Bug in CGEMM produces incorrect results

Accelerated Computing CUDA CUDA Programming and Performance
1

Edited: I’m an idiot.
cgemmBug.cu (3.52 KB)

2

Should this line:

cublasCgemm(‘n’, ‘n’, 3, 3, 1, make_float2(1,1), ca_ptr, 3, cb_ptr, 1, make_float2(0, 0), cab_ptr, 3);

actually be:

cublasCgemm(‘n’, ‘n’, 3, 3, 1, make_float2(1,0), ca_ptr, 3, cb_ptr, 1, make_float2(0, 0), cab_ptr, 3);

?

3

Ahhhhhhhhhhhhh. Thank you.