Abstract: A method and apparatus for hybrid pre and post-retire tentative access tracking is herein described. Access tracking is often performed during execution of critical sections, which may be defined by traditional locks or transactional memory instructions. Pre-retire accesses to memory are performed to update tracking information for access during execution of a critical section. However, post-retire updates to tracking information are performed for subsequent consecutive critical section accesses in a pipeline when a previous end critical section operation is retired.

Abstract: A method and apparatus for registering a user-handler in hardware for transactional memory is herein described. A user-accessible register is to hold a reference to a transactional handler. An event register may also be provided to specify handler events, which may be done utilizing user-level software, privileged software, or by hardware. When an event is detected execution vectors to the transaction handler based on the reference to the transactional handler held in the user-accessible register. The transactional handler handles the event and then execution returns to normal flow.

Abstract: A system for optimizing cache coherence message traffic volume is disclosed. The system includes a plurality of caches in a multi-level memory hierarchy and a plurality of agents. Each agent is associated with a cache. The system includes one or more monitoring engines. Each agent in the plurality of agents is associated with a monitoring engine. The agents can execute a processor level software instruction causing a memory region to be private to the agent. Each of the agents is configured to execute a memory access for data on an associated cache and to send a request for data up the hierarchy on a cache miss. The monitoring engine is configured to intercept request for data from an agent and to prevent snooping for the cache line in peer caches when the cache line associated with a memory region represented as private to the agent.

Abstract: A computer system may support one or more techniques to allow dynamic pinning of the memory pages accessed by a non-CPU device, such as a graphics processing unit (GPU). The non-CPU may support virtual to physical address mapping and may thus be aware of the memory pages, which may not be pinned but may be accessed by the non-CPU. The non-CPU may notify or send such information to a run-time component such as a device driver associated with the CPU. The device driver may, dynamically, perform pinning of such memory pages, which may be accessed by the non-CPU. The device driver may even unpin the memory pages, which may be no longer accessed by the non-CPU. Such an approach may allow the memory pages, which may be no longer accessed by the non-CPU to be available for allocation to the other CPUs and/or non-CPUs.

Abstract: Handling garbage collection and exceptions in hardware assisted transactions. Embodiments are practiced in a computing environment including a hardware assisted transaction system. A method includes beginning a hardware assisted transaction, raising an exception while in the hardware assisted transaction, including creating an exception object, determining that the transaction should be rolled back, and as a result of determining that the transaction should be rolled back, marshaling the exception object out of the hardware assisted transaction.

Abstract: In one embodiment, the present invention includes a method for directly communicating between an accelerator and an instruction sequencer coupled thereto, where the accelerator is a heterogeneous resource with respect to the instruction sequencer. An interface may be used to provide the communication between these resources. Via such a communication mechanism a user-level application may directly communicate with the accelerator without operating system support. Further, the instruction sequencer and the accelerator may perform operations in parallel. Other embodiments are described and claimed.

Abstract: Technologies are generally described herein for writing data to either volatile or nonvolatile memory. An estimated time for a last write operation to occur and an estimated time to a power down event are determined. A threshold time is generated from the estimated time for a last write operation to occur and the estimated time to a power down event. The threshold time represents time at which a cost to write to volatile memory may become greater than a cost to write to nonvolatile memory. The cost may be based at least in part on the need to copy data stored in a volatile memory to be persisted after a power down event from the volatile to the nonvolatile memory.

Abstract: Methods, apparatus and systems for memory access obscuration are provided. A first embodiment provides memory access obscuration in conjunction with deadlock avoidance. Such embodiment utilizes processor features including an instruction to enable monitoring of specified cache lines and an instruction that sets a status bit responsive to any foreign access (e.g., write or eviction due to a read) to the specified lines. A second embodiment provides memory access obscuration in conjunction with deadlock detection. Such embodiment utilizes the monitoring feature, as well as handler registration. A user-level handler may be asynchronously invoked responsive to a foreign write to any of the specified lines. Invocation of the handler more frequently than expected indicates that a deadlock may have been encountered. In such case, a deadlock policy may be enforced. Other embodiments are also described and claimed.

Abstract: In one embodiment, the present invention includes a method for receiving control in a kernel mode via a ring transition from a user thread during execution of an unbounded transactional memory (UTM) transaction, updating a state of a transaction status register (TSR) associated with the user thread and storing the TSR with a context of the user thread, and later restoring the context during a transition from the kernel mode to the user thread. In this way, the UTM transaction may continue on resumption of the user thread. Other embodiments are described and claimed.

Abstract: Technologies are generally described herein for writing data to either volatile or nonvolatile memory. An estimated time for a last write operation to occur and an estimated time to a power down event are determined. A threshold time is generated from the estimated time for a last write operation to occur and the estimated time to a power down event. The threshold time represents time at which a cost to write to volatile memory may become greater than a cost to write to nonvolatile memory. The cost may be based at least in part on the need to copy data stored in a volatile memory to be persisted after a power down event from the volatile to the nonvolatile memory.

Abstract: In one embodiment, the present invention includes a method for directly communicating between an accelerator and an instruction sequencer coupled thereto, where the accelerator is a heterogeneous resource with respect to the instruction sequencer. An interface may be used to provide the communication between these resources. Via such a communication mechanism a user-level application may directly communicate with the accelerator without operating system support. Further, the instruction sequencer and the accelerator may perform operations in parallel. Other embodiments are described and claimed.

Abstract: In one embodiment, the present invention includes a method for receiving control in a kernel mode via a ring transition from a user thread during execution of an unbounded transactional memory (UTM) transaction, updating a state of a transaction status register (TSR) associated with the user thread and storing the TSR with a context of the user thread, and later restoring the context during a transition from the kernel mode to the user thread. In this way, the UTM transaction may continue on resumption of the user thread. Other embodiments are described and claimed.

Abstract: A computer system may support one or more techniques to allow dynamic pinning of the memory pages accessed by a non-CPU device, such as a graphics processing unit (GPU). The non-CPU may support virtual to physical address mapping and may thus be aware of the memory pages, which may not be pinned but may be accessed by the non-CPU. The non-CPU may notify or send such information to a run-time component such as a device driver associated with the CPU. The device driver may, dynamically, perform pinning of such memory pages, which may be accessed by the non-CPU. The device driver may even unpin the memory pages, which may be no longer accessed by the non-CPU. Such an approach may allow the memory pages, which may be no longer accessed by the non-CPU to be available for allocation to the other CPUs and/or non-CPUs.

Abstract: A method and apparatus for accelerating a Software Transactional Memory (STM) system is herein described. A data object and metadata for the data object may each be associated with a filter, such as a hardware monitor or ephemerally held filter information. The filter is in a first, default state when no access, such as a read, from the data object has occurred during a pendancy of a transaction. Upon encountering a first access to the metadata, such as a first read, access barrier operations, such as logging of the metadata; setting a read monitor; or updating ephemeral filter information with an ephemeral/buffered store operation, are performed. Upon a subsequent/redundant access to the metadata, such as a second read, access barrier operations are elided to accelerate the subsequent access based on the filter being set to the second state to indicate a previous access occurred.

Abstract: Technologies are generally described herein for writing data to either volatile or nonvolatile memory. An estimated time for a last write operation to occur and an estimated time to a power down event are determined. A threshold time is generated from the estimated time for a last write operation to occur and the estimated time to a power down event. The threshold time represents time at which a cost to write to volatile memory may become greater than a cost to write to nonvolatile memory. The cost may be based at least in part on the need to copy data stored in a volatile memory to be persisted after a power down event from the volatile to the nonvolatile memory.

Abstract: In one embodiment, the present invention includes a method for executing a transactional memory (TM) transaction in a first thread, buffering a block of data in a first buffer of a cache memory of a processor, and acquiring a write monitor on the block to obtain ownership of the block at an encounter time in which data at a location of the block in the first buffer is updated. Other embodiments are described and claimed.

Abstract: A computer system may support one or more techniques to allow dynamic pinning of the memory pages accessed by a non-CPU device (e.g., a graphics processing unit, GPU). The non-CPU may support virtual to physical address mapping and may thus be aware of the memory pages, which may not be pinned but may be accessed by the non-CPU. The non-CPU may notify or send such information to a run-time component such as a device driver associated with the CPU. In one embodiment, the device driver may, dynamically, perform pinning of such memory pages, which may be accessed by the non-CPU. The device driver may even unpin the memory pages, which may be no longer accessed by the non-CPU. Such an approach may allow the memory pages, which may be no longer accessed by the non-CPU to be available for allocation to the other CPUs and/or non-CPUs.

Abstract: In one example of digital relay for out of network devices, a communications device includes a detecting component configured to detect that one or more communication endpoints is within communication range, a message receiving component configured to receive a digital message from a first communication endpoint, and a message relaying component configured to transmit the digital message to the second communication endpoint, wherein the first communication endpoint and the second communication endpoint are not communicatively coupled to each other.

Abstract: Monitoring performance of one or more architecturally significant processor caches coupled to a processor. The methods include executing an application on one or more processors coupled to one or more architecturally significant processor caches, where the application utilizes the architecturally significant portions of the architecturally significant processor caches. The methods further include at least one of generating metrics related to performance of the architecturally significant processor caches; implementing one or more debug exceptions related to performance of the architecturally significant processor caches; or implementing one or more transactional breakpoints related to performance of the architecturally significant processor caches as a result of utilizing the architecturally significant portions of the architecturally significant processor caches.

Abstract: In one embodiment, the present invention includes a method for executing a transactional memory (TM) transaction in a first thread, buffering a block of data in a first buffer of a cache memory of a processor, and acquiring a write monitor on the block to obtain ownership of the block at an encounter time in which data at a location of the block in the first buffer is updated. Other embodiments are described and claimed.