Commit Graph

Select branches

Hide Pull Requests

abstractions_4

add_local_buffer

after_qol

algebraic_upat

all_bv_no_idx

all_sink

amd_isel

amd_refac

amd_smu_reset_patch

amd_sqtt

amd_uop

amd_uop_cleanups

assign_copy

assign_try_3

before2

better_spec

better_syn

bitcast_spec

broadcast_shape_2

bt_ei

buf_heur

bump_nv_2404

cache_reshape

call_inline

call_is_kernels

callify

cast_vec_as_cast

cat_mop

changes_from_rangify

check_clsetarg

check_dims

cifar_30

cifar_30_clean

clean_load

clean_spec

clone_tg

codegen_try_2

contig_on_after

continue_mat

deepsek

default_late_allreduce

delete_kernel_py

delete_ones

delete_slow_rangeify

delta_attention

delta_attention_scan

demote_op

disable_rules

dsp_search

dsp_search_merged

dtype_shape

earlier_gater

emu_cdna

emulate_ctx

end_all_ranges

endrange

er_prioity

execution_unit

failing_assign

fancy_usb

fast_llm

fix_fa_fwd_clang_22

fix_fuse

fix_llvm_cast

fix_range_merging_sd

fix_rangeify_tests

fix_rc

fix_sched_copy

fix_slice_store

fix_src_spec

fixup_fix_kernel

full_spec

fun_w_egraphs

gate-on-load-store

gate_rewrite

gemma4_gpt

gen_pdf_fast

gh-pages

gl_dims

gldims

glm_flash

grad_outerworld

grok_x86_novec

group_div_rules

hcq2_vram_copies

heur_4096

heur_group

hk_gemm

http_threading

hwtest_fixes

id_store

if_range_cleanups

ignore_oob

image_idx_xy

image_in_mc

image_no_vec

image_refactors_stack

index_to_slice

invalid_try_3

jinja_template

kernel4

kernel_is_call

kernelless

kimi_slop

kitten_matmul

late_Locals

late_add_load

late_allreduce

late_codegen

late_decomp_is_late

late_lower_idx

late_view

late_weakint

lds_double_buffering

less_wmma_arg

lil_dg_change

lil_rangeify_td

llama_trainer

llm_262k_minimal

llm_flash_attn

llm_qwen

llm_serve_fixes

load_view_late

local_cache

locals_have_device

loop_end_op

loop_is_rng

lowerer_hacks

mac_pytest

manual_kernel_3

master

mega_lowerer

min_new_lin

min_outer_range

minor_views

mixins7

mlperf_training_v4.1

mlperf_training_v5.0

mlperf_training_v6.0

modern_scan

more_apps

more_early_comps

more_folding_rules

more_kernelcount

more_pcode

more_rangeify_pm_tests

more_ren_clean

more_uop_gc_test

more_uop_mm

more_vmap

move_devec

move_dtype_decomps

move_expander

move_gates_to_load_store

move_mop_cleanup

move_schedule

move_simp_views

moveleftright

mt_typing

multi_range

multioutput

new_ck

new_devec

new_invalid_and_image

new_linearizer

new_shape

new_test

no_assign_2

no_buf_0

no_decimals

no_late_simplify_marg

no_merge_ends

no_merge_views

no_mstack_in_allreduce

no_one_ranges

no_sip_usbgpu

no_tsink_base

no_tuplize

no_user_ck_contig

no_vec_x86_4

no_weakint_variable

noop_index

novecimg

num_batches_tracked_shape

olmo3_support

one_compiler

one_hcopt

only_assign_buffers

only_loop_backends

only_reg_emu

only_reg_emu2

outer_range

outerworld_work

pad_invalid_work

pad_mixin

param_call_is_function

parse_sqtt_more

pci_iommu

port_nir

postrange_hcopts

precompile_backward

prefill_qwen_fast

print_special_ops

ps_clean

pyremu_try

python_prg

python_speed

q4k_support

qcom-openpilot-13ms

qcom_830

qcom_mmapeak

qcom_speed_codex

qol_string

qwen36_27b_amd_900

qwen36_27b_amd_minimal

qwen36_amd_bugfixes

qwen_mergable

r_cleanups

r_new_opt

ranged_call

rangeify_2

rangeify_const_folding

rdna3_vibes

rdna4_asm

rdna4_gemm

real_pyrender

recurse_sub

reg_coalese

relu_pattern

remove_assign

remove_cast_hack

remove_ck_contig

remove_define_var

remove_forced_re

remove_on_stack

remove_programspec

remove_pylint

remove_slow_tests

remove_vec_2

remove_vec_3

renderers_use_shapes_2

replace_if_with_range

reprocess_node_experiments

reshape_trait

resnet-ci

revert-11547-fix_mismatch_red

revert-11715-revert-11711-amd_llvm_tc_cast

revert-11809-no_ki_in_gpudims

revert-11811-better_div_nesting

revert-11876-memplan_vbig

revert-12104-simplify_valid_in_symbolic

revert-12308-bufferize_early

revert-12731-jitlink_init_check

revert-12813-after_cleanups

revert-13046-better_cleanup_arange_buffers

revert-13311-prof_link_sqtt

revert-13583-rocr_desc_2

revert-14734-tload

revert-17141-full_rewrite_to_sink

revert-17408-_broadcasted-weak

revert-9438-simple_matmul_np_rand

revert_wrong_ck_fix

rewrite_rangeify2

rtoposort

scan_assign

scan_op_8

sched_cache

sched_lin

sd_load_simple

self_type

set_testpath

shrink_in_render

simp_spec

simple_gidxs

simple_pool

simple_priority

simple_rangeify

simpler-sd-compile

simpler_end

simpler_fusion

simpler_image

simpler_postrange

simpler_reduce

slice_like_shrink

spec_3

spec_cleanup

spec_ren

speed_one_linear

split_buf_idx

split_decomp_pass

split_sub

stack_0

store_is_void

string_ranges

support_opts_in_contig

swizzle_in_kernel

sym_llm

sym_work

symbolic-chunk-split

szdiff_bot_test

tc_is_heur

tensor_graph_single_sink

test_amd_crash

test_fa

test_from_postopt

test_hgemm

test_unit

tiny_scan

tmr

tool_call_support

topoprop

ttomsa-new_linearizer

tuplegtuple

two_stage_remove

typed_checks

typing_work

unused_permute_arg

uop_prg

upd_test4

update_benchmark

update_benchmark_search

update_mlperf

user_failure

vecless_x86_glm

verify_sink_dims_clean

vf_ish

view_after_const

view_in_codegen

viz_cbroadcast

warp_fun

wgpu-f16-packed

working_postopt

x86_moves

x86_numel

v0.10.0

v0.10.1

v0.10.2

v0.10.3

v0.11.0

v0.12.0

v0.13.0

v0.4.0

v0.5.0

v0.6.0

v0.7.0

v0.8.0

v0.9.0

v0.9.1

v0.9.2

Mono Color

• 6da956b9fa that should be right Comma Device 2022-07-19 19:47:37 -07:00
• f4ed837f2f float16 fixups Comma Device 2022-07-19 19:44:40 -07:00
• aa00a3948e needs_load in image correct Comma Device 2022-07-19 19:25:47 -07:00
• 314d70ff17 zero out the buffer Comma Device 2022-07-19 19:17:47 -07:00
• b8a67905e5 save weights Comma Device 2022-07-19 19:14:14 -07:00
• 46e7dfade1 REQUIRES_SIMPLE_REDUCE geohot 2022-07-19 11:42:14 -07:00
• acbeaf0ba9 adam in benchmark_train_efficientnet geohot 2022-07-19 09:33:07 -07:00
• ef1100fdff touchups geohot 2022-07-19 09:30:06 -07:00
• 5d45c6e516 Fold reduce (#362) George HotzandGitHub 2022-07-19 09:24:02 -07:00
• 2d402d1135 buffer_id is 8 bytes Comma Device 2022-07-18 20:27:45 -07:00
• 577c23731e outputs with size Comma Device 2022-07-18 20:21:33 -07:00
• 29581b5c85 inputs and outputs Comma Device 2022-07-18 20:17:26 -07:00
• ae30641b0d fix row pitch Comma Device 2022-07-18 19:48:19 -07:00
• 02f23e526c output file to disk Comma Device 2022-07-18 19:23:22 -07:00
• ef4afdb5d2 tests maybe geohot 2022-07-18 08:23:41 -07:00
• a2c4bcf313 disable opencl tests geohot 2022-07-18 08:17:21 -07:00
• 5093455166 don't shuffle if there's children involved geohot 2022-07-17 21:15:40 -07:00
• f76d41812b prune graph geohot 2022-07-17 15:38:43 -07:00
• eda6f071b2 default opt level 2 geohot 2022-07-17 14:54:40 -07:00
• 73b0471b25 join expands geohot 2022-07-17 13:42:05 -07:00
• cfabbbd6bb more crap to remove without convs geohot 2022-07-17 13:02:27 -07:00
• 5e96ed523a fix opencl bug, no training on opencl geohot 2022-07-17 12:55:26 -07:00
• f93e297804 fix bug caused by rounding geohot 2022-07-17 12:49:58 -07:00
• cff297ef9d w/e, that's a later prob geohot 2022-07-17 12:32:50 -07:00
• 4bc07326d4 we need that opt to make gpu decent speed geohot 2022-07-17 12:26:18 -07:00
• 6375e7129a opencl not imported geohot 2022-07-17 12:14:39 -07:00
• bf299802f8 fixup tests geohot 2022-07-17 12:11:53 -07:00
• 762e859089 testopencl geohot 2022-07-17 11:56:40 -07:00
• 608e2431f7 test opencl, commit to removing the crap conv code from GPU geohot 2022-07-17 11:55:37 -07:00
• 3c4565fa21 SLICE -> PAD,SHRINK geohot 2022-07-17 11:33:59 -07:00
• 9574dd8559 some permutes are reshapes geohot 2022-07-17 10:34:24 -07:00
• 77806e0d64 fix permute stacking geohot 2022-07-17 10:24:57 -07:00
• c28a99087b add PAD movementop geohot 2022-07-17 10:22:03 -07:00
• 4527c0453d get_movementroot geohot 2022-07-17 09:37:27 -07:00
• b00cc93102 bugfixes geohot 2022-07-17 09:19:22 -07:00
• d07f379038 don't merge movement ops geohot 2022-07-17 09:09:50 -07:00
• f6ea7c022a Revert "EXPAND -> REPEAT" geohot 2022-07-17 08:42:10 -07:00
• cca089b11d Revert "more expand -> repeat" geohot 2022-07-17 08:41:48 -07:00
• 2e7b1630a8 more expand -> repeat geohot 2022-07-17 08:40:49 -07:00
• 115d2eadf5 EXPAND -> REPEAT geohot 2022-07-17 08:38:54 -07:00
• 1eb247f823 contiguous and same length geohot 2022-07-16 08:49:07 -07:00
• d04b274cd2 noop removal can replace with reshape geohot 2022-07-16 08:32:42 -07:00
• d985217fa4 skip reduce noops geohot 2022-07-16 07:47:43 -07:00
• bcf422dfdd Device2 (#358) 3 people 2022-07-16 07:26:19 -07:00
• d651caa864 fixup openpilot/compile.py George Hotz 2022-07-11 13:59:09 -07:00
• 5e46561f7e no_grad = NOT backward geohot 2022-07-10 20:54:57 -07:00
• b34ae7876f lol chr(10) not chr(13) geohot 2022-07-10 20:03:11 -07:00
• 817b64f5e5 A conv is a reduce op (#356) George HotzandGitHub 2022-07-10 19:58:50 -07:00
• 057e4f5aa5 a little faster and cleaner geohot 2022-07-09 08:14:01 -07:00
• c77ba7fa3f ops_cpu readibility geohot 2022-07-09 07:48:54 -07:00
• 0a36475700 no einsum for now geohot 2022-07-09 00:04:40 -07:00
• c39a245696 that's not where i thought we'd lose lines... geohot 2022-07-08 23:52:38 -07:00
• 75e1848b09 always SHUFFLE_RESHAPE_OPS geohot 2022-07-08 23:19:39 -07:00
• 44848ee5dc prints show we can precompute from the outside geohot 2022-07-08 10:59:20 -07:00
• 68959be05d precompute weights for opencl geohot 2022-07-08 10:56:48 -07:00
• d8e7f1f8bc opencl type ignore geohot 2022-07-08 10:33:55 -07:00
• ae335b6d3e opencl works, but tons of kernels geohot 2022-07-08 10:22:04 -07:00
• 5b66d1bb0b begin fixing up opencl geohot 2022-07-08 10:20:14 -07:00
• 7e17f2ae8d fix mypy, add TODOs geohot 2022-07-08 09:57:22 -07:00
• 8557ed88df use ast engine for merged reduceop geohot 2022-07-08 09:37:40 -07:00
• 3656a5615a MERGE_ELEMENTWISE_INTO_REDUCE geohot 2022-07-08 09:32:28 -07:00
• ca9532ce29 less lines, and typing found a bug geohot 2022-07-08 08:57:12 -07:00
• 2035b89e54 wooo 1k lines geohot 2022-07-08 08:44:57 -07:00
• 2a8c1071d9 cleanups geohot 2022-07-08 08:36:31 -07:00
• e6733286df unify conv and reduce geohot 2022-07-08 08:27:30 -07:00
• 9c34b3eef3 tighten up gpu kernels geohot 2022-07-08 07:59:04 -07:00
• 563bf2d8e8 force input/weight to be contiguous (uncached) geohot 2022-07-08 07:40:30 -07:00
• 1cf805a56a fix no MERGE_MOVEMENT_OPS bug geohot 2022-07-08 07:27:53 -07:00
• c0ef998b48 remove finished todo geohot 2022-07-07 11:39:25 -07:00
• 715e335c60 fix types geohot 2022-07-07 11:36:09 -07:00
• 9ee8426c51 much better cache geohot 2022-07-07 11:32:00 -07:00
• eb6696c3a5 only childless elementwise ops get merged geohot 2022-07-07 11:13:25 -07:00
• 04e7e4104c track graph children and make lazycache use weak references geohot 2022-07-07 11:01:18 -07:00
• 001cfe83a2 local geohot 2022-07-07 10:05:26 -07:00
• 2720ef49ca extra and test and tuple geohot 2022-07-07 10:01:33 -07:00
• 059fe94700 junk import geohot 2022-07-06 21:47:38 -07:00
• a61a4d09ad merge conv and binary op geohot 2022-07-06 08:27:26 -07:00
• 6e0015095f LBCACHE geohot 2022-07-04 16:05:19 -07:00
• 7a5acd3ace cache geohot 2022-07-04 16:04:48 -07:00
• d5d9cffe7c training param for batchnorm geohot 2022-07-04 13:28:03 -07:00
• 21c78b9316 can be v slow geohot 2022-07-04 13:23:34 -07:00
• 46bce4156f CL profiling geohot 2022-07-04 13:22:12 -07:00
• 34f43ea10e LAZY and CLCACHE are defaults geohot 2022-07-04 13:09:15 -07:00
• 425b0dcd58 sorry linecount, CLCACHE geohot 2022-07-04 12:52:04 -07:00
• b7afd83267 track cl mem used geohot 2022-07-04 12:19:00 -07:00
• 5ef62c33a1 SHUFFLE_MOVEMENT_OPS is OPT=3 geohot 2022-07-04 09:55:30 -07:00
• d5de8452c6 dashed loadops geohot 2022-07-04 09:50:56 -07:00
• e74adcce5c refactoring geohot 2022-07-04 09:25:19 -07:00
• 0bdb021880 separate realize functions for different ops geohot 2022-07-04 09:07:22 -07:00
• 81b73f97a3 Optiimzation (#355) George HotzandGitHub 2022-07-04 08:58:57 -07:00
• df7976248b be lazy with the gpubuffer copies for host for constant folding geohot 2022-07-03 23:04:14 -07:00
• 4d4ea47ca7 one more line geohot 2022-07-03 17:28:42 -07:00
• 02cd8510cb cleanups geohot 2022-07-03 17:23:20 -07:00
• d89542640a hmm, typechecker isn't checking everything geohot 2022-07-03 17:12:51 -07:00
• 6b0aa2a902 sorry about the line count, this is a good optimization geohot 2022-07-03 17:11:13 -07:00
• 748618530b tests will run at okay speed now? geohot 2022-07-03 16:41:52 -07:00
• c3d13893f9 add SHUFFLE_MOVEMENT_OPS, exactly 1000 lines geohot 2022-07-03 16:30:42 -07:00
• e6e43e820e should fix tests geohot 2022-07-03 16:06:11 -07:00
• 71a812fbf2 elementwise_ops geohot 2022-07-03 15:29:38 -07:00
• d7aad46758 test lazy also, make TestMNIST faster geohot 2022-07-03 15:18:00 -07:00