 Samuel Thibault
|
7d2bf0c630
Enable asynchronous CUDA kernels wherever possible
|
11 anos atrás |
 Nathalie Furmento
|
98d9eb5c64
Functions starpu_insert_task and starpu_mpi_insert_task are renamed
|
12 anos atrás |
|
Nathalie Furmento
|
30860cac02
Renaming of diverse functions and datatypes.
|
12 anos atrás |
|
Samuel Thibault
|
02d4168c57
Fix reduction codelets modes further
|
12 anos atrás |
|
Samuel Thibault
|
ad0f7bbef7
Set reduction codelet modes in codelet, not at execution
|
12 anos atrás |
|
Nathalie Furmento
|
5e28436be9
examples: field where of struct starpu_codelet is optional, no longer set it
|
12 anos atrás |
|
Samuel Thibault
|
77fb61de8d
Use starpu_data_invalidate_submit, not starpu_data_invalidate, to interlace computations with invalidations
|
12 anos atrás |
|
Samuel Thibault
|
84152479db
destroy data instead of setting it to zero
|
12 anos atrás |
|
Nathalie Furmento
|
c2dbabb2e3
examples: cg example needs a CUDA card which supports double
|
13 anos atrás |
|
Samuel Thibault
|
1ae1ac96cf
Use starpu_data_cpy in CG
|
13 anos atrás |
|
Nathalie Furmento
|
0cbe489754
Function renaming:
|
13 anos atrás |
|
Nathalie Furmento
|
f766606c07
examples: when submission fails because it does not find a suitable worker, it is not an error, just skip the example
|
14 anos atrás |
|
Nathalie Furmento
|
572473b193
examples: check return values of starpu functions
|
14 anos atrás |
|
Nathalie Furmento
|
b9476e6d4a
update code w.r.t coding style
|
14 anos atrás |
|
Nathalie Furmento
|
61940dab45
examples: inside struct starpu_codelet, use xxx_funcs fields instead of the xxx_func ones
|
14 anos atrás |
|
Nathalie Furmento
|
b3a6c1f52b
do not use typedef for starpu_codelet, use it directly as a struct
|
14 anos atrás |
|
Nathalie Furmento
|
0ebf7ba326
Rename public opaque type starpu_data_handle as starpu_data_handle_t
|
14 anos atrás |
|
Nathalie Furmento
|
c80a8acea1
Datatypes renaming
|
14 anos atrás |
|
Nathalie Furmento
|
7517525097
examples/cg/cg_kernels.c: include limits.h which defines INT_MAX
|
14 anos atrás |
 Cédric Augonnet
|
339892a2a6
Use a memcpy operation that relies on a stream
|
14 anos atrás |
|
Cédric Augonnet
|
48aea5e024
Bug fix: use a kernel to actually put zeros in the vector that is initialized
|
14 anos atrás |
|
Cédric Augonnet
|
1618f94c1f
Instead of using cublasSdot or cublasDdot, we reimplement scalar product in
|
14 anos atrás |
|
Cédric Augonnet
|
2c390b92c0
Use the starpu_insert_task helper that makes it easier to deal with a stack of
|
14 anos atrás |
|
Samuel Thibault
|
0d9f08ae8d
Fix copyrights, thanks Ludo!
|
14 anos atrás |
|
Cédric Augonnet
|
e5d5a53202
Thanks to implicit deps, there is no need for a barrier anymore.
|
15 anos atrás |
|
Cédric Augonnet
|
ef29dd6da4
Fix compilation on machines without CUDA
|
15 anos atrás |
|
Cédric Augonnet
|
c750225a2b
Fix the CG example:
|
15 anos atrás |
|
Cédric Augonnet
|
14ebc94929
Implement cg with filters
|
15 anos atrás |
|
Cédric Augonnet
|
9c959490e7
Start to implement a conjugate gradient
|
15 anos atrás |
