https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949794#M15146 I'm trying to measure the speed of convolution/correlation in MKL and getting a result which seems to be too low: 0.5 gigaflop.<BR /><BR />I'm computing the correlation of a short sequence (n=16) with a long sequence (m=8016) in single precision. I force MKL to use MODE_DIRECT. Using MODE_DIRECT, the op count should be 2*m*n. I do 8000 such correlations and measure the overall time, so time quantization is not an issue. <BR /><BR />I seem to be getting the correct answer, so I'm not grossly misusing the function calls.<BR /><BR />Does anyone have an idea what the problem could be, or have sample code where they measure speed of convolution/correlations? Tue, 24 Jan 2006 09:16:14 GMT wyc 2006-01-24T09:16:14Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949794#M15146 I'm trying to measure the speed of convolution/correlation in MKL and getting a result which seems to be too low: 0.5 gigaflop.<BR /><BR />I'm computing the correlation of a short sequence (n=16) with a long sequence (m=8016) in single precision. I force MKL to use MODE_DIRECT. Using MODE_DIRECT, the op count should be 2*m*n. I do 8000 such correlations and measure the overall time, so time quantization is not an issue. <BR /><BR />I seem to be getting the correct answer, so I'm not grossly misusing the function calls.<BR /><BR />Does anyone have an idea what the problem could be, or have sample code where they measure speed of convolution/correlations? Tue, 24 Jan 2006 09:16:14 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949794#M15146 wyc 2006-01-24T09:16:14Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949795#M15147 <DIV></DIV> <P>The direct convolution/correlation code for IA32shipped with MKL 8.0 and 8.0.1isn't very fast. We are currently working on this performance issue.</P> <P>As a workaround, I would recommend using MODE_FFT or MODE_AUTO to involve the Fourier algorithm whichmustprovide better performance.</P> Tue, 24 Jan 2006 19:34:31 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949795#M15147 Yevgeny_Latkin__Inte 2006-01-24T19:34:31Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949796#M15148 Can you estimate how fast the correlation/convolution should be when it is implemented? I thought convolution was a perfect example of an algorithm which would work well out of cache (assuming that the shorter vector is small). Thu, 26 Jan 2006 10:29:34 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949796#M15148 wyc 2006-01-26T10:29:34Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949797#M15149 <DIV></DIV> <P>Sorry; I cannot make a commitment about future MKL releases.</P> <P>But,we could probably pay more attention to the use-case which you mention -- filtering longer data samples with a very short sequence. Currently, performance looks better for a longer filter, like n=100 and above.</P> <P>Besides, did you try signal processing (FIR) functions from the Intel IPP library?</P> <DIV></DIV> Thu, 26 Jan 2006 17:19:34 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/speed-of-MKL-convolution-correlation/m-p/949797#M15149 Yevgeny_Latkin__Inte 2006-01-26T17:19:34Z