Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactor: fold-constant without recursive function call #178

Merged
merged 19 commits into from
Nov 8, 2024

Conversation

hikettei
Copy link
Owner

@hikettei hikettei commented Nov 8, 2024

  • optimize: optimize-aasm
    • no recursive function calling in fold-constant
  • bugfix on defsimplifier
    • node-class
    • recurisve node creation
    • <node>
    • ids appeared in the outputs should be simplified
    • optimize pattern matcher, constant folding < 1e-5 sec
    • Full optimization on %tensor->aasm (make it zero cost for smaller graph)
    • !view < 1e-4sec without using fixnum

@hikettei
Copy link
Owner Author

hikettei commented Nov 8, 2024

CATEN> (let ((g (%tensor->aasm (!add (iconst 0) (iconst 'x))))) (time (caten/aasm::apply-fold-constant g)))
Evaluation took:
  0.000 seconds of real time
  0.000016 seconds of total run time (0.000014 user, 0.000002 system)
  100.00% CPU
  0 bytes consed
  

Graph[seen=NIL, outputs=NIL] {
    <Node[BUFFER] ALLOCATE(NID744) : SID739 <- () where :nrank=0 :dtype=INT32>
    <Node[BUFFER] LOAD(NID746) : BID747 <- (SID739) where :value=X>
}

@hikettei hikettei changed the title Opt: Fast ViewRange Creation Refactor: fold-constant Nov 8, 2024
@hikettei hikettei marked this pull request as ready for review November 8, 2024 07:45
@hikettei hikettei changed the title Refactor: fold-constant Refactor: fold-constant without recursive function call Nov 8, 2024
@hikettei hikettei merged commit 9157d35 into main Nov 8, 2024
5 checks passed
@hikettei hikettei deleted the fast-viewrange branch November 8, 2024 07:56
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Development

Successfully merging this pull request may close these issues.

1 participant