github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-04-29 03:00:14 -04:00
Code Issues Packages Projects Releases Wiki Activity
12,045 Commits 248 Branches 14 Tags
f9cfb64cd94dea5b734f492922064722cbdf4880
Commit Graph

12045 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
qazal
6c487656f9 viz: kernel metadata from rodata entry (#14487) 2026-02-02 15:41:42 +09:00
Robbe Derks
d75a1b0d5a usbgpu: use BOT interface for patch.py (#13644) 
* BOT usage

* cleanup

* fix lint

* fix ruff

* fix -7?
 2026-02-02 11:54:46 +08:00
Christopher Milan
2931b52875 skip autogen if MTLCompiler is loaded (#14466) 2026-02-01 22:12:27 -05:00
George Hotz
9a32d6e090 add depth limit for SPEC=2 (#14485) 
* make SPEC=2 work for everything

* that's a horrible fix

* add depth limit
 2026-02-02 10:43:28 +08:00
George Hotz
368a692e1a make SPEC=2 work for everything (#14476) 
* make SPEC=2 work for everything

* that's a horrible fix
 2026-02-02 10:30:56 +08:00
chenyu
ea1f1d2b9d test_assign_to_bitcast_view (#14483) 
currently disk allows assign same size dtype into a bitcasted view
 2026-02-01 16:46:04 -05:00
chenyu
6deeccc192 fix RING with single dest (#14482) 2026-02-01 12:14:46 -05:00
chenyu
3ff390159b don't implicitly change dtype in assign (#14481) 
broadcast shape is fine, but implicitly cast dtype is hard to find
 2026-02-01 11:48:54 -05:00
imaolo
2111762a48 failed test case for RING output device (#14191) 
* Add enable/disable scheduler cache ContextVar

* add allreduce ring and naive to() tests

* clearer test comparing native vs ring allreduce

* split tests, add helper

* removing trailing whitespace

---------

Co-authored-by: chenyu <chenyu@fastmail.com>
 2026-02-01 11:48:43 -05:00
chenyu
02afae04f4 atol in test_call_gemm (#14480) 
flaky
 2026-02-01 11:24:58 -05:00
chenyu
5705398a1f assign cleanup [pr] (#14479) 
share more code path between disk and non-disk. also raise RuntimeError instead of Assert for mismatches
 2026-02-01 09:10:22 -05:00
chenyu
da500dbe06 simplify late_buffer_view [pr] (#14478) 
check the only allowed Ops in the chain, and offset cannot be negative
 2026-01-31 22:38:40 -05:00
chenyu
b4f96301e0 remove unused rules [pr] (#14477) 2026-01-31 21:29:30 -05:00
qazal
54e78dbec8 viz: remove hardcoded strings in cfg tests (#14462) 2026-02-01 09:30:43 +09:00
chenyu
5d38db9da6 generic bitcast assign (#14474) 
a.bitcast(X).assign(src) -> a.assign(src.bitcast(a.dtype))
 2026-01-31 17:29:20 -05:00
chenyu
b38fc43b07 assert assign dtype mismatch for disk [pr] (#14473) 
the disk hack is generally wrong, now force bitcast on the source before assign
 2026-01-31 17:08:54 -05:00
chenyu
ced886f26c failed test case for assign into bitcast (#14469) 
* failed test case for assign into bitcast

DISK assign has custom hack for this. need to fix before we can unify assign

* test_assign_bitcast_different_size
 2026-01-31 14:26:47 -05:00
chenyu
81eee5b30a unused spec [pr] (#14468) 
no BUFFER_VIEW in tensor, and no CONTIGUOUS in KERNEL
 2026-01-31 13:53:16 -05:00
nimlgen
f873c7b6c5 amd: fetch_name is file_name (#14465) 2026-01-31 20:11:07 +03:00
chenyu
c765641215 remove unused allow_any_len [pr] (#14464) 
STORE has 2 src, RESHAPE has 2 src, BUFFER has 2 src
added some tests for the untested allow_any_len
 2026-01-31 11:05:42 -05:00
chenyu
b4f5a51ebb move tests to unit (#14463) 
test_uop_graph does not need device, test_memory_planner can use NULL
 2026-01-31 10:49:31 -05:00
qazal
616e9c1483 CDNA assembly gemm in tensor.py with flag (#14310) 
* work

* work

* the assembly

* remove the old one

* remove ws bufs, assert splitk

* notes cleanup

* work

* gemm args

* gemm in mixins would be nice

* add gemm gradient

* print counters

* the realize is for DEBUG=2 aesthetics

* dedup

* rewrite to python dsl, no list copies

* leave that

* add B, M, N, K to gemm name

* it's M0 not NULL

* fp16 support

* test cleanup + more gemms

* work from viz

* more work

* gemm batch_size

* xccg path work

* tiny comments on the label naming

* s_waitcnt
 2026-01-31 22:34:14 +09:00
chenyu
55f806b713 tighter late_buffer_view match [pr] (#14456) 
src must be len 2 at that point
 2026-01-31 07:28:26 -05:00
qazal
d69bc5aa1a make DEV=NULL EMULATE=AMD amd_asm_matmul run (#14460) 2026-01-31 20:45:24 +09:00
qazal
4976544bf9 multi ram usage tests on the NULL device (#14457) 2026-01-31 14:14:53 +09:00
chenyu
99b44121bc failed test case for non-consecutive disk read (#14455) 
silently fail now
 2026-01-30 23:44:04 -05:00
George Hotz
b705c9143c assembly/amd: test more instructions (#14365) 
* assembly/amd: test more instructions

* more

* passing

* revert

* no const fold

* remove junk

* cleaner
 2026-01-31 12:40:22 +08:00
George Hotz
c9a3ddb341 benchmark llama walltime script (#14454) 
* benchmark llama walltime script

* adj layers
 2026-01-31 10:21:54 +08:00
George Hotz
f5346d6a1a fix USE_ATOMICS for non float dtypes and make it the default (#14444) 
* embedded multistep test

* complex test

* with jit

* fix dtypes and reenable USE_ATOMICS

* that test didn't catch anything
 2026-01-31 09:44:16 +08:00
Christopher Milan
e575dd8275 prevent UB in long decomp and more emulated tests (#14447) 2026-01-30 19:38:41 -05:00
chenyu
3204f94454 correct var_vals schedule filter (#14451) 
complete_create_schedule_with_vars returns var_vals that's used in schedule
 2026-01-30 17:10:07 -05:00
chenyu
cfcd1debb5 test schedule with multiple AFTER (#14449) 2026-01-30 15:59:00 -05:00
nimlgen
486d53d646 device: call free for external_ptr (#14448) 
* device: call free for external_ptr

* lin
 2026-01-30 23:53:17 +03:00
nimlgen
e0978498dc amd: read_ptr/write_ptr/doorbells are not lists (#14445) 2026-01-30 23:11:57 +03:00
Christopher Milan
1803ee939d EMULATED_DTYPES=long works with CPU_LLVM (#14446) 2026-01-30 13:54:43 -05:00
chenyu
03613e83ad update TestTensorMetadata (#14443) 
run with SCACHE=0 some more TODOs
 2026-01-30 12:39:01 -05:00
George Hotz
cbb1eed57b hotfix: partial revert of 9eb449f88, caused llama NaN 2026-01-30 17:19:27 +00:00
chenyu
26f5c00265 move TestTensorMetadata to unit (#14442) 2026-01-30 12:14:21 -05:00
chenyu
c05a0b85ae flip unique const src order [pr] (#14441) 
* flip unique const src order [pr]

matches buffer, simplifies replace_input_buffer

* combine rules
 2026-01-30 11:44:18 -05:00
George Hotz
ee2c78709d mlperf/llama: disable USE_ATOMICS for now 2026-01-31 00:42:08 +08:00
chenyu
beecac4d85 expand ranges -> unroll outer ranges [pr] (#14440) 2026-01-30 11:26:05 -05:00
chenyu
9eb449f882 clean up toposort sched_sink [pr] (#14439) 2026-01-30 10:18:28 -05:00
George Hotz
838cd078bc use atomics for embedding backward (#14400) 
* embedding is slow

* failing

* float is fine

* null

* it fails

* simplify embedding with broadcasting

* ATOMIC_ADD incoming

* min change

* simpler test

* better test

* fix test

* real test

* simpler

* cleanups

* types and names

* _zero_kernel

* grad multi

* hack

* none

* multi unshard

* more for call

* don't tag in call

* good

* call_multi

* call_multi wow claude is useless

* embedding backward mutli test

* test passes

* fix as_param

* shape_to_shape_arg

* add clip

* before cast

* fix spec=2, use atomics
 2026-01-30 18:10:59 +08:00
nimlgen
1998e0bb28 nv: add prof props to dev (#14437) 2026-01-30 12:51:43 +03:00
George Hotz
7a9dee4e50 add call/param UOps (#14433) 
* add call/param UOps

* resolve call

* skip that for now

* grad on call

* fix tests
 2026-01-30 14:51:45 +08:00
qazal
66d6a68016 viz: sqtt work from cdna gemm (#14434) 
* it's the tag

* initialize rows based on the disasm

* test_cfg with Ops.BINARY

* pyremu wants s_code_end?

* test_diamond

* diff cleanup
 2026-01-30 14:00:56 +09:00
Christopher Milan
88caf57ef4 ci: unify python versions (#14430) 2026-01-29 21:42:03 -05:00
chenyu
86a204d22a allow Tensor setitem input to be list/tuple (#14432) 
matches assign, and generally matches numpy
 2026-01-29 21:26:58 -05:00
chenyu
4a80319093 clean up split_store final logic [pr] (#14429) 
explicitly check the structure
 2026-01-29 18:40:07 -05:00
Christopher Milan
e47f12f671 ci: replace testing_minimal with testing_unit (#14427) 2026-01-29 18:02:43 -05:00