Commit Graph
13471 Commits
Author SHA1 Message Date
qazalandGitHub 8ba58304f7 viz: reenable tests (#15626) 2026-04-07 07:52:44 +09:00
chenyuandGitHub 2f7d085450 shared _normalize_indices for getitem (#15625)
* shared _normalize_indices for getitem

* list
2026-04-06 17:45:36 -04:00
chenyuandGitHub 66ec188d50 more activations to mixin (#15624) 2026-04-06 15:41:41 -04:00
chenyuandGitHub 1483f7e71c support shift by Tensor (#15623)
* support shift by Tensor

* use mixin
2026-04-06 15:14:57 -04:00
chenyuandGitHub 6e30a5f5ea update shifts in torch backend (#15622) 2026-04-06 14:08:33 -04:00
chenyuandGitHub a444be172d lower fuzz_symbolic_symbolic_div timeout (#15619)
mitigate timeout crash due to high total time
2026-04-06 12:58:29 -04:00
chenyuandGitHub 01b49c8647 support int operand for shifts (#15618)
matches torch/jax, also symbolic rule to remove mask
2026-04-06 12:32:12 -04:00
nimlgenandGitHub e2700475cf mlx: cleaner (#15617)
* mlx: cleaner

* x
2026-04-06 17:49:47 +03:00
Valtteri ValoandGitHub 86c4431d74 add gpu_family detection to Metal, target MSL 4.0 on macOS 26+ (#15079)
use supportsFamily API to detect GPU generation instead of parsing
ICB debug description strings. also adds metal4.0 compiler target.
2026-04-06 06:51:38 +08:00
13PerriusandGitHub ff0c941548 remove redundant iteration and toposort in _deepwalk (#15532) 2026-04-06 06:38:45 +08:00
Andrew CappelliandGitHub e39cfe685a validate lr, momentum, weight_decay in optimizers (#15576) 2026-04-06 06:37:34 +08:00
nimlgenandGitHub 6a334ceb27 hotfix: fix bert (#15613) 2026-04-05 23:41:21 +03:00
nimlgenandGitHub e3986a6b74 mlx: init runtime (#15612)
* mlx: init

* x

* swap
2026-04-05 22:52:29 +03:00
nimlgenandGitHub e0988dbae5 hcq: support non for signal_t and compute_t (#15611)
* hcq: support non for signal_t and compute_t

* revert

* x
2026-04-05 18:56:47 +03:00
nimlgenandGitHub 5e134aa087 hcq: add write/poll_bit commands (#15610)
* hcq: add write/poll_bit commands

* x
2026-04-05 18:09:44 +03:00
nimlgenandGitHub 604cdbf2f7 am: large allocs aligned to 2mb to use 2mb pages (#15609) 2026-04-05 18:01:31 +03:00
qazalandGitHub b2d5b29f45 assembly/amd: validate dsl keyword args (#15608)
* assembly/amd: validate dsl keyword args

* hm, this should use the SOP2 s_waits

* use the sop2 s_waits
2026-04-05 23:00:24 +09:00
qazalandGitHub 056fcd7758 viz: web work from rdna4 gemm (#15607)
* add rdna4 barrier

* fix realtime
2026-04-05 19:14:16 +09:00
7e54992bf6 fp8 llama (#15588)
Co-authored-by: qazal <[email protected]>
2026-04-04 18:24:57 -07:00
qazalandGitHub 4d36366717 assembly/amd: match rdna4 hw gidx init in emulator (#15604)
* simple rdna4 copy kernel with hw fault

* the trivial fix: use ttmp instead of s

* now copy kernel fails in mockgpu

* rm crashing kernel
2026-04-05 02:28:18 +09:00
chenyuandGitHub 2ba5a6ddc8 remove detach in selu (#15602)
UOp does not have detach. this does not change behavior
2026-04-04 11:04:29 -04:00
qazalandGitHub f7aed180e4 viz/cli: add Other row in profiler (#15600) 2026-04-04 22:40:53 +09:00
sirhcmandGitHub 74ecf6d3e6 opaque structs are also c.Struct (#15596) 2026-04-03 19:40:43 -04:00
645d45d968 DEV has arch (#15577)
Co-authored-by: Comma Device <[email protected]>
2026-04-03 19:17:19 -04:00
nimlgenandGitHub 902edc3781 hcq: hcqbuf in copy (#15595) 2026-04-03 22:47:36 +03:00
nimlgenandGitHub 2c4271209e hcq: peer groups for remote (#15594)
* hcq: set real peer group

* x

* x

* x
2026-04-03 19:03:07 +03:00
chenyuandGitHub 8fdef2d3e4 mean/std/var to mixin (#15593) 2026-04-03 10:42:41 -04:00
qazalandGitHub 9920b42b5e hotfix: renderer.target.arch in disasm (#15592) 2026-04-03 22:23:51 +09:00
nimlgenandGitHub 237084b276 remote: support several hosts (#15585)
* remote: support several hossts

* f
2026-04-03 11:22:15 +03:00
sirhcmandGitHub 0ed8d9271d Renderers accept Target or nothing (#15590) 2026-04-03 01:09:41 -04:00
wozeparrotandGitHub 3a26920141 feat: framework ci (#15589) 2026-04-02 22:03:51 -07:00
sirhcmandGitHub 736fea8412 select_first_inited cleanup and better errors (#15587) 2026-04-02 19:27:58 -04:00
sirhcmandGitHub 8c50da800d [pr] cleanup unused ctx's in codegen (#15586) 2026-04-02 19:06:58 -04:00
nimlgenandGitHub 694dc5a717 install script in benchmark (#15584) 2026-04-02 18:15:58 +03:00
nimlgenandGitHub 046c3f1240 mlx: add loopback with send/recv (#15583) 2026-04-02 18:15:46 +03:00
chenyuandGitHub c64226e97c fix CreationMixin doc (#15582) 2026-04-02 09:46:28 -04:00
qazalandGitHub 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
chenyuandGitHub 61bc91aa8c Tensor cumalu cleanups (#15579)
* Tensor cumalu cleanups

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

* fix

* fix null
2026-04-01 16:32:51 -04:00
nimlgenandGitHub 148ad09559 am: do not use dbell for ih (#15571) 2026-04-01 21:34:21 +03:00
nimlgenandGitHub 93a85c7348 am: raise when using more sdma engines (#15569) 2026-04-01 21:33:42 +03:00
nimlgenandGitHub da12c2ea16 better install msg (#15570) 2026-04-01 20:09:37 +03:00
b1tgandGitHub 20497f2840 fold BIND to CONST when min==max (#15568) 2026-04-01 11:19:04 -04:00
qazalandGitHub 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
chenyuandGitHub 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
qazalandGitHub 09f60d80fd llama: fix FP8=1 FAKEDATA=1 (#15564) 2026-04-01 20:53:03 +09:00