wozeparrot
473e5e4368
feat: make USE_ATOMICS embedding bwd faster ( #15151 )
2026-03-14 21:21:10 -07:00
qazal
3858bfc83d
sqtt: CDNA inst decodes ( #15274 )
...
* sqtt: CDNA inst decodes
* JUMP packets other way
* cdna insts
* r3
* r4
* lds from simd1 and simd2
2026-03-14 21:03:46 +09:00
Christopher Milan
d753c5d7e5
IMAGE=1 image_conv2d pads for bank conflicts ( #15252 )
2026-03-14 07:59:16 -04:00
Christopher Milan
9047249a7c
m.where(x.pad_to(m.shape), Invalid) ranges shrink ( #15275 )
2026-03-14 07:26:36 -04:00
nimlgen
f392c53c66
system: merge remote into pciiface ( #15273 )
...
* system: merge remote into pciiface
* clenaer
* move
* mypy
* fix
2026-03-14 18:44:20 +08:00
chenyu
13eec8fbe8
remove unused assign rules [pr] ( #15268 )
2026-03-14 05:37:49 -04:00
Christopher Milan
dabdc986df
shrink guarded ranges, try 2 ( #15272 )
2026-03-14 04:24:05 -04:00
Christopher Milan
7cf4b16c91
Revert "shrink guarded ranges" ( #15271 )
2026-03-14 03:44:38 -04:00
Christopher Milan
d9951e2f8e
shrink guarded ranges ( #15263 )
2026-03-14 03:38:48 -04:00
qazal
43ffd66fda
viz: oneline inst list ( #15269 )
...
* viz: oneline inst list
* save 5 chars
* gradual padding
2026-03-14 15:37:18 +09:00
George Hotz
86f17468ed
store in spec + USB BOT fix ( #15265 )
...
* move spec to store
* usb bot flag
* Revert "usb bot flag"
This reverts commit 7b8b7824f0
2026-03-14 13:25:05 +08:00
George Hotz
06d7cddb33
amd_copy_matmul is cleaner ( #15248 )
...
* amd_copy_matmul is cleaner
* it runs
* replicated stuff
* add tid there
* it runs
* cleanup
* x.src[1]
* flatten
* move that
* keep that assert
2026-03-14 12:56:09 +08:00
chenyu
b3600e4774
don't emit assign in transform_precompiled_call [pr] ( #15262 )
2026-03-13 22:42:35 -04:00
qazal
4d60312f7f
viz: asm python dsl syntax highlighting ( #15259 )
2026-03-14 06:37:43 +09:00
qazal
6209ddfc90
viz: improve disasm of s_code_end ( #15258 )
...
* viz: improve amd disasm of s_code_end
* better tests
* order was good
2026-03-14 03:31:14 +09:00
wozeparrot
a191ac0566
llama: use mlperf model ( #15257 )
2026-03-13 08:08:32 -07:00
Sieds Lykles
4b59083d7c
assign into empty works ( #15256 )
2026-03-13 10:24:29 -04:00
qazal
60b1b908c6
sqtt: CDNA layout header packet is the same size ( #15255 )
2026-03-13 22:28:24 +09:00
nimlgen
4e21735f31
system: update tinygpu app ( #15247 )
2026-03-13 20:36:57 +08:00
nimlgen
1fbe1fef2c
move write_configs to drivers ( #15253 )
2026-03-13 19:02:34 +08:00
chenyu
018c01508d
test case for call precompile multi ( #15254 )
2026-03-13 06:28:43 -04:00
nimlgen
bc16f80b50
am: remove dma_regions param ( #15251 )
...
* am: remove dma_regions param
* linter
2026-03-13 18:12:48 +08:00
chenyu
576e7f985f
remove handle_assign_mops [pr] ( #15249 )
2026-03-13 01:53:21 -04:00
Christopher Milan
c251fc67c5
ci: consider arch in venv and apt caches and go back to 3.12 ( #15250 )
2026-03-13 00:36:49 -04:00
Christopher Milan
d4b947ea9a
ci: explicitly request python 3.12.10 instead of 3.12 ( #15246 )
...
3.12.10 is the most recent 3.12 version that has toolcache builds for linux, macos, and windows
2026-03-12 23:00:46 -04:00
George Hotz
a7d2429c21
amd_uop_matmul more cleanups ( #15240 )
2026-03-13 10:24:43 +08:00
qazal
d893b14193
sqtt: update cdna packet names ( #15243 )
...
* sqtt: update cdna packet names
* change
* order
2026-03-13 08:49:09 +09:00
wozeparrot
749162bd2f
llama memory tweaks ( #15223 )
2026-03-12 12:36:23 -07:00
qazal
9a7173b7a0
viz: visualize full range of shader clock frequency, auto zoom to kernel range ( #15225 )
...
* start this
* work
* rm those
* relative to start cycle
* cleanup
* cover the full range of packets
* correct event type
* start the ui change
* fit=true
* better
* always the zoom identity
* diff cleanup
* shader engine itrace can be turned off
2026-03-13 00:07:31 +09:00
chenyu
d9c09397c0
Ops.STORE is shapeless [pr] ( #15239 )
2026-03-12 09:05:30 -04:00
nimlgen
d746ccb791
system: fix vfio ( #15235 )
2026-03-12 18:31:00 +08:00
nimlgen
d104a903f8
system: print output when err ( #15230 )
2026-03-12 18:30:49 +08:00
George Hotz
e560a46f59
update amd_uop_matmul ( #15236 )
...
* update amd_uop_matmul
* use custom kernel
* simpler
* ignore
2026-03-12 17:33:12 +08:00
chenyu
90b7f4341d
failed two level divmod recombine case ( #15233 )
2026-03-12 04:04:36 -04:00
chenyu
8b8d9a443c
remove unused invalid rules [pr] ( #15231 )
2026-03-12 03:10:34 -04:00
George Hotz
bdd62fd484
remove unneeded realize map entries ( #15229 )
...
* remove unneeded realize map entries
* not that
2026-03-12 14:23:19 +08:00
chenyu
842c978df3
remove staticmethod dtypes.max/min ( #15227 )
...
always use x.dtype.max/min
2026-03-11 23:11:24 -04:00
b1tg
18dc77ccab
add fp8 fnuz dtypes with PYTHON backend support ( #14945 )
...
* add fp8 fnuz dtypes with PYTHON backend support
* rm emu related change
* clarify fp8 fnuz zero handling
* Revert "rm emu related change"
This reverts commit efa4763c22b1tg@users.noreply.github.com >
Co-authored-by: chenyu <chenyu@fastmail.com >
2026-03-11 22:30:18 -04:00
George Hotz
4f3f55328b
do not patch on invalid tensor tests ( #15226 )
...
* do not patch on invalid tensor tests
* cleanup
2026-03-12 09:35:20 +08:00
wozeparrot
4fab320abe
llama: clean ( #15224 )
2026-03-11 13:33:59 -07:00
wozeparrot
05d6d9120a
llama offload null ( #15222 )
2026-03-11 10:04:31 -07:00
qazal
d3eef70162
viz: render shader clock frequency graph ( #15197 )
2026-03-12 01:32:49 +09:00
chenyu
39b0f4bcc1
remove Ops.THREEFRY in remove_bufferize [pr] ( #15220 )
2026-03-11 05:30:33 -04:00
chenyu
6489a6f212
Revert "remove mop_cleanup [pr] ( #15217 )" ( #15218 )
...
This reverts commit 6b50df940a
2026-03-11 04:17:56 -04:00
chenyu
6b50df940a
remove mop_cleanup [pr] ( #15217 )
...
no kernel diff, i think this was needed due to force_reshape?
test/external/external_benchmark_schedule.py is about the same speed
2026-03-11 03:54:42 -04:00
Christopher Milan
2fb8a7f60f
fix test_invalid_tensor when before values are nan ( #15215 )
2026-03-10 23:51:19 -04:00
chenyu
fce87f19a8
better fold_add_divmod_recombine ( #15214 )
2026-03-10 23:24:22 -04:00
chenyu
df8deec949
test for nest_by_factor selection ( #15213 )
2026-03-10 22:41:31 -04:00
chenyu
be6b0bce1f
variations of (x%c)+(x//c)*c ( #15212 )
...
put those into one function
2026-03-10 22:41:14 -04:00
qazal
a408d90f4f
viz: always detect sqtt packet overlaps, add timeline tests ( #15211 )
...
* test
* work
* it's called CALL, better assert
* qol
* row_ends
2026-03-11 05:32:38 +09:00
