|
@yoshuawuyts | |||||
|
After talking with @fitzgen this week, I'm now really excited about superoptimizers!
It's like: what if we could fuzz a code — but instead of yielding bugs, we search for optimizations.
When a new optimization is found it's saved, so that it can later be integrated in compilers
|
||||||
|
||||||
|
Jonathan Turner
@jntrnr
|
2. velj |
|
I remember reading some papers on this many moons ago, so there might be some academic literature worth a skim.
Makes me think, too, of the search space exploration from generative programming. You could find the permutations for how to express the algorithm, then cost them.
|
||
|
|
||
|
Gudmund
@gudmvatn
|
3. velj |
|
I believe cranelift-dan talked about this
|
||
|
|
||
|
Hassen
@hassen94025
|
2. velj |
|
Makes you think about the cost of optimization.
|
||
|
|
||
|
infinitesimal
@infinitesimal_p
|
3. velj |
|
Any literature recommendations on this? Seems interesting!
|
||
|
|
||