Patent Grant
7581139
The technical field of this invention is emulation hardware particularly for highly integrated digital signal processing systems.
Advanced wafer lithography and surface-mount packaging technology are integrating increasingly complex functions at both the silicon and printed circuit board level of electronic design. Diminished physical access to circuits for test and emulation is an unfortunate consequence of denser designs and shrinking interconnect pitch. Designed-in testability is needed so the finished product is both controllable and observable during test and debug. Any manufacturing defect is preferably detectable during final test before a product is shipped. This basic necessity is difficult to achieve for complex designs without taking testability into account in the logic design phase so automatic test equipment can test the product.
In addition to testing for functionality and for manufacturing defects, application software development requires a similar level of simulation, observability and controllability in the system or sub-system design phase. The emulation phase of design should ensure that a system of one or more ICs (integrated circuits) functions correctly in the end equipment or application when linked with the system software. With the increasing use of ICs in the automotive industry, telecommunications, defense systems, and life support systems, thorough testing and extensive real-time debug becomes a critical need.
Functional testing, where the designer generates test vectors to ensure conformance to specification, still remains a widely used test methodology. For very large systems this method proves inadequate in providing a high level of detectable fault coverage. Automatically generated test patterns are desirable for full testability, and controllability and observability. These are key goals that span the full hierarchy of test from the system level to the transistor level.
Another problem in large designs is the long time and substantial expense involved in design for test. It would be desirable to have testability circuitry, system and methods that are consistent with a concept of design-for-reusability. In this way, subsequent devices and systems can have a low marginal design cost for testability, simulation and emulation by reusing the testability, simulation and emulation circuitry, systems and methods that are implemented in an initial device. Without a proactive testability, simulation and emulation plan, a large amount of subsequent design time would be expended on test pattern creation and upgrading.
Even if a significant investment were made to design a module to be reusable and to fully create and grade its test patterns, subsequent use of a module may bury it in application specific logic. This would make its access difficult or impossible. Consequently, it is desirable to avoid this pitfall.
The advances of IC design are accompanied by decreased internal visibility and control, reduced fault coverage and reduced ability to toggle states, more test development and verification problems, increased complexity of design simulation and continually increasing cost of CAD (computer aided design) tools. In the board design the side effects include decreased register visibility and control, complicated debug and simulation in design verification, loss of conventional emulation due to loss of physical access by packaging many circuits in one package, increased routing complexity on the board, increased costs of design tools, mixed-mode packaging, and design for produceability. In application development, some side effects are decreased visibility of states, high speed emulation difficulties, scaled time simulation, increased debugging complexity, and increased costs of emulators. Production side effects involve decreased visibility and control, complications in test vectors and models, increased test complexity, mixed-mode packaging, continually increasing costs of automatic test equipment and tighter tolerances.
Emulation technology utilizing scan based emulation and multiprocessing debug was introduced more than 10 years ago. In 1988, the change from conventional in circuit emulation to scan based emulation was motivated by design cycle time pressures and newly available space for on-chip emulation. Design cycle time pressure was created by three factors. Higher integration levels, such as increased use of on-chip memory, demand more design time. Increasing clock rates mean that emulation support logic causes increased electrical intrusiveness. More sophisticated packaging causes emulator connectivity issues. Today these same factors, with new twists, are challenging the ability of a scan based emulator to deliver the system debug facilities needed by today's complex, higher clock rate, highly integrated designs. The resulting systems are smaller, faster, and cheaper. They have higher performance and footprints that are increasingly dense. Each of these positive system trends adversely affects the observation of system activity, the key enabler for rapid system development. The effect is called “vanishing visibility.”
With today's highly integrated System-On-a-Chip (SOC) technology, the visibility and control gap has widened dramatically over time. Traditional debug options such as logic analyzers and partitioned prototype systems are unable to keep pace with the integration levels and ever increasing clock rates of today's systems. As integration levels increase, system buses connecting numerous subsystem components move on chip, denying traditional logic analyzers access to these buses. With limited or no significant bus visibility, tools like logic analyzers cannot be used to view system activity or provide the trigger mechanisms needed to control the system under development. A loss of control accompanies this loss in visibility, as it is difficult to control things that are not accessible.
To combat this trend, system designers have worked to keep these buses exposed. Thus the system components were built in a way that enabled the construction of prototyping systems with exposed buses. This approach is also under siege from the ever-increasing march of system clock rates. As the central processing unit (CPU) clock rates increase, chip to chip interface speeds are not keeping pace. Developers find that a partitioned system's performance does not keep pace with its integrated counterpart, due to interface wait states added to compensate for lagging chip to chip communication rates. At some point, this performance degradation reaches intolerable levels and the partitioned prototype system is no longer a viable debug option. In the current era production devices must serve as the platform for application development.
Increasing CPU clock rates are also limiting availability of other simple visibility mechanisms. Since the CPU clock rates can exceed the maximum I/O state rates, visibility ports exporting information in native form can no longer keep up with the CPU. On-chip subsystems are also operated at clock rates that are slower than the CPU clock rate. This approach may be used to simplify system design and reduce power consumption. These developments mean simple visibility ports can no longer be counted on to deliver a clear view of CPU activity. As visibility and control diminish, the development tools used to develop the application become less productive. The tools also appear harder to use due to the increasing tool complexity required to maintain visibility and control. The visibility, control, and ease of use issues created by systems-on-a-chip tend to lengthen product development cycles.
Even as the integration trends present developers with a tough debug environment, they also present hope that new approaches to debug problems will emerge. The increased densities and clock rates that create development cycle time pressures also create opportunities to solve them. On-chip, debug facilities are more affordable than ever before. As high speed, high performance chips are increasingly dominated by very large memory structures, the system cost associated with the random logic accompanying the CPU and memory subsystems is dropping as a percentage of total system cost. The incremental cost of several thousand gates is at an all time low. Circuits of this size may in some cases be tucked into a corner of today's chip designs. The incremental cost per pin in today's high density packages has also dropped. This makes it easy to allocate more pins for debug. The combination of affordable gates and pins enables the deployment of new, on-chip emulation facilities needed to address the challenges created by systems-on-a-chip.
When production devices also serve as the application debug platform, they must provide sufficient debug capabilities to support time to market objectives. Since the debugging requirements vary with different applications, it is highly desirable to be able to adjust the on-chip debug facilities to balance time to market and cost needs. Since these on-chip capabilities affect the chip's recurring cost, the scalability of any solution is of primary importance. “Pay only for what you need” should be the guiding principle for on-chip tools deployment. In this new paradigm, the system architect may also specify the on-chip debug facilities along with the remainder of functionality, balancing chip cost constraints and the debug needs of the product development team.
Host computer 120 is generally a personal computer. Host computer 120 provides access the debug capabilities through emulator controller 130. Debugger application program 110 presents the debug capabilities in a user-friendly form via host computer 120. The debug resources are allocated by debug application program 110 on an as needed basis, relieving the user of this burden. Source level debug utilizes the debug resources, hiding their complexity from the user. Debugger application program 110 together with the on-chip trace and triggering facilities provide a means to select, record, and display chip activity of interest. Trace displays are automatically correlated to the source code that generated the trace log. The emulator provides both the debug control and trace recording function.
The debug facilities are preferably programmed using standard emulator debug accesses through a JTAG or similar serial debug interface. Since pins are at a premium, the preferred embodiment of the invention provides for the sharing of the debug pin pool by trace, trigger, and other debug functions with a small increment in silicon cost. Fixed pin formats may also be supported. When the pin sharing option is deployed, the debug pin utilization is determined at the beginning of each debug session before target system 140 is directed to run the application program. This maximizes the trace export bandwidth. Trace bandwidth is maximized by allocating the maximum number of pins to trace.
The debug capability and building blocks within a system may vary. Debugger application program 110 therefore establishes the configuration at runtime. This approach requires the hardware blocks to meet a set of constraints dealing with configuration and register organization. Other components provide a hardware search capability designed to locate the blocks and other peripherals in the system memory map. Debugger application program 110 uses a search facility to locate the resources. The address where the modules are located and a type ID uniquely identifies each block found. Once the IDs are found, a design database may be used to ascertain the exact configuration and all system inputs and outputs.
Host computer 120 generally includes at least 64 Mbytes of memory and is capable of running Windows 95, SR-2, Windows NT, or later versions of Windows. Host computer 120 must support one of the communications interfaces required by the emulator. These may include: Ethernet 10T and 100T, TCP/IP protocol; Universal Serial Bus (USB); Firewire IEEE 1394; and parallel port such as SPP, EPP and ECP.
Host computer 120 plays a major role in determining the real-time data exchange bandwidth. First, the host to emulator communication plays a major role in defining the maximum sustained real-time data exchange bandwidth because emulator controller 130 must empty its receive real-time data exchange buffers as fast as they are filled. Secondly, host computer 120 originating or receiving the real-time data exchange data must have sufficient processing capacity or disc bandwidth to sustain the preparation and transmission or processing and storing of the received real-time data exchange data. A state of the art personal computer with a Firewire communication channel (IEEE 1394) is preferred to obtain the highest real-time data exchange bandwidth. This bandwidth can be as much as ten times greater performance than other communication options.
Emulation controller 130 provides a bridge between host computer 120 and target system 140. Emulation controller 130 handles all debug information passed between debugger application program 110 running on host computer 120 and a target application executing on target system 140. A presently preferred minimum emulator configuration supports all of the following capabilities: real-time emulation; real-time data exchange; trace; and advanced analysis.
Emulation controller 130 preferably accesses real-time emulation capabilities such as execution control, memory, and register access via a 3, 4, or 5 bit scan based interface. Real-time data exchange capabilities can be accessed by scan or by using three higher bandwidth real-time data exchange formats that use direct target to emulator connections other than scan. The input and output triggers allow other system components to signal the chip with debug events and vice-versa. Bit I/O allows the emulator to stimulate or monitor system inputs and outputs. Bit I/O can be used to support factory test and other low bandwidth, non-time-critical emulator/target operations. Extended operating modes are used to specify device test and emulation operating modes. Emulator controller 130 is partitioned into communication and emulation sections. The communication section supports host communication links while the emulation section interfaces to the target, managing target debug functions and the device debug port. Emulation controller 130 communicates with host computer 120 using one of industry standard communication links outlined earlier herein. The host to emulator connection is established with off the shelf cabling technology. Host to emulator separation is governed by the standards applied to the interface used.
Emulation controller 130 communicates with the target system 140 through a target cable or cables. Debug, trace, triggers, and real-time data exchange capabilities share the target cable, and in some cases, the same device pins. More than one target cable may be required when the target system 140 deploys a trace width that cannot be accommodated in a single cable. All trace, real-time data exchange, and debug communication occurs over this link. Emulator controller 130 preferably allows for a target to emulator separation of at least two feet. This emulation technology is capable of test clock rates up to 50 MHZ and trace clock rates from 200 to 300 MHZ, or higher. Even though the emulator design uses techniques that should relax target system 140 constraints, signaling between emulator controller 130 and target system 140 at these rates requires design diligence. This emulation technology may impose restrictions on the placement of chip debug pins, board layout, and requires precise pin timings. On-chip pin macros are provided to assist in meeting timing constraints.
The on-chip debug facilities offer the developer a rich set of development capability in a two tiered, scalable approach. The first tier delivers functionality utilizing the real-time emulation capability built into a CPU's mega-modules. This real-time emulation capability has fixed functionality and is permanently part of the CPU while the high performance real-time data exchange, advanced analysis, and trace functions are added outside of the core in most cases. The capabilities are individually selected for addition to a chip. The addition of emulation peripherals to the system design creates the second tier functionality. A cost-effective library of emulation peripherals contains the building blocks to create systems and permits the construction of advanced analysis, high performance real-time data exchange, and trace capabilities. In the preferred embodiment five standard debug configurations are offered, although custom configurations are also supported. The specific configurations are covered later herein.
The trace data for an application running on the CPU is referred to as background code. If there is an emulation halt, the application is paused. In the paused state, the following can happen. The data processor may receive a run command and resume running in the background code. The data processor may receive a single step command and run for only one cycle and pause again. The data processor may receive a real time interrupt (RTI). In this case the data processor will also run to service the real time interrupt. In this case this trace data is referred to as foreground trace data. The data processor sends an Interrupt During Suspend (IDS) bit to the emulation hardware. This IDS bit indicates to the emulation hardware that an interrupt happened while the data processor was suspended.
These and other aspects of this invention are illustrated in the drawings, in which:
The emulation user needs to be able to distinguish from the transmitted trace data when there is a real time interrupt while the program is paused. This would be in contrast to an interrupt that occurred while the program was still in background code. Otherwise the trace data would be misleading. The user could not determine when the trace data switched from tracing background code to tracing foreground code.
The size of the debug function and its associated capabilities for any particular embodiment of a system-on-chip may be adjusted by either deleting complete functions or limiting the number of event detectors and trigger builders deployed. Additionally, the trace function can be incrementally increased from program counter trace only to program counter and data trace along with ASIC and CPU generated data. The real-time data exchange function may also be optionally deployed. The ability to customize on-chip tools changes the application development paradigm. Historically, all chip designs with a given CPU core were limited to a fixed set of debug capability. Now, an optimized debug capability is available for each chip design. This paradigm change gives system architects the tools needed to manage product development risk at an affordable cost. Note that the same CPU core may be used with differing peripherals with differing pin outs to embody differing system-on-chip products. These differing embodiments may require differing debug and emulation resources. The modularity of this invention permits each such embodiment to include only the necessary debug and emulation resources for the particular system-on-chip application.
The real-time emulation debug infrastructure component is used to tackle basic debug and instrumentation operations related to application development. It contains all execution control and register visibility capabilities and a minimal set of real-time data exchange and analysis such as breakpoint and watchpoint capabilities. These debug operations use on-chip hardware facilities to control the execution of the application and gain access to registers and memory. Some of the debug operations which may be supported by real-time emulation are: setting a software breakpoint and observing the machine state at that point; single step code advance to observe exact instruction by instruction decision making; detecting a spurious write to a known memory location; and viewing and changing memory and peripheral registers.
Real-time emulation facilities are incorporated into a CPU mega-module and are woven into the fabric of CPU core 201. This assures designs using CPU core 201 have sufficient debug facilities to support debugger application program 110 baseline debug, instrumentation, and data transfer capabilities. Each CPU core 201 incorporates a baseline set of emulation capabilities. These capabilities include but are not limited to: execution control such as run, single instruction step, halt and free run; displaying and modifying registers and memory; breakpoints including software and minimal hardware program breakpoints; and watchpoints including minimal hardware data breakpoints.
Alignment is implemented in 2 steps. First, the data collected in early stages of the pipeline is aligned in a per case basis in order to account for the differences in the data collection behavior. This presents a simpler group of data to the second processing step. Heterogeneous stage aligner 510 performs this initial alignment step. Second, the data collected in the first step presents a single type of behavior. The 3-stage delay pipeline 530 aligns this data from the first stage as a group to the last arriving memory access read data (rd_data).
The point of collection of the last arriving memory access read data (rd_data) is the target point of alignment. In this example this point of collection is stage 5 of the pipeline (e5). As a first step towards the final alignment goal, the early arriving data is processed in various ways and aligned via heterogeneous stage aligner 510 to the second stage of the pipeline (e2). In order to be considered fully aligned to e2, the data should not be updated at the beginning of the clock cycle if the pipeline did not advance in the previous cycle. This is indicated by cpu_stall=1 in previous cycle. For the example illustrated in
The pipeline-flow control information (pctl) data group is collected in pipeline stage e1. This data has a data retention policy similar to the policy of any stage in the architectural pipeline. Thus all that is required to align pipeline-flow control information (pctl) to pipeline stage e2 is the single stage pipeline delay element 511. Pipeline delay element 511 is implemented by a single register stage that updates when the pipeline advances (cpu_stall=0).
A second set of early collected data is the program counter (pc). The program counter is generated in pipeline stage e0. The program counter is delayed 1 clock cycle via a single register stage (not shown) and then presented at the input of heterogeneous stage aligner 510 as the signal pc_e0+1 clock delay. Program counter (pc) data is aligned to pipeline stage e2 via a single register stage in pipeline delay element 512. Pipeline delay element 512 updates only when the pipeline advances (cpu_stall=0) and only if the current instruction in pipeline state e1 is a new instruction (inst_exe=1). OR gate 513 receives the cpu_stall signal and the inst_exe signal and insures pipeline delay element 512 advances only under these conditions. Enforcing these 2 conditions ensures that the aligned program counter (pc) value in pipeline stage e2 during multicycle instructions remains the same during all the cycles it takes to execute the instruction. This retention is in spite of the fact that the program counter (pc) retention policy will overwrite the program counter (pc) value presented after the first clock cycle of the instruction in pipeline stage e1.
The three remaining sets of early collected data are related to memory accesses. These are memory access control (mem_acc_ctl), memory access address (mem_addr) and memory access write data (wr_data). For the particular implementation illustrated in
Memory access elastic buffer 520 copes with these alternatives. Received memory access control data (mem_acc_ctl) supplies the input to two stage pipeline delay element 521, the input to multiplexer 522 and an input to elastic buffer control 523. The memory access address (mem_addr) and memory access write data (wr_data) supply the input to pipeline delay element 521 and multiplexer 522. It should be understood that the memory access control data (mem_acc_ctl), the memory access address (mem_addr) and memory access write data (wr_data) are handled in parallel in pipeline delay element 521 and multiplexer 522.
The memory access control data (mem_acc_ctl) indicates the pipeline stage of collection of the memory access signals. Elastic buffer control 523 uses this indication to control pipeline delay element 521 and multiplexer 522. If the memory access data was collected during pipeline stage e2, then elastic buffer control 523 sends a select signal to multiplexer 522 to select the directly received memory access signals. If the memory access data was collected during pipeline stage e1, then elastic buffer control 523 sends a select signal to multiplexer 522 to select memory access signals from pipeline delay element 521. Elastic buffer control 523 also controls pipeline delay element 521 to insert one pipeline stage delay. If the memory access data was collected during pipeline stage e0, then elastic buffer control 523 sends a select signal to multiplexer 522 to select memory access signals from pipeline delay element 521. Elastic buffer control 523 also controls pipeline delay element 521 to insert two pipeline stage delays. This behavior is summarized in Table 1.
The 3-stage delay pipeline 530 takes the homogeneously behaved data at its input already aligned to the second pipeline stage e2. Three-stage delay pipeline 530 includes pipeline delay element 531 for the memory access data, pipeline delay element 532 for the program counter data and pipeline delay element 533 for the pipeline-flow control information. Three-stage delay pipeline 530 outputs this data at pipeline stage e5. This is the same stage as the arrival of the read data (rd_data). Three-stage delay pipeline 530 sends every bit of input data through 3 serially connected registers that update its content every clock cycles that the pipeline is not stalled (cpu_stall=0). The clock signal clk1 is supplied to pipeline delay elements 511 and 512 and to every register of pipeline delay elements 521, 531, 532 and 533. The cpu_stall signal stalls pipeline delay elements 511, 512, 531, 532 and 533 when the central processing unit is stalled. Since the memory access data is not updated by heterogeneous stage aligner 510 during pipeline stall cycles, no data is lost during pipeline stalls. Pipeline flattener 501 effectively aligns the program counter (pc), pipeline-flow control information (pctl), memory access control (mem_acc_ctl), memory access address (mem_addr), memory access write data (wr_data) to the late received read data (rd_data) in pipeline stage e5.
In
In order to prevent the potential lost of the read data, additional registering stage is inserted in the path of the data. This one pipeline stage delay is implemented via pipeline delay elements 601, 602, 603, 604 and 605. The pipeline delay element 605 provides storage to capture the read data (rd_data) and eliminates the loss of read data associated with the instruction in pipeline state e5 being overwritten when the read data in pipeline stage e4 slides into pipeline stage e5 during the first cycle of a CPU stall window. Pipeline delay elements 601, 602, 603 and 604 do not hold data and have been added as delay elements to compensate for the delay of pipeline delay register 605, which captures and holds the read data. In order to remove the 1 clock slide in the read data, the hold signal supplied to pipeline delay register 605 is a 1 clock delayed version of the pipeline stall signal (cpu_stall) provided by delay element 606.
The correction to the N-bit sliding on the memory data is done via an N-bit slide operation in the opposite direction to the slide of the data. The data bus is assumed to be 2 words wide in this embodiment. The sliding of data at the input is limited to a swapping between the upper and lower words of the bus. Shift correction circuit 630 receives the memory access control signal and detects the sliding condition. Shift correction circuit 630 controls multiplexers 631, 632, 633, and 634 to enable or disable a swap of the most significant and least significant bits. In order to restore the architectural view of the data it is necessary to align the least significant bits of the write data and the read data to the least significant bits of the data bus. On a normal state of the multiplexer control signal from shift control circuit 630 multiplexer 631 selects the most significant bits from pipeline delay element 601 to output to the most significant bits of pipeline delay element 611, multiplexer 632 selects the least significant bits from pipeline delay element 601 output to the least significant bits of pipeline delay element 611, multiplexer 633 selects the most significant bits from pipeline delay element 605 to output to the most significant bits of pipeline delay element 615, multiplexer 634 selects the least significant bits from pipeline delay element 605 output to the least significant bits of pipeline delay element 611. In the opposite swap state multiplexer 631 selects the least significant bits from pipeline delay element 601 to output to the most significant bits of pipeline delay element 611, multiplexer 632 selects the most significant bits from pipeline delay element 601 output to the least significant bits of pipeline delay element 611, multiplexer 633 selects the least significant bits from pipeline delay element 605 to output to the most significant bits of pipeline delay element 615, multiplexer 634 selects the most significant bits from pipeline delay element 605 output to the least significant bits of pipeline delay element 611. This swaps the most significant bits with the least significant bits of both the write data and the read data.
If no special information is provided to the user regarding a real time interrupt in the trace stream, the user will not know that the interrupt occurred while CPU core 201 is paused. Therefore when the emulation hardware detects that the stream is switching to a real time interrupt it transmits special information in the trace stream. Thus the emultation hardware informs the user of this transition.
CPU core 201 sends an interrupt during suspend (IDS) bit to the emulation hardware upon detection of a real time interrupt during an emulation halt. The IDS bit indicates that an interrupt happened while CPU core 201 was suspended. Trace logic 403 uses this information to supply to the user data indicating the interrupt. This information also permits the user to distinguish between the background and foreground trace data. This is routed through the emulation logic as shown in
When the emulation halt occurs, a stall signal is transmitted to CPU core 201 and the emulation hardware such as trace collection 230. Pipeline flattener 401 and alignment logic 402 stall in response to this stall signal. Trace logic 403 is not held by this signal. The sequence of trace data from CPU core 201 from prior instructions are held in pipeline flattener 401. The trace data sequence is as follows:
1. Background trace data.
2. Background trace data held in pipeline flattener 401. The timing bits indicate stall cycles corresponding to the duration of the emulation halt.
3. The real time interrupt causes CPU core 201 to resume running. This real time interrupt is indicated by the IDS bit. This first flushes the remaining background trace data, which had been stuck in pipeline flattener 401 due to the stall.
4. When the IDS bit reaches trace logic 403, trace information is sent to the user (DATA OUT) indicating that the current trace data is from a real time interrupt.
5. The foreground trace data starts. The IDS bit must also propagated through pipeline flattener 401 even though CPU core 201 is already servicing the real time interrupt.
When the real time interrupt completes, CPU core 201 goes back to being stalled due to the original emulation halt. At the beginning of the stall, even though CPU core 201 has finished servicing the real time interrupt, the emulation logic produces the following trace data stream.
1. Foreground trace data.
2. Foreground trace data held in pipeline flattener 401. The IDS becomes zero. The timing bits indicate stall cycles that correspond to the duration of the emulation halt.
3. This may be followed by another real time interrupt.
4. If the foreground trace data may be followed not by a real time interrupt but a run command.
This technique enables user to distinguish between the two streams using the special information sent in the trace stream in response to the IDS bit.
This application is a divisional of U.S. patent application Ser. No. 10/301,968 filed Nov. 22, 2002 now U.S. Pat. No. 7,127,639.
| Number | Name | Date | Kind |
|---|---|---|---|
| 6185731 | Maeda et al. | Feb 2001 | B1 |
| 6260150 | Diepstraten et al. | Jul 2001 | B1 |
| 6324684 | Matt et al. | Nov 2001 | B1 |
| 6553513 | Swoboda et al. | Apr 2003 | B1 |
| 6557116 | Swoboda et al. | Apr 2003 | B1 |
| 6564339 | Swoboda et al. | May 2003 | B1 |
| 6567933 | Swoboda et al. | May 2003 | B1 |
| 7047178 | Agarwala | May 2006 | B2 |
| 7099817 | Agarwala et al. | Aug 2006 | B2 |
| 7237151 | Swoboda et al. | Jun 2007 | B2 |
| 7310749 | Swoboda et al. | Dec 2007 | B2 |
| 7392431 | Swoboda | Jun 2008 | B2 |
| 7444504 | Agarwala et al. | Oct 2008 | B2 |
| 7464018 | Agarwala et al. | Dec 2008 | B2 |
| 20030204707 | Liang | Oct 2003 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 20070011663 A1 | Jan 2007 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10301968 | Nov 2002 | US |
| Child | 11467746 | US |
