1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
|
// Decide branch direction for GPU use for the dot-product MxM
extern "C"
{
#include "GB_mxm.h"
}
#include "GB_cuda.h"
#include <cuda_runtime.h>
bool GB_AxB_dot3_cuda_branch
(
const GrB_Matrix M, // mask matrix
const bool Mask_struct, // if true, use the only structure of M
const GrB_Matrix A, // input matrix
const GrB_Matrix B, // input matrix
const GrB_Semiring semiring, // semiring that defines C=A*B
const bool flipxy, // if true, do z=fmult(b,a) vs fmult(a,b)
GB_Context Context
)
{
// very rough estimate of the work to do
double adeg = ((double) GB_nnz (A)) / ((double) GB_IMAX (1, A->nvec)) ;
double bdeg = ((double) GB_nnz (B)) / ((double) GB_IMAX (1, B->nvec)) ;
double work = GB_nnz (M) * GB_IMIN (adeg, bdeg) ;
// TODO if A or B are not accessed (first, 2nd, or pair ops)
// then the type if A can be user-defined here, for CUDA.
// The code generation process currently does not support user-defined
// types and operators, but this needs to be handled.
int ngpus_to_use = GB_ngpus_to_use (work) ;
GBURBLE (" work:%g GPUs:%d ", work, ngpus_to_use) ;
if (ngpus_to_use > 0
// FIXME: FUTURE: user-defined and complex types and operators
// FIXME: guard against user-defined ADD and MULT
// && (A->type->code < GB_FC32_code)
// && (B->type->code < GB_FC32_code)
)
{
return true;
}
else
{
// FIXME: remove debug outout here:
std::cout << "Not using cuda path. M_is_hypersparse: " << GB_IS_HYPERSPARSE(M) <<
", A->iso: " << A->iso << ", B->iso: " << B->iso << ", A_BITMAP: " << GB_IS_BITMAP(A) <<
", B_BITMAP: " << GB_IS_BITMAP(B) << ", GB_IS_FULL(A): " << GB_IS_FULL(A)
<< ", GB_IS_FULL(B): " << GB_IS_FULL(B) << ", semiring header size: " << semiring->header_size <<
std::endl;
return false;
}
}
|
