That's certainly a lot higher than I would expect. Are you sure you're getting the ARMvTE inline assembly optimizations?
On Feb 8, 3:04 am, Andy Quan <[email protected]> wrote: > Hi, > I am working on android multimedia on an ARMv5TE compatible chip with > hardware SIMD acceleration (CPU 600Mhz, 32K I$/D$, no L2$, 130Mhz external > DDR, 128MB DRAM). With some performance calibration codes, I find that > "audio flinger" thread costs around 50Mhz - 60Mhz when music player is > playing any 44Khz - 48Khz sample rate stereo audio (single track). This is > much higher beyond my expectation (20Mhz - 30Mhz). > > In my understanding, major cost of audio flinger should be resampling, > mixing, alsa rendering and some data copy. > 1. And I have confirmed that my testing case does not require resampling, > that is, a simple mixing and rendering will lead to the end. > 2. I have confirmed that on this platform alsa rendering does not need so > many cycles as I observed. > > Anybody can help me understand why this audio flinger thread costs so many > Mhz? Or did I misunderstand anything? I guess I did not catch the bottleneck > of this audio flinger:) > > BTW, my code base is based on the one by the end of last year, so I am not > sure if there is any performance related update in the current release. > > -- > Thanks, > Andy --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "android-framework" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/android-framework?hl=en -~----------~----~----~----~------~----~------~--~---
