|
Hi there, Just found this forum. Very glad to know there are so many friends discussing C6x DSP. We have been using C6416 to develop a video encoder for a long time. A problem kept confusing us, which is the execution time on simulator is quite different from the result on emulator. We use the timer in C6416 to get the execution time. For encoding a D1 size image, the gap can even over 60%, in another word, if we can get 30fps on simulator but actually result is less than 15fps on emulator. The key issue here is we even don't know where the gap comes from. As a result, we don't know how to further improve the encoder. Does anybody have rough idea about that? Thanks a lot in advance. Best, Quentin |
|
|
gap between emulator and simulator?
Started by ●August 11, 2004
Reply by ●August 11, 20042004-08-11
|
qing_lee73 wrote: > > Hi there, > > Just found this forum. Very glad to know there are so many friends > discussing C6x DSP. > > We have been using C6416 to develop a video encoder for a long time. > A problem kept confusing us, which is the execution time on simulator > is quite different from the result on emulator. We use the timer in > C6416 to get the execution time. For encoding a D1 size image, the > gap can even over 60%, in another word, if we can get 30fps on > simulator but actually result is less than 15fps on emulator. The key > issue here is we even don't know where the gap comes from. As a > result, we don't know how to further improve the encoder. > > Does anybody have rough idea about that? Thanks a lot in advance. > > Best, > Quentin > > _____________________________________ Depends on which simulator you are using. If you use the device simulator, it is as close as the actual end hardware traget as possible. The more variables you model, the slower the speed of the simulation. Certain things like EMIF, throughput on buses with multiple requestors are hard to model. Bank openings and closings are hard to model as well. Use device simulator and this should be fairly close. Regds Jagadeesh Sankaran |
