WO2007133559A9 - Traitement de différences fonctionnelles compactes - Google Patents

Traitement de différences fonctionnelles compactes

Info

Publication number
WO2007133559A9
WO2007133559A9 PCT/US2007/011132 US2007011132W WO2007133559A9 WO 2007133559 A9 WO2007133559 A9 WO 2007133559A9 US 2007011132 W US2007011132 W US 2007011132W WO 2007133559 A9 WO2007133559 A9 WO 2007133559A9
Authority
WO
WIPO (PCT)
Prior art keywords
new
file
common
data
original
Prior art date
Application number
PCT/US2007/011132
Other languages
English (en)
Other versions
WO2007133559A3 (fr
WO2007133559A2 (fr
Inventor
Matthias Reinsch
Original Assignee
Innopath Software
Matthias Reinsch
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US11/432,984 external-priority patent/US7890464B2/en
Application filed by Innopath Software, Matthias Reinsch filed Critical Innopath Software
Publication of WO2007133559A2 publication Critical patent/WO2007133559A2/fr
Publication of WO2007133559A9 publication Critical patent/WO2007133559A9/fr
Publication of WO2007133559A3 publication Critical patent/WO2007133559A3/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F8/00—Arrangements for software engineering
    • G06F8/70—Software maintenance or management
    • G06F8/71—Version control; Configuration management

Definitions

  • the disclosed embodiments relate to updating of electronic files using difference files.
  • embedded software components hosted on mobile processing devices for example mobile wireless devices, often include numerous software bugs that require correction.
  • Software includes one or more files in the form of human-readable American Standard Code for Information Interchange (ASCII) plain text files or binary code.
  • ASCII American Standard Code for Information Interchange
  • Software files can be divided into smaller units that are often referred to as modules or components.
  • Portable processor-based devices like mobile processing devices typically include a real-time operating system (RTOS) in which all software components of the device are linked as a single large file.
  • RTOS real-time operating system
  • Another typical solution for updating files uses difference programs to generate a description of how a revised file differs from an original file.
  • difference programs that produce such difference data.
  • the difference files produced using these difference programs can sometimes be too large for transfer via the associated communication protocols.
  • Figure 1 is a block diagram showing file differencing and updating, under an embodiment.
  • Figure 2 is a flow diagram for generation of a delta file, under the embodiment of Figure 1.
  • Figure 3 is a block diagram showing an original and a new version of an executable file.
  • Figure 4 is a flow diagram for pre-processing different versions of an electronic file, under the embodiment of Figure 1 and Figure 2.
  • Figure 5 is a flow diagram for pre-processing code (text) sections of different versions of an electronic file, under the embodiment of Figure 4.
  • Figure 6 is a flow diagram for generating new target address values for the code (text) sections of an original version of an electronic file, under the embodiment of Figure 5.
  • Figure 7 is a block diagram of an original version and a new version of a file, under an embodiment, showing the different addresses (and corresponding notations) described with reference to generation of new target address values for the code (text) sections of an original version of an electronic file.
  • Figure 8 is a flow diagram for pre-processing data sections of original and new versions of an electronic file, under the embodiment of Figure 4.
  • Figure 9 is a flow diagram for generating new data pointer values for the data sections of an original version of an electronic file, under the embodiment of Figure 8.
  • Figure 10 is a block diagram of an original version and a new version of a file, under an embodiment, showing the different addresses (and corresponding notations) used in generating new data pointer values for an original version of an electronic file.
  • Figure 11 is a flow diagram of hint merging of common function units, under the embodiment of Figure 4.
  • Figure 12 is a flow diagram of hint merging of common data units, under the embodiment of Figure 4.
  • Figure 13 shows post-processing of preprocessing data that generates one or more new code blocks, under an embodiment.
  • Figure 14 shows post-processing of preprocessing data that increases the size of one or more previously identified common code blocks, while also generating at least one new code block, under an embodiment.
  • Figures 15A-15F show graphical representations which provide example results of the preprocessing and post-processing methodologies.
  • Figure 16 is a sample plot of a piece-wise constant function, under an embodiment.
  • Figure 17 is a diagram illustrating a typical table of integers, as operated on by an embodiment.
  • Figure 18 is a diagram illustrating code, such as ARM code, as operated on by an embodiment.
  • Figure 19 is a diagram illustrating a more complicated example of a table of values, as operated on by an embodiment.
  • Figure 20 is a diagram illustrating an example that shows the importance of the endian being correct in values, as operated on by an embodiment.
  • Figure 21 is a diagram illustrating another ARM code example, as operatedn embodiment.
  • Systems and methods are provided for pre-processing original and new versions of files as part of difference file generation between the original and new file versions. This pre-processing supports a further reduction in the size of the difference file.
  • Software/executable changes between file versions include primary changes/logical changes, which are defined to be the source code changes, and secondary changes.
  • the secondary changes generally result from the primary changes and are generated by the software compiler/linker utilities.
  • the secondary changes include address changes, pointer target address changes, and changes in address offsets resulting from the primary changes and generated by the software compiler/linker utilities.
  • the pre-processing systems and methods provided use approximation rules between file versions to remove/reduce the secondary changes and encode information relating to the removal of these changes in information of the corresponding difference file.
  • the systems and methods of an embodiment include a set of algorithms that reduce the difference package size by preprocessing a variety of regions in software images for embedded computing devices, an example of which is the compressed read-only memory (ROM) file system (CRAMFS) image. These regions are collectively referred to herein as “Compact Functional Differences (CFD) preprocessing regions.”
  • CFD preprocessing regions One of the main types of CFD preprocessing regions includes the various large tables of pointers that are common in execute in place (XIP) files (pre-linked "execute in place” files).
  • XIP execute in place
  • the algorithms are not limited to these tables, and they also treat a variety of other types of regions that are created by the compiler. Preprocessing systems and methods described below, and referred to herein as
  • XIP preprocessing perform tasks on both the client side as well as the server side.
  • the XIP preprocessing generates CFD hint data directly from a pair of CRAMFS images, without the use of symbol files or log files generated by compiler/linker utilities.
  • One example of how to apply the XIP preprocessing is to look at the XIP files in the old and new CRAMFS file systems, and look for files with matching paths. Approximate matching techniques are used to find regions in these files that are in correspondence and have the necessary alignment properties. Then the differences between these regions are analyzed in order to describe them using a transformation that depends only on the values of groups of N bytes, not their addresses.
  • the value of N can for example be four (4), but it is not limited to this value.
  • G(x) x + f(x), where f is a piece-wise constant function.
  • the algorithms take into account the tradeoff between increasing the preprocessing success rate and keeping the overall amount of preprocessing data small.
  • the XIP post processing strives to apply the changes in a way that is fast and uses minimal memory.
  • the improvements due to the preprocessing described herein are dramatic, because a large number of changes can be encoded using just a few bytes of data to store a simple f-function.
  • the best cases are the regions of pure data pointers, which were the original motivation for the CFD preprocessing techniques.
  • the XIP preprocessing of an embodiment includes client-side (referred to herein as "DUC-side”) topics and server-side (referred to herein as "NSCD-side”) topics.
  • the DUC-side topics include but are not limited to the following: definitions of the f-function; XIP hint file format; how to apply; faster implementations; and ways to reduce DUC memory requirements.
  • the NSCD-side topics include but are not limited to the following: how to create XIP hint data; work on making the package smaller (speed of applying the hints is less important); and speed of creating the hints is less important.
  • Step 1 On the NSCD-side, there are two steps, "Step 1" and “Step 2" which are described below. There are two steps because of use of the idea of preprocessing the new image rather than the old, and there is a need to know the list of postprocessor calls to calculate the fix-up data.
  • XIP preprocessing a (mandatory) scenario in which this is relevant is when the start address or end address of a postprocessor call is not 4-byte aligned. Then the 4-byte value that is split cannot be treated as one of the 4-byte values that are transformed by the post-processing algorithms. Therefore, any changes that need to be encoded here would have to be encoded as fix-up data.
  • CRAMFS block dependency algorithms might decide not to use a preprocessed portion. Then this portion would not be present in the list of postprocessor calls, and therefore it would not require any fix-up data, so it would be wasteful to store fix-up data for this portion.
  • XIP post-processing is based on the transformation (x + f(x)), where "f()" is a piece-wise constant function that can be stored with a small amount of data (which is part of the XIP hint data; the other main part of the XIP hint data is the specification of the regions where the post-processing is to be applied). If the region being post-processed includes two identical 4-byte values at different locations, and the result after post-processing is supposed to have two different values, then this cannot be accomplished by the transformation (x + f(x)); the region would be split, which might use more hint data than storing the fix-up data.
  • FIG. 1 is a block diagram showing file differencing and updating provided under an embodiment.
  • the file differencing and updating includes a differencing component and an updating component.
  • the differencing component referred to herein as the file difference generator, generates a difference file in a first processor-based or computer system from an original version and a new version of an electronic file.
  • the updating component referred to herein as the update generator, generates a copy of the new file on a second processor-based or computer system using the difference file and the hosted copy of the original file.
  • a first computer system 102 and a second computer system 104 communicate via a communication path 106.
  • These computer systems 102 and 104 include any collection of computing components and devices operating together, as is known in the art.
  • the computer systems 102 and 104 can also be components or subsystems within a larger computer system or network.
  • the communication path 106 includes any medium by which files are communicated or transferred between the computer systems 102 and 104. Therefore, this path 106 includes wireless connections, wired connections, and hybrid wireless/wired connections.
  • the communication path 106 also includes couplings or connections to networks including local area networks (LANs), metropolitan area networks (MANs), wide area networks (WANs), proprietary networks, interoffice or backend networks, and the Internet.
  • LANs local area networks
  • MANs metropolitan area networks
  • WANs wide area networks
  • proprietary networks interoffice or backend networks
  • the Internet and the Internet.
  • the communication path 106 includes removable fixed mediums like floppy disks, hard disk drives, and CD-ROM disks, as well as telephone lines, buses, and electronic mail messages.
  • the first communication system 102 receives an original, or old, version 110 and a new version 112 of an electronic file.
  • the new file 112 is generally an updated or revised version of the original file 110, but is not so limited.
  • the electronic files 110 and 112 include software files including dynamic link library files, shared object files, embedded software components (EBSCs), firmware files, executable files, data files including hex data files, system configuration files, and files including personal use data, but are not so limited. Since any type of file can be regarded as a byte stream, hereinafter a file can be described as a byte stream, depending on the context.
  • Components of the file difference generator 114 receive the new file 112, compare it to the original file 110, and calculate the differences between the compared files, as described below. These differences include byte-level differences between the compared files, but are not so limited.
  • the file difference generator 114 generates a difference file 116, referred to herein as a delta file 116, during the comparison.
  • the file difference generator 114 of an embodiment couples among components of the host computer system 102, where the components include at least one of a processor, at least one controller, at least one memory device, and at least one bus, but is not so limited.
  • the components of the file difference generator 114 of an embodiment include at least one pre-processing subsystem 124 and at least one differencing subsystem 134.
  • the pre-processing subsystem 124 also referred to as the pre-processor 124, includes at least one processor running under control of at least one pre-processing algorithm, program, or routine.
  • the differencing subsystem 134 includes at least one processor running under control of at least one differencing algorithm, program, or routine.
  • the delta file 116 provides an efficient representation of the differences between the new files 112 and the original files 110.
  • the delta file 116 includes meta-data along with actual data of replacement and/or insertion operations that represent the differences between the new or current version of the associated file and previous versions of the file, as described in the United States Patent Application entitled “Byte-Level File Differencing and Updating Algorithms," Application number 10/146,545, filed on May 13, 2002.
  • the file difference generator 114 provides any differences between the original 110 and the new 112 files in the delta file 116 using a minimum number of bytes and a pre-defined format or protocol, thereby providing a delta file optimized in space.
  • the delta file 116 is transferred or transmitted to another processing system 104 via the communication path 106. Prior to transfer, the delta file 116 may be compressed using compression techniques known in the art, but is not so limited.
  • the update generator 118 hosted on the receiving system 104 uses the delta file 116 along with the hosted original file 11OH to generate or create a copy of the new file 112C. This copy of the new file 112C is then used to update the original file 11OH hosted on the client device that is targeted for revision or updating. Upon completion of this update process, the new file now stored on the second computer system is identical to the new file 112 received in the first computer system 102.
  • the differences between an original file and a new file are typically smaller than the new file, leading to significant storage and transmission savings if the differences are transmitted and stored instead of the entire new file. This is particularly important for mobile electronic devices (client devices) hosting programs that are updated via connections that typically can be slow and expensive, for example wireless or cellular connections.
  • the reduced size of the delta file provides numerous improvements, one of which includes a reduction in bandwidth required for transmission of the delta file to the client device; the smaller file means less bandwidth is required for the transfer. Also, smaller files require less time for transmission and, therefore, decrease the probability that the file transfer will be interrupted and simultaneously reduce transmission errors in the received file. In addition, it is safer to transmit the delta files than the new software images via a non- secure connection. All of these improvements increase customer satisfaction.
  • FIG. 2 is a flow diagram for generation of a delta file, under the embodiment of Figure 1.
  • Operation begins when a new file and an original file are received in a first processing or computer system t at block 204.
  • the map files corresponding to the new and original files are also received.
  • the map file is a high- level text file that includes the start address and size of each symbol of a software image, with symbol examples including function and global variables.
  • the map file is output by compiler/linker utilities, and is also known as a log file, symbol file, and/or list file.
  • Pre-processing operations are performed between contents of the new file and the original file in order to identify common segments and simple patterns among contents of the two files, at block 206.
  • the pre-processing uses identified common segments and patterns to reduce/remove the secondary changes between the new and original files.
  • the pre-processing of an embodiment includes reducing and/or removing changes between the original and new files resulting from address shifts associated with logical changes, as described below, but is not so limited. Thus, this pre-processing reduces the differences among common segments of the files, including secondary changes, thereby increasing the efficiency of the difference calculation.
  • the byte-level differences are calculated between the new file and the modified original file, at block 208.
  • the calculated differences are coded and merged, and the delta file is generated by following the pre-defined encoding format, at block 210.
  • the delta file is then optimized as is known in the art to further reduce the file size, when possible, at block 212, and the optimized delta file is provided as an output, at block 214.
  • pre-processing operations are performed between the contents of the new file and the original file in order to identify common segments and simple patterns among contents of the two files. The knowledge of common segments and simple patterns is used to reduce/remove the secondary changes, thereby resulting in an overall performance gain.
  • the transmission of electronic file or software upgrades between a system and a client device can take a significant amount of time, especially when done via low bandwidth channels.
  • An example is a cellular telephone software upgrade. It has become typical practice to send the byte-level file differences or changes between the new and original software versions over the cellular wireless couplings. The significant transfer time arises because the differences between the new and original versions of the executable files are more complex than the differences between their corresponding source files.
  • the changes introduced in the executable files include two main types of changes: primary changes and secondary changes.
  • the primary changes also referred to as logical changes, are source code changes arising from source code line deletion from the original file, source code line addition to the new file, and source code line modifications.
  • the secondary changes are defined to include, but not limited to, address changes, pointer target address changes, and changes in address offsets resulting from the primary changes and generated by the software compiler/linker utilities.
  • the pre-processing routines described below remove/reduce the secondary changes and encode information relating to the removal of these changes in information of the corresponding delta file.
  • FIG. 3 is a block diagram 300 showing an original Vl and a new V2 version of an executable file.
  • the original Vl and new V2 versions both include a code (text) section 310 and 320 and a data section 312 and 322.
  • the new version V2 differs from the original Vl in that the new version V2 includes an additional block of new code 302 of size 0x500.
  • the presence of the block of new code 302 introduces two types of secondary changes.
  • a first type of secondary change occurs in the code (text) section 320 of the new version V2 and results in the branch instruction BRANCH having a different branch displacement, also referred to as the relative address or branch offset, resulting from the addition of the block of new code 302 between the address of the branch instruction BRANCH and the target instruction address TARGET.
  • the target instruction address is 0x1000
  • the branch instruction address is 0x3000, resulting in a branch displacement of 0x2000 (0x3000 - 0x1000) in the original version.
  • the addition of the block of new code (size 0x500) between the branch instruction address and the target instruction address changes the branch instruction address to 0x3500 (0x3000 + 0x500).
  • the branch displacement in the new version V2 changes to 0x2500 (0x3500 - 0x1000).
  • a second type of secondary change occurs in the data section 322 of the new version V2 and results in a change of the value stored in the data pointer POINTER, or the data pointer value, which stores the absolute address of a corresponding data area.
  • the change of the data pointer value results from the addition of the new code 302 in the code (text) section.
  • the new code 302 is inserted at a point in the original version that precedes the data section 322.
  • the data pointer value which is 0x8000 in the original version, changes to 0x8500 (0x8000 + 0x500) in the new version.
  • the secondary changes spread throughout the executable files to the point that the secondary changes can outnumber the primary changes when considered in the context of byte-level file differencing.
  • the pre-processing routines described below use relationships between the original and new versions of files to reduce the amount of information encoded in the delta file as to differences relating to the secondary changes. With the minimum information, the effect of cascading computations can be achieved when doing byte-level differencing and reconstructing, thereby reducing the size of the delta file.
  • the pre-processing routine of an embodiment includes at least one approximation routine and at least one merging routine for use in minimizing the delta file size.
  • the approximation routines function to reduce/remove secondary changes according to text (code) and data model assumptions.
  • the merging routines also referred to as a hint merging routines, encode the model information at a least-cost level for transfer to devices receiving the new versions. The model information is used in the recovery of the new version in the device, as described above, but is not so limited.
  • Figure 4 is a flow diagram 206 for pre-processing different versions of an electronic file, under the embodiment of Figure 1 and Figure 2.
  • the common function units and common data units are extracted from the associated map files, at block 402.
  • the common function units are merged to form common function blocks, at block 404, as described in detail below and with reference to Figure 11.
  • the common data units are merged to form common data blocks, at block 406, as described in detail below and with reference to Figure 12.
  • the pre-processing routines of a pre-processing system pre-process the code (text) sections of the new and original files, at block 408, as described below with reference to Figures 5, 6, and 7.
  • the pre-processing routines subsequently or simultaneously pre-process the data sections of the new and original files, at block 410, as described below with reference to Figures 8, 9 and 10.
  • the common function blocks and common data blocks are encoded, at block 412, and a modified version of the original file is output for use in performing byte-level file differencing.
  • Figure 5 is a flow diagram 408 for pre-processing code (text) sections of different versions of an electronic file, under the embodiment of Figure 4.
  • the code (text) sections of the files are made up of one or more function units or blocks.
  • the pre-processing of code (text) sections of an embodiment begins with the identification of start and end addresses of each function unit of the original and the new files, at block 502, where the function units of the original files are referred to herein as "original function units" and the function units of the new files are referred to herein as "new function units.”
  • the start and end addresses of the function units are identified using map files, but are not so limited.
  • the pre-processor next assigns a unique index or index value to each unique function unit, at block 504.
  • This assignment of index values to unique function units supports identification of function units that are common to both the original and new file versions. Consequently, when the same function unit is found in both the original and the new versions of the file, that function is assigned a common index value, but the embodiment is not so limited.
  • startAddress is generally defined as the starting address of the corresponding function unit; therefore "start AddressV 1" is the startAddress of a function unit of the original file and "startAddress V2" is the startAddress of a function unit of the new file.
  • the pre-processor generates a HintTable of common function units using information of the index value, starting address, and ending address from the tables above, at block 506.
  • the HintTable includes information of the common function units assembled in one table, including index value, the starting and ending addresses of the original function units of the original file version (Vl), and the starting and ending addresses of the new function units of the new file version (V2).
  • the information of the HintTable is arranged using any number of techniques known in the art. Using the information above, the HintTable of an embodiment is generated as follows:
  • Figure 6 is a flow diagram 508 for generating new target address values for the code (text) sections of an original version of an electronic file, under the embodiment of Figure 5.
  • the instruction address is generally referred to herein using the notation "addr”; therefore, “addrVl” refers to an instruction address in the original function unit, while “addrV2" refers to an instruction address in the new function unit.
  • targetAddr refers to a target address in the original function unit
  • targetAddrV2 refers to the corresponding target address in the new function unit.
  • start address of the original function unit that includes targetAddrVl is referenced herein using the notation “targetStartAddrVl ”
  • start address of the new function unit that includes targetAddrV2 is referenced herein using the notation "targetStartAddrV2.”
  • Figure 7 is a block diagram of an original version Vl and a new version V2 of a file, under an embodiment, showing the different addresses (and corresponding notations) described with reference to generation of new target address values for the code (text) sections of an original version of an electronic file.
  • a first common function unit CFU A and a second common function unit CFU B are common to both the original Vl and the new V2 versions.
  • Common function unit CFU B of original version Vl has a starting address startAddrVl, and includes a calculable instruction cal_insA located at instruction address addrVl. The calculable instruction is described below.
  • Common function unit CFU A of original version V 1 has a starting address targetStartAddrVl, and includes the target address targetAddrVl of the calculable instruction cal_insA.
  • common function unit CFU B of new version V2 has a starting address startAddrV2, and includes a calculable instruction cal_insB located at instruction address addrV2.
  • Common function unit CFU A of new version V2 has a starting address targetStartAddrV2, and includes the target address targetAddrV2 of the calculable instruction cal_insB.
  • the current target address of the calculable instruction is then generated or calculated, at block 604, as follows.
  • the target address is computed using the current instruction address and the corresponding instruction decoding.
  • instruction encode(target AddrVl — addrVl).
  • This type of instruction is referred to as a calculable instruction, referred to herein as "cal_ins' ⁇ and the value (targetAddrVl - addrVl) is referred to as the calculable instruction's value or the "instruction value.”
  • the pre-processor of an embodiment generates the instruction value in the new version of a common function unit using the instruction value in the original version along with the original function unit starting addresses and the new function unit starting addresses.
  • the instruction value in the new version (targetAddrV2 — addrV2) is generated or computed using the instruction value in the original version (targetAddrVl - addrVl), start AddrVl, targetStartAddrVl, startAddrV2, and targetStartAddrV2.
  • the pre-processor accesses the HintTable to identify k, the function unit of the original file version that includes the target address targetAddrVl, at block 606. With original function unit j and the identity of its targeted function unit k, the pre-processor reads startAddrVl, startAddrV2 > targetStartAddrVl , and targetStartAddrV2 from the HintTable, at block 608. Continuing, the pre-processor now generates the instruction value (targetAddrV2 — addrV2) for cal_insB to replace the instruction value of cal_insA 5 at block 610, as follows.
  • the pre-processor of an embodiment operates under at least two assumptions, but is not so limited.
  • the first assumption assumes that a common function unit having the same size in both the original version and the new version has the same instruction structure across both the original and new versions, where the instruction structure includes the instruction type and instruction order.
  • the original instruction at address addrVl of the original version, cal_insA is then replaced with the new instruction.
  • the pre-processor subsequently determines if any calculable instructions of the original function unit j remain un-preprocessed, at block 614. When calculable instructions remain un-preprocessed, the pre-processor returns to read another, or alternatively the next, un-preprocessed calculable instruction from the original function unit j, at block 602, and pre-processing continues as described above. When all calculable instructions of original function unit j have been pre- processed, as determined at block 614, the preprocessor determines if the value in counter j is greater than a value n, at block 616. A determination that j is not greater than n indicates that there are common function units that have not been pre- processed, so the value in counter j is incremented by one, at block 618, and preprocessing continues as described above.
  • a determination that j is greater than n indicates that all function units of the original file version have been pre-processed. Consequently, the pre-processor outputs a modified version of the original file that includes an approximated version of the new file, at block 620.
  • FIG. 8 is a flow diagram 410 for pre- processing data sections of original and new versions of an electronic file, under the embodiment of Figure 4.
  • the data sections of the files are made up of one or more global variable units or blocks, referred to herein as "data units.”
  • the pre-processing of data sections of an embodiment begins with the identification of start and end addresses of each data unit of the original and new files, at block 802, where the data units of the original files are referred to herein as
  • original data units and the data units of the new files are referred to herein as “new data units.”
  • new data units The start and end addresses of the data units are identified using map files, but are not so limited.
  • the pre-processing next assigns a unique index or index value to each unique data unit, at block 804.
  • This assignment of index values to unique data units supports identification of data units that are common to both the original and new file versions. Consequently, when the same data unit is found in both the original and the new versions of the file, that data unit is assigned a common index value, but the embodiment is not so limited.
  • the assignment of index values to unique data units is the same as the assignment of index values to unique function units described above, but is not so limited.
  • the generation of tables organized according to index value and including the starting address and ending address for each data unit is the same as the generation of tables described above for the function units of the original and new file versions.
  • Alternative embodiments can assign index values and generate tables using any number of techniques known in the art.
  • the pre-processor generates a HintTable of common data units using information of the index value, starting address, and ending address from the tables above, at block 806.
  • the HintTable includes information of the common data units assembled in one table, including index value, the starting and ending addresses of the original data units of the original file version (Vl), and the starting and ending addresses of the new data units of the new file version (V2).
  • Generation of the HintTable is as described above with reference to the HintTable of common function units, but the HintTable can be generated in alternative embodiments using any number of techniques known in the art.
  • Figure 9 is a flow diagram 808 for generating new data pointer values for the data pointers in the original version of an electronic file, under the embodiment of Figure 8.
  • a data pointer is an instruction that points to some data in a data unit.
  • the notation used herein in describing generation of new data pointer values follows.
  • the address of an instruction that includes a data pointer is generally referred to herein using the notation "dataPointer”; therefore, “dataPointerVl” refers to a dataPointer in the original version, while “dataPointerV2" refers to a dataPointer in the new version.
  • FIG. 10 is a block diagram of an original version Vl and a new version V2 of a file, under an embodiment, showing the different addresses (and corresponding notations) used in generating new data pointer values for an original version of an electronic file.
  • the original Vl and new V2 versions both include a code (text) segment and a data segment.
  • the data segment includes a common data unit CDU A that is common to both the original Vl and the new V2 file versions.
  • the original version Vl includes a data pointer dataPointerVl in the code (text) segment which points to some data in common data unit CDU A of the original version, where the start address of CDU A in the original version Vl is dataStartAddrVl.
  • the new version V2 includes a data pointer dataPointerV2 in the code (text) segment which points to some data in common data unit CDU A of the new version V2, where the start address of CDU A in the new version V2 is dataStartAddrV2.
  • the dataStartAddr addresses are from map files associated with the software images of the original Vl and new V2 versions of the file.
  • generation of new data pointer values begins by identifying an un-preprocessed data pointer in the original version, at block 902.
  • the pre-processor of an embodiment generates the data pointer value (dataPointerV2) of the new version using the pointer value in the original version (dataPointerVl) along with the start addresses of the data unit including data pointed to by dataPointerVl (dataStartAddrVl) and the start address of its corresponding data unit in the new version (dataStartAddr V2), as described below. This is done by accessing the HintTable to identify the k, the data unit in the original version which includes data pointed to by dataPointerVl, at block 904. With the identity of the data unit k, the pre-processor reads dataStartAddrVl and dataStartAddrV2 from the HintTable, at block 906.
  • the pre-processor subsequently generates the data pointer value (dataPointerV2) to replace dataPointerVl , at block 908, as follows.
  • the preprocessor of an embodiment operates under at least two additional assumptions with regard to data units, referred to herein as the third and fourth assumptions, but is not so limited.
  • the third assumption assumes that a common data unit having the same size in both the original version and the new version has the same data structure across both the original and new versions, where the data structure includes the type, size, and order of the data variables.
  • the pre-processor uses the data pointer value dataPointerV2, at block 910, the pre-processor replaces dataPointerVl with dataPointerV2 at the corresponding address in the original version. The pre-processor subsequently determines if any data pointers of the original data unit remain un-preprocessed, at block 912.
  • the preprocessor When data pointers of the original data unit remain un-preprocessed, the preprocessor returns to read another, or alternatively the next, un-preprocessed data pointer from the original data unit, at block 902, and pre-processing continues as described above. When all data pointers of the original data unit have been pre- processed, as determined at block 912, the preprocessor outputs a modified version of the original file including an approximated version of the new file, at block 914.
  • the starting and ending addresses of the common function units and the common data units, as used above in the HintTables, are collectively referred to as
  • the hints are transferred to client devices receiving the difference or delta file for use in recovering the new file version from the delta file and the original file, as described above. Because transfer of the hints is performed via the same low bandwidth channel used for transfer of the delta file and other information to the client device, the pre-processor of an embodiment performs hint merging to merge common function units and common data units of the HintTable without affecting the performance of the pre-processing algorithms or routines. This merging reduces the size of the file that includes the hints and, consequently, the amount of information transferred to the client devices.
  • FIG. 11 is a flow diagram of hint merging 404 of common function units, under the embodiment of Figure 4.
  • Figure 12 is a flow diagram of hint merging 406 of common data units, under the embodiment of Figure 4. Hint merging of the common function units and common data units is described in turn below.
  • startAddress Vl refers to a start address of an original function or data unit
  • startAddress V2 refers to the corresponding start address in the new function or data unit.
  • endAddress Vl refers to an ending address of an original function or data unit
  • endAddress V2 refers to the corresponding ending address in the new function or data unit.
  • HintTable of common function units was generated that includes information of the common function units assembled in one table, including index value, the starting and ending addresses of the original function units of the original file version Vl , and the starting and ending addresses of the new function units of the new file version V2.
  • the HintTable provided above is used as an example in describing the hint merging of common function units, but the embodiment is not so limited:
  • the ending address of the original version Vl of common function unit j is different from the starting address of the original version Vl of common function unit (j + 1), operation proceeds to determine whether the value of j is less than the quantity (n - I) 3 at block 1112.
  • the pre-processor determines whether the ending address of the new version
  • V2 of common function unit j is the same as the starting address of the new version
  • the ending address of the new version V2 of common function unit j is different from the starting address of the new version V2 of common function unit (J + I) 5 operation proceeds to determine whether the value of j is less than the quantity (n - 1), at block 1112.
  • the pre-processor merges the information of common function unit j and common function unit (j + 1) to form a common function block to replace the entry for unit (j +
  • the pre-processor determines, at block 1112, whether the value of j is less than the quantity (n — 1). When the value of j equals the quantity (n — 1), indicating ' that all function units have been preprocessed, operation proceeds to output all usable entries as common function blocks, at block 1116, and operation returns. When the value of j is less than the quantity (n — 1), indicating that function units remain un- preprocessed, the value of j is incremented, at block 1114, and operation proceeds to read information of common function units corresponding to the new value of j, at block 1 102. Pre-processing then continues as described above. An example involving hint merging of common function units is described below, with reference to Figure 11 and the HintTable.
  • Operation begins with the preprocessor marking all three entries usable and reading the original and new versions of function units 1 and 2 from the HintTable.
  • the pre-processor merges the information of common function unit 1 and common function unit 2 to form a common function block as follows:
  • the pre-processor next reads the original and new versions of function units 2 and 3 from the HintTable.
  • the pre-processor of an embodiment performs hint merging of common data units in a manner similar to that described above with regard to the common function units.
  • the pre-processor determines, at block 1204, whether the original version Vl and new version V2 of common data unit j are equal in size. This determination is made by taking a difference between the starting and ending addresses of the original Vl and new V2 versions as endAddrVl(j) - startAddrVIG) - endAddrV2(j) - startAddrV2(j), but is not so limited. When the original Vl and new V2 versions have different size files, operation proceeds to determine whether the value of j is less than the quantity (n - 1), at block 1212.
  • the ending address of the original version V 1 of common data unit j is different from the starting address of the original version Vl of common data unit (j + 1), operation proceeds to determine whether the value of j is less than the quantity (n - 1), at block 1212.
  • the ending address of the new version V2 of common data unit j is different from the starting address of the new version V2 of common data unit (J + I) 5 operation proceeds to determine whether the value of j is less than the quantity (n - 1), at block 1212.
  • the pre-processor merges the information of common data unit j and common data unit (j
  • the pre-processor determines, at block 1212, whether the value of j is less than the quantity (n - 1). When the value of j equals the quantity (n — 1), indicating that all function units have been preprocessed, operation proceeds to output usable entries in the HintTable as common data blocks, at block 1216, and operation returns.
  • the systems and methods described for pre-processing different versions of an electronic file can be applied to software and executable files of any number of processing and/or processor-based systems using any number of instruction architectures.
  • the systems and methods herein can be used in the ARM® architecture, as described in the "ARM Architecture Reference Manual," 2 nd Edition, by D. Jagger and D. Seal.
  • the ARM® architecture is a microprocessor architecture based on a 16/32-bit embedded Reduced Instruction Set Computer (RISC) core, and incorporating the Thumb 16-bit instruction set.
  • RISC Reduced Instruction Set Computer
  • the calculable instructions referred to above would include “branch with link (BL)” and “branch with link and change mode to ARM/Thumb (BLX)" instructions of the ARM/Thumb instruction set.
  • the data pointer referred to above includes the DCD instruction of the ARM/Thumb instruction set.
  • the processing of an embodiment described herein includes post-processing in addition to the pre-processing that generates common code/data tables directly from a pair of software executables, without the use of symbol information from map files, symbol files or log files generated by compiler/linker utilities.
  • the post-processing of pre-processed data includes post-processing of an array of common code blocks as described above.
  • Figure 13 shows post-processing that generates one or more new code blocks, under an embodiment.
  • Figure 14 shows post-processing that increases the size of one or more previously identified common code blocks, while also generating at least one new code block, under an embodiment.
  • systems and methods are provided for pre-processing original and new versions of files as part of difference file generation between the original and new file versions.
  • Various embodiments, as part of the pre-processing use approximation rules between file versions to remove/reduce the secondary changes and encode information relating to the removal of these changes in information of the corresponding difference file.
  • the difference file can be sent "over-the-air" (OTA) and used to generate a copy of the new file version on a different computer system.
  • OTA over-the-air
  • the vertical line to the left represents an original version of an electronic file.
  • the vertical line to the right designated, "new image V2" represents a new version of the electronic file.
  • Each vertical line represents units of code, such as a byte sequence.
  • the old image Vl and new image V2 will be used in describing embodiments of systems and methods for post-processing that improve the performance of preprocessed instructions, but is not so limited.
  • the post-processing examines instruction characteristics to determine pairs of instructions having a specified relationship, described below.
  • the post-processing processing examines instruction characteristics to determine pairs of aligned instructions.
  • the postprocessing For each pair of aligned instructions, the postprocessing generates a pair of target addresses.
  • post-processing algorithms may be used to extend some of the existing (known) common code blocks and/or create new common code blocks.
  • the post-processing algorithms account for any trade-off between increasing the preprocessing success rate and keeping the overall amount of preprocessing data small. This is an important consideration, because one may readily improve the preprocessing success rate by introducing large amounts of new preprocessing data, which is undesirable.
  • any data blocks may then be re-calculated which improves the preprocessing performance of instructions that have data pointers.
  • Such instructions include but are not limited to the familiar ARM LDR instruction.
  • Post-processing of preprocessed data described herein receives preprocessed data as an array of common code blocks.
  • Calculable instructions such as Thumb BL instructions for example, may be analyzed in the one or more common code blocks to identify instruction characteristics which may be used to determine whether to extend and/or add common code blocks.
  • calculable instructions may be analyzed to identify instructions that have target addresses in the image but not in the known common code blocks. Since both the old and the new images are available at this time, a determination may be made as to whether there are instructions having a specified relationship, such as aligned instructions, in the code blocks. An evaluation as to whether to add and/or extend may then be made based in part on the relationship and/or the starting addresses and/or target addresses of the instructions. Based on the collection of this type of data, including the starting and target addresses, the existing or known common code blocks may be extended and/or new common code blocks may be generated.
  • the post-processing of an embodiment collects data by creating or generating two (2) arrays of integers, with size equal to the number of bytes in the old image. All of the integers are initialized to zero, and int *backward, * forward.
  • backward[i] counts how many instructions, such as BL instructions, would be preprocessed correctly if the previous code block were extended forward to include address "i".
  • Forward[i] counts how many instructions would be preprocessed correctly if the following code block were extended backward to include address "i”.
  • the post-processing also creates a list of addrjpair structures, wherein the structures have an allocated size equal to about half the number of bytes in the old image which should be sufficient to account for the instructions. For example, since the relevant bits are different for the first and second parts of a Thumb BL instruction, there should not be more BL instructions in an image than about one quarter the number of bytes in the image.
  • a variable keeps track of the number of addr__pair's stored in the list, which is initially set to zero.
  • An addr_pair structure of an embodiment includes a pair of values as
  • addr__pair *addrArray These values represent a target address in the old image, and a target address in the new image, but the embodiment is not so limited. A pair of target addresses are added to this list when the pair includes characteristics that result in the instruction not being preprocessed correctly using either the preceding or following code block. Generally, the "backward” and “forward” arrays may be used when attempting to extend known code blocks, and the addr_pair list may be used when attempting to create new code blocks.
  • the old image Vl and the new image V2 include a number of known common code blocks 1300, 1302, 1304, and a new common code block 1307.
  • known common code block 1300 may represent a JPEG or MPEG file.
  • Known common code block 1302 may represent a WAV file.
  • Known common code block 1304 may represent an address book.
  • Common code block 1307 is a "new" common code block which is generated by the post-processing of an embodiment (the "dashed" boundary of common code block 1307 represents it as a new common code block). While a certain number of common code blocks are depicted, that there may be greater or fewer numbers of common code blocks.
  • the post-processing of an embodiment improves the performance of the preprocessing of instructions that use the concept of relative addressing.
  • Such instructions may be referred to as "beacons” or “generalized beacons.” They include but are not limited to calculable instructions, such as the familiar ARM BL instruction. Since both the old and the new images are available when the preprocessing data is created, it is possible to test if there are instructions at aligned locations in the common code blocks. It is also possible to examine the associated starting and/or target addresses. Based on the collection of this type of data about the instructions, including the starting and/or target addresses, it is possible to extend the existing common code blocks and/or create new common code blocks.
  • a function/algorithm uses the addr_pair list to try to create a new code block.
  • addr_pair list is used by this function to try to create new code blocks as:
  • Figure 13 depicts the post-processing which results in the addition of a new common code block 1307.
  • the post-processing identifies instructions that have target addresses in the image but not in the known code blocks.
  • the post-processing has identified instructions 1308, 1310, 1312, and 1314 as instructions which include characteristics that may lead to the generation of one or more new common code blocks.
  • Each instruction includes an associated starting address and a target address (1308a-b, 1310a-b, 1312a-b, and 1314a-b).
  • the postprocessing of the instructions 1308, 1310, 1312, and 1314 results in the addition of the new common code block 1307.
  • the post-processing has identified instructions 1308,
  • the post-processing has identified that the target addresses 1308b, 1310b, 1312b, and 1314b of instructions 1308, 1310, 1312, and 1314 do not exist in a known code block and are spaced apart according to a defined minimal spacing. Additionally, the starting addresses 1308a, 1310a, 1312a, and 1314a of instructions 1308, 1310, 1312, and 1314 exist within one or more known common code blocks, such as known common code block 1306. The spacing between the target addresses 1308b, 131 Ob, 1312b, and 1314b is but one factor which may be used in the determination as to whether to create a new common code block.
  • the number of instructions that include target addresses that point to a common area is yet another factor which may be used in the determination as to whether to create a new common code block.
  • the frequency of instructions that include target addresses that point to a common area is another factor which may be used in the determination as to whether to create a new common code block.
  • one or more new code blocks are created based on at least one of a number of instructions, a frequency of instructions, and a spacing between instructions that may point to one or more common code areas in the old and new images.
  • post-processing is used to increase the size or expand one or more previously identified common code blocks, while also generating a new common code block, under an embodiment.
  • the "backward” and “forward” arrays may be used by this function/algorithm to extend a code blocks as:
  • the old image Vl and the new image V2 include a number of known common code blocks 1400, 1402, 1404, 1406, and a new common code block 1408.
  • the post-processing may increase the size or expand one or more previously identified common code blocks, such as one or more of the known common code blocks 1400-1406.
  • the post- processing identifies instructions that have target addresses in the image but not in the known code blocks.
  • the post-processing also identifies instructions having starting addresses which have a calculable relationship with respect to the instructions and/or the known common code blocks.
  • the post-processing has identified instructions 1410 and 1412 as instructions which include characteristics that may lead to the expansion of one or more known common code blocks, such as known common code block 1402. That is, the identified instructions 1410 and 1412 have corresponding target addresses 1410b and 1412b that are not in a known common code block. Since the target addresses 1410b and 1412b are not in a known common code block, the post-processing may then evaluate one or more characteristics of the instructions 1410 and 1412 to determine whether to expand the known common code block 1402.
  • the post-processing may examine each starting address 1410a and 1412a of instructions 1410 and 1412 to determine whether to expand known common code block 1402 to include the target addresses 1410b and 1412b.
  • the post-processing since starting address 1410a is within the known common code block 1400, the post-processing examines the offset or difference between the starting address 1410a of instruction 1410 and the starting address 1400a of the known common code block 1400. This difference is represented as "delta-a" in Figure 14.
  • the post-processing examines the offset or difference between the starting address 1412a of instruction 1412 and the starting address 1400b of the known common code block 1400. This difference is represented as "delta-b".
  • delta-a is equal to delta-b
  • starting addresses 1410a and 1412a are aligned, and the post-processing expands the known common code block 1402 to include the target addresses 1410b and 1412b.
  • the post-processing may increase the size of common code block 1402 so that the target addresses 1410b and 1412b are now encompassed by the expanded common code block 1402,. Consequently, the extended common code block 1402 includes a new boundary 1418 which corresponds to the increase in the size of the common code block 1402.
  • the post-processing may extend the size of a common code block by using the backward and/or forward arrays to thereby change two (2) numbers in the hint table.
  • Moving the boundary may also change the SIZE variable, described above.
  • the post-processing has identified instructions 1414 and 1416 as instructions which include characteristics that may lead to the expansion of one or more known common code blocks, such as known common code block 1406. That is, the identified instructions 1414 and 1416 have corresponding target addresses 1414b and 1416b that are not in a known common code block. Since the target addresses 1414b and 1416b are not in a known common code block, the postprocessing may then evaluate one or more characteristics of the one or more instructions 1414 and 1416 to determine whether to expand known common code block 1406.
  • the post-processing may examine each starting address 1414a and 1416a of instructions 1414 and 1416 to determine whether to expand known common code block 1406 to include the target addresses 1414b and 1416b. As described above, in accordance with an embodiment, the post-processing examines the offset or difference between the starting address 1414a of instruction 1414 and the starting address 1404a of the known common code block 1404. This difference is represented as "delta-c" in Figure 14. Likewise, the post-processing examines the offset or difference between the starting address 1416a of instruction 1416 and the starting address 1404b of the known common code block 1404. This difference is represented as "delta-d".
  • delta-c is equal to delta-d 5 then starting addresses 1414a and 1416a are aligned, and the post-processing expands the known common code block 1404 to include the target addresses 1414b and 1416b. Stated a different way, based on one or more characteristics of the instructions 1414 and 1416, the post-processing may increase the size of common code block 1406 so that the target addresses 1414b and 1416b are now encompassed by the expanded common code block 1406,. Consequently, the extended common code block 1406 includes a new boundary 1420 which corresponds to the increase in the size of the common code block 1406. In this example, the common code block 1406 has been extended backward, whereas in the first example the common code block 1402 was extended forward. The determination as to whether to extend backwards or forwards is based at least in part on how many instructions would be preprocessed correctly, as described above.
  • Common code block 1408 is a "new" common code block which is generated by the post-processing of an embodiment (the "dashed" boundary of the common code block 1408 represents it as a new common code block).
  • the post-processing identifies instructions that have target addresses in the image but not in the known code blocks. For example, the post-processing has identified instructions 1422, 1424, 1426, and 1428 as instructions which include characteristics that may lead to the generation of one or more new common code blocks. The post-processing of the instructions 1422, 1424, 1426, and 1428 resulted in the addition of the new common code block 1408. As described above, the post-processing has identified instructions 1422,
  • the post-processing has identified that the target addresses 1422b, 1424b, 1426b, and 1428b do not exist in a known code block and are spaced apart according to a defined minimal spacing. Additionally, the starting addresses 1422a, 1424a, 1426a, and 1428a exist within known common code blocks 1402 and 1404. The spacing between the target addresses 1422b, 1424b, 1426b, and 1428b is but one factor which may be used in the determination as to whether to create a new common code block, such as the new common code block 1408.
  • the number of instructions that include target addresses that point to a common area is yet another factor which may be used in the determination as to whether to create a new common code block.
  • the frequency of instructions that include target addresses that point to a common area is another factor which may be used in the determination as to whether to create a new common code block.
  • one or more new code blocks are created based on at least one of a number of instructions, a frequency of instructions, and a spacing between instructions that may point to one or more common areas in the old and new versions.
  • Parameters that may be changed include but are not limited to: /* maximum number of blocks that can be added.
  • the overall improvements realized using the postprocessing techniques are dramatic. This is due, in part, because the embodiments are able to find common code blocks that are small in size but very important in terms of the overall graph of instructions within the images.
  • One such example would be the veneers that are common in ARM code. The veneers are often just a few bytes, and are consequently difficult to detect using conventional techniques.
  • the "data arrays" of an embodiment include but are not limited to the following: int *backward, *forward; addr_pair *addr Array;
  • two different post-processing procedures may be executed, and a procedure is selected based on the number of instructions that are correctly preprocessed.
  • the two different procedures include a first procedure and a second procedure described below.
  • the first procedure includes but is not limited to: clearing the data arrays; based on the input code blocks, calculating the data arrays; modify__tryToAddArray(); clearing the data arrays; based on the modified code blocks, calculating the data arrays; modify_tryToExtend(); and based on the modified code blocks, calculating a number of instructions correctly preprocessed.
  • the second procedure includes but is not limited to: clearing the data arrays; based on the input code blocks, calculating the data arrays; rnodify_tryToExtend(); clearing the data arrays; based on the modified code blocks, calculating the data arrays; modify__tryToAddArray(); based on the modified code blocks, calculating a number of instructions correctly preprocessed.
  • the first procedure and the second procedure differ in that they use the algorithms (modify_tryToAddArray() and modify_tryToExtend()) in the opposite sequence.
  • the selection of the first or second procedure is made based at least in part on a number of instructions correctly preprocessed.
  • Figures 15A-15F are graphical representations which provide example results of the preprocessing and post-processing methodologies discussed above.
  • the sample data includes "vl .bin” which represents the "old image”, and "v2.bin” which represents the "new image”.
  • the two images have approximately the same number of BL instructions. There are 259,653 BL instructions in vl .bin, and 261,283 BL instructions in v2.bin.
  • Figure 15A depicts the results after pre-processing.
  • a number of common code blocks appear as vertical strips, having varying widths.
  • 1500, 1502, 1504, 1506, 1508, 1510, and 1512 are exemplary common code blocks.
  • Figure 15A includes other common code blocks.
  • the horizontal axis at the bottom of Figure 15 A is the address space of the old image (vl .bin) after the removal of any padding regions.
  • the horizontal axis in the middle of the diagram is the address space of the new image (v2.bin) after the removal of any padding regions.
  • the size of the padding region is 245,076 bytes. Addresses that are less than 9,585,324 are not affected by the padding, and the address in the diagram is equal to the actual address in the image.
  • each vertical strip represents one common code block
  • a first part identifies a region in the old image (this is the extent of the strip that is in contact with the lower axis in the diagram)
  • a second part identifies a region in the new image (this is the extent of the strip that is in contact with the middle axis in the diagram).
  • Most of the strips in the diagram appear to be nearly vertical, because the horizontal scale of the diagram is more than nine (9) megabytes, and small shifts in the code blocks are not easily seen on such a large scale.
  • BL instructions in vl .bin there are 258,822 BL instructions in vl .bin that are in the common code blocks. Of these, 187,162 have target addresses that are also in the common code blocks. Of these target addresses, 186,058 are predicted correctly by the common code blocks. A correct prediction generally means that the spacing between the target addresses differs between the old and new images in a manner that agrees with the relevant common code block shifts.
  • the post-processing methodologies analyze instructions, such as BL instructions, that do not have target addresses in the common code blocks. If the instructions do not have target addresses in the common code blocks, then the post-processing attempts to improve the common code blocks, by extension of known common code blocks and/or the addition of new common code blocks, but is not so limited.
  • the post-processing attempts the following procedures: 1. modify_tryToAddArray
  • the post-processing results in a count of 242,198 BL instructions having correct predictions.
  • This post-processing results in a count of 257,217 BL instructions having correct predictions. Since the second procedure results in a higher number of instructions having correct predictions, for this example, the second procedure is used. Note that the count from the second procedure is fairly close to the total number of BL instructions in vl.bin, 258,822. Also, recall that the count of BL instructions with correct predictions was 186,058 before post-processing. Thus, the delta file is significantly smaller as a result of the post-processing methodology.
  • the size of the delta file without post-processing is 274,259 bytes.
  • the size of the delta file with post-processing is 131,799 bytes.
  • the post-processing results in a delta file that is less than half the size of the delta file without using postprocessing.
  • Figure 15B depict the results after post-processing.
  • a number of common code blocks appear as vertical strips, having varying widths.
  • 1514, 1516, 1518, 1520, 1522, 1524, and 1526 are exemplary common code blocks.
  • Figure 15B includes other common code blocks.
  • the upper plot is the density of bytes affected by postprocessing. We note that the curve tends to be higher than the corresponding curve in Figure 1. This is especially noticeable on the left, in the address range from 0 to about 2,000,000 (addresses are referenced to the horizontal axis at the bottom of the figure).
  • FIG. 15C shows the BL instructions in the v2.bin that have target addresses within the image.
  • a similar graph could be made for vl.bin. Each dot is marked with a diamond to make it easier to see.
  • a dot represents a BL instruction. The value on the horizontal axis is the address of the instruction, and the value on the vertical axis is its target address. Looking at the lower-left part of the diagram, we see that quite a few of the BL instructions that have addresses in the range from 0 to 2,000,000 have target addresses around 1,000,000. This corresponds to the horizontal grouping of dots in the lower left (surrounded by a dotted line).
  • Figures 15D and 15E are the result of the magnification of the relevant regions of Figures 15A and 15B, respectively. Only the horizontal axes have been magnified (the vertical axes are unchanged).
  • Figure 15D is a zoom in on Figure 15 A, expanding the horizontal axis to show the range from 900,000 to 1,100,000 which highlights common code blocks 1500, 1502, and 1504.
  • Figure 15E is a zoom in on Figure 15B, expanding the horizontal axis to show the range from 900,000 to 1,100,000 which highlights common code blocks 1514, 1516, 1518, and 1520.
  • Figure 15D which is for the calculation without the post-processing methodologies, there are some visible gaps.
  • the post-processing has mostly filled the gaps, either by extending (e.g. using modify_tryToExtend) neighboring common code blocks (e.g. 1514, 1516, and 1520) and/or by introducing new (e.g. using rnodify__tryToAddArray) common code blocks (e.g. 1518).
  • each common code block is described by three numbers: a, b and c.
  • the prefix "a” represents the start address of the block in vl .bin.
  • the prefix "b” represents the start address of the block in v2.bin.
  • the prefix "c” represents the length of a block (in both vl .bin and v2.bin).
  • XIP processing Embodiments of the electronic file updating system and method generally referred to herein as XIP processing will now be described with reference to Figures 16-21.
  • Execute-in-place (XIP) files are only one type of file appropriate to the system and method, and there are many other examples of files that would benefit from the preprocessing described.
  • f-function is a piece-wise constant function, that takes an integer as an input and returns an integer, making use of a table of pairs of integers, which must also be specified. Alternative embodiments may use unsigned integers instead. Under one embodiment, one f-function is defined for each XIP preprocessing region.
  • each interval of constancy has a well-defined right-hand endpoint and function value there.
  • An f- array holds the data for the piece-wise constant function.
  • the function is defined by this information for the right-hand endpoint of each interval: (input, output). In the code, such a pair is called (a, b).
  • Figure 16 is a sample plot of the function, under an embodiment, with 'x' indicating a point stored in the array. As is standard, the input axis is horizontal and the output axis is vertical.
  • a format of an embodiment includes one f-array for each region. Note that one f-array may be used for more than one region ("universality").
  • the XIP hint data is also called “intHint” data (preprocessing data for the "integer table hinting").
  • the f-array is an array of integer pairs, (a, b).
  • the arrays are encoded as an initial value followed by subsequent differences.
  • the reason why the .b values with even and odd indices are stored separately is that the .b values often alternate between two values (such as 0 and a nonzero value) for a part of the array, so this is better after compression.
  • the endian is an input quantity.
  • the ARM heuristics for determining the endian should not be used here because there is no guarantee that enough ARM code can be found for the statistics.
  • the endian could be stored in the device.xml file. Alternatively, since this is a value that probably will not be changed too much, it can be stored in the NSCDCLI_CONFIG.txt file, so no changes are required to the device.xml files.
  • stepl_XI P ( ) The main application programming interface ("API") for Step 1 is stepl_XI P ( ) :
  • This API creates the preprocessed new image and the XIP hint data.
  • arguments imagel old image (will not be changed) image2 new image (will be changed) image2b output buffer for preprocessed new image.
  • Memory is allocated here.
  • Memory is allocated here for this buffer.
  • ebsc_list_input The input EBSC list. This will not be changed. endian The endian of the device.
  • stepl_XIP const dt_buffer_s *imagel, dt_buffer_s *image2, dt_buffer_s *image2b, dt_buffer_s *hint_XIP, const dt_EbscInfoList_s *ebsc_list, int endian
  • the stepl_XIP()function carries out these steps (after creating an EBSC list that is the same as the input list, except that the types of some components can be changed from CRAMFS_TYPE to BINARYJT YPE if they are really not CRAMFSJT YPE in at least one of the images):
  • call get_of f set ( ) For each of the two links, call get_of f set ( ) to get the offset of the real part relative to the beginning of the CRAMFS component.
  • step (7) Now the actual preprocessed-new can be created in a manner similar to that in step (4), except that now only "useful" common code blocks are used, as determined in step (6). This is done using the copy of the new image, created in step (3) of this algorithm.
  • Diagnostic information is produced during the calculation. This is useful to check the assumptions that are part of the foundation of the algorithms. This information can also be used to predict how successful the preprocessing is expected to be for an individual case.
  • frac_diff size 0.000310 2401352
  • frac diff size 0.000406 2386696 total_copied 22325964 Parameters
  • a list of parameters is set in the code.
  • o n 1 y_mu 11 i p 1 e s_o f _ 4 is a boolean implemented as an integer, currently set to 0. A nonzero value would mean that only values that are multiples of 4 are processed.
  • This function may be thought of as a generalized parameter.
  • An ordinary parameter is just one number.
  • the function useful ( ) is a boolean-valued function of several integers .
  • this is an extension of the Step 2 without the XIP preprocessing.
  • the new functions are in the DUC-side post-processing functions.
  • Examples XIP preprocessing is useful for a variety of regions created by the compiler.
  • heuristics are used to determine whether the XIP preprocessing is profitable for any given region, without knowing exactly what is in it.
  • Figures 17-21 illustrate examples from the NEC_09_28 ROOT data set. For all of these examples: Number different, after preprocessing: 0. For each example, the data includes an Image 1, and Image 2, and Differences of 4-byte values.
  • Figure 17 is a diagram illustrating a typical table of integers, as operated on by an embodiment.
  • the integers could be some sort of pointers.
  • the number of bytes different before preprocessing 25 (a large number here is good).
  • the number of elements in the preprocessing data array 2 (a small number here is good).
  • Figure 18 is a diagram illustrating code, such as ARM code, as indicated by the familiar e's at the beginning of most of the 32-bit values in the images.
  • the number of elements in the preprocessing data array 8.
  • Figure 19 is a diagram illustrating a more complicated example of a table of values, as operated on by an embodiment.
  • the number of bytes different before preprocessing 44.
  • the number of elements in the preprocessing data array 6
  • Figure 20 is a diagram illustrating an example that shows the importance of the endian being correct in values, as operated on by an embodiment.
  • the number of elements in the preprocessing data array 3.
  • Figure 21 is a diagram illustrating another ARM code example, as operated on by an embodiment. It appears that the data includes many 32-bit ARM LDR instructions (e59 ), showing that not only changes in data pointer are preprocessed according to the embodiments described.
  • the number of bytes different before preprocessing 25.
  • the computing devices receiving and using the delta file may be client devices that host corresponding software applications in need of updating, for example cellular telephones, mobile electronic devices, mobile communication devices, personal digital assistants, and other processor-based devices.
  • client devices that host corresponding software applications in need of updating
  • This support is provided for all mobile device software ranging from firmware to embedded applications by enabling carriers and device manufacturers to efficiently distribute electronic file content and applications via their wireless infrastructure.
  • Another example of systems that benefit from the pre-processing and/or postprocessing described above includes systems using wired serial connections to transfer the delta file from a device hosting the file difference generator to a device hosting the file update generator. These systems typically have slow transfer rates and, because the transfer rates are slow, a reduction in the size of the delta file is a way to realize faster transfer times.
  • Yet another example of systems that benefit from use of the pre-processing and/or post-processing includes wireless systems using radio communications to transfer the delta file from a device hosting the file difference generator to a device hosting the file update generator. While suffering from low reliability associated with the wireless connections, these systems also have slow transfer rates.
  • the use of a smaller delta file in these systems provides several advantages. For example, the smaller file size results in a faster delta file transfer time. The faster transfer time, while saving time for the device user, reduces the opportunity for the introduction of errors into the delta file, thereby increasing system reliability. Also, with cellular communications, the reduced transfer time results in a cost savings for the consumer who is typically charged by the minute for service. As another advantage, the smaller delta file reduces the bandwidth required to transfer the delta files to client devices.
  • aspects of the invention may be implemented as functionality programmed into any of a variety of circuitry, including programmable logic devices (PLDs), such as field programmable gate arrays (FPGAs), programmable array logic (PAL) devices, electrically programmable logic and memory devices and standard cell-based devices, as well as application specific integrated circuits (ASICs).
  • PLDs programmable logic devices
  • FPGAs field programmable gate arrays
  • PAL programmable array logic
  • ASICs application specific integrated circuits
  • microcontrollers with memory such as electronically erasable programmable read only memory (EEPROM)
  • embedded microprocessors firmware, software, etc.
  • aspects of the invention may be embodied in microprocessors having software-based circuit emulation, discrete logic (sequential and combinatorial), custom devices, fuzzy (neural) logic, quantum devices, and hybrids of any of the above device types.
  • the underlying device technologies may be provided in a variety of component types, e.g., metal-oxide semiconductor field-effect transistor (MOSFET) technologies like complementary metal-oxide semiconductor (CMOS), bipolar technologies like emitter-coupled logic (ECL), polymer technologies (e.g., silicon-conjugated polymer and metal-conjugated polymer-metal structures), mixed analog and digital, etc.
  • MOSFET metal-oxide semiconductor field-effect transistor
  • CMOS complementary metal-oxide semiconductor
  • ECL emitter-coupled logic
  • polymer technologies e.g., silicon-conjugated polymer and metal-conjugated polymer-metal structures
  • mixed analog and digital etc.
  • the functions described herein as the invention can be performed by programs or sets of program codes, including software, firmware, executable code or instructions running on or otherwise being executed by one or more general-purpose computers or processor-based systems.
  • the computers or other processor-based systems may include one or more central processing units for executing program code, volatile memory, such as RAM for temporarily storing data and data structures during program execution, non- volatile memory, such as a hard disc drive or optical drive, for storing programs and data, including databases and other data stores, and a network interface for accessing an intranet and/or the Internet.
  • volatile memory such as RAM for temporarily storing data and data structures during program execution
  • non- volatile memory such as a hard disc drive or optical drive
  • programs and data including databases and other data stores
  • a network interface for accessing an intranet and/or the Internet.
  • the functions described herein may also be implemented using special purpose computers, wireless computers, state machines, and/or hardwired electronic circuits.
  • components of the various systems and methods disclosed herein may be described using computer aided design tools and expressed (or represented), as data and/or instructions embodied in various computer-readable media, in terms of their behavioral, register transfer, logic component, transistor, layout geometries, and/or other characteristics.
  • Formats of files and other objects in which such circuit expressions may be implemented include, but are not limited to, formats supporting behavioral languages such as C, Verilog, and HLDL, formats supporting register level description languages like RTL, and formats supporting geometry description languages such as GDSII, GDSIII, GDSIV, CIF, MEBES and any other suitable formats and languages.
  • Computer-readable media in which such formatted data and/or instructions may be embodied include, but are not limited to, non-volatile storage media in various forms (e.g., optical, magnetic or semiconductor storage media) and carrier waves that may be used to transfer such formatted data and/or instructions through wireless, optical, and/or wired signaling media. Examples of transfers of such formatted data and/or instructions by carrier waves include, but are not limited to, transfers (uploads, downloads, e-mail, etc.) over the Internet and/or other computer networks via one or more data transfer protocols (e.g., HTTP, FTP, SMTP, etc.).
  • Such data and/or instruction-based expressions of the above described systems and methods may be processed by a processing entity (e.g., one or more processors) within the computer system in conjunction with execution of one or more other computer programs.
  • a processing entity e.g., one or more processors

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Security & Cryptography (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Stored Programmes (AREA)

Abstract

Des modes de réalisation de la présente invention concernent des systèmes et procédés destiné au prétraitement et au post-traitement de versions d'origine et de nouvelles versions de fichiers dans le cadre de la génération de fichiers différentiels entre les versions d'origine et les nouvelles versions. Les systèmes et les procédés d'un mode de réalisation comprennent un jeu d'algorithmes qui réduisent la taille du fichier différentiel en prétraitant un ensemble de régions d'images logicielles pour des dispositifs de calcul embarqués, dont un exemple est l'image du système de fichiers de mémoire à lecture seule (ROM) comprimé ou CRAMFS. Les algorithmes traitent un ensemble de types de régions de données créées par le compilateur. Les modes de réalisation fonctionnent du côté serveur et du côté client. Du côté serveur, le prétraitement génère directement des données indices de différences fonctionnelles compactes (CFD) à partir d'une paire d'images CRAMFS sans utiliser de fichiers de symboles ni de fichiers journaux générés par des utilitaires de compilation/édition de liens.
PCT/US2007/011132 2006-05-10 2007-05-09 Traitement de différences fonctionnelles compactes WO2007133559A2 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US79957806P 2006-05-10 2006-05-10
US60/799,578 2006-05-10
US11/432,984 2006-05-12
US11/432,984 US7890464B2 (en) 2003-06-20 2006-05-12 Processing software images and generating difference files

Publications (3)

Publication Number Publication Date
WO2007133559A2 WO2007133559A2 (fr) 2007-11-22
WO2007133559A9 true WO2007133559A9 (fr) 2008-05-15
WO2007133559A3 WO2007133559A3 (fr) 2008-09-25

Family

ID=38694437

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2007/011132 WO2007133559A2 (fr) 2006-05-10 2007-05-09 Traitement de différences fonctionnelles compactes

Country Status (1)

Country Link
WO (1) WO2007133559A2 (fr)

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6816872B1 (en) * 1990-04-26 2004-11-09 Timespring Software Corporation Apparatus and method for reconstructing a file from a difference signature and an original file
US6230316B1 (en) * 1998-04-17 2001-05-08 Symantec Corporation Patching rebased and realigned executable files
US6216175B1 (en) * 1998-06-08 2001-04-10 Microsoft Corporation Method for upgrading copies of an original file with same update data after normalizing differences between copies created during respective original installations
US6466999B1 (en) * 1999-03-31 2002-10-15 Microsoft Corporation Preprocessing a reference data stream for patch generation and compression
US6748584B1 (en) * 1999-12-29 2004-06-08 Veritas Operating Corporation Method for determining the degree to which changed code has been exercised
US7058941B1 (en) * 2000-11-14 2006-06-06 Microsoft Corporation Minimum delta generator for program binaries
US7523098B2 (en) * 2004-09-15 2009-04-21 International Business Machines Corporation Systems and methods for efficient data searching, storage and reduction

Also Published As

Publication number Publication date
WO2007133559A3 (fr) 2008-09-25
WO2007133559A2 (fr) 2007-11-22

Similar Documents

Publication Publication Date Title
US7676506B2 (en) Differential file compression of software image versions
US7890464B2 (en) Processing software images and generating difference files
US7089270B2 (en) Processing software images for use in generating difference files
EP0870235B1 (fr) Appareil et procede pour pretraiter des programmes d'ordinateur avant leur transmission par un reseau
Debray et al. Profile-guided code compression
KR100949855B1 (ko) 파일 시스템 이미지를 디퍼런싱하기 위한 장치, 시스템, 방법 및 컴퓨터 판독가능형 매체
US20050010576A1 (en) File differencing and updating engines
US20060101040A1 (en) Algorithms for block-level code alignment of software binary files
US7661102B2 (en) Method for reducing binary image update package sizes
US8032873B2 (en) Computer program code size partitioning system for multiple memory multi-processing systems
US20050010870A1 (en) Post-processing algorithm for byte-level file differencing
WO2004015570A1 (fr) Generation de code optimise
EP1550047A1 (fr) Algorithmes de differenciation et d'actualisation de fichiers au niveau octet
CN101026825A (zh) 用于补丁最小化工具的系统和方法
CN111176717B (zh) 生成安装包的方法、装置及电子设备
EP1691282A1 (fr) Outil d'optimisation de construction pour la gestion efficace de constructions logicielles pour des dispositifs mobiles
CN105049514A (zh) 一种文件下载方法和装置
US20110264669A1 (en) method for compressing a .net file
US20110107046A1 (en) Binary-level update of compressed read-only file systems
CN1208481A (zh) 分布式处理
WO2007133559A9 (fr) Traitement de différences fonctionnelles compactes
CN101297275A (zh) 处理软件镜像并产生差异文件
Lu et al. METHOD FOR COMPRESSING A .NET FILE

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 07794668

Country of ref document: EP

Kind code of ref document: A2

NENP Non-entry into the national phase in:

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 07794668

Country of ref document: EP

Kind code of ref document: A2