Commits

Show all
Author Commit Message Labels Comments Date
Misha Brukman
Merge from trunk
Branches
parallel
convert-repo
This commit was manufactured by cvs2svn to create branch 'parallel'.
Branches
parallel
Chris Lattner
Finegrainify namespacification
Tags
RELEASE_11
Chris Lattner
Fixes for PR114: Thanks to Reid Spencer!
Chris Lattner
Update for final version of paper
Chris Lattner
Pass zero when we have a collapsed node, so the PA runtime can be a bit more thoughtful
Chris Lattner
Don't bother destroying any pools allocated in main
Chris Lattner
count the number of typesafe pools
John T. Criswell
Do not set the number of nodes per slab to the maximum if we're allocating an array that fits within a page. The array might have larger nodes.
Chris Lattner
Remvoe the haxor out of assert. The release version won't have them.
Chris Lattner
While we're at it, allocate space for the links list as well. This fixes em3d in large problem size
Chris Lattner
Make sure to reserve space for the slabheader header as well
Chris Lattner
Do not export poolallocarray, make it static and move it before poolalloc instead
John T. Criswell
Fixed the faster allocation of arrays. With Chris's new numbers, performance is looking better. Half of the Olden tests show a speed improvement.
John T. Criswell
Disable faster array allocation until it works.
Chris Lattner
Compute speedup, not whatever it was I had before
Chris Lattner
Don't force Large Problem Size
Chris Lattner
Completely rewrite handling of singlearrays
John T. Criswell
Paranoia: Ensure that we're always allocating something.
Chris Lattner
Indirect call sites that only call external functions, they don't/can't pass extra arguments.
Chris Lattner
Uncomment out my assertions, fix buggy assertion
John T. Criswell
Simplified code that scans the free array list looking for a slab. Corrected the free array scanning code to update fields when an array slab is allocated.
John T. Criswell
Added code to allocate multiple small arrays within a single slab (assuming the slab is a single page). This slightly improves half of the Olden programs from their previous settings.
Chris Lattner
Implement a small optimization to handling of GEP's that are equivalent to casts. This results in substantially reduced collapsing for some testcases
John T. Criswell
Removed commented out macros that mess with our beloved asserts.
Chris Lattner
Default to disabling the optimization
John T. Criswell
Calculate the maximum number of nodes per page once for each pool instead of every time we create a slab. This provides minor performance improvement, I think.
John T. Criswell
Fixed a bug in the code. Unfortunetly, this seems to make it a little slower in some spots.
Chris Lattner
Fix spelling, remove bogus assertion
John T. Criswell
Removed needless pointer deferencing. This seems to speed up some of the Olden benchmarks.
  1. Prev
  2. Next