US 9,465,632 B2Granted

Parallel hardware hypervisor for virtualizing application-specific supercomputers

AssigneeGlobal Supercomputing Corporation

InventorsKemal Ebcioglu, Atakan Dogan, Reha Oguz Altug, Mikko Herman Lipasti, Eray Özkural

Life of this patent

Estimating the term…

Loading patent…

Abstract

A parallel hypervisor system for virtualizing application-specific supercomputers is disclosed. The hypervisor system comprises (a) at least one software-virtual hardware pair consisting of a software application, and an application-specific virtual supercomputer for accelerating the said software application, wherein (i) The virtual supercomputer contains one or more virtual tiles; and (ii) The software application and the virtual tiles communicate among themselves with messages; (b) One or more reconfigurable physical tiles, wherein each virtual tile of each supercomputer can be implemented on at least one physical tile, by configuring the physical tile to perform the virtual tile's function; and (c) A scheduler implemented substantially in hardware, for parallel pre-emptive scheduling of the virtual tiles on the physical tiles.

Claims16

Dependent claims

A hypervisor system for virtualizing application-specific supercomputers, the system comprising:

(a) at least one software-virtual hardware pair consisting of a software application, and an application-specific virtual supercomputer for accelerating the software application, where:
i. the application-specific virtual supercomputer comprises a plurality of virtual tiles;
and
ii. the software application and the virtual tiles communicate among themselves with messages;
(b) a plurality of reconfigurable physical tiles, where each virtual tile of each application-specific virtual supercomputer can be implemented on at least one reconfigurable physical tile, by configuring the reconfigurable physical tile to perform the virtual tile's function;
and
(c) a scheduler consisting of a plurality of interconnected special-purpose hardware components, communicating and synchronizing to perform parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles;
where none among the plurality of interconnected special-purpose hardware components of the scheduler is a processor executing software instructions;
and
where the plurality of interconnected special-purpose hardware components of the scheduler includes two or more owner units for managing virtual tiles, where each virtual tile is assigned to exactly one owner unit and where at least one virtual tile is assigned to each owner unit;
and
where the parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles comprises:
simultaneously performing by each owner unit:
pre-empting a virtual tile v1 assigned to v1's respective owner unit and operating on a reconfigurable physical tile p1, where said pre-empting is done by v1's respective owner unit;
letting v1 remain pre-empted for a period of time, by v1's respective owner unit;
receiving at v1's respective owner unit, from a virtual tile v0 operating on a reconfigurable physical tile p0, an indication of an attempt to send a communication message from v0 to the pre-empted virtual tile v1;
and
in response to said indication, resuming operation of v1 on a reconfigurable physical tile p2, by v1's respective owner unit, where p2 is determined based on statistics collected from the hypervisor system and is not necessarily the same as p1, and enabling delivery of the communication message directly from p0 to p2;
where said simultaneous performance by owner units is subject to resource constraints and dependency constraints.

16 more elements

A method for automatically converting a single-threaded software application into a software-virtual hardware pair consisting of a software application and an application-specific virtual supercomputer, integrated into a hypervisor system for virtualizing application specific supercomputers, where the method comprises:

(a) automatically converting a code fragment from the single-threaded software application into customized non-virtual supercomputer hardware, whose hardware execution is functionally equivalent to software execution of the code fragment;
(b) generating interfaces on hardware and software parts of the single-threaded software application, where the interfaces, at run time:
i. perform a software-to-hardware program state transfer upon entry to the code fragment;
ii. perform a hardware-to-software program state transfer upon exit from the code fragment;
and
iii. maintain memory coherence between software and hardware memories of the single threaded software application;
(c) partitioning the customized non-virtual supercomputer hardware obtained in steps (a) and (b) into multiple modules;
(d) creating a union module which is capable of realizing any of the modules created by step (c) depending on configuration parameters provided to the union module;
(e) creating a software-virtual hardware pair consisting of a software application and an application-specific virtual supercomputer, where:
i. the software application part of the software-virtual hardware pair is the single-threaded software application as modified in step (b);
ii. the application-specific virtual supercomputer part of the software-virtual hardware pair consists of virtual tiles, each of which is a module obtained in step (c);
(f) adding functionality to stop, start, read internal state of, and write internal state of the union module of step (d);
to create a reconfigurable physical tile capable of realizing any among the virtual tiles of step (e);
and
(g) integrating at least one copy of the reconfigurable physical tile obtained in step (f) within the hypervisor system, to realize the software-virtual hardware pair consisting of the software application and the application-specific virtual supercomputer, within the hypervisor system,
where the hypervisor system comprises:
(h) at least one software-virtual hardware pair consisting of a software application, and an application-specific virtual supercomputer for accelerating the software application, where:
i. the application-specific virtual supercomputer comprises one or more virtual tiles;
and
ii. the software application and the virtual tiles communicate among themselves with messages;
(i) one or more reconfigurable physical tiles, where each virtual tile of each application-specific virtual supercomputer can be implemented on at least one reconfigurable physical tile, by configuring the reconfigurable physical tile to perform the virtual tile's function;
and
(j) a scheduler implemented substantially in hardware, for parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles.

20 more elements

A hypervisor system for virtualizing application-specific supercomputers, the system comprising:

(a) at least one software-virtual hardware pair consisting of a software application, and an application-specific virtual supercomputer for accelerating the software application, where:
i. the application-specific virtual supercomputer comprises one or more virtual tiles;
and
ii. the software application and the virtual tiles communicate among themselves with messages;
(b) one or more reconfigurable physical tiles, where each virtual tile of each application-specific virtual supercomputer can be implemented on at least one reconfigurable physical tile, by configuring the reconfigurable physical tile to perform the virtual tile's function;
and
(c) a scheduler implemented substantially in hardware, for parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles;
(d) a table mapping each virtual tile
i. to NULL, if the virtual tile is not currently contained in any reconfigurable physical tile;
ii. to reconfigurable physical tile the virtual tile is currently contained in, otherwise;
(e) a first level cache for the table near each reconfigurable physical tile, mapping message destination virtual tiles to message destination reconfigurable physical tiles, such that when sending a message from a message source virtual tile v0 contained in a message source reconfigurable physical tile p0, to a message destination virtual tile v contained in a message destination reconfigurable physical tile p:
the message source reconfigurable physical tile p0 containing the message source virtual tile v0
can send a message directly and efficiently to
the message destination reconfigurable physical tile p containing the message destination virtual tile v
by
adding a field containing p to the message as it leaves the source virtual tile v0 contained in p0,
using the p field for routing the message toward the destination virtual tile v contained in p, through a scalable network, and
reverting the message back to its original form when the destination virtual tile is reached,
whenever v is found to be mapped to p in the first level cache near the message source reconfigurable physical tile p0.

16 more elements

An adaptive method for allocating resources to ASIC reconfigurable physical tiles in a hypervisor system for virtualizing application specific supercomputers, the method comprising:

(a) determining frequency of use of each software-virtual hardware pair consisting of a software application and an application-specific virtual supercomputer in the hypervisor system, over a recent time interval;
(b) searching for a software-virtual hardware pair consisting of a software application and an application-specific virtual supercomputer in the hypervisor system, where the software-virtual hardware pair has greatest frequency of use as determined in step (a), such that:
i. the software-virtual hardware pair is not already mapped to an ASIC reconfigurable physical tile;
and
ii. frequency of use of the software-virtual hardware pair exceeds a threshold;
(c) if such a software-virtual hardware pair could be found in the hypervisor system in step (b), creating a new ASIC reconfigurable physical tile for this software-virtual hardware pair;
(d) for each ASIC reconfigurable physical tile in the hypervisor system:
i. increasing or decreasing resources allocated to this ASIC reconfigurable physical tile, in order to make allocated resources proportional to average frequency of use of software-virtual hardware pair consisting of a software application and an application-specific virtual supercomputer, implemented by this ASIC reconfigurable physical tile;
(e) repeating all of the steps (a)-(d) above, periodically;
where the hypervisor system comprises:
(f) at least one software-virtual hardware pair consisting of a software application, and an application-specific virtual supercomputer for accelerating the software application, where:
i. the application-specific virtual supercomputer comprises one or more virtual tiles;
and
ii. the software application and the virtual tiles communicate among themselves with messages;
(g) one or more reconfigurable physical tiles, where each virtual tile of each application-specific virtual supercomputer can be implemented on at least one reconfigurable physical tile, by configuring the reconfigurable physical tile to perform the virtual tile's function;
and
(h) a scheduler implemented substantially in hardware, for parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles;
where at least one reconfigurable physical tile of the hypervisor system is an ASIC reconfigurable physical tile, which is defined as a reconfigurable physical tile implemented in ASIC technology, capable of realizing any one of the virtual tiles of an application-specific virtual supercomputer, based on configuration parameters provided to the reconfigurable physical tile.

15 more elements

A hypervisor system for virtualizing application-specific supercomputers, the system comprising:

(a) at least one software-virtual hardware pair consisting of a software application, and an application-specific virtual supercomputer for accelerating the software application, where:
i. the application-specific virtual supercomputer comprises a plurality of virtual tiles;
and
ii. the software application and the virtual tiles communicate among themselves with messages;
(b) a plurality of reconfigurable physical tiles, where each virtual tile of each application-specific virtual supercomputer can be implemented on at least one reconfigurable physical tile, by configuring the reconfigurable physical tile to perform the virtual tile's function;
and
(c) means for parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles,
where a scheduler consists of a plurality of means for communicating and synchronizing to perform parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles;
and
where the plurality of means for communicating and synchronizing to perform parallel pre-emptive scheduling of the virtual tiles on physical tiles includes two or more owner units for managing virtual tiles, where each virtual tile is assigned to exactly one owner unit and where at least one virtual tile is assigned to each owner unit;
and
where the parallel pre-emptive scheduling of the virtual tiles on the reconfigurable physical tiles comprises:
simultaneously performing by each owner unit:
pre-empting a virtual tile v1 assigned to v1's respective owner unit and operating on a reconfigurable physical tile p1, where said pre-empting is done by v1's respective owner unit;
letting v1 remain pre-empted for a period of time, by v1's respective owner unit;
receiving at v1's respective owner unit, from a virtual tile v0 operating on a reconfigurable physical tile p0, an indication of an attempt to send a communication message from v0 to the pre-empted virtual tile v1;
and
in response to said indication, resuming operation of v1 on a reconfigurable physical tile p2, by v1's respective owner unit, where p2 is determined based on statistics collected from the hypervisor system and is not necessarily the same as p1, and enabling delivery of the communication message directly from p0 to p2;
where said simultaneous performance by owner units is subject to resource constraints and dependency constraints.

16 more elements

Dependencies are inferred from the claim text.

Facts

Application: 13/366,318Filed 4 Feb 2012
Published: US 2013/0205295 A18 Aug 2013
Granted: 11 Oct 2016
Term ends: Not estimatedThe source records don’t support an estimate.

Classification8 codes

G06F 9/45533
Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines:…
G06F 9/45558
Hypervisors; Virtual machine monitors: hypervisor-specific management and integration aspects
G06F 9/5077
Partitioning or combining of resources: logical partitioning of resources; Management or configuration of virtualized resources
G06F 15/7871
With reconfigurable architecture: reconfiguration support, e.g. configuration loading, configuration switching, or hardware OS