diff options
| author | Ivan Maidanski <ivmai@mail.ru> | 2011-12-22 13:08:10 +0400 |
|---|---|---|
| committer | Ivan Maidanski <ivmai@mail.ru> | 2011-12-22 13:08:10 +0400 |
| commit | 2dd56eb49583be85fa44b9975fd63b0df85e8c15 (patch) | |
| tree | bbd828be3bea9b4271c0f1c8c9cd2ddf1a1e1f77 /doc/scale.html | |
| parent | a8c3d5aa2da414b8d47f5688dd4972c62ee28d1b (diff) | |
| download | bdwgc-2dd56eb49583be85fa44b9975fd63b0df85e8c15.tar.gz | |
Remove trailing spaces at EOLn in all files
Diffstat (limited to 'doc/scale.html')
| -rw-r--r-- | doc/scale.html | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/doc/scale.html b/doc/scale.html index b9de2e40..feb14f13 100644 --- a/doc/scale.html +++ b/doc/scale.html @@ -31,7 +31,7 @@ outperforms native queue-based mutual exclusion implementations in most cases, sometimes drastically so. <H2>Options for enhanced scalability</h2> Version 6.0 of the collector adds two facilities to enhance collector -scalability on multiprocessors. As of 6.0alpha1, these are supported +scalability on multiprocessors. As of 6.0alpha1, these are supported only under Linux on X86 and IA64 processors, though ports to other otherwise supported Pthreads platforms should be straightforward. They are intended to be used together. @@ -60,7 +60,7 @@ with that returned by the standard allocators. It may be used by other threads. The only difference is that, if the thread allocates enough memory of a certain kind, it will build a thread-local free list for objects of that kind, and allocate from that. This greatly reduces -locking. The thread-local free lists are refilled using +locking. The thread-local free lists are refilled using <TT>GC_malloc_many</tt>. <P> An important side effect of this flag is to replace the default @@ -129,7 +129,7 @@ Note that setting GC_NPROCS to 1 also causes some lock acquisitions inside the collector to immediately yield the processor instead of busy waiting first. In the case of a multiprocessor and a client with multiple simultaneously runnable threads, this may have disastrous performance -consequences (e.g. a factor of 10 slowdown). +consequences (e.g. a factor of 10 slowdown). <H2>Performance</h2> We conducted some simple experiments with a version of <A HREF="gc_bench.html">our GC benchmark</a> that was slightly modified to @@ -200,7 +200,7 @@ since the memory system is a major constraint for the garbage collector, the processors usually share a single memory bus, and thus the aggregate memory bandwidth does not increase in -proportion to the number of processors. +proportion to the number of processors. <P> These results are likely to be very sensitive to both hardware and OS issues. Preliminary experiments with an older Pentium Pro machine running |