For Q1, perhaps the Codec Engine's Engine_getCpuLoad() might help at a high level? It's not very precise, but could help with a gross understanding of how loaded the DSP is.
Chris > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] > ] On Behalf Of Lee Dong Jun > Sent: Sunday, January 27, 2008 9:47 PM > To: Davinci-linux-open-source@linux.davincidsp.com > Subject: davinci performance test > > dear all > > I have 3 questions about performace. > > Q1: Could you know me how to measure the performace of dsp or where to > get the code for performace of dsp using the codec engine? > > Q2: To measure the performace of arm, I tested using "lmbench", > "whetstone", "Arithmetic" and "dhrystone". > the following is one of them, pls look at float and double part in the > Davinci ARM, why multiplication is fatster than adder? > > [x86] > =============== Performance Test =============== > Loop Count : 10000000 > > [Integer] > Add: 0.022108 seconds a: 1, b: 1, result: 10000001 > Sub: 0.022264 seconds a: 1, b: 1, result: -9999999 > Mul: 0.047729 seconds a: 1, b: 1, result: 1 > Div: 0.238955 seconds a: 1, b: 1, result: 1 > > [Float] > Add: 0.084810 seconds a: 1.000000, b: 1.000000, result: > 10000001.000000 > Sub: 0.084962 seconds a: 1.000000, b: 1.000000, result: > -9999999.000000 > Mul: 0.091097 seconds a: 1.000000, b: 1.000000, result: 1.000000 > Div: 0.207312 seconds a: 1.000000, b: 1.000000, result: 1.000000 > > [Double] > Add: 0.084813 seconds a: 1.000000, b: 1.000000, result: > 10000001.000000 > Sub: 0.084825 seconds a: 1.000000, b: 1.000000, result: > -9999999.000000 > Mul: 0.091098 seconds a: 1.000000, b: 1.000000, result: 1.000000 > Div: 0.207333 seconds a: 1.000000, b: 1.000000, result: 1.000000 > > [Davinci ARM] > =============== Performance Test =============== > Loop Count : 10000000 > > [Integer] > Add: 0.815467 seconds a: 1, b: 1, result: 10000001 > Sub: 0.816182 seconds a: 1, b: 1, result: -9999999 > Mul: 0.886441 seconds a: 1, b: 1, result: 1 > Div: 2.870055 seconds a: 1, b: 1, result: 1 > > [Float] > Add: 4.052900 seconds a: 1.000000, b: 1.000000, result: > 10000001.000000 > Sub: 4.097681 seconds a: 1.000000, b: 1.000000, result: > -9999999.000000 > Mul: 3.019289 seconds a: 1.000000, b: 1.000000, result: 1.000000 > Div: 2.320593 seconds a: 1.000000, b: 1.000000, result: 1.000000 > > [Double] > Add: 5.885883 seconds a: 1.000000, b: 1.000000, result: > 10000001.000000 > Sub: 5.981949 seconds a: 1.000000, b: 1.000000, result: > -9999999.000000 > Mul: 4.667958 seconds a: 1.000000, b: 1.000000, result: 1.000000 > Div: 3.563422 seconds a: 1.000000, b: 1.000000, result: 1.000000 > > > > Q3 : i tested using "CrossCorrelation", the following is the result of > that, the time in [Davinci ARM] is strange. could you tell me why ARM > is very very slow? > > =============== Performance Test =============== > > Loop count : 100000 > > [x86] > 0.079904 seconds > > [Davinci ARM] > 12.728044 seconds > > [Davinci DSP] > 0.055818 seconds > > > > > -- > Thank you and Best regards, > Lee > > Lee Dong Jun (李 東 俊) > Assistant Manager > Daesung Industrial Corp > tel : +82-2-3408-1477 > fax : +82-2-3409-0310 > cell : +82-10-4879-0917 > [EMAIL PROTECTED] > [EMAIL PROTECTED] > _______________________________________________ Davinci-linux-open-source mailing list Davinci-linux-open-source@linux.davincidsp.com http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
