{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

Differential Equations Solutions 6

Differential Equations Solutions 6 - α ≈ 3 ns The block...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
16 Chapter 3. Solutions: Computer Memory and Arithmetic CHALLENGE 3.5. The data cannot be fully explained by our simple model, since, for example, this machine uses prefetching, two levels of cache, and a more complicated block replacement strategy than the least-recently-used one that we discussed. Some of the parameters can be extracted using our model, though. The discontinuity in times as we pass from m = 2 14 to m = 2 15 indicates that the capacity of the cache is 2 14 (single-precision) words (2 16 bytes). The discontinuity between stride s = 2 3 and s = 2 4 says that = 2 3 words, so b = 2 11 words. The elements toward the bottom left corner of the table indicate that
Background image of page 1
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: α ≈ 3 ns. The block of entries for m ≥ 2 15 and 2 6 ≤ s ≤ 2 10 indicates that perhaps μ ≈ 18 − 3 = 15 ns. To further understand the results, consult a textbook on computer organiza-tion and the UltraSPARC III Cu User’s Manual at http://www.sun.com/processors/manuals/USIIIv2.pdf . CHALLENGE 3.6. On my machine, the time for floating-point arithmetic is of the order of the time μ for a cache miss penalty. This is why misses noticeably slow down the execution time for matrix operations. CHALLENGE 3.7. No answer provided. CHALLENGE 3.8. No answer provided....
View Full Document

  • Fall '11
  • Dr.Robin
  • CPU cache, Web server, Untouchable number, SPARC, Sun Microsystems microprocessors, block replacement strategy

{[ snackBarMessage ]}

Ask a homework question - tutors are online