| Pre-indexer? [message #104468] |
Wed, 02 June 2004 22:09 |
Eclipse User |
|
|
|
Originally posted by: keithen.hayenga.palmsource.com
Our company's developers typically start their programs with a single
include that in turn includes all our other header files. We notice that a
number of CDT tasks start off by reparsing all of our headers and that
adds tens of seconds to every operation. It seems as though the equivalent
of precompiled headers could result in significant time savings.
My first thought was that index files might be the way to attack this, so
instead of precompiling, this would be preindexing. I started by creating
an index file that had only the symbols from our SDK headers. (Since we
only index translation units, there is a dummy c file involved.) The index
file has the meaningful name like 1706641915.index, which somehow would
have to be linked in to the Index constructors in order to copy it over.
It looked like i would be able to just rename the copy to the indexFile
and create the other tables.
However, the scanner has it's own set of definitions that i have now
bypassed and i was counting on those ifndef statements being recognized so
the scanner would skip over all our header files. So maybe i do need a
custom pre-indexed file format. I could find no discussions about such
topics in the archives and wondered if someone had already thought about
some form of preprocessing. Any hints, guidance, or warnings of pitfalls
would be welcome. While i am still working on 1.2.1 sources, i would hope
that we could roll this forward to CDT 2 and gain us all some performance.
|
|
|
Powered by
FUDForum. Page generated in 0.27000 seconds
Home