Alternative VE Offloading  2.7.6
Introduction

VE offloading is a framework to provide accelerator-style programming on Vector Engine (VE).Using VEO, a programmer can execute code on VE and can control the execution from VH main program.

This document describes public APIs for VEO. The page "Modules" shows a list of VEO API functions. The page "Examples" shows a list of all example.

Author
Erich Focht
NEC Corporation
Revision History
Revision, Date, Updates/Remarks
Rev.6, Mar. 2021,
This revision covers veoffload-aveo-2.7.6 or later.
Added desctiptions of heterogeneous memory(HMEM) API.
Generated an extra page containing the source of examples.
Added desctiorions of MPI+VEO examples.
Rev.5, Jan. 2021,
This revision covers veoffload-aveo-2.7.5 or later.
Added desctiptions of environment variables to optimize the performance of data transfer.
Deleted restrictions that these APIs could return a success value if veo_write_mem() or veo_async_write_mem() fails.
Rev.4, Dec. 2021,
This revision covers veoffload-aveo-2.7.3 or later.
Added a workaround for slow veo_write_mem() or veo_async_write_mem() transfer speeds depending on memory.
Fixed API reference of veo_get_context().
Added restrictions that these APIs could return a success value if veo_write_mem() or veo_async_write_mem() fails.
Rev.3, Oct. 2020,
This revision covers veoffload-aveo-2.7.1 or later.
Updated the version.
Rev.2, Sep. 2020,
This revision covers veoffload-aveo-2.7.0 or later.
Updated installation instructions.
Updated the size limit for arguments passed to functions.
Added restrictions on using multiple contexts with multiple OpenMP threads.
Fixed minor errors.
Rev.1, Jul. 2020,
This revision covers veoffload-aveo-2.6.2 or later.
First release.