Changes between Version 20 and Version 21 of GpuWorkFetch


Ignore:
Timestamp:
Dec 29, 2008, 4:13:14 PM (16 years ago)
Author:
davea
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • GpuWorkFetch

    v20 v21  
    5353'''double cpu_req_seconds''': number of CPU seconds requested
    5454
    55 '''double cuda_req_seconds''': number of CUDA seconds requested
    56 
    57 '''double ninstances_cpu''': send enough jobs to occupy this many CPUs
    58 
    59 '''double ninstances_cuda''': send enough jobs to occupy this many CUDA devs
     55'''double cpu_req_ninstances''': send enough jobs to occupy this many CPUs
     56
     57And for each coprocessor type:
     58
     59'''double req_seconds''': number of instance-seconds requested
     60
     61'''double req_ninstances''': send enough jobs to occupy this many instances
    6062
    6163For compatibility with old servers, the message still has '''work_req_seconds''',
    62 which is the max of (cpu,cuda)_req_seconds.
     64which is the max of the req_seconds.
    6365
    6466The semantics are: a scheduler should send jobs for a resource type
    6567only if the request for that type is nonzero.
    66 
    67 New fields in the scheduler reply message (these are not currently used):
    68 
    69 '''double have_cpu_jobs''': this project sometimes has CPU jobs for this platform (although this reply may not include any).
    70 
    71 '''double have_cuda_jobs''': same, for CUDA jobs.
    7268
    7369== Client ==