Abstract: Systems and methods are disclosed for error correction control (ECC) for a memory device comprising a data portion and an ECC portion, the memory device coupled to a system on a chip (SoC). The SoC includes an ECC cache. On receipt of a request to write a line of data to the memory, a determination is made if the data is compressible. If so, the data line is compressed. ECC bits are generated for the compressed or uncompressed data line. A determination is made if an ECC cache line is associated with the received data line. If the data line is compressible, the ECC bits are appended to the compressed data line and the appended data line is stored in the data portion of the memory. Otherwise, the ECC bits are stored in the ECC cache and the data line is stored in the data portion of the memory.

Abstract: Certain aspects of the present disclosure provide methods and apparatus for operating a wearable display device. Certain aspects of the present disclosure provide a method for operating a wearable display device. The method includes determining a position of the wearable display device based on a motion sensor. The method includes rendering, by a graphics processing unit, an image based on the determined position. The method includes determining a first updated position of the wearable display device based on the motion sensor. The method includes warping, by a warp engine, a first portion of the rendered image based on the first updated position. The method includes displaying the warped first portion of the rendered image on a display of the wearable display device.

Abstract: Certain aspects of the present disclosure provide methods and apparatus for operating a wearable display device. Certain aspects of the present disclosure provide a method for operating a wearable display device. The method includes determining a position of the wearable display device based on a motion sensor. The method includes rendering, by a graphics processing unit, an image based on the determined position. The method includes determining a first updated position of the wearable display device based on the motion sensor. The method includes warping, by a warp engine, a first portion of the rendered image based on the first updated position. The method includes displaying the warped first portion of the rendered image on a display of the wearable display device.

Abstract: Systems and methods are disclosed for implementing error correction control regions (ECC) in a memory device without the need to ECC protect the entire memory device. An exemplary method comprises defining one or more ECC regions in a memory device, the memory device coupled to a system on a chip (SoC). An ECC block is provided on the SoC, the ECC block in communication with the one or more ECC regions in the memory device. A determination is made with the ECC block whether to store data in a first of the one or more ECC regions. Responsive to the determination ECC bits are generating for, and interleaved with, the received data and interleaved ECC bits and data are caused to be written to the first ECC region. Otherwise, received data is sent to a non-ECC region of the memory device.

Abstract: Systems and methods are disclosed for implementing error correction control regions (ECC) in a memory device without the need to ECC protect the entire memory device. An exemplary method comprises defining one or more ECC regions in a memory device, the memory device coupled to a system on a chip (SoC). An ECC block is provided on the SoC, the ECC block in communication with the one or more ECC regions in the memory device. A determination is made with the ECC block whether to store data in a first of the one or more ECC regions. Responsive to the determination ECC bits are generating for, and interleaved with, the received data and interleaved ECC bits and data are caused to be written to the first ECC region. Otherwise, received data is sent to a non-ECC region of the memory device.

Abstract: Systems, methods, and computer programs are disclosed for reducing motion-to-photon latency and memory bandwidth in a virtual reality display system. An exemplary method involves receiving sensor data from one or more sensors tracking translational and rotational motion of a user for a virtual reality application. An updated position of the user is computed based on the received sensor data. The speed and acceleration of the user movement may be computed based on the sensor data. The updated position, the speed, and the acceleration may be provided to a warp engine configured to update a rendered image before sending to a virtual reality display based on one or more of the updated position, the speed, and the acceleration.

Abstract: Systems and methods are disclosed for error correction control (ECC) for a memory device comprising a data portion and an ECC portion, the memory device coupled to a system on a chip (SoC). The SoC includes an ECC cache. On receipt of a request to write a line of data to the memory, a determination is made if the data is compressible. If so, the data line is compressed. ECC bits are generated for the compressed or uncompressed data line. A determination is made if an ECC cache line is associated with the received data line. If the data line is compressible, the ECC bits are appended to the compressed data line and the appended data line is stored in the data portion of the memory. Otherwise, the ECC bits are stored in the ECC cache and the data line is stored in the data portion of the memory.

Abstract: One disclosed computing system comprises a x86 processor, memory, a PCIe root complex (RC), a PCIe bus, and an interconnect chip having a PCIe endpoint (EP) that is connected to the PCIe RC through a PCIe link, the PCIe EP being connected to an AMBA® bus. The interconnect chip may communicate with the IO device via the AMBA® bus in an AMBA® compliant manner and communicate with the host system in a PCIe compliant manner. This communication may include receiving a command from the processor, sending the command to the IO device over the AMBA® bus, receiving a response from the IO device over the AMBA® bus, and sending over the AMBA® bus and the PCIe link one or more DMA operations to the memory. Further communication may include sending an IOAPIC interrupt to the processor of the host system according to PCIe ordering rules.

Abstract: One disclosed computing system comprises a x86 processor, memory, a PCIe root complex (RC), a PCIe bus, and an interconnect chip having a PCIe endpoint (EP) that is connected to the PCIe RC through a PCIe link, the PCIe EP being connected to an AMBA® bus. The interconnect chip may communicate with the IO device via the AMBA® bus in an AMBA® compliant manner and communicate with the host system in a PCIe compliant manner. This communication may include receiving a command from the processor, sending the command to the IO device over the AMBA® bus, receiving a response from the IO device over the AMBA® bus, and sending over the AMBA® bus and the PCIe link one or more DMA operations to the memory. Further communication may include sending an IOAPIC interrupt to the processor of the host system according to PCIe ordering rules.

Abstract: A system and method of accounting for lost clock cycles in a microprocessor. A method includes detecting a first reason which prevents exit of an entry from an instruction retirement queue, and incrementing a first count corresponding to the first reason, wherein the first count is incremented while the first reason prevents exit of the entry from the queue. A first point in time is determined when said first reason no longer prevents exit of the entry from the queue. A second reason which prevents exit of the entry from the queue is detected, wherein the second reason came into existence prior to said first point in time. A second count corresponding to the second reason is incremented, wherein incrementing the second count begins at the first point in time.

Abstract: A processor core includes an instruction decode unit that may dispatch a same integer instruction stream to a plurality of integer execution units operating in lock-step. The processor core also includes signature generation logic that may generate, concurrently with execution of the integer instructions, a respective signature from result signals conveyed on respective result buses in one or more pipeline stages within each of the integer execution units in response to the result signals becoming available. The processor core also includes compare logic that may detect a mismatch between signatures from each of the integer execution units. Further, in response to the compare logic detecting any mismatch, the compare logic may cause instructions causing the mismatch to be re-executed.

Abstract: A processor core includes an instruction decode unit that may dispatch a same integer instruction stream to a plurality of integer execution units operating in lock-step. The processor core also includes signature generation logic that may generate, concurrently with execution of the integer instructions, a respective signature from result signals conveyed on respective result buses in one or more pipeline stages within each of the integer execution units in response to the result signals becoming available. The processor core also includes compare logic that may detect a mismatch between signatures from each of the integer execution units. Further, in response to the compare logic detecting any mismatch, the compare logic may cause instructions causing the mismatch to be re-executed.

Abstract: A processor core includes an instruction decode unit that may dispatch a same integer instruction stream to a plurality of integer execution units and may consecutively dispatch a same floating-point instruction stream to a floating-point unit. The integer execution units may operate in lock-step such that during each clock cycle, each respective integer execution unit executes the same integer instruction. The floating-point unit may execute the same floating-point instruction stream twice. Prior to the integer instructions retiring, compare logic may detect a mismatch between execution results from each of the integer execution units. In addition, prior to the results of the floating-point instruction stream transferring out of the floating-point unit, the compare logic may also detect a mismatch between results of execution of each consecutive floating-point instruction stream. Further, in response to detecting any mismatch, the compare logic may cause instructions causing the mismatch to be re-executed.

Abstract: A system and method of accounting for lost clock cycles in a microprocessor. A method includes detecting a first reason which prevents exit of an entry from an instruction retirement queue, and incrementing a first count corresponding to the first reason, wherein the first count is incremented while the first reason prevents exit of the entry from the queue. A first point in time is determined when said first reason no longer prevents exit of the entry from the queue. A second reason which prevents exit of the entry from the queue is detected, wherein the second reason came into existence prior to said first point in time. A second count corresponding to the second reason is incremented, wherein incrementing the second count begins at the first point in time.

Abstract: A processor is provided that implements a replay mechanism to recover from soft errors. The processor includes a protected execution unit, a check unit to detect errors in results generated by the protected execution unit, and a replay unit to track selected instructions issued to the protected execution unit. When the check unit detects an error, it triggers the replay unit to reissue the selected instructions to the protected execution unit. One embodiment of the replay unit provides an instruction buffer that includes pointers to track issue and retirement status of in-flight instructions. When the check unit indicates an error, the replay unit resets a pointer to reissue the instruction for which the error was detected.

Abstract: The present invention relates to the design of highly reliable high performance microprocessors, and more specifically to designs that use cache memory protection schemes such as, for example, a 1-hot plus valid bit scheme and a 2-hot vector cache scheme. These protection schemes protect the 1-hot vectors used in the tag array in the cache and are designed to provide hardware savings, operate at higher speeds and be simple to implement. In accordance with an embodiment of the present invention, a tag array memory including an input conversion circuit to receive a 1-hot vector and to convert the 1-hot vector to a 2-hot vector. The tag array memory also including a memory array coupled to the input conversion circuit, the memory array to store the 2-hot vector; and an output conversion circuit coupled to the memory array, the output conversion circuit to receive the 2-hot vector and to convert the 2-hot vector back to the 1-hot vector.

Abstract: A computer system includes processor having dual execution cores and a non-volatile memory that stores an error recovery routine. The processor's execution cores operate in lock step when the processor is in a redundant execution mode, and they operate independently when the processor is in a split execution mode. The error recovery routine is invoked when the processor detects a soft error while operating in the redundant execution mode. The error recovery routine switches the processor to split execution mode. In split mode, each execution core saves uncorrupted processor state data to a designated memory location and updates any corrupted data with corresponding processor state data from the other execution core. The error recovery routine returns the processor to redundant mode, initializes each execution core with the recovered processor state data, and returns control of the processor to the program thread that was executing when the soft error was detected.

Abstract: A signaling mechanism associated with errors in a processor are promoted or demoted based on a set of stored values.

Abstract: The present invention relates to the design of highly reliable high performance microprocessors, and more specifically to designs using a blind invalidate circuit in high-speed memories. In accordance with an embodiment of the present invention, a tag array memory circuit including a plurality of memory bit circuits coupled together to form an n-bit memory cell; and a blind invalidate circuit coupled to a memory bit circuit in the n-bit memory cell, the blind invalidate circuit to clear a bit in the memory bit circuit, if a primary clear bit line is asserted and a received bit value of a right-adjacent memory bit circuit is zero.

Abstract: The present invention relates to the design of highly reliable high performance microprocessors, and more specifically to designs that use cache memory protection schemes such as, for example, a 1-hot plus valid bit scheme and a 2-hot vector cache scheme. These protection schemes protect the 1-hot vectors used in the tag array in the cache and are designed to provide hardware savings, operate at higher speeds and be simple to implement. In accordance with an embodiment of the present invention, a tag array memory including an input conversion circuit to receive a 1-hot vector and to convert the 1-hot vector to a 2-hot vector. The tag array memory also including a memory array coupled to the input conversion circuit, the memory array to store the 2-hot vector; and an output conversion circuit coupled to the memory array, the output conversion circuit to receive the 2-hot vector and to convert the 2-hot vector back to the 1-hot vector.