OpenCL Intel FPGA shift and masking

[DRuiz9]

Hi. I'm trying to perform shift right operations via bit masking. I want realize a 16 bit right shift, therefore I have to take only the 16 MSB from the result. value variable is a 32 bits integer. One way to achieve this that works (obviously) is apply :

result = value >> 16

 
The Intel OpenCL documentation puts an example to use masking:

__kernel fixed_point_add (__global const unsigned int * restrict a,
__global const unsigned int * restrict b,
__global unsigned int * restrict result)
{
size_t gid = get_global_id(0);
unsigned int temp;
temp = 0x3_FFFF & ((0x1_FFFF & a[gid]) + ((0x1_FFFF & b[gid]));
result[gid] = temp & 0x3_FFFF;
}

 I understand that:

• 0x1_FFFF := 0000 0000 0000 0001 1111 1111 1111 1111
• 0x3_FFFF := 0000 0000 0000 0011 1111 1111 1111 1111

Thus, to get the 16 MSB bits I will have to set a mask like this:

• 0xFFFF0000 := 1111 1111 1111 1111 0000 0000 0000 0000

result = (0xFFFF0000 & value)

But, the result that I get is always 0...
To sum up:

1. How it actually works bit masking in OpenCL?
2. Is possible get specific bits from a int data type and assign them to other int? 

[AnilErinch_A_Intel]

Hi ,

You can try the different bit wise operations as a sample code then print the results to understand their operations.

Thanks and Regards

Anil