|
QUDA
0.9.0
|


Public Member Functions | |
| void | operator() (DslashCuda &dslash, cudaColorSpinorField *in, const int volume, const int *faceVolumeCB, TimeProfile &profile) |
Public Member Functions inherited from anonymous_namespace{dslash_policy.cuh}::DslashPolicyImp | |
| virtual | ~DslashPolicyImp () |
Variation of multi-gpu dslash where the packing kernel writes buffers directly to host memory
Definition at line 1595 of file dslash_policy.cuh.
|
inlinevirtual |
Implements anonymous_namespace{dslash_policy.cuh}::DslashPolicyImp.
Definition at line 1597 of file dslash_policy.cuh.
References dslash::aux_worker, comm_peer2peer_enabled(), DslashCommsPattern::commDimTotal, anonymous_namespace{dslash_policy.cuh}::commsComplete(), DslashCommsPattern::commsCompleted, DslashCommsPattern::completeSum, dslash_comms, dslash_exterior_compute, dslash_interior_compute, DslashCommsPattern::dslashCompleted, dslash::dslashStart, anonymous_namespace{dslash_policy.cuh}::getStreamIndex(), quda::Host, fused_exterior_ndeg_tm_dslash_cuda_gen::i, in, INTERIOR_KERNEL, anonymous_namespace{dslash_policy.cuh}::issuePack(), anonymous_namespace{dslash_policy.cuh}::issueRecv(), quda::Nstream, DslashCommsPattern::previousDir, PROFILE, quda::QUDA_PROFILE_COMMS_START, quda::QUDA_PROFILE_DSLASH_KERNEL, quda::QUDA_PROFILE_EVENT_RECORD, quda::QUDA_PROFILE_STREAM_WAIT_EVENT, quda::QUDA_PROFILE_TOTAL, quda::qudaEventRecord(), quda::qudaStreamSynchronize(), quda::qudaStreamWaitEvent(), quda::Remote, anonymous_namespace{dslash_policy.cuh}::setMappedGhost(), and streams.

1.8.14