Changes between Version 66 and Version 67 of DataParallel/BenchmarkStatus


Ignore:
Timestamp:
Dec 2, 2010 7:46:57 AM (5 years ago)
Author:
benl
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • DataParallel/BenchmarkStatus

    v66 v67  
    177177----
    178178= Key =
    179 dph.<benchmark>.<version>.<parallelism>.[threads][[br]]
    180 repa.<benchmark>.[version].[threads][[br]]
    181 
    182 Project
    183  * Either ''dph'' or ''repa''. Repa programs use the same parallel array library as DPH, but do not go through the vectorising transform.
    184 
    185 Version
     179dph.<''benchmark''>.<''version''>.<''parallelism''>.[''threads''][[br]]
     180repa.<''benchmark''>.[''version''].[''threads''][[br]]
     181
     182''version''
    186183 * ''vectorised'' means it's been through the DPH vectorising transform.
    187184 * ''vector'' is a hand written version using immutable Data.Vectors
     
    189186 * ''vector-immutable'' means the same as ''vector'' and is used when there is also an mutable version.
    190187
    191 Parallelism
     188''parallelism''
    192189 * Whether a benchmark is natively parallel or sequential.
    193190 * Parallel versions are also run single threaded (with -N1) and sequential versions are also run with (-N4) so we get the parallel GC.
    194191 * Parallel versions with -N1 will tend to be slower than natively sequential versions due to overheads for supporting parallelism.
    195192
    196 Speedup
     193''threads''
     194 * Value passed to Haskell Runtime with -N threads flag.
     195 * Number of Haskell Execution Contexts (HECs) used when running the benchmark.
     196 * Can be less than the number of hardware threads / cores in the physical machine.
     197
     198'''speedup'''
    197199 * Runtime of reference / runtime of benchmark.
    198200 * Measures how much faster a benchmark is relative to the reference.
    199201
    200 Relative Efficiency.
     202(relative) '''efficiency'''
    201203 * Speedup / number of threads.
    202204 * Indicates the communication overhead involved with running something in parallel.
    203205 * Can be > 1 if the parallel version running with a single thread is faster than the sequential reference version.
    204206
    205 Status
     207'''Status:'''
    206208 * '''BROKEN''': Benchmark doesn't compile, or crashes when run.
    207209 * '''SLOWDOWN''': Benchmark gets slower as number of threads increases.