US 9,792,044 B2Granted

Decompression history buffer read/write pipelines

AssigneeOracle International Corporation

InventorMilton Shih

Life of this patent

Estimating the term…

Loading patent…

Abstract

A method for a hardware decompression read pipeline, the method including determining a length and a distance of a first entity from a buffer; launching a first read request for reading a first data from the buffer; obtaining a second entity from the buffer; determining a distance of the second entity; calculating a new distance for the second entity; and decreasing a first counter by one for each cycle that data is read and returned from the buffer, wherein, when a first number of pending read data clock cycles in the first counter is less than a predetermined number of clock cycles and there is no read-after-write conflict: launching a second read request prior to completion of the first read request. In other aspects, a method for a hardware decompression write pipeline and systems for a hardware decompression read pipeline and a hardware decompression write pipeline are provided.

Claims20

Dependent claims

A method for a hardware decompression read pipeline, the method comprising:

obtaining a first entity from a buffer;
determining a length and a distance of the first entity;
converting the length and distance for the first entity into a first number of pending read data clock cycles required to read a first data from the buffer corresponding to the first entity;
converting the length of the first entity into a second number of pending read data bytes that will be returned by the buffer;
storing the first number of pending read data clock cycles in a first counter;
storing the second number of pending read data bytes in a second counter;
launching a first read request for reading the first data from the buffer corresponding to the first entity;
obtaining a second entity from the buffer;
determining a distance of the second entity;
calculating a new distance for the second entity using the second number of pending read data bytes that need to be read from the buffer for the first entity;
attempting to launch a second read request, prior to completion of the first read request, to read a second data from the buffer corresponding to the second entity;
and
decreasing the first counter by one for each cycle that data is read and returned from the buffer;
and
decreasing the second counter by a number of valid data bytes returned by the buffer;
wherein, when the first number of pending read data clock cycles in the first counter is less than a predetermined number of clock cycles and there is no read-after-write conflict:
launching the second read request prior to completion of the first read request.

14 more elements

A system for a hardware decompression read pipeline, the system comprising:

a first buffer configured to store a plurality of data corresponding to a plurality of entities;
an entity processing circuit configured to:
obtain a first entity from the compressed file data stream,
determine a length and a distance of the first entity,
launch a first read request and to start to read a first data from the first buffer corresponding to the first entity,
obtain a second entity from the compressed file data stream, to determine a length of the second entity, and
attempt to launch a second read request, prior to completion of the first read request, to read a second data from the first buffer corresponding to the second entity;
a new distance calculator circuit configured to calculate a new distance for the second entity;
a data cycle calculator circuit configured to convert the first entity distance into a first number of pending read data clock cycles;
a first counter configured to store the first number of pending read data clock cycles and to decrease the first counter by one for each cycle that read data is returned from the first buffer;
a second counter configured to store a second number of pending read data bytes that need to be read from the first buffer and to decrease the second counter by a number of valid bytes returned from the first buffer;
and
a read-after-write conflict detection circuit that stalls all reads to the first buffer while data from a second buffer is being transferred to the first buffer,
wherein, when the first number of pending read data clock cycles in the first counter is less than a predetermined number of clock cycles, and there is no read-after-write conflict, the second read request is launched prior to completion of the first read request.

11 more elements

Dependencies are inferred from the claim text.

Facts

Application: 15/042,793Filed 12 Feb 2016
Published: US 2017/0235497 A117 Aug 2017
Granted: 17 Oct 2017
Term ends: Not estimatedThe source records don’t support an estimate.

Classification12 codes

G06F 3/0608
Specifically adapted to achieve a particular effect: saving storage space on storage systems
G06F 3/0656
Vertical data movement, i.e. input-output transfer; data movement between one or more hosts and one or more storage devices: data buffering…
G06F 3/0661
Vertical data movement, i.e. input-output transfer; data movement between one or more hosts and one or more storage devices: format or protocol…
G06F 3/0683
In-line storage system: plurality of storage devices