optimization that doesn't make recompilation be needed more often?
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Could there be a flag weaker than -O that doesn't cause recompilation any more than -O0 does -- would that provide any worthwhile optimizations? (an intermediate speed-tradeoff option for haskell developers.) Dependency on the details of modules never expected to be recompiled (e.g. base library, or any package not currently being compiled, but those are just heuristics) should be fine. Wondering, Isaac -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFGTh/PHgcxvIWYTTURAk6bAJwJm9p9A2z2oyRdHd3YsqmEGdd7MQCfd7LV E3FgIi+XUZNsRlzgxKPhIGI= =OzsN -----END PGP SIGNATURE-----
On Fri, May 18, 2007 at 05:51:12PM -0400, Isaac Dupree wrote:
Could there be a flag weaker than -O that doesn't cause recompilation any more than -O0 does -- would that provide any worthwhile optimizations? (an intermediate speed-tradeoff option for haskell developers.) Dependency on the details of modules never expected to be recompiled (e.g. base library, or any package not currently being compiled, but those are just heuristics) should be fine.
I have wanted something like this too. I find I can get a similar effect via careful use of NOINLINE pragmas. like I will often have a huge complicated module like module Grin.EvalAnalysis(grinEvalAnalysis) where ... and if I put a {-# NOINLINE grinEvalAnalysis #-} in there then changes to the module don't cause other stuff to be recompiled. John -- John Meacham - ⑆repetae.net⑆john⑈
Try -fomit-interface-pragmas Regardless of optimisation level, this tells GHC to generate only *essential* info in interface files (notably type signatures). That will kill all cross-module inlinings, but it will also dramatically reduce cross-module recompilation dependencies; in fact it'll reduce them back to the -Onot level. Simon | -----Original Message----- | From: glasgow-haskell-users-bounces@haskell.org [mailto:glasgow-haskell-users-bounces@haskell.org] On | Behalf Of Isaac Dupree | Sent: 18 May 2007 22:51 | To: GHC Users | Subject: optimization that doesn't make recompilation be needed more often? | | -----BEGIN PGP SIGNED MESSAGE----- | Hash: SHA1 | | Could there be a flag weaker than -O that doesn't cause recompilation | any more than -O0 does -- would that provide any worthwhile | optimizations? (an intermediate speed-tradeoff option for haskell | developers.) Dependency on the details of modules never expected to be | recompiled (e.g. base library, or any package not currently being | compiled, but those are just heuristics) should be fine. | | Wondering, | | Isaac | -----BEGIN PGP SIGNATURE----- | Version: GnuPG v1.4.6 (GNU/Linux) | Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org | | iD8DBQFGTh/PHgcxvIWYTTURAk6bAJwJm9p9A2z2oyRdHd3YsqmEGdd7MQCfd7LV | E3FgIi+XUZNsRlzgxKPhIGI= | =OzsN | -----END PGP SIGNATURE----- | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users@haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users
participants (3)
-
Isaac Dupree -
John Meacham -
Simon Peyton-Jones