- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Here is a simple (well, I hope so...) question about profile-guided optimizations with ECL 7.1 (IPF targets) :
Does the platform where the prof_gen compilation and the instrumented execution are made has any influence on the binary obtained later with the final feedback compilation ? Particularly: does the cache capacity of the MPU has any impact on the code generated when using PGOs ?
Does the platform where the prof_gen compilation and the instrumented execution are made has any influence on the binary obtained later with the final feedback compilation ? Particularly: does the cache capacity of the MPU has any impact on the code generated when using PGOs ?
Link Copied
2 Replies
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Bronx,
No. PGO really has no awareness of cache sizes or hit/miss rates. Its instrumentation is mainly measuring control flow and basic block counts. It does not measure cache hit or any other events you could measure using say for example, the processor's performance counters.
Max
No. PGO really has no awareness of cache sizes or hit/miss rates. Its instrumentation is mainly measuring control flow and basic block counts. It does not measure cache hit or any other events you could measure using say for example, the processor's performance counters.
Max
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
thanks a lot for your answer
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page