| commit | b42923e75b2a7f3e09c4a1b51bad160bc1f9130b | |
| author | Stas Boukarev <stassats@gmail.com> | |
| Wed, 14 Dec 2016 15:32:01 +0000 (14 18:32 +0300) | ||
| committer | Stas Boukarev <stassats@gmail.com> | |
| Wed, 14 Dec 2016 15:32:01 +0000 (14 18:32 +0300) | ||
| tree | d3d69964d2a8fa13197b213839d792ec2c69bef9 | treesnapshot (tar.gz zip) |
| parent | 2de7232a6f9ab6d60152b53aaed5a4066299d2c1 | commitdiff |
Optimize full calls to FILL on specialized vectors.
Build a widetag table of bashers and value transfomers. That only
leaves (complex double-float) and (complex single-float) on #-64-bit
to do by hand.
Filling large arrays is now just as fast as the inlined variant.
Small arrays still win from inlining due to overhead.
Build a widetag table of bashers and value transfomers. That only
leaves (complex double-float) and (complex single-float) on #-64-bit
to do by hand.
Filling large arrays is now just as fast as the inlined variant.
Small arrays still win from inlining due to overhead.
| NEWS | diffblobblamehistory | |
| src/code/seq.lisp | diffblobblamehistory | |
| src/compiler/seqtran.lisp | diffblobblamehistory |