Reply by bharath April 28, 20062006-04-28
Hello,
As per my knowledge,the difference is due to CPU Memory stalls(Memory Bank Conflict). If two values which have to be loaded simultaneously are in the same memory bank,there will be single CPU STALL. By using the pragma DATA_MEM_BANK we can avoid these STALLS to some extent. Another reason is Cache misses. L1D and L1P cache misses also will increase the cycles. When cache profiling is done on simulator ,u can find the number of data and program cache misses. By doing cache optimization these misses can be reduced.

regards
bharath

Swaminathan Sathappan wrote:

Hi everyone,

When I profile the same code in simulator(6416 Functional Simulator) and DSK (TMS320c6416), the difference in MIPS(Million Instructions Per Cycle) between the two is 1.0.Hardware board is consuming more cycles (higher MIPS) than the simulator. What could be the reason??

Swaminathan.S
India.
______________________________________________________________________

---------------------------------

Send instant messages to your online friends http://uk.messenger.yahoo.com
Reply by Swaminathan Sathappan April 28, 20062006-04-28
Hi everyone,

When I profile the same code in simulator(6416 Functional Simulator)
and DSK (TMS320c6416), the difference in MIPS(Million Instructions Per
Cycle) between the two is 1.0.Hardware board is consuming more cycles
(higher MIPS) than the simulator. What could be the reason??

Swaminathan.S
India.

______________________________________________________________________