Intel Corporation

Inte Corporation

Yao Zu Dong

Kun Tian

Zhiyuan Lv

Yunhong Jiang

Zhen Zhou

Bin Yang

Fengguang Wu

Jingqi Liu

Zhi Hu Wang

Mingqiu Sun

Ke Yu

Xiao Mei Zheng

Chao Peng

Yuankai Guo

Haozhong Zhang

Lei Li

Shoumeng Yan

Tao Xu

Mohammad R Haghighat

Xiao Dong Lin

Jia Bao

Di Zhang

Jinkui Ren

Shuo Liu

Yuyang Du

Chao Xie

Yong Yao

Yulei Zhang

Qiming Shi

Fengqian Gao

Xiantao Zhang

Tian Zhang

Systems, apparatuses and methods may provide for technology that detects a creation of a thread, dedicates a memory region to objects associated with the thread, and conducts a reclamation of the memory region in response to a termination of the thread. In one example, the memory region is a heap region and the reclamation bypasses at least a pause phase and a copy phase of a garbage collection process with respect to the heap region.

Avoidance of garbage collection in high performance memory management systems

Apparatuses, methods and storage medium associated with in-vehicle computing, are disclosed herein. In embodiments, an in-vehicle system computing platform having a hypervisor to host one or more virtual machines (VMs) includes a memory shrink manager, and a memory snapshot manager. The memory shrink manager is configured to orchestrate shrinking a memory footprint of one of the one or more VMs for a suspend process invoked in response to the computing platform being powered off. The memory snapshot manager is configured to save the shrunken memory footprint of the one VM to the persistent storage during the suspend process, and to reload a subset of the saved shrunken memory footprint during a resume process to resume the one VM from suspension to the persistent storage. The resume process is invoked in response to the computing platform being powered on, cold booted.

Power off and power on method and apparatus for an in-vehicle system

Various embodiments are generally directed to virtualized systems. A first guest memory page may be identified based at least in part on a number of accesses to a page table entry for the first guest memory page in a page table by an application executing in a virtual machine (VM) on the processor, the first guest memory page corresponding to a first byte-addressable memory. The execution of the VM and the application on the processor may be paused. The first guest memory page may be migrated to a target memory page in a second byte-addressable memory, the target memory page comprising one of a target host memory page and a target guest memory page, the second byte-addressable memory having an access speed faster than an access speed of the first byte-addressable memory.

Method and apparatus to use DRAM as a cache for slow byte-addressible memory for efficient cloud applications

Methods, apparatus, systems and articles of manufacture to deduplicate duplicate memory in a cloud-computing environment are disclosed herein. An example apparatus to deduplicate duplicate memory comprises a parser to parse process information corresponding to instances of an application, a group generator to group process information into application groups based on the process information indicating instances corresponding to the same directory paths and application names, a data structure generator to generate a pair of binary search trees for an application group, and a merge controller to deduplicate duplicate memory contents detected in the application group.

Methods and apparatus to deduplicate duplicate memory in a cloud computing environment

A disclosed example includes accessing, by a backend block service driver in an input/output virtual machine executing on one or more processors, a first command submitted to a buffer by a paravirtualized input/output frontend block driver executing in a guest virtual machine; generating, by the backend block service driver, a translated command based on the first command by translating a virtual parameter of the first command to a physical parameter associated with a physical resource; submitting, by the backend block service driver, the translated command to an input/output queue to be processed by the physical resource based on the physical parameter; and submitting, by the backend block service driver, a completion status entry to the buffer, the completion status entry indicative of completion of a direct memory access operation that copies data between the physical resource and a guest memory buffer corresponding to the guest virtual machine.

Methods and apparatus to process commands from virtual machines

Method and apparatus to use DRAM as a cache for slow byte-addressable memory for efficient cloud applications

A disclosed example includes accessing, by virtual nonvolatile memory in a virtual machine monitor executing on one or more processors, a first command submitted to a guest queue by a native nonvolatile memory driver executing in a guest virtual machine; generating, by the virtual nonvolatile memory, a translated command based on the first command by translating a virtual parameter of the first command to a physical parameter associated with physical nonvolatile memory; submitting, by the virtual nonvolatile memory, the translated command to a shadow queue to be processed by the physical nonvolatile memory based on the physical parameter; and submitting, by the virtual nonvolatile memory, a completion status entry to the guest queue, the completion status entry indicative of completion of a direct memory access operation that copies data between the physical nonvolatile memory and a guest memory buffer corresponding to the guest virtual machine.

Systems and methods for container access to graphics processing unit (GPU) resources are disclosed herein. In some embodiments, a computing system may include a physical GPU and kernel-mode driver circuitry, to communicatively couple with the physical GPU to create a plurality of emulated GPUs and a corresponding plurality of device nodes. Each device node may be associated with a single corresponding user-side container to enable communication between the user-side container and the corresponding emulated GPU. Other embodiments may be disclosed and/or claimed.

Container access to graphics processing unit resources

Method and apparatus to use dram as a cache for slow byte-addressible memory for efficient cloud applications

Method, systems and apparatuses may provide for technology that divides an application into a plurality of portions that are each associated with one or more functions of the application and determine a plurality of transition probabilities between the plurality of portions. Some technology may also receive at least a first portion of the plurality of portions, and receive a relation file indicating the plurality of transition probabilities between the plurality of portions.

Scenario profile based partitioning and management of application code

A host machine includes a guest machine, a device emulator, and a hypervisor communicably coupled to the guest machine and the device emulator. The guest machine executes a non-real time thread that causes a non-real time I/O emulation by the device emulator. Responsive to receipt of a real time thread by the guest machine, the hypervisor determines whether the non-real time I/O emulation is abortable or non-abortable. If abortable, the hypervisor aborts the non-real time thread and causes the guest machine to execute the real time thread. Upon completing the execution of the real time thread, the hypervisor causes the guest machine to revert to a non-real time context based on a previous system snapshot. Upon establishing the non-real time context, the hypervisor causes the guest machine to execute the previously aborted non-real time thread.

I/O emulation with abortion in virtualized environments including transfer of portions of non-real time I/O emulation to work threads prior to switching contexts

A virtual machine migration controller may perform the live migration of a plurality of virtual machines from a first physical host system to a second physical host system. The virtual machine migration controller may determine a memory page dirty rate for each of a plurality of virtual machines. The virtual machine migration controller may additionally identify virtual machines that share memory pages and/or map to different memory pages having, at least in part, identical data or information. The virtual machine migration controller may group virtual machines demonstrating commonality among mapped memory pages. The virtual machine migration controller may determine a projected migration time based on the dirtying rate, the commonality of memory pages, and the available bandwidth. The virtual machine migration controller orders and transfers virtual machine groups based on the projected migration time.

Virtual machine batch live migration

Examples may include intercepting packets outputted from a primary virtual machine (PVM) hosted by a first server and converting one or more fields of protocol headers for each intercepted packet such that output-packet-similarity may be increased between the PVM outputted packets and packets outputted by a secondary virtual machine (SVM) hosted by a second server.

Techniques for improving output-packet-similarity between primary and secondary virtual machines

An apparatus and method are described for implementing a hybrid layer of address mapping for an IOMMU implementation. For example, one embodiment of a graphics processing apparatus comprises: virtualization circuitry to implement a virtualized execution environment in which a plurality of guest virtual machines (VMs) are to execute and share execution resources of the graphics processing apparatus; an input/output (I/O) memory management unit (IOMMU) to couple the VMs to one or more I/O devices; a hybrid layer address mapping (HLAM) module to combine entries from a per-process graphics translation table (PPGTT) with entries from a global graphics translation table (GGTT) into a first integrated page table, the first integrated page table mapping PPGTT guest page numbers (GPNs) to host page numbers (HPNs) and mapping GGTT virtual GPNs to HPNs; the HLAM to transform a GGTT GPN into a virtual GPN usable to access a corresponding HPN within the first integrated page table in response to a GGTT read/write operation generated by a first guest virtual machine (VM).

Apparatus and method for a hybrid layer of address mapping for a virtualized input/output (I/O) implementation

Technologies for scheduling workload submissions for a graphics processing unit (GPU) in a virtualization environment include a GPU scheduler embodied in a computing device. The virtualization environment includes a number of different virtual machines that are configured with a native graphics driver. The GPU scheduler receives GPU commands from the different virtual machines, dynamically selects a scheduling policy, and schedules the GPU commands for processing by the GPU.

Growing community of inventors

Shanghai, China

Yao Zu Dong