Opened 14 years ago
Closed 12 years ago
#73 closed task (wontfix)
"normalize-type" caching
| Reported by: | ehuelsmann | Owned by: | nobody |
|---|---|---|---|
| Priority: | major | Milestone: | unscheduled |
| Component: | libraries | Version: | |
| Keywords: | performance bite-size | Cc: | |
| Parent Tickets: |
Description
Peter Graves points out that with his "canonicalize-type" buffering, he achieved measurable performance gain in XCL. We want to do the same in ABCL.
Attachments (1)
Change History (6)
comment:1 Changed 14 years ago by
| Summary: | "normalize-type" buffering → "normalize-type" caching |
|---|
comment:2 Changed 14 years ago by
| Keywords: | bite-size added |
|---|
comment:3 Changed 14 years ago by
| Component: | CLOS → libraries |
|---|---|
| Keywords: | performance, bite-size → performance bite-size |
| Milestone: | → 0.24 |
| Owner: | changed from somebody to nobody |
Setting a milestone; based on Peter's feedback, this should be a real performance saver.
comment:4 Changed 13 years ago by
| Milestone: | 0.24 → unscheduled |
|---|
Changed 12 years ago by
| Attachment: | cache-normalized-types.patch added |
|---|
Normaziled types caching patch
comment:5 Changed 12 years ago by
| Resolution: | → wontfix |
|---|---|
| Status: | new → closed |
Using the attached patch, I see a 95% cache hit rate while compiling some checkout of Maxima. However, I see no speed-ups in build times for ABCL or Maxima nor do I see significant improvement of the execution of the (compiled) ANSI test suite.
Makes me wonder why Peter thinks he saw large performance improvements with XCL.
Recording the patch here so it gets recorded. However, closing this issue for lack of evidence of improvement; no additional complexity seems to be warranted based on these findings.
Result caching is what I mean, ofcourse.