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
12,835 Commits 248 Branches 14 Tags
fefb0ebc2abbb162be9447224880e7ef977f79ec
Commit Graph

12835 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
qazal
fefb0ebc2a gemm/asm: fp8 cleanups (#15580) 
* normal gemm here

* s/dtypes.fp8e4m3/FP8_DTYPE

* gemm_bw

* device UOp stays NULL
 2026-04-02 19:02:38 +09:00
chenyu
61bc91aa8c Tensor cumalu cleanups (#15579) 
* Tensor cumalu cleanups

* happy
 2026-04-02 05:23:22 -04:00
chenyu
1aa04eab08 simple CreationMixin (#15567) 
start with full_like, zeros_like, ones_like
 2026-04-01 23:00:56 -04:00
wozeparrot
5b2a3251c4 mlperf system json for mi350 (#15575) 2026-04-01 15:30:33 -07:00
Christopher Milan
6c67bd4c14 better error message when invalid renderer is specified (#15573) 2026-04-01 17:12:55 -04:00
Christopher Milan
0d6fbc2355 remove flaky and redundant image test (#15574) 2026-04-01 16:33:13 -04:00
Christopher Milan
20f7f0be8e nir renderers use arch (#15556) 
* nir renderers use arch

* fix

* fix null
 2026-04-01 16:32:51 -04:00
nimlgen
148ad09559 am: do not use dbell for ih (#15571) 2026-04-01 21:34:21 +03:00
nimlgen
93a85c7348 am: raise when using more sdma engines (#15569) 2026-04-01 21:33:42 +03:00
nimlgen
da12c2ea16 better install msg (#15570) 2026-04-01 20:09:37 +03:00
b1tg
20497f2840 fold BIND to CONST when min==max (#15568) 2026-04-01 11:19:04 -04:00
qazal
9275f283e5 viz: update flag and display names (#15566) 
* rename to occ, other_simd

* se pkts

* match viz cli tool in names
 2026-04-01 21:48:37 +09:00
chenyu
f5c0794df2 fix Tensor.const_like (#15565) 
used to always return a 0-d tensor, now returns an expanded Tensor based on self.shape and matches UOp
 2026-04-01 08:35:19 -04:00
qazal
09f60d80fd llama: fix FP8=1 FAKEDATA=1 (#15564) 2026-04-01 20:53:03 +09:00
nimlgen
6d1e992e89 copyout sharded w/o ioring (#15562) 
* copyout sharded w/o ioring

* x

* x

* f
 2026-04-01 14:47:29 +03:00
nimlgen
150c456977 add OSError to suppress_finalizing (#15558) 2026-04-01 12:33:59 +03:00
chenyu
fc5b94b902 fix UOp.where(const, const) (#15560) 
* fix UOp.where(const, const)

* fix
 2026-04-01 05:28:49 -04:00
chenyu
5aeb2273db add amd_copy_matmul.py to CI (#15555) 
more tests before cleanup
 2026-03-31 22:39:18 -04:00
Christopher Milan
034f617971 NVCCRenderer is separate from CUDARenderer (#15554) 2026-03-31 21:26:13 -04:00
wozeparrot
8b5b9a0e90 llama: run_and_time (#15533) 2026-03-31 15:46:16 -07:00
Christopher Milan
acf239e4d2 specify renderer in DEV, <dev>_<ren>=1 is deprecated (#15551) 2026-03-31 18:35:14 -04:00
nimlgen
5181c8e23a llm: fix nan in kvcache (#15552) 2026-04-01 00:38:45 +03:00
nimlgen
3af25ccdb4 docs: minor tinygpu changes (#15550) 2026-03-31 21:29:15 +03:00
nimlgen
477d194630 hipcomgr and tinygpu scripts (#15549) 2026-03-31 20:07:52 +03:00
nimlgen
83085f103c tinygpu docs (#15545) 
* tinygpu docs

* x

* x

* fix
 2026-03-31 19:49:38 +03:00
nimlgen
ca89215a59 nv: use nvcc over nak by default (#15547) 2026-03-31 18:54:56 +03:00
qazal
a15345a53e viz/cli: improve --help message (#15546) 
* viz/cli: improve --help message

* not the default

* more work

* -s

* respect colored
 2026-03-31 22:31:33 +09:00
nimlgen
10d570b3d5 signed tinygpu (#15541) 2026-03-31 14:55:09 +03:00
chenyu
4ac2552642 improve ReduceMixin.all (#15544) 
use prod instead of min since `mul` lowered to `and` directly
 2026-03-31 07:54:27 -04:00
chenyu
89ec22131a tests to show double negation in min is not cancelled (#15543) 2026-03-31 06:59:13 -04:00
qazal
8feb8edc68 gemm/asm: add fp8 support to cdna asm_gemm (#15542) 
* work

* hmm, mixins

* rhs_transposed

* also fix the dtype

* check for hipcc

* Exception

* select dev

* default
 2026-03-31 19:32:54 +09:00
chenyu
2939ae8b22 more mixin (#15540) 
isclose is elementwise, min, any, all to OpMixin
 2026-03-31 05:46:55 -04:00
chenyu
e69f5f9f69 more movement methods to mixin (#15536) 
* more movement methods to mixin

* cleanups
 2026-03-31 05:16:47 -04:00
nimlgen
ceb63c8c2f new bundle id (#15307) 
* new bundle id

* new profiles
 2026-03-31 12:16:03 +03:00
qazal
467c0af8aa viz: skip flaky sever tests (#15538) 2026-03-31 17:20:30 +09:00
qazal
f88e255cea gemm/asm: split and parameterize dtype in llama gemm tests (#15408) 
* gemm/asm: more tests for emulator, parameterize llama gemm tests

* bf16 atol
 2026-03-31 17:12:44 +09:00
b1tg
a63392a565 llm: pairwise ranking topk for MoE expert selection (#15499) 2026-03-31 12:46:39 +08:00
wozeparrot
79cccf3003 write sz output to file (#15534) 2026-03-30 20:16:17 -07:00
Christopher Milan
6fb038d109 replace CompilerSet with list (#15530) 
* replace CompilerSet with list

* oops

* default Renderer list
 2026-03-30 23:07:52 -04:00
qazal
bc866a93f0 viz: rename exec to sqtt (#15527) 
* viz: rename exec to sqtt

* more
 2026-03-31 08:06:51 +09:00
Christopher Milan
adbfd82d1d DEV is ContextVar, setting Device.DEFAULT is deprecated (#15508) 2026-03-30 17:10:49 -04:00
nimlgen
9583489068 add mlx driver to extra (#15526) 
* mlx driver

* x

* simpler
 2026-03-30 20:28:49 +03:00
qazal
ad6347f6d8 sqtt: allow mapping sopk to IMMEDIATE packets (#15525) 
* work

* with s_waitcnt

* with the sopp variants, increase threads

* remove that

* sdst=NULL produces IMMEDIATE, otherwise is SALU
 2026-03-30 23:12:17 +09:00
chenyu
301b2cea57 move matmul to mixin (#15524) 2026-03-30 07:39:09 -04:00
chenyu
f0eaac4235 reduce mixin (#15523) 2026-03-30 05:23:58 -04:00
chenyu
f485d0b664 UOp.sum -> usum, prod -> uprod [pr] (#15522) 
rename to prep reduce mixin
 2026-03-29 04:51:55 -04:00
qazal
36a925e2a2 viz: color wmma, one color map for cli and web (#15519) 
* viz: color wmma, one color map for cli and web

* op_type

* like uops

* mypy cli
 2026-03-29 04:53:01 +09:00
wozeparrot
0c3e438229 llama: mllog (#15502) 2026-03-28 11:18:25 -07:00
nimlgen
7e57e101d5 better oor message in profiles (#15516) 
* better oor message

* x
 2026-03-28 20:25:07 +03:00
qazal
266fb07721 viz: show exec duration (#15484) 
* duration

* handwritten tests

* rdna3 pickle

* rdna4 pickle

* asserts

* rm that

* wmma work

* r4

* this shows the overlap well

* ohh okay it goes back

* are ds_load and ds_store different queues on RDNA4?

* print msg, v_mul_lo_u32 is 4 cycles?

* discover

* wmma something

* wmma comment

* less

* less

* better comments

* work

* inst st

* delay column

* better cli

* emit_alt

* update test_handwritten

* work
 2026-03-28 22:48:59 +09:00