Help with ippmL2Norm and zero vectors

[samueleforconi]

Hi, I was trying to compute the norm of some vector and vector arrays.
My question is: why the function "ippmL2Norm" returns -1.#IO when
in input there is a vector with all 0 components?
According to the formula on the manual I think that it should return 0.

This is the piece of code that I use for test:
Ipp32f c[5*3] = { 1.0f, 2.0f, 3.0f,
		  0.0f, 0.0f, 0.0f,
		  7.0f, 8.0f, 9.0f,
		  3.0f, 1.0f, 8.0f,
		  9.0f, 6.0f, 4.0f }; 
Ipp32f d[5*3] = { 3.0f, 5.0f, 6.0f,
		  5.0f, 1.0f, 1.0f,
		  7.0f, 8.0f, 3.0f,
		  1.0f, 9.0f, 5.0f,
		  2.0f, 2.0f, 6.0f };
Ipp32f e[3]   = { 0.0f, 0.0f, 0.0f };

int cstride0 = sizeof(Ipp32f)*3;
int cstride2 = sizeof(Ipp32f);
int dstride0 = sizeof(Ipp32f)*3;
int dstride2 = sizeof(Ipp32f);
int estride2 = sizeof(Ipp32f);
	
int len   = 3;
int count = 5;
int i     = 0;
	
Ipp32f dst[5];
Ipp32f c_nor[5];
Ipp32f d_nor[5];
Ipp32f e_nor;

ippmDotProduct_vava_32f(c, cstride0, cstride2, d, dstride0, dstride2, dst, len, count);

ippmL2Norm_va_32f(c, cstride0, cstride2, c_nor, len, count);
ippmL2Norm_va_32f(d, dstride0, dstride2, d_nor, len, count);
ippmL2Norm_v_32f(e, estride2, &e_nor, len);

printf("Dot Prod
");
for(i=0;i<5;i++)
	printf("[%d] = %.3f
",i,dst);
printf("
Norm of vector array c:
");
for(i=0;i<5;i++)
	printf("[%d] = %.3f
",i,c_nor);	
printf("
Norm of vector array d:
");
for(i=0;i<5;i++)
	printf("[%d] = %.3f
",i,d_nor);
printf("
Norm of vector e: %.3f", e_nor);


And in output I have:
Dot Prod
[0] = 31.000
[1] = 0.000
[2] = 140.000
[3] = 52.000
[4] = 54.000

Norm of vector array c:
[0] = 3.742
[1] = -1.#IO
[2] = 13.928
[3] = 8.602
[4] = 11.533

Norm of vector array d:
[0] = 8.367
[1] = 5.196
[2] = 11.045
[3] = 10.344
[4] = 6.633

Norm of vector e: -1.#IO

I can't understand where is my error, thank you for your help.
Samuele.

[Vladimir_Dudnik]

Hi Samuele,

could you please specify what version of IPP do you use, how do you link that and what is your target platform and OS?

Regards,
Vladimir

[samueleforconi]

Hi,
thank you for your answer.

I'm actually using Intel IPP v5.2 for Windows on IA-32.
I use dynamic linkage (to ippm.lib from directory stublib).

Regards,
Samuele.

[Vladimir_Dudnik]

Hello,

there is comment from our expert:

Function ippmL2Norm really returns -1.#IO in case of input vector with all 0 components for vector length =3,4,5,6 on IA32&EM64t architecture.

This implementation error will be corrected to IPP 6.0 version.

Regards,
Vladimir