Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
37ef_ced3
on Oct 9, 2022
|
parent
|
context
|
favorite
| on:
Decompiling x86 Deep Neural Network Executables
Convolutions are fused into convolutions, elementwise operations are fused into convolutions, everything is inlined except where function calls are needed for pthread work units (and those work units are all custom/arbitrary).
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: