 Samuel Thibault
|
57e59bf2d9
provide good examples by always using cudaMemcpyAsync, not cudaMemcpy
|
лет назад: 14 |
|
Samuel Thibault
|
f18180c477
synchronize only the transfer stream
|
лет назад: 14 |
|
Samuel Thibault
|
3e58d92492
Add 2d shadow filters
|
лет назад: 14 |
|
Samuel Thibault
|
cb492ad84b
Keep documentation coherent with C convention
|
лет назад: 14 |
|
Samuel Thibault
|
972caa3d96
Fix build without OpenGL headers
|
лет назад: 14 |
|
Samuel Thibault
|
70864cd9ea
there are some optimizations with eager, just not wise scheduling
|
лет назад: 14 |
|
Samuel Thibault
|
1d7dcdc5b6
fix compilation warning
|
лет назад: 14 |
|
Samuel Thibault
|
c3610e1434
drop unused factor
|
лет назад: 14 |
|
Samuel Thibault
|
9803389d55
fix number
|
лет назад: 14 |
|
Samuel Thibault
|
fd06e119aa
more comments
|
лет назад: 14 |
|
Samuel Thibault
|
cf5341ae88
more details on the shadow at work
|
лет назад: 14 |
|
Samuel Thibault
|
6bc0d82b7c
fix non-cuda/opencl build
|
лет назад: 14 |
|
Samuel Thibault
|
a88b26f1a7
update changelog
|
лет назад: 14 |
|
Samuel Thibault
|
ee1e7926a3
Put single-combined-worker in nice order for starpu_machine_display
|
лет назад: 14 |
|
Samuel Thibault
|
3b540f359f
document that starpu_machine_display shows combined workers
|
лет назад: 14 |
|
Samuel Thibault
|
65688ef4d8
print topology before bandwidth
|
лет назад: 14 |
|
Samuel Thibault
|
1aeeab7392
mergeinfo for backport
|
лет назад: 14 |
|
Samuel Thibault
|
844951370b
Show topology in starpu_machine_display
|
лет назад: 14 |
|
Samuel Thibault
|
bb9da9185b
Fix output alignment
|
лет назад: 14 |
|
Samuel Thibault
|
8f8194dfa6
forwardport r6996 from 1.0: fix bus calibration for more than 32 cpus
|
лет назад: 14 |
|
Samuel Thibault
|
ab9b3fcfbd
forwardport r6997 from 1.0: fix bus calibration for more than 32 cpus
|
лет назад: 14 |
|
Samuel Thibault
|
f33693b5d4
Add actual bus measurements in starpu_machine_display output
|
лет назад: 14 |
|
Samuel Thibault
|
f30a1170aa
Replace starpu_force_bus_sampling with bus_calibrate configuration option, as calibration should done carefully during starpu initialization, and not after initialization
|
лет назад: 14 |
|
Samuel Thibault
|
09dab6fd3e
Drop spurious nbsp
|
лет назад: 14 |
|
Samuel Thibault
|
c39f87d9cd
clearly separate CUDA and OpenCL
|
лет назад: 14 |
|
Samuel Thibault
|
473617f4c0
Do not print CPU%d, it is misleading, explain that it is CPUs in preferrence order, instead
|
лет назад: 14 |
|
Samuel Thibault
|
75ad46f0de
document temporary buffer optimizations
|
лет назад: 14 |
|
Samuel Thibault
|
a7aa151601
revert r6895, it perturbs measurements
|
лет назад: 14 |
|
Samuel Thibault
|
226d8973c7
Do not use starpu_cuda_set_device while calibrating the bus: since we clear things with cudaThreadExit, opengl interoperability can be enabled at proper starpu initialization
|
лет назад: 14 |
|
Samuel Thibault
|
9b373e7f9c
Fix initialization for CUDA bus calibration: we use cudaThreadExit to clean our environment before starting starpu
|
лет назад: 14 |
