github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-09 15:08:02 -05:00
Code Issues Packages Projects Releases Wiki Activity
11,158 Commits 203 Branches 13 Tags
423b76a8523e75470ef7579e14f84d4f097b3a9b
Commit Graph

1371 Commits

Author SHA1 Message Date
George Hotz
423b76a852 improve sqtt format parser (saturday coffee shop project) (#13419) 
* improve sqtt format parser

* actually read the trash code ChatGPT wrote

* cleanups

* hand written parser

* quality

* more

* was missing first packet

* maybe

* filt

* fixups

* label the waves

* progress
 2025-11-22 15:04:10 -08:00
qazal
c14033e10f viz: faster startup time with SQTT=1 (#13337) 
* roc.py cleanups

* direct append

* viz index cleanup

* simd row details

* add kernel arg

* late instructions decode

* more instruction decode to sep server request

* 200ms startup, 6 second to waves timeline

* sort units

* creating new http paths is easy now

* instructions unpacker

* min diff, use hyphens

* summary table
 2025-11-22 22:02:30 +08:00
George Hotz
dabb02767f set AMD profile mode with sudo on SQTT or PMC (#13403) 
* require profile mode

* add mode setter

* cleanup

* not needed

* SQTT_LIMIT_SE
 2025-11-20 23:19:11 -08:00
b1tg
91e289cb14 amd fp8 llvm (#13186) 
* amd fp8 llvm support

* fix max

* clean

* add test_mi350.sh

---------

Co-authored-by: chenyu <chenyu@fastmail.com>
 2025-11-20 12:35:57 -05:00
Roelof van Dijk
1058748440 torch backend: no aten.detach for torch 2.10 compat (#13381) 
* this works, less cpp?

* simpler = better

* keep torch 2.9 working as well
 2025-11-20 09:12:15 -08:00
qazal
9dbc550692 roc: map disassembly to prog name (#13384) 2025-11-20 23:47:19 +08:00
Roelof van Dijk
0dc2ff431d fix: revive torch backend (#13280) 
* fix: revive torch backend

* as_strided view vs copy

* Revert "as_strided view vs copy"

This reverts commit 82a61223f2.

* add extra tests (move inplace, add fusion tests)

* better fusion with inplace_op

* no optimizer hooks (break mnist training fusion)

* split off fusion tests in separate file, assert on resnet fusion

fix: remove comments

* cleanup, reduce diff

* reduce diff

* better fusion and identity checks

---------

Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
 2025-11-19 15:26:50 -08:00
wozeparrot
56b2540349 tk: keep extra tile data by replacing uop (#13370) 2025-11-19 15:11:43 -08:00
nimlgen
0c9fbf87e1 nvioctl: classes (#13346) 2025-11-19 16:14:15 +03:00
wozeparrot
be72b78dcb tk: small fixes (#13345) 
* fix: handle case where final uop isn't a tk wrapped one

* clean: remove after from mma
 2025-11-19 00:58:50 -08:00
qazal
a647c9eca6 sqtt ui minor fixes (#13335) 
* roc.py cleanups

* direct append

* viz index cleanup

* simd row details
 2025-11-19 01:27:56 +08:00
nimlgen
331f70aa75 roc: ctrlc (#13255) 
* roc: ctrl-c works

* rm
 2025-11-18 19:29:28 +08:00
George Hotz
6d3385c284 print special ops in postrange (#13318) 
* print special ops in postrange

* fix on OSX
 2025-11-17 14:43:23 -08:00
George Hotz
98e9e73286 hotfix: amd_uop_matmul getenvs 2025-11-17 13:26:01 -08:00
qazal
e7e1935225 cleanup sqtt/test_timing (#13315) 2025-11-18 04:28:05 +08:00
wozeparrot
33773fda87 tk initial mi350 (#13289) 2025-11-17 11:46:32 -08:00
nimlgen
e2cee64050 Revert "hcq: add tag to exec events (#13311)" (#13314) 
This reverts commit f63ded5817.
 2025-11-17 22:15:31 +03:00
nimlgen
f63ded5817 hcq: add tag to exec events (#13311) 
* hcq: add tag to exec events

* f

* fix

* fix
 2025-11-17 16:59:30 +03:00
qazal
50a443f558 viz: add shader engine to wave exec payload (#13310) 
* viz: show sqtt shader engine

* order it from smallest unit

* easier to config
 2025-11-17 19:11:34 +08:00
George Hotz
55be95da15 cleanup sqtt raw parser (#13309) 
* cleanup sqtt raw parser

* better names (don't merge yet)

* clean up amd

* a few more names

* one more filter
 2025-11-16 13:11:51 -08:00
George Hotz
cabd4add48 more work parsing SQTT, separate VIZ/PROFILE (#13308) 
* more work parsing SQTT

* more minimal runner

* sep VIZ/PROFILE

* parse print new

* improve parser

* more filter

* that

* split them

* lil cleanup

* skip flaky test

* AQL in mmapeak
 2025-11-16 10:40:39 -08:00
qazal
13efdf8c31 test s_nop stall (#13307) 2025-11-17 00:59:39 +08:00
George Hotz
295600dc5a saturday coffee shop work parsing the att format (#13295) 
* saturday coffee shop work parsing the att format

* add examples

* parser

* classes of packets

* fully vibe coded parser

* vibing

* empty

* some vibe names

* vibes

* most of these are wrong

* more vibes

* better names

* parsing

* parse

* cleanup parser

* touchups
 2025-11-16 08:25:51 -08:00
qazal
c70b06ec19 sqtt test_timing work (#13304) 
* sqtt test_timing cleanups

* only the instruction

* v_mfma_f32_16x16x32_f16 16 cycles, only after second one though
 2025-11-16 23:49:24 +08:00
wozeparrot
ef42334239 tk: load store cleanup (#13290) 2025-11-15 17:08:23 -08:00
qazal
7c110e1a57 viz: minor cleanups for sqtt (#13275) 
* small prg cleanup

* test_timing
 2025-11-15 01:08:56 +08:00
qazal
2ee701a009 roc: fix CEnum access (#13270) 
* roc: add decoder to ci

* also add installer

* use CEnum syntax

* try 2

* add to setup

* revert ci change

* the other enum too
 2025-11-14 21:41:24 +08:00
nimlgen
14eb48b13a autogen: rename nv_gpu to nv_570 (#13273) 
* autogen: rename nv_gpu to nv_570

* rename
 2025-11-14 20:07:19 +08:00
Christopher Milan
09f3aae169 In-tree autogen: all C libraries (#13220) 
* checkout files from autogen branch

* ioctl with payload

* fix am generations

* properly fix generations

This reverts commit b2a54f4f41.

* revert discovery.h

* support pragma pack(1)

* typo

* better getter

* typo

* NVCEC0_QMDV05_00_RELEASE[01]_ENABLE

* align support

* anon handling fix

---------

Co-authored-by: George Hotz <72895+geohot@users.noreply.github.com>
 2025-11-13 18:57:44 -08:00
wozeparrot
777cbec5b3 tk: rename rt tile dims to base (#13265) 2025-11-13 18:43:02 -08:00
wozeparrot
7eb0d8e744 feat: mixins on tiles (#13246) 2025-11-13 16:52:52 -08:00
George Hotz
ba84d415fe work from benchmarking tinybox red v2 (#13264) 
* work from benchmarking tinybox red v2

* gpuburn
 2025-11-13 16:38:40 -08:00
wozeparrot
547304c471 tk: group cleanup (#13262) 2025-11-13 14:19:51 -08:00
wozeparrot
4ada51618f tk: don't flatten in clear (#13249) 2025-11-13 13:38:01 -08:00
George Hotz
faf68c03a8 more mi350x matmul work (#13138) 
* more mi350x matmul work

* broken compute
 2025-11-13 09:09:28 -08:00
alpharush
7e0aaadecd feat: add repro command to summary (#10930) 2025-11-13 08:52:27 -08:00
nimlgen
f9b7586e08 roc: fix blob gc (#13256) 2025-11-13 23:38:35 +08:00
qazal
006dea4c3e roc: only save instruction execs (#13254) 2025-11-13 21:28:40 +08:00
George Hotz
17aa3379e9 hotfix: improve self_tokenize 2025-11-13 00:18:57 -08:00
qazal
be2e24cb25 roc: requires sudo to install (#13237) 2025-11-12 16:59:22 -05:00
George Hotz
8f1f195b6d hotfix: no hexdump for usbgpu patch.py 2025-11-12 12:05:37 -08:00
qazal
8b26cf2b3d sqtt: update rcp timing test (#13231) 
* sqtt: assert correct output in timing test

* found why
 2025-11-13 02:01:54 +08:00
nimlgen
af17e07251 viz: sqtt touchups (#13228) 
* viz: sqtt touchups

* revert

* matches
 2025-11-12 22:40:37 +08:00
nimlgen
fcd8d0751a test_timing for hip (#13229) 2025-11-12 20:28:58 +08:00
wozeparrot
371c1f2355 tk: move tiles to class (#13224) 2025-11-11 21:53:46 -08:00
wozeparrot
787f0070ed feat: don't use output reg as local reduce reg (#13203) 2025-11-11 14:35:16 -08:00
George Hotz
0c978d45e6 stub attention (#13196) 
* stub attention

* name the kernels
 2025-11-10 13:48:38 -08:00
qazal
50934050bc sqtt: append all wave execs (#13190) 2025-11-10 23:50:08 +08:00
qazal
38a24731a1 cleanup sqtt tooling (#13188) 
* cleanup viz/serve.py

* use latest profile in rgptool.py

* unwrap nullable in roc.py, fix disasms typing
 2025-11-10 20:52:57 +08:00
wozeparrot
6252831ceb feat: initial tk library (#13160) 2025-11-09 22:54:29 -08:00