Abstract: Techniques and circuits are provided for stochastic rounding. In an embodiment, a circuit includes carry-save adder (CSA) logic having three or more CSA inputs, a CSA sum output, and a CSA carry output. One of the three or more CSA inputs is presented with a random number value, while other CSA inputs are presented with input values to be summed. The circuit further includes adder logic having adder inputs and a sum output. The CSA carry output of the CSA logic is coupled with one of the adder inputs of the adder logic, and the CSA sum output of the CSA logic is coupled with another input of the adder inputs of the adder logic. A particular number of most significant bits of the sum output of the adder logic represent a stochastically rounded sum of the input values.

Abstract: A technique for use in a memory system includes swapping a first plurality of pages of a first memory of the memory system with a second plurality of pages of a second memory of the memory system. The first memory has a first latency and the second memory has a second latency. The first latency is less than the second latency. The technique includes updating a page table and triggering a translation lookaside buffer shootdown to associate a virtual address of each of the first plurality of pages with a corresponding physical address in the second memory and to associate a virtual address for each of the second plurality of pages with a corresponding physical address in the first memory.

Abstract: An electronic device has an activation function functional block that implements an activation function. During operation, the activation function functional block receives an input including a plurality of bits representing a numerical value. The activation function functional block then determines a range from among a plurality of ranges into which the input falls, each range including a separate portion of possible numerical values of the input. The activation function functional block next generates a result of a linear function associated with the range. Generating the result includes using a separate linear function that is associated with each range in the plurality of ranges to approximate results of the activation function within that range.

Abstract: In one form, a computer system includes a central processing unit, a memory controller coupled to the central processing unit and capable of accessing non-volatile random access memory (NVRAM), and an NVRAM-aware operating system. The NVRAM-aware operating system causes the central processing unit to selectively execute selected ones of a plurality of application programs, and is responsive to a predetermined operation to cause the central processing unit to execute a memory persistence procedure using the memory controller to access the NVRAM.

Abstract: A processor modifies memory management mode for a range of memory locations of a multilevel memory hierarchy based on changes in an application phase of an application executing at a processor. The processor monitors the application phase (e.g., computation-bound phase, input/output phase, or memory access phase) of the executing application and in response to a change in phase consults a management policy to identify a memory management mode. The processor automatically reconfigures a memory controller and other modules so that a range of memory locations of the multilevel memory hierarchy are managed according to the identified memory management mode. By changing the memory management mode for the range of memory locations according to the application phase, the processor improves processing efficiency and flexibility.

Abstract: A method includes monitoring, at a cache coherence directory, states of cachelines stored in a cache hierarchy of a data processing system using a plurality of entries of the cache coherence directory. Each entry of the cache coherence directory is associated with a corresponding cache page of a plurality of cache pages, and each cache page representing a corresponding set of contiguous cachelines. The method further includes selectively evicting cachelines from a first cache of the cache hierarchy based on cacheline utilization densities of cache pages represented by the corresponding entries of the plurality of entries of the cache coherence directory.

Abstract: A processing system employs an expandable memory buffer that supports enlarging the memory buffer when the processing system generates a large number of long latency memory transactions. The hybrid structure of the memory buffer allows a memory controller of the processing system to store a larger number of memory transactions while still maintaining adequate transaction throughput and also ensuring a relatively small buffer footprint and power consumption. Further, the hybrid structure allows different portions of the buffer to be placed on separate integrated circuit dies, which in turn allows the memory controller to be used in a wide variety of integrated circuit configurations, including configurations that use only one portion of the memory buffer.

Abstract: A processing system includes at least one central processing unit (CPU) core, at least one graphics processing unit (GPU) core, a main memory, and a coherence directory for maintaining cache coherence. The at least one CPU core receives a CPU cache flush command to flush cache lines stored in cache memory of the at least one CPU core prior to launching a GPU kernel. The coherence directory transfers data associated with a memory access request by the at least one GPU core from the main memory without issuing coherence probes to caches of the at least one CPU core.

Abstract: Cluster manager functional blocks perform operations for migrating pages in portions in corresponding migration clusters. During operation, each cluster manager keeps an access record that includes information indicating accesses of pages in the portions in the corresponding migration cluster. Based on the access record and one or more migration policies, each cluster manager migrates pages between the portions in the corresponding migration cluster.

Abstract: A method includes monitoring, at a cache coherence directory, states of cachelines stored in a cache hierarchy of a data processing system using a plurality of entries of the cache coherence directory. Each entry of the cache coherence directory is associated with a corresponding cache page of a plurality of cache pages, and each cache page representing a corresponding set of contiguous cachelines. The method further includes selectively evicting cachelines from a first cache of the cache hierarchy based on cacheline utilization densities of cache pages represented by the corresponding entries of the plurality of entries of the cache coherence directory.

Abstract: The described embodiments include a computing device with two or more translation lookaside buffers (TLB). During operation, the computing device updates an entry in the TLB based on a virtual address to physical address translation and metadata from a page table entry that were acquired during a page table walk. The computing device then computes, based on a lease length expression, a lease length for the entry in the TLB. Next, the computing device sets, for the entry in the TLB, a lease value to the lease length, wherein the lease value represents a time until a lease for the entry in the TLB expires, wherein the entry in the TLB is invalid when the associated lease has expired. The computing device then uses the lease value to control operations that are allowed to be performed using information from the entry in the TLB.

Abstract: Systems, apparatuses, and methods for implementing a hybrid cache. A processor may include a hybrid L2/L3 cache which allows the processor to dynamically adjust a size of the L2 cache and a size of the L3 cache. In some embodiments, the processor may be a multi-core processor and there may be a single cache partitioned into a logical L2 cache and a logical L3 cache for use by the cores. In one embodiment, the processor may track the cache hit rates of the logical L2 and L3 caches and adjust the sizes of the logical L2 and L3 cache based on the cache hit rates. In another embodiment, the processor may adjust the sizes of the logical L2 and L3 caches based on which application is currently being executed by the processor.

Abstract: Systems, apparatuses, and methods for sorting memory pages in a multi-level heterogeneous memory architecture. The system may classify pages into a first “hot” category or a second “cold” category. The system may attempt to place the “hot” pages into the memory level(s) closest to the systems' processor cores. The system may track parameters associated with each page, with the parameters including number of accesses, types of accesses, power consumed per access, temperature, wearability, and/or other parameters. Based on these parameters, the system may generate a score for each page. Then, the system may compare the score of each page to a threshold. If the score of a given page is greater than the threshold, the given page may be designated as “hot”. If the score of the given page is less than the threshold, the given page may be designated as “cold”.

Abstract: A system, method, and computer program product are provided for a memory device system. One or more memory dies and at least one logic die are disposed in a package and communicatively coupled. The logic die comprises a processing device configurable to manage virtual memory and operate in an operating mode. The operating mode is selected from a set of operating modes comprising a slave operating mode and a host operating mode.

Abstract: A technique for use in a memory system includes swapping a first plurality of pages of a first memory of the memory system with a second plurality of pages of a second memory of the memory system. The first memory has a first latency and the second memory has a second latency. The first latency is less than the second latency. The technique includes updating a page table and triggering a translation lookaside buffer shootdown to associate a virtual address of each of the first plurality of pages with a corresponding physical address in the second memory and to associate a virtual address for each of the second plurality of pages with a corresponding physical address in the first memory.

Abstract: A technique for source-side memory request network admission control includes adjusting, by a first node, a rate of injection of memory requests by the first node into a network coupled to a memory system. The adjusting is based on an injection policy for the first node and memory request efficiency indicators. The method may include injecting memory requests by the first node into the network coupled to the memory system. The injecting has the rate of injection. The technique includes adjusting the rate of injection by the first node. The first node adjusts the rate of injection according to an injection policy for the first node and memory request efficiency indicators. The injection policy may be based on an injection rate limit for the first node. The injection policy for the first node may be based on an injection rate limit per memory channel for the first node.

Abstract: A processing system includes at least one central processing unit (CPU) core, at least one graphics processing unit (GPU) core, a main memory, and a coherence directory for maintaining cache coherence. The at least one CPU core receives a CPU cache flush command to flush cache lines stored in cache memory of the at least one CPU core prior to launching a GPU kernel. The coherence directory transfers data associated with a memory access request by the at least one GPU core from the main memory without issuing coherence probes to caches of the at least one CPU core.

Abstract: The described embodiments include a computing device with two or more translation lookaside buffers (TLB) that performs operations for handling entries in the TLBs. During operation, the computing device maintains lease values for entries in the TLBs, the lease values representing times until leases for the entries expire, wherein a given entry in the TLB is invalid when the associated lease has expired. The computing device uses the lease value to control operations that are allowed to be performed using information from the entries in the TLBs. In addition, the computing device maintains, in a page table, longest lease values for page table entries indicating when corresponding longest leases for entries in TLBs expire. The longest lease values are used to determine when and if a TLB shootdown is to be performed.

Abstract: A system, method, and computer program product are provided for a memory device system. One or more memory dies and at least one logic die are disposed in a package and communicatively coupled. The logic die comprises a processing device configurable to manage virtual memory and operate in an operating mode. The operating mode is selected from a set of operating modes comprising a slave operating mode and a host operating mode.

Abstract: The described embodiments include a computer system having a multi-level memory hierarchy with two or more levels of memory, each level being one of two or more types of memory. The computer system handles storing objects in the multi-level memory hierarchy. During operation, a system runtime in the computer system identifies an object to be stored in the multi-level memory hierarchy. The system runtime then determines, based on one or more attributes of the object, that the object is to be pinned in a level of the multi-level memory hierarchy. The system runtime then pins the object in the level of the multi-level memory hierarchy. In the described embodiments, the pinning includes hard pinning and soft pinning, which are each associated with corresponding retention policies for pinned objects.