← Back to context Comment by peterashford 3 months ago You think the JVM is slow? 7 comments peterashford Reply mceachen 3 months ago IME large linear algebra algos run like molasses in a jvm compared to compiled solutions. You're always fighting the gc. za3faran 3 months ago Do you have any benchmarks to show, out of curiosity? light_hue_1 3 months ago Ok. But we have plenty of C libraries to bind to that for.They're far slower in Python but that hasn't stopped anyone. bluGill 3 months ago Depends. JVM is fast once hotspot figures things out - but that means the first level is slow and you lose your users. vips7L 3 months ago You can always load JIT caches if you can’t wait for warm up. palata 3 months ago What about AOT?
mceachen 3 months ago IME large linear algebra algos run like molasses in a jvm compared to compiled solutions. You're always fighting the gc. za3faran 3 months ago Do you have any benchmarks to show, out of curiosity? light_hue_1 3 months ago Ok. But we have plenty of C libraries to bind to that for.They're far slower in Python but that hasn't stopped anyone.
light_hue_1 3 months ago Ok. But we have plenty of C libraries to bind to that for.They're far slower in Python but that hasn't stopped anyone.
bluGill 3 months ago Depends. JVM is fast once hotspot figures things out - but that means the first level is slow and you lose your users. vips7L 3 months ago You can always load JIT caches if you can’t wait for warm up. palata 3 months ago What about AOT?
IME large linear algebra algos run like molasses in a jvm compared to compiled solutions. You're always fighting the gc.
Do you have any benchmarks to show, out of curiosity?
Ok. But we have plenty of C libraries to bind to that for.
They're far slower in Python but that hasn't stopped anyone.
Depends. JVM is fast once hotspot figures things out - but that means the first level is slow and you lose your users.
You can always load JIT caches if you can’t wait for warm up.
What about AOT?