nimlgenandGitHub bd580d8ea4 hcq: use mmio interface in nv (#9986) ...

* hcq: start mmio interface

* allow double cast

* revert

* faster?

* simpler, not needed more now

* dd

* types

* fix

2025-04-22 21:58:12 +03:00

..

codegen

faster block reorder (#9990)

2025-04-22 19:18:57 +01:00

engine

refactor schedule linearize to defaultdict [pr] (#9984)

2025-04-23 00:00:23 +08:00

frontend

add onnx frontend stub [pr] (#9558)

2025-03-24 12:24:34 +08:00

nn

nn.state docs cleanup (#8332)

2025-03-18 17:16:40 -04:00

renderer

don't use ops name, just keep sink (#9922)

2025-04-18 08:59:18 +01:00

runtime

hcq: use mmio interface in nv (#9986)

2025-04-22 21:58:12 +03:00

shape

cache folded upcast [pr] (#9733)

2025-04-04 11:23:19 +08:00

viz

don't use ops name, just keep sink (#9922)

2025-04-18 08:59:18 +01:00

__init__.py

UOp.variable (#7010)

2024-10-12 18:20:44 +08:00

device.py

make DISABLE_COMPILER_CACHE a ContextVar [pr] (#9983)

2025-04-22 10:32:54 -04:00

dtype.py

FP8s part 1 (#9887)

2025-04-15 11:20:02 -04:00

gradient.py

fix kernelize usage with pm_gradient (#9953)

2025-04-22 17:26:05 +08:00

helpers.py

make DISABLE_COMPILER_CACHE a ContextVar [pr] (#9983)

2025-04-22 10:32:54 -04:00

ops.py

non recursive toposort is 2x+ faster (#9979)

2025-04-22 13:59:38 +01:00

py.typed

add a single py.typed (#6083)

2024-08-14 17:31:46 -07:00

spec.py

microbenchmarks + micro speed ups (#9972)

2025-04-22 11:30:46 +01:00

tensor.py

move metadata set to tensor [pr] (#9976)

2025-04-22 12:30:35 +01:00

upat.py

don't use ops name, just keep sink (#9922)

2025-04-18 08:59:18 +01:00