An apparatus is described. The apparatus can include non-volatile memory, an embedded processor, and a memory controller. The memory controller can access data from the byte addressable non-volatile memory using at least one of: a first addressing scheme or a second addressing scheme. The memory controller can provide the data to a host system over a first interface when the data is accessed using the first addressing scheme. The memory controller can provide the data to the embedded processor over a second interface when the data is accessed using the second addressing scheme.

In one embodiment, a method for tensor data distribution using a direct-memory access agent includes generating, by a first controller, source addresses indicating locations in a source memory where portions of a source tensor are stored. A second controller may generate destination addresses indicating locations in a destination memory where portions of a destination tensor are to be stored. The direct-memory access agent receives a source address generated by the first controller and a destination address generated by the second controller and determines a burst size. The direct-memory access agent may issue a read request comprising the source address and the burst size to read tensor data from the source memory and may store the tensor data into an alignment buffer. The direct-memory access agent then issues a write request comprising the destination address and the burst size to write data from the alignment buffer into the destination memory.

A method for writing to a non-volatile electronic memory with data words and assigned pieces of index information. The non-volatile electronic memory is initially filled exclusively with empty data frames. The empty data frames are overwritable with multi-data frames and/or individual data frames. A multi-data frame includes a selectable number of sequentially stored data words, and a multi-data frame header. A frame-type marker, the number of data words, and a selectable start index are stored in the multi-data frame header so that each data word is assignable a unique index value from an index interval by incrementing or decrementing. An individual data frame includes one data word and an individual data frame header. A frame-type marker and a selectable index value for the one data word of the individual data frame are stored in the individual data frame header.

Example implementations disclosed herein can be used to build, maintain, and use a hash table distributed across the plurality multiple nodes in a multi-node computing system. The hash table can include data pages associated by corresponding pointers according to a tree data structure. The data pages include leaf data pages. Each leaf data page can be associated with a corresponding hash value and include a tag bitmap. When a transaction associated with a key is executed, a hash value and a tag value are generated based on the key. The leaf data pages can be searched using the hash value. A probability that a leaf data page includes the key can be determined based on a comparison tag value with the tag bitmap.

N-way associative cache pools can be implemented in an N-way associative cache. Different cache pools can be indicated by pool values. Different processes running on a computer can use different cache pools. An N-way associative cache circuit can be configured to have one or more stripe mode cache pools that are N-way associative. A cache control circuit can receive a physical address for a memory location and can interpret the physical address as fields including a tag field that contains a tag value and a set field that contains a set value. The physical address can also be used to determine a pool value that identifies one of the stripe mode cache pools. A set of N cache entries in the one of the stripe mode cache pools can be concurrently searched for the tag value. The set of N cache entries is determined using the set value.

A memory system includes a nonvolatile memory and a controller. The nonvolatile memory has first regions in which data writes and data reads can be executed in parallel. Each of the first regions has second regions which are each a data write/read unit. The controller acquires first values indicating a data write load for each of the first regions, detects a first region having a first value greater than or equal to a first threshold, acquires second values indicating a data write load for each of the plurality of second regions in the detected first region, detects a second region having a second value greater than or equal to a second threshold but less than or equal to a third threshold that is higher than the second threshold, and then move data from the detected second region to a second region in another first region.

Methods, systems, and devices for multi-stage memory device performance notification are described. A memory system may include a first set of memory cells of a first type associated with a first performance level and a second set of memory cells of a second type associated with a second performance level. The memory system may have an interface and a control circuit coupled with the first and second sets of memory cells. The control circuit may be configured to determine a first parameter associated with a transition between the first performance level and the second performance level. The control circuit may also be configured to store the first parameter in a first register based at least in part on determining the first parameter.

Described herein is a graphics processing unit (GPU) comprising a first processing cluster to perform parallel processing operations, the parallel processing operations including a ray tracing operation and a matrix multiply operation; and a second processing cluster coupled to the first processing cluster, wherein the first processing cluster includes a floating-point unit to perform floating point operations, the floating-point unit is configured to process an instruction using a bfloat16 (BF16) format with a multiplier to multiply second and third source operands while an accumulator adds a first source operand with output from the multiplier.

Methods, apparatus, systems and articles of manufacture are disclosed facilitate read-modify-write support in a coherent victim cache with parallel data paths. An example apparatus includes a random-access memory configured to be coupled to a central processing unit via a first interface and a second interface, the random-access memory configured to obtain a read request indicating a first address to read via a snoop interface, an address encoder coupled to the random-access memory, the address encoder to, when the random-access memory indicates a hit of the read request, generate a second address corresponding to a victim cache based on the first address, and a multiplexer coupled to the victim cache to transmit a response including data obtained from the second address of the victim cache.

A caching system including a first sub-cache and a second sub-cache in parallel with the first sub-cache, wherein the second sub-cache includes: line type bits configured to store an indication that a corresponding cache line of the second sub-cache is configured to store write-miss data, and an eviction controller configured to evict a cache line of the second sub-cache storing write-miss data based on an indication that the cache line has been fully written.