cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Visitor
Visitor
846 Views
Registered: ‎05-21-2018

SDSOC and sds++ for half-precision floating-point

Jump to solution

I am trying to implementation the example "Matrix Multiply and ADD" using half float point. When all the code was executed in CPU, the result is right. When I switch "mmult" and "madd" from CPU to FPGA, although it can successfully generate the bit stream and image file, it still runs with error "ERROR: DMASIMPLE could not find physical address for buffer - make sure the buffer is allocated with sds_alloc". In my code, I do use "sds_alloc" to allocate the memory. Should I add some options to sds++?

0 Kudos
Reply
1 Solution

Accepted Solutions
Visitor
Visitor
1,021 Views
Registered: ‎05-21-2018

I tried modify the function declaration, changing the argument from array to pointer, it works finally.

View solution in original post

0 Kudos
Reply
2 Replies
Visitor
Visitor
1,022 Views
Registered: ‎05-21-2018

I tried modify the function declaration, changing the argument from array to pointer, it works finally.

View solution in original post

0 Kudos
Reply
Visitor
Visitor
766 Views
Registered: ‎05-01-2018

hello,I have the same problem with you. I try it according to your method, but it will still be wrong.Can you show me the screenshots of the change?

0 Kudos
Reply