Hi Marc, Is there a document anywhere that describes how to performance tune Gambit ? I've been writing some genetic programming code which Gambit has been great for for small examples but now I'm running into severe performance problems. I'm tryin to load a large CSV file of approximately 40MB into a list of records but I gave up on it when it still hadn't finished after 30 minutes. When its loaded, I'll need to iterate over the full set around 500k times which I'm expecting taking a while but hopefully no more than a couple of days. I'm using the interpreter at the moment so can try compiling but is there anything else that I can do ? Thanks Andrew
Afficher les réponses par date
On Sep 22, 2009, at 4:48 AM, Andrew Whaley wrote:
Hi Marc,
Is there a document anywhere that describes how to performance tune Gambit ? I've been writing some genetic programming code which Gambit has been great for for small examples but now I'm running into severe performance problems.
I'm tryin to load a large CSV file of approximately 40MB into a list of records but I gave up on it when it still hadn't finished after 30 minutes. When its loaded, I'll need to iterate over the full set around 500k times which I'm expecting taking a while but hopefully no more than a couple of days. I'm using the interpreter at the moment so can try compiling but is there anything else that I can do ?
1. Compile. Profile. Use better algorithms. 2. Use declarations, typically (declare (standard-bindings)(extended-binding)(block)(not safe)) (if these are applicable) and flonum and fixnum-specific operations (also if applicable) Brad
On Sep 22, 2009, at 4:48 AM, Andrew Whaley wrote:
I'm tryin to load a large CSV file of approximately 40MB into a list of records but I gave up on it when it still hadn't finished after 30 minutes.
There was a long thread (split into several subthreads in the archive) on improving the speed of reading CSV files beginning about here: https://mercure.iro.umontreal.ca/pipermail/gambit-list/2007-February/ 001112.html I don't think anyone started with interpreted code ... Brad
participants (2)
-
Andrew Whaley -
Bradley Lucier