Michal Mrozek fb8706d2b4 Free GPU Virtual Address Range in case of pin failure. ...

-Fixes VA range leak in case memory is not GPU accessible.
Signed-off-by: Michal Mrozek <michal.mrozek@intel.com>

Change-Id: I74d7f43ec7f10ba2734c7d55b335c189229d71b5

2020-01-13 17:24:57 +01:00

..

linux

Free GPU Virtual Address Range in case of pin failure.

2020-01-13 17:24:57 +01:00

windows

Change memory interface makeResident call to accept multiple allocations

2020-01-13 12:50:23 +01:00

CMakeLists.txt

Move files to core

2020-01-07 16:53:35 +01:00

device_factory.cpp

Make MemoryOperationsHandler adapter specific

2019-12-30 19:26:19 +01:00

device_factory.h

Move HardwareInfo ownership to ExecutionEnvironment [1/n]

2019-05-08 16:11:01 +02:00

hw_info_config_bdw_plus.inl

clGetDeviceInfo support unified shared memory 2/n.

2019-07-15 18:24:46 +02:00

hw_info_config.h

Move headers to core

2019-12-02 14:20:45 +01:00

hw_info_config.inl

Change namespace from OCLRT to NEO

2019-03-26 15:48:19 +01:00

metrics_library.cpp

Performance Counters: gen12 support

2019-11-19 16:58:04 +01:00

metrics_library.h

Move files to core

2019-11-19 16:19:52 +01:00

ocl_reg_path.h

Fix OCL specific registry path in a core dir

2019-09-02 17:33:50 +02:00

os_context.h

Check Windows OsContext is properly initialized

2019-12-20 12:25:43 +01:00

os_inc_base.h

Move linux files to core

2019-12-13 09:23:33 +01:00

os_interface.h

Remove not needed header inclusions

2019-12-11 16:44:12 +01:00

os_time.cpp

Move headers to core

2019-12-02 14:20:45 +01:00

os_time.h

Change namespace from OCLRT to NEO

2019-03-26 15:48:19 +01:00

performance_counters.cpp

Refactor PerfCounters initialization. Allow only 1 engine type at once

2019-12-19 11:50:31 +01:00

performance_counters.h

Refactor PerfCounters initialization. Allow only 1 engine type at once

2019-12-19 11:50:31 +01:00