Readme: | =========================================================== This is recompile V5.10 of: STREAM: Sustainable Memory Bandwidth in High Performance Computers https://www.cs.virginia.edu/stream/
=========================================================== In package there are two versions, both compilled with -O3: stream_efika - stream for testing Efika onboard RAM stream_efika_VRAM - stream for testing Efika VRAM->RAM feature ===========================================================
Requirements: -Efika5200b ( theoretically any MorphOS machine with CPU 5200b, 603e, e300 ) -stream_efika_VRAM version requires graphics card with at least 128 MB VRAM and enabled Radeon tooltypes for using VRAM as Slow RAM: MEM_LIMIT=32 MEM_SPARE2SYS=YES
Both stream_efika and stream_efika_VRAM are compilled from C source with -mcpu=603e parameter and unmodified ruleset and should be compared with Standart results table: The "standard" set of results presents the results of the C or Fortran versions of the STREAM benchmark running with 64-bit data types on production hardware.
-stream_efika measures operations speed in standart RAM. This version uses non-modified algorithm, with STREAM_ARRAY_SIZE=1000000 ( = 1/10 of default size ), which requires only 22.9 MiB of RAM. -stream_efika_VRAM measures operation speed in Radeon VRAM. This version uses also non-modified algorithm, with STREAM_ARRAY_SIZE=4000000, which requires 91.6 MiB of RAM. It allocate more RAM than Efika 5200b has, thus task run in VRAM memory. If you are irish luck and has Efika with 256 MB onboard RAM or more, this benchmark returns same results like stream_efika.
This benchmark is originally designed for multi-CPU, vectorized, 64-bit floating-point data streams. With this benchmark we can compare how our single-CPU, 32-bit machines without vectorization are able to manage such data.
Stream works on top of arrays with 10 million ( Efika has array only 1 mil or 4 mil ) double-precision floating-pont values. Copy: a(i) = b(i) Scale: a(i) = q * b(i) Add: a(i) = b(i) + c(i) Triad: a(i) = b(i) + q*c(i)
Results of stream v5.10 cannot be compared with results from previous versions.
January 19th 2025 Sailor
|