topic Hi, in Intel® oneAPI Math Kernel Library

Element-wise vector comparison

Mikhail_K_ — Fri, 17 Mar 2017 15:50:54 GMT

Hi,

I have a huge mxn matrix A where I need to compute A(i,j) > 0, ie. 1 for all values which are positive, and 0 otherwise

I would be better off writing it in C++, but let's assume I'm in a situation where I don't have admin rights and cannot install/compile anything. I'm using a single-threaded scripting language (VBA) to wrap calls to mkl_rt.dll, and any combination of MKL functions is still faster.

One way to solve this is to compute 1-0^(x+abs(x)), ie something like

vdabs m*n, A(1, 1), temp(1, 1)
vdadd m*n, A(1, 1), temp(1, 1), temp(1, 1)
vdpow m*n, res(1, 1), temp(1, 1), res(1, 1) /***** (res() is an array full of zeros)
vdlinearfrac m*n, res(1, 1), res(1, 1), 0, -1, 0, 1, res(1, 1)

But the vdpow function is quite costly computation-wise (overall speed difference is ca x4 in favor of mkl, but I expect more from one single thread in VBA vs compiled multi-threaded code)

Another solution is to find the maximum value in the array, divide by max+1 to force all values into <-1, 1> and apply vdceil, <-1, 0] becomes 0, and <0, 1> becomes 1, but there can be some problems if the numbers are huge and end up as [-1, 1]

Can anyone think of a simpler way to do it?

Hi Mikhail,

Ying_H_Intel — Mon, 27 Mar 2017 01:19:11 GMT

Hi Mikhail,

It seems no simple function in MKL can do such operation

If possible, could you use other library like IPP library, which can do such operation by one function

for example

Conversion of a grayscale image to a bitonal image
IppStatus ippiGrayToBin_<srcDataType>1u_C1R (please search IPP's developer guide)

ippiThreshold_GTVal etc.

You mentioned you can't compile on one machine, Could it possible for you to wrap IPP dll on other machine with admin right , then copy the dll to VBA, thus you can use IPP dll?

Best Regards,

Ying

Hi,

Mikhail_K_ — Tue, 11 Apr 2017 09:07:06 GMT

Hi,

Yep, thanks, that's what I ended up doing.

Wasted whole evening trying to export some functions in ipps.h as stdcall to a custom dll... But then realized stdcall is default in IPP... Is it?

In MKL its cdecl (though most functions have stdcall)?

By the way, why are multi

Mikhail_K_ — Wed, 12 Apr 2017 09:59:48 GMT

By the way, why are multi-threaded vector arithmetic functions (like ippsMul_64f) depreciated in IPP, but its seemingly ok to have multi-threaded vdMul in MKL?

>>...

SergeyKostrov — Thu, 13 Apr 2017 22:49:15 GMT

>>... >>...stdcall is default in IPP... Is it? Yes for Windows platforms. ( It is not clear what platform you're using / I think Windows... ) [ ippdefs.h ] ... #if defined( _WIN32 ) || defined ( _WIN64 ) #define __STDCALL __stdcall #define __CDECL __cdecl #define __INT64 __int64 #define __UINT64 unsigned __int64 #else #define __STDCALL #define __CDECL #define __INT64 long long #define __UINT64 unsigned long long #endif #if !defined( IPPAPI ) #if defined( IPP_W32DLL ) && (defined( _WIN32 ) || defined( _WIN64 )) #if defined( _MSC_VER ) || defined( __ICL ) #define IPPAPI( type,name,arg ) \ __declspec(dllimport) type __STDCALL name arg; #else #define IPPAPI( type,name,arg ) type __STDCALL name arg; #endif #else #define IPPAPI( type,name,arg ) type __STDCALL name arg; #endif #endif ...

Hi MK, Sergey,

Ying_H_Intel — Mon, 17 Apr 2017 01:39:15 GMT

Hi MK, Sergey,

Yes, IPP custom dll is using default stdcall.

Thanks

Ying