github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-14 17:38:06 -05:00
Code Issues Packages Projects Releases Wiki Activity
10,877 Commits 203 Branches 14 Tags
self_type
Commit Graph

150 Commits

Author SHA1 Message Date
George Hotz
8cbef912d2 move reshape to MathTraits (#13054) 
* move reshape to MathTraits

* confirm it works in amd_uop_matmul
 2025-11-02 12:56:15 +08:00
George Hotz
267be7fc5e fp16 acc 2025-11-02 12:53:04 +08:00
George Hotz
e98506735b add CONTRACT support to UOp programs (#13043) 
* add contract support

* use contract

* 342 tflops
 2025-11-01 19:11:32 +08:00
George Hotz
65a0a31475 AMD mi350x matmul from stream (#13040) 
* works

* working mfma

* 120 TFLOPS

* regs

* 192 TFLOPS

* try pipelining

* something

* notes

* contract

* linter to 3.11

* that was a bug
 2025-11-01 17:55:19 +08:00
George Hotz
bc178d14a9 matmul example on metal showing off tensor core (#13033) 
* matmul example on metal showing off tensor core

* flip the args of placeholder

* mat_idx

* imp
 2025-10-31 19:40:36 +08:00
George Hotz
b46229ca51 use shrink in amd_matmul_uop (#13026) 
* use shrink in amd_matmul_uop

* colors
 2025-10-31 10:43:41 +08:00
George Hotz
512513c403 cleanup amd uop matmul (#13025) 
* cleanup amd uop matmul

* remove mod

* move that out

* better variable names

* var names

* more

* render fallback

* colors
 2025-10-31 10:04:45 +08:00
George Hotz
4a741e8364 modernize amd uop matmul (#13011) 
* modernize amd uop matmul

* progress

* comment

* more comments

* revert that

* mac cleanups

* fix estimates

* format
 2025-10-30 17:02:38 +08:00
George Hotz
25c2da1579 check SPEC=2 in CI (#12945) 
* check SPEC=2 in CI

* split SPEC=2

* fast enough
 2025-10-27 21:53:57 +08:00
chenyu
c5cee74706 remove BLOCK_REORDER (#12854) 
not used
 2025-10-21 19:10:14 -04:00
b1tg
60d7e232f2 cuda fp8 (#12782) 
* cuda fp8

* tensor core

* tc test

* clean

* clean pm
 2025-10-21 15:05:25 -04:00
chenyu
ae51bdd06a remove trivial use of RANGEIFY flag (#12550) 
some tests need update still
 2025-10-09 02:29:38 -04:00
chenyu
0e266f376c ops_gpu -> ops_cl (#12103) 2025-09-10 15:15:48 -04:00
nimlgen
fb96394ff5 auto-select available compilers (#12094) 
* device: auto select compilers

* fix

* metal+opencl

* nv/cuda

* test without ptx

* ptx

* fix tests

* fix

* fix test

* rename

* test + cleaner

* xx

* ops

* better test

* win?

* um?

* types

* debug

* win??

* sep rung

* wtf?

* debug

* skip win

* revert this

* types
 2025-09-10 19:52:01 +03:00
George Hotz
38dcadf07b delete kernel.py (#12040) 
* delete kernel.py

* delete that file

* rip and tear

* don't test search

* imports

* fix torch frontend

* not a part of regen
 2025-09-05 15:52:07 -07:00
George Hotz
afad7d0cd1 remove dtype from range, it will be dtypes.index soon [pr] (#11914) 
* remove dtype from range, it will be dtypes.index soon [pr]

* a few more
 2025-08-29 09:52:07 -07:00
George Hotz
394c2d1db1 update Kernel API in tests + move optimize_local_size (#11907) 2025-08-28 15:12:47 -07:00
George Hotz
27701ef823 add locals support to rangeify (#11826) 2025-08-24 14:03:12 -07:00
qazal
793ace530e update amd_uop_matmul.py import (#11581) 
Using this for testing SQTT
 2025-08-08 17:07:35 +03:00
George Hotz
82be8abfd2 move opt under codegen (#11569) 2025-08-07 14:19:17 -07:00
George Hotz
4f26a9ad32 check elements_per_thread in tensorcore [pr] (#11435) 2025-07-30 11:55:48 -07:00
George Hotz
1bef2d80c1 unrolls are all in the same scope (#11429) 
* unrolls are all in the same scope

* fix that import
 2025-07-29 16:55:37 -07:00
George Hotz
03909f2772 permute locals for HL uop matmul (#11412) 
* permute locals for HL uop matmul

* parens fix that

* permutes

* 20 TFLOPS
 2025-07-29 08:19:59 -07:00
George Hotz
735ad5f10d kernel4 and 5 in uops (#11411) 
* move simplify views to merge views

* add amd kernel 4

* Revert "move simplify views to merge views"

This reverts commit 1e07dff384.

* k4 in python

* kernel4 written in uops

* k5 support

* cleanups
 2025-07-28 19:35:48 -07:00
George Hotz
fddc645668 HL=2 top matmul (#11406) 
* HL=2 top matmul

* top colored
 2025-07-28 12:32:38 -07:00
George Hotz
dfeee63d30 uop matmul work (#11388) 
* uop matmul work

* works with locals
 2025-07-26 21:23:55 -07:00
George Hotz
2c70eaf18c fix load / barrier (#11386) 
* fix load / barrier

* cleanups

* fix CI
 2025-07-26 10:27:37 -07:00
George Hotz
466ab5a3f2 store/load not pass through index (#11381) 
* noop

* fix noop

* store cat is NOOP

* store dtype is void

* stores aren't passed through anymore

* meh, skip those for ptx

* correct ptx skip

* hl runs
 2025-07-25 21:01:47 -07:00
George Hotz
490a93902c define reg doesn't have init anymore (#11365) 
* define reg doesn't have init anymore

* remove that

* no special logic for dr

* fix amd uop matmul
 2025-07-24 19:15:49 -07:00
George Hotz
0602b22086 kernel spec (#11359) 
* kernel spec

* ops.VIEW

* work
 2025-07-24 12:45:38 -07:00
George Hotz
b0dc97d1f7 write out kernel 3 in uops (#11352) 
* write out kernel 3 in uops

* matmul is correct

* gemm passes spec

* bugfix to match speed

* cleanups
 2025-07-23 17:32:38 -07:00
George Hotz
108aac8af4 use AddrSpace instead of local (#11314) 
* use AddrSpace instead of local

* addrspace in test
 2025-07-21 14:00:06 -07:00
George Hotz
842184a1ab rename kernelize to schedule, try 2 (#11305) 2025-07-21 11:18:36 -07:00
chenyu
a0438012af remove Kernel.get_program [pr] (#11203) 2025-07-12 20:50:29 -04:00
George Hotz
d67c8e7b42 local metal on metal in uop syntax (#11185) 
* local metal on metal in uop syntax

* TODO: just put the axis_info in the kernelinfo

* local

* amd_matmul works @ 28 TFLOPS

* clean up matmul

* kernel8 works

* remove that

* locals

* axistype innovation

* work

* cleanup

* kernel3 regs

* cleanup kernel3

* work

* why is it broken

* no beam

* reenable

* permutes
 2025-07-12 16:31:19 -07:00
chenyu
6283d50224 DEPRECATED_linearize -> to_program [pr] (#11198) 2025-07-12 13:46:20 -04:00
George Hotz
2893feb9f6 cleanups for kernel.py (#11143) 
* cleanups for kernel.py

* fixups
 2025-07-08 18:10:25 -07:00
George Hotz
856759c79c add halide example (#10980) 
* add halide example

* upd halide gemm

* partial works

* touchups
 2025-06-26 16:14:57 -07:00
George Hotz
92678e59ee move kernel to opt (#10899) 2025-06-20 15:22:28 -07:00
Sidharth N. Babu
ef14dfb277 compile fixes (#10442) 2025-06-06 18:38:37 -04:00
George Hotz
411392dfb7 move files into uop dir (#10399) 
* move files into uop dir [pr]

* tinygrad.uop is a thing

* fix uop docs, no pr

* fix viz
 2025-05-18 11:38:28 -07:00
chenyu
720f20865b remove required_optimizations (#9848) 2025-04-19 16:51:16 -04:00
chenyu
f5256e0020 Kernel.apply_opts [pr] (#9917) 
* Kernel.apply_opts [pr]

updated all `for opt in`. also updated a few test_liinearizer tests to not implcitly depend on hand_coded_optimization

* not you yet
 2025-04-17 08:00:56 -04:00
chenyu
8c6299bced move hand_coded_optimizations to heuristic.py [pr] (#9844) 
* move hand_coded_optimizations to heuristic.py [pr]

also folded all long lines

* make a copy and rename self -> k

* fix test
 2025-04-10 23:40:16 -04:00
chenyu
c5db5b83b9 add SHOULD_USE_TC=1 check to simple_matmul (#9802) 
* add SHOULD_USE_TC=1 check to simple_matmul

also zero centered the random input and update atol for tf32

* ATOL=2e-2 for HALF
 2025-04-09 02:24:42 -04:00
George Hotz
78caf55154 Revert "FP8 support on NVIDIA (#8631)" 
This reverts commit 2c8e4ea865.
 2025-04-09 12:27:41 +08:00
George Hotz
14928fecff Revert "fix TF32 tensor core dropped in tc_sm89 (#9798)" 
This reverts commit 7c9a96824f.
 2025-04-09 12:27:39 +08:00
chenyu
7c9a96824f fix TF32 tensor core dropped in tc_sm89 (#9798) 
also add `SHOULD_USE_TC=1` to verify TC is applied in simple_matmul
 2025-04-08 23:20:50 -04:00
pkotzbach
2c8e4ea865 FP8 support on NVIDIA (#8631) 
* squashed fp8 commits

* tensorcore start

* minor changes

* pre-commit

* pylint

* Delete fp8mul.cu

* clean

* small bugfix

* fix test_dtype

* fix test_dtype_alu

* add EMULATE_CUDA_SM89

* fix ci

* fix test_linearizer

* fix test_linearizer

* fix swizzle

* add debug to simple_matmul

* fixed swizzle

* python emulator

* refactor python emulator

* setup fix

* numpy setup

* ml_dtypes only in emulate_cuda_sm89

* fix pylint

* fix tests

* fix mypy

* fix mypy

* fix ruff

* done python emulator

* add acc type

* tests

* mypy

* clean code

* add cuda tensor core tests to CI

* minor fix

* clean test_dtype.py

* clean cstyle.py

* clean test_ops.py

* fix test

* fix test

* whitespaces

* pylint

* pylint

* amd?

* amd?

* amd

* reduce lines

* mockgpu remove

* fix

* ruff

* ruff

* fix mypy

* ruff

* test only for cuda

* fixed formatting

* small fixes

* small fix

* least_upper_dtype if fp8s not supported

* log and reciprocal are supported for fp8s

* ops python fixes

* dtypes.fp8s use

* e4m3 + e5m2 result dtype test

* truncate linter fix

---------

Co-authored-by: pkotzbach <pawkotz@gmail.com>
Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
Co-authored-by: chenyu <chenyu@fastmail.com>
 2025-04-08 21:54:04 -04:00
Francis Lam
1e5d9ad8f7 extra/gemm/max_matmul: start of custom kernels for GEMM (#6926) 
* extra/gemm/max_matmul: start of custom kernels for GEMM

* add an unoptimized FP16/FP16 MMA example

* add slow 3-stage fp16 acc example

* add correct 3-stage pipeline with unswizzled/flat smem input (slow)

* add acc fp16 example with 3 stages and swizzle (no bank conflicts)

* add max version of NV fp16_fp16_fp16

* fix up comments and removed unused code in max variations

* add start of no_xor example

* fix to account for UOps to Ops
 2025-03-19 15:04:57 +08:00