High bandwidth memory system
    41.
    发明授权

    公开(公告)号:US11436165B2

    公开(公告)日:2022-09-06

    申请号:US16569657

    申请日:2019-09-12

    Abstract: A high-bandwidth memory (HBM) includes a memory and a controller. The controller receives a data write request from a processor external to the HBM and the controller stores an entry in the memory indicating at least one address of data of the data write request and generates an indication that a data bus is available for an operation during a cycle time of the data write request based on the data write request comprising sparse data or data-value similarity. Sparse data includes a predetermined percentage of data values equal to zero, and data-value similarity includes a predetermined amount of spatial value locality of the data values. The predetermined percentage of data values equal to zero of sparse data and the predetermined amount of spatial value locality of the special-value pattern are both based on a predetermined data granularity.

    Mitigating DRAM cache metadata access overhead with SRAM metadata cache and bloom filter

    公开(公告)号:US10282294B2

    公开(公告)日:2019-05-07

    申请号:US15587286

    申请日:2017-05-04

    Abstract: A system and method for mitigating overhead for accessing metadata for a cache in a hybrid memory module are disclosed. The method includes: providing a hybrid memory module including a DRAM cache, a flash memory, and an SRAM for storing a metadata cache; obtaining a host address including a DRAM cache tag and a DRAM cache index; and obtaining a metadata address from the DRAM cache index, wherein the metadata address includes a metadata cache tag and a metadata cache index. The method further includes determining a metadata cache hit based on a presence of a matching metadata cache entry in the metadata cache stored in the SRAM; in a case of a metadata cache hit, obtaining a cached copy of data included in the DRAM cache and skipping access to metadata included in the DRAM cache; and returning the data obtained from the DRAM cache to a host computer. The SRAM may further store a Bloom filter, and a potential DRAM cache hit may be determined based on a result of a Bloom filter test. A cache controller of the hybrid memory module may disable the Bloom filter when a metadata cache hit ratio is higher than a threshold.

    Software stack and programming for DPU operations

    公开(公告)号:US10180808B2

    公开(公告)日:2019-01-15

    申请号:US15426015

    申请日:2017-02-06

    Abstract: A system includes a library, a compiler, a driver and at least one dynamic random access memory (DRAM) processing unit (DPU). The library may determine at least one DPU operation corresponding to a received command. The compiler may form at least one DPU instruction for the DPU operation. The driver may send the at least one DPU instruction to at least one DPU. The DPU may include at least one computing cell array that includes a plurality of DRAM-based computing cells arranged in an array having at least one column in which the at least one column may include at least three rows of DRAM-based computing cells configured to provide a logic function that operates on a first row and a second row of the at least three rows and configured to store a result of the logic function in a third row of the at least three rows.

Patent Agency Ranking