b6d hotfix: no hexdump for usbgpu patch.py George Hotz 2025-11-12 12:05:37 -08:00
  • 9a53fcbde4 amd: sqtt on rdna3.5 (#13233) nimlgen 2025-11-13 03:30:42 +08:00
  • 13f10a31dc AMD_LLVM default off (#13232) George Hotz 2025-11-12 11:06:33 -08:00
  • 8b26cf2b3d sqtt: update rcp timing test (#13231) qazal 2025-11-13 02:01:54 +08:00
  • bc8e537423 Add NONZERO op to onnx backend (#13211) Jan Akhremchik 2025-11-12 20:55:51 +04:00
  • af17e07251 viz: sqtt touchups (#13228) nimlgen 2025-11-12 22:40:37 +08:00
  • 7a6853fa40 viz: show python callstack in the first graph (#13218) qazal 2025-11-12 20:52:28 +08:00
  • 82eb63d3ad qcom: auto switch idle timer when profiling (#13230) nimlgen 2025-11-12 20:31:24 +08:00
  • fcd8d0751a test_timing for hip (#13229) nimlgen 2025-11-12 20:28:58 +08:00
  • 74b9d33acb viz: direct link to program source (#13227) qazal 2025-11-12 16:27:13 +08:00
  • 371c1f2355 tk: move tiles to class (#13224) wozeparrot 2025-11-11 21:53:46 -08:00
  • c793a08fbf Merge branch 'master' into remove_assign George Hotz 2025-11-11 19:27:07 -08:00
  • 41a098a82d In-tree autogen: libc.py (#13217) Christopher Milan 2025-11-11 22:13:48 -05:00
  • 222bb12ddf tk softmax (#13205) wozeparrot 2025-11-11 15:13:16 -08:00
  • 787f0070ed feat: don't use output reg as local reduce reg (#13203) wozeparrot 2025-11-11 14:35:16 -08:00
  • ece1415def clean up image_dot and image_conv2d (#13222) chenyu 2025-11-11 12:53:03 -08:00
  • 2f0ea29b34 qcom: 48bit timestamps (#13214) nimlgen 2025-11-12 04:14:33 +08:00
  • bc55bc4849 cleanup test_viz profiler tests (#13221) qazal 2025-11-11 21:46:48 +02:00
  • 23b90945c3 add a benchmark for openpilot vision with DEBUG=2 (#13219) chenyu 2025-11-11 11:41:52 -08:00
  • c2075f3613 gc disable during big rewrites (#13215) George Hotz 2025-11-11 10:30:47 -08:00
  • e59313da08 migrate pytest and ruff (#13216) Roelof van Dijk 2025-11-11 19:27:51 +01:00
  • 6fd7ce3832 migrate to pyproject.toml (#13189) Gaétan Lepage 2025-11-11 18:09:27 +01:00
  • 8002921a04 viz: improve the program run tooltip (#13212) qazal 2025-11-11 18:56:03 +02:00
  • 038f8a6c2d use linearizer in schedule George Hotz 2025-11-10 23:42:33 -08:00
  • f91e366a17 viz: display the graph layout recursion error (#13194) qazal 2025-11-11 09:25:12 +02:00
  • 73497af4c0 clean: use np for allclose (#13204) wozeparrot 2025-11-10 23:02:43 -08:00
  • a6360fd94d store can have shape (#13202) George Hotz 2025-11-10 22:16:47 -08:00
  • f3692b7406 clean up hip renderer (#13063) b1tg 2025-11-11 13:44:24 +08:00
  • 22b8579234 one last regressed dm kernel (#13201) chenyu 2025-11-10 20:30:52 -08:00
  • 58b7e4fab3 GROUPTOP heuristic on more axes (#13206) chenyu 2025-11-10 20:30:37 -08:00
  • a6913b9add replace ASSIGN with STORE/AFTER George Hotz 2025-11-10 19:05:40 -08:00
  • db8c6d9a04 work outer_range George Hotz 2025-11-10 15:26:16 -08:00
  • 0647f87bf8 outer range runs in the scheduler George Hotz 2025-11-10 14:49:42 -08:00
  • 829cdafccc update openpilot slow conv uop ast (#13197) chenyu 2025-11-10 14:03:20 -08:00
  • 0c978d45e6 stub attention (#13196) George Hotz 2025-11-10 13:48:38 -08:00
  • 58c30fc7ce minor image_conv2d cleanup (#13193) chenyu 2025-11-10 13:05:40 -08:00
  • 60e55d9a2d line count 18500 (#13191) chenyu 2025-11-10 10:52:13 -08:00
  • 09a59c2203 qcom: support new chip versioning (#13185) nimlgen 2025-11-10 23:57:29 +08:00
  • 50934050bc sqtt: append all wave execs (#13190) qazal 2025-11-10 23:50:08 +08:00
  • 38a24731a1 cleanup sqtt tooling (#13188) qazal 2025-11-10 20:52:57 +08:00
  • 845a24dcc6 viz: group sqtt waves by program (#13187) qazal 2025-11-10 19:25:23 +08:00
  • fd6803000e mutmut cfg (#13184) George Hotz 2025-11-09 23:29:29 -08:00
  • 6252831ceb feat: initial tk library (#13160) wozeparrot 2025-11-09 22:54:29 -08:00
  • 925231aec1 repeat does less reshape for 1s (#13183) George Hotz 2025-11-09 19:43:02 -08:00
  • d7369de048 hotfix: update weekly commits table George Hotz 2025-11-09 19:37:06 -08:00
  • 6c48c87e51 improved ASSERT_MIN_STEP_TIME (#13182) chenyu 2025-11-09 13:41:12 -08:00
  • 17715688c7 system: validate vendor for APLPCIIfaceBase (#13181) nimlgen 2025-11-10 02:49:21 +08:00
  • 614783693e nv: remove hardcoded expansion_rom_off (#13180) nimlgen 2025-11-09 21:43:19 +08:00
  • e1d46de8f8 update GROUPTOP heuristic more (#13178) chenyu 2025-11-08 23:31:12 -08:00
  • 41e45c20ff minor stuff reading the printed code [pr] (#13177) chenyu 2025-11-08 21:58:51 -08:00
  • 8e868dced8 only GROUPTOP one reduce kernel (#13176) chenyu 2025-11-08 19:38:44 -08:00
  • 834067d91f move onnx import in compile3 (#13172) chenyu 2025-11-08 12:44:34 -05:00
  • 7f3240dbfe nv: cleanup alloc (#13170) nimlgen 2025-11-09 00:14:46 +08:00
  • 7250fc0354 viz: double click on kernel run goes to codegen (#13147) qazal 2025-11-08 23:40:50 +08:00
  • 8a7fa9e7b4 sqtt: show total cycles of kernel in viz (#13169) qazal 2025-11-08 21:00:40 +08:00
  • 2ba8b4946f external_benchmark_op_cat.py (#13168) chenyu 2025-11-08 01:54:10 -05:00
  • a62496cb3d clean up get_grouped_dims [pr] (#13159) chenyu 2025-11-08 01:53:54 -05:00
  • eb0192b0bb feat: print ranges that aren't ended (#13167) wozeparrot 2025-11-07 22:01:29 -08:00
  • b41541bc44 bounty: Remove Tensor._pool alternative implementation and verify kernels remain the same (#13164) George Hotz 2025-11-07 16:59:48 -08:00
  • ffb9e8396f fix indexing bug with convs George Hotz 2025-11-07 16:45:19 -08:00
  • 6a509da7f3 Scheduler.reduceops helper [pr] (#13162) chenyu 2025-11-07 18:59:46 -05:00
  • 2413311289 make _pool simpler (#13161) George Hotz 2025-11-07 15:58:44 -08:00
  • 7c4971f345 ONE_POOL simple_pool George Hotz 2025-11-07 15:56:11 -08:00
  • a5fd297df5 Revert "try this now" George Hotz 2025-11-07 15:52:33 -08:00
  • 607cdc2164 try this now George Hotz 2025-11-07 15:50:08 -08:00