HIPSYCL_DEBUG_LEVEL
: if set, overrides the output verbosity.0
: none,1
: error,2
: warning,3
: info,4
: verbose, default is the value ofHIPSYCL_DEBUG_LEVEL
macro.HIPSYCL_VISIBILITY_MASK
: can be used to activate only a subset of backends. Syntax:backend;backend2;..
. Possible values areomp
,cuda
,hip
andze
(as level zero is the backend andspirv
inHIPSYCL_TARGETS
just the target format).omp
will always be active as a CPU backend is required. Device level visibility has to be set via vendor specific variables for now, including{CUDA,HIP}_VISIBLE_DEVICES
andZE_AFFINITY_MASK
.HIPSYCL_RT_DAG_REQ_OPTIMIZATION_DEPTH
: maximum depth when descending the DAG requirement tree to look for DAG optimization opportunities, such as eliding unnecessary dependencies.HIPSYCL_RT_MQE_LANE_STATISTICS_MAX_SIZE
: For themulti_queue_executor
, the maximum size of entries in the lane statistics, i.e. the maximum number of submissions to retain statistical information about. This information is used to estimate execution lane utilization.HIPSYCL_RT_MQE_LANE_STATISTICS_DECAY_TIME_SEC
: The time in seconds (floating point value) after which to forget information about old submissions.HIPSYCL_RT_SCHEDULER
: Set scheduler type. Allowed values:direct
is a low-latency direct-submission scheduler.unbound
is the default scheduler and supports automatic work distribution across multiple devices. If theHIPSYCL_EXT_MULTI_DEVICE_QUEUE
extension is used, the scheduler must beunbound
.
HIPSYCL_DEFAULT_SELECTOR_BEHAVIOR
: Set behavior of default selector. Allowed values:strict
(default): Strictly behave as defined by the SYCL specificationmultigpu
: Makes default selector behave like a multigpu selector from theHIPSYCL_EXT_MULTI_DEVICE_QUEUE
extensionsystem
: Makes default selector behave like a system selector from theHIPSYCL_EXT_MULTI_DEVICE_QUEUE
extension
HIPSYCL_HCF_DUMP_DIRECTORY
: If set, hipSYCL will dump all embedded HCF data files in this directory. HCF is hipSYCL's container format that is used by all compilation flows that are fully controlled by hipSYCL to store kernel code.HIPSYCL_PERSISTENT_RUNTIME
: If set to 1, hipSYCL will use a persistent runtime that will continue to live even if no SYCL objects are currently in use in the application. This can be helpful if the application consists of multiple distinct phases in which SYCL is used, and multiple launches of the runtime occur.HIPSYCL_RT_MAX_CACHED_NODES
: Maximum number of nodes that the runtime buffers before flushing work.HIPSYCL_SSCP_FAILED_IR_DUMP_DIRECTORY
: If non-empty, hipSYCL will dump the IR of code that fails SSCP JIT into this directory.