Cyclops Tensor Framework
parallel arithmetic on multidimensional arrays
offload.h File Reference
This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Data Structures

class  CTF_int::offload_arr
 offloaded array/buffer More...
 
class  CTF_int::offload_tsr
 offloaded and serialized tensor data More...
 

Namespaces

 CTF_int
 

Functions

void CTF_int::offload_init ()
 initialize offloading, e.g. create cublas More...
 
void CTF_int::offload_exit ()
 exit offloading, e.g. destroy cublas More...
 
double CTF_int::estimate_download_time (int64_t size)
 estimate time it takes to upload More...
 
double CTF_int::estimate_upload_time (int64_t size)
 estimate time it takes to download More...
 
void CTF_int::host_pinned_alloc (void **ptr, int64_t size)
 allocate a pinned host buffer More...
 
void CTF_int::host_pinned_free (void *ptr)
 free a pinned host buffer More...
 
template<typename dtype >
void CTF_int::offload_gemm (char tA, char tB, int m, int n, int k, dtype alpha, offload_tsr &A, int lda_A, offload_tsr &B, int lda_B, dtype beta, offload_tsr &C, int lda_C)
 
template<typename dtype >
void CTF_int::offload_gemm (char tA, char tB, int m, int n, int k, dtype alpha, dtype const *dev_A, int lda_A, dtype const *dev_B, int lda_B, dtype beta, dtype *dev_C, int lda_C)