The present invention relates generally to graphics processing. More particularly, this invention relates to graphics scaling operations.
As more and more media processors are coupled through a common memory system to share graphics content for performing graphics operations separately, it is becoming more and more expensive to allow one media processor, such as a graphics processing unit (GPU), to consume the result produced by another media processor. Usually, these media processors may be capable of handling 2D, 3D, Video and/or display command streams driven by API (application programming interface) calls from a graphics application. A host processor such as a central processing unit (CPU) is typically required to explicitly synchronize these graphics operations by stopping one media processor and starting another one when switching between APIs. Such synchronization is usually very costly and prevents simultaneous, parallel execution of a separate media processors.
In particular, a media processor may be driven by a system on a chip requiring a host processor included in the chip to respond to an interrupt signifying the completion of individual hardware graphics operations such as a single copy operation or a single solid color fill operation. However, responding to a single interrupt per operation can be expensive. Furthermore, interrupts may prevent simultaneous execution of media processors and a host processor. Consequently, the overall performance is degraded with lower parallelism.
On the other hand, multiple media processors and a display device coupled with a common memory system may require synchronization. For example, more than one component of a graphics content may arrive asynchronously from separate media processors to be displayed in a display device for a single application. Executing a graphics command to display the graphics content may depend on when each component is properly rendered and ready to display. Apparently, it is necessary to ensure maximum parallelism among multiple media processors to allow rendering different components for the same graphics content in a synchronous manner.
Additionally, parallel operations between a host processor and coupled media processors may be limited by a bottleneck introduced when deleting commonly shared graphics resources. Typically, media processor drivers ensure the media processors are idle prior to deleting graphics resources such as allocated memories, memory management unit (MMU) entries, the textures etc., that might otherwise be in use by pending graphics operations. This, however, prevents parallel operations by the host processor and media processors.
Furthermore, graphics rendering operations such as scaling may be limited by a fixed number of fractional precision in media processor hardware for arithmetic representations. Often, the bit-precision of the scale factor is limited to optimize mathematical operations inside the media processor hardware to allow a multiplication instead of a division which may be more expensive. For example, a scale factor may be inversely represented through a limited bit-precision fixed-point arithmetic. As a result, a limiting factor is introduced that certain scale factors cannot be represented accurately.
An embodiment of the present invention includes a method and apparatus that determine an up scale factor and a down scale factor according to a scale factor received from a graphics application program interface (API) to scale a graphics data in a graphics processing unit (GPU). The up scale factor and the down scale factor may be precisely stored in the GPU based on a fixed number of bits. An actual scale factor which can be precisely stored in the GPU corresponding to the scale factor may differ from the scale factor with a difference. Graphics commands may be sent to the GPU to scale the graphics data according to the up scale factor and the down scale factor separately. A combined scale factor corresponding to a combination of the up scale factor and the down scale factor may differ from the scale factor less then the difference between the actual scale factor and the scale factor.
Other features of the present invention will be apparent from the accompanying drawings and from the detailed description that follows.
The present invention is illustrated by way of example and not limitation in the figures of the accompanying drawings, in which like references indicate similar elements and in which:
A method and an apparatus for processing graphics commands in graphics processors are described herein. In the following description, numerous specific details are set forth to provide thorough explanation of embodiments of the present invention. It will be apparent, however, to one skilled in the art, that embodiments of the present invention may be practiced without these specific details. In other instances, well-known components, structures, and techniques have not been shown in detail in order not to obscure the understanding of this description.
Reference in the specification to “one embodiment” or “an embodiment” means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the invention. The appearances of the phrase “in one embodiment” in various places in the specification do not necessarily all refer to the same embodiment.
The processes depicted in the figures that follow, are performed by processing logic that comprises hardware (e.g., circuitry, dedicated logic, etc.), software (such as is run on a general-purpose computer system or a dedicated machine), or a combination of both. Although the processes are described below in terms of some sequential operations, it should be appreciated that some of the operations described may be performed in different order. Moreover, some operations may be performed in parallel rather than sequentially.
According to one embodiment, graphics driver 155 fetches graphics commands from a command stream buffer (not shown) that receives the graphics commands from applications 151-153. In response to the received graphics commands, the graphic driver 155 may determine the resources required by the graphics commands and their relationships such as graphics resources dependencies. The graphics driver 155 may send multiple graphics commands to a GPU if there is no resource conflict among the graphics commands; otherwise, the graphics driver 155 may send one command at a time to a GPU, dependent upon the availability of hardware resources of the GPU. For example, graphics driver 155 may send multiple 2D commands to a GPU. However, graphics driver 155 can only send one 3D command at a time or alternatively, any 2D commands would not be sent until a previously sent 3D command has been executed.
In addition, graphics driver 155 further includes a resource management functionality to keep track what graphics commands have been sent to a GPU and how many of those commands have been executed by the GPU. The graphics driver 155 further keep track of the graphics resources (e.g., memory) used by the commands. In one embodiment, the graphics resources used by a graphics command would not be freed or released until the graphics command has been executed by the GPU.
Further the graphics driver 155 and/or graphics APIs 154 may communicate with display controller or display driver 159, for example, via an interrupt. In one embodiment, each of the GPUs 156-157 processes the graphics data concurrently and asynchronously. For example, multiple GPUs may process multiple components of a video frame concurrently. However, the video frame would not be displayed or updated (e.g., refresh) until all of the components have been processed by the GPUs. The communications between the graphics APIs 154 and graphics driver 155 with the display controller 159 would allow the display controller 159 to synchronize, for example, via a swap command or a swap ID, all of the processed components before display the video frame in the display 160.
Serializing Command Streams For Graphics Processors
According to certain embodiments of the invention, graphics commands may be serialized to allow one or more media processors to process graphics data in an optimal and efficient manner. Each media processor may be a dedicated hardware chip to, for example, process graphics data according to a stream of graphics commands. A driver for a media processor may serialize graphics commands for the media processor to execute. In one embodiment, a stream of graphics commands may be generated from multiple graphics applications via one or more graphics APIs (Application Programming Interface). The stream of graphics commands received in a media processor may include multiple types of graphics commands, such as 2D graphics commands or 3D graphics commands. In one embodiment, serializing the stream of graphics commands may allow the media processor to execute graphics commands of the same type in an order in which these graphics commands have been received without requiring synchronization interrupts. In another embodiment, serializing graphics commands may prevent one graphics API from running while another is constructing a new graphics command stream. Multiple media processors may execute serialized graphics command streams substantially concurrently with minimum synchronization interrupts from an associated host processor.
Referring back to
According to one embodiment, 2D, 3D, GL and notification graphics commands may be stored in a command stream buffer 101. One or more graphics commands may be arranged in the command stream buffer 101 according to a sequence of graphics API calls from applications. A command stream buffer 101 may be associated with a graphics processing unit 119 through a graphics driver module 117. For example, a graphics API may request for a command stream buffer. In response to the request, the associated graphics driver allocates a segment of memory or buffer and passes the handle or pointer back to the graphics API such that the graphics API can store the graphics commands received from an application in the allocated command stream buffer. A graphics command generator may be coupled with more than one command stream buffer associated with more than one graphics processing units. In one embodiment, the graphics processing unit 119 may be a media or graphics processor separate from the host processor hosting the command stream buffer 101. The graphics driver module 117 may retrieve graphics commands from the command stream buffer 101 in an FIFO (First In First Out) order as stored in the command stream buffer 101. In one embodiment, a graphics driver module 117 may schedule processing a retrieved graphics command according to a header data, such as header data 203 in
A command fetching module 309 in the graphics driver module 117 may retrieve graphics commands from the command stream buffer 101 according to an order the graphics commands have been stored, such as FIFO. In one embodiment, the command fetching module 309 may retrieve the next graphics command from the command stream buffer according to a status of the associated graphics processing unit 119 and/or dependency conditions identified from previously received graphics commands. The status of a graphics processing unit may include an indication on the availability of associated hardware resources, such as a hardware FIFO buffer for receiving processing instructions. In one embodiment, the status of a graphics processing unit may be obtained based on values of predetermined memory locations, such as a status register associated with the graphics processing unit. The command fetching module 309 may proceed fetching the next graphics command as instructed by a dependency checking module 307.
A dependency checking module 307 may extract header data from a retrieved graphics command to determine if an associated graphics resource has been allocated in a memory coupled to the graphics processing unit 119. In one embodiment, a memory for allocating graphics resources may be coupled to both the graphics processing unit 119 and the host processor hosting the graphics driver module 117. The dependency checking module may notify a memory management module to allocate required graphics resources. In one embodiment, the dependency checking module may compare the graphics instructions carried by the retrieved graphics command with graphics instructions from previously retrieved graphics commands to determine if a dependency condition exists for the current graphics command. In one embodiment, forwarding graphics instructions of a 2D graphics command to a graphics processing unit may depend on when graphics instructions of a previous 3D graphics command are completely executed in the same graphics processing unit.
A command dispatching module 327 may forward hardware graphics commands based on graphics instructions from a retrieved graphics command to the graphics processing unit 119. In one embodiment, the command dispatching module 327 may send the graphics processing unit 119 with additional kernel level data associated with the hardware graphics commands, such as interrupt addresses or callback addresses. The command dispatching module 327 may send notification messages to a graphics display driver module 125 for a retrieved graphics notification command. In one embodiment, a notification message may inform the graphics display driver module 125 readiness of one condition for displaying a frame buffer. The command dispatching module may not initiate sending commands or messages if corresponding dependency conditions are not satisfied according to a dependency checking module 307. In other embodiments, a command dispatching module may generate a synchronization instruction appended to hardware graphics commands for a processing unit to notify a graphics driver module the completion of executing the hardware graphics commands. The notification may signify that a hardware buffer for an IO port of a graphics processing unit is emptied.
In one embodiment, a graphics processing unit 119 may include IO ports 323 coupled to graphics core logics, such as, for example, a 2D graphics processing module 305 and a 3D graphics processing module 303. IO ports 323 may buffer hardware graphics commands received. The size of a hardware buffer associated with IO ports 323 may limit the maximum number of hardware graphics commands allowed at a time. In one embodiment, a 2D graphics processing module and a 3D graphics processing module may be based on separate hardware components for a graphics processing unit 119. IO ports 323 may forward 2D and 3D graphics instructions received from hardware graphics commands to 2D graphics processing module 305 and 3D graphics processing module 303 respectively. In one embodiment, a 2D graphics processing module 305 and a 3D graphics processing module 303 may not operate at the same time in a graphics processing unit 119. In another embodiment, a 2D graphics processing module may process multiple 2D graphics instructions from one or more 2D graphics commands at one time (e.g. a batch). In other embodiments, a 3D graphics processing module may not process 3D graphics instructions from another 3D graphics command before finishing processing 3D graphics instructions for a current 3D graphics command (e.g. one graphics command at a time).
At block 403, in one embodiment, process 400 may extract header data from a currently retrieved graphics command. A header data in a graphics command, such as header data 203 of
At block 405, in one embodiment, process 400 may determine if there are dependency conditions associated with a currently retrieved graphics command. When dependency conditions are satisfied or removed, process 400 may proceed to send hardware graphics commands or notification messages associated with a graphics command. Dependency conditions for a retrieved graphic command may be determined based on header data included in the retrieved graphic command and/or previous retrieved graphics commands. For example, if it is determined that a current command depends on a previous command (e.g., graphics resource dependency), the current command will be held off until the previous command has been processed by the hardware. Otherwise, both current and previous commands may be sent down to the hardware for executions. In a particular embodiment, when a first command is a 3D graphics command and a second command received subsequently is a 2D graphics command, the second command will not be sent to the hardware until the first command has been processed by the hardware. In another particular embodiment, if both first and second commands are 2D graphics commands, both the first and second commands may be sent to the hardware as a batch for execution, where the hardware may buffer both commands during the execution. Other configurations may exist dependent upon a specific design of a graphics processing unit.
At block 507, process 500 may determine if a currently retrieved graphics command depends on previously retrieved graphics commands. A command dependency condition may exist for a currently retrieved graphics command depending on a previously retrieved graphics command being completely processed 509. In one embodiment, sending a current command to a graphics processing unit or another graphics driver module may require a graphics processing unit to complete processing previously retrieved graphics commands. For example, a currently retrieved graphics command may have a dependency on a previous retrieved 3D graphics command. A currently retrieved 3D graphics command may have a dependency condition on a previously retrieved 2D graphics command. A currently retrieved graphics command may have a dependency on a previously retrieved OpenGL graphics command. At block 511, process 500 may determine a capacity dependency condition 513 on an associated graphics processing unit if there are graphics instructions included in a currently retrieved graphics command to send to the graphics processing unit. In one embodiment, a capacity dependency condition may be associated with available processing capacity of a graphics processing unit. A status of processing capacity may be related to number of empty slots of a hardware buffer for an IO port in a graphics processing unit, such as IO port 323 of
Referring back to
At block 411, when the determined dependency conditions are satisfied or removed, process 400 may proceed to send hardware graphics commands based on a currently retrieved graphics command to an associated graphics processing unit. In one embodiment, process 400 may send a notification message to another graphics driver according to a header data extracted from a currently retrieved graphics command. Process 400 may retrieve next graphics command from a graphics command buffer at block 413 and continue at block 403 to process the next graphics command. In one embodiment, a stream of 2D graphics commands may be continuously retrieved from a command buffer to be forwarded to a graphics processing unit subject to the capacity of a graphics processing unit if associated graphics resources are resident.
Batch Processing Graphics Operations
In one embodiment, a single notification such as an interrupt may be generated for a batch of multiple graphics commands to reduce an expensive processing overhead requiring one interrupt per graphics command. Tracking which graphics commands have been submitted to determine when these graphics commands are completed may provide information to guide memory management system to free graphics resource no longer in use. Additionally, whether a graphics processing unit is idle may be determined to track which graphics resources can be freed or released when they are no longer in use.
A hardware command queue 601 may be coupled to a graphics processor 607 in a graphics processing unit, such as graphics processing unit 119 of
In one embodiment, a graphics resource status data 613 may provide a store for status information about a graphics resource, such as time stamps 617 and flags 619. Time stamps 617 may store values indicating whether a corresponding graphics resource is currently being operated on. Flags 619 may indicate if there are outstanding requests made against a corresponding graphics resource, such as a request to delete from a process initiated by a user. A graphics resource status data 613 may be updated by a graphics driver module 117. In one embodiment, time stamps 617 of a graphics resource may be updated by a graphics driver module 117 when submitting hardware graphics commands for a corresponding graphics command associated with the graphics resource. In other embodiments, time stamps 617 of a graphics resource may be updated by a graphics driver module 117 when a graphics processor 607 has completed processing graphics instructions corresponding to the graphics command associated with a graphics resource. A graphics resource handler module 615 may manage graphics resources, such as allocating, mapping and deleting corresponding memory space. In one embodiment, a graphics driver module 117 may update a graphics resource status data 613 via a graphics resource handler module 615.
In another embodiment, process 700 may generate a hardware synchronization command when a hardware command queue is full, such as hardware command queue 601 of
At block 705, according to one embodiment, process 700 may submit a batch of hardware graphics commands followed by a hardware synchronization command to a graphics processing unit for execution or processing. In one embodiment, process 700 may submit a batch of hardware graphics commands together with the hardware synchronization command to a graphics processing unit in one single transaction. At block 707, in one embodiment, process 700 may proceed to generate a next batch of hardware graphic commands from subsequent graphics commands in a command stream buffer, such as the command stream buffer at block 701. At block 709, process 700 may generate a next hardware synchronization command for the next batch of hardware graphics commands.
In one embodiment, at block 711, process 700 may receive a single notification message subsequent to complete execution of a batch of hardware graphics commands in a graphics processing unit. In another embodiment, process 700 may receive a single notification message when a hardware command queue associated with a graphics processing unit is emptied. A notification message may be generated based on a hardware interrupt caused by a graphics processing unit when finishing executing the batch of graphics commands. In one embodiment, a graphics processing unit may execute a hardware synchronization command to perform a hardware interrupt to a host processor for sending a notification message. Process 700 may be performed in a host processor. At block 713, process 713 may determine the next batch of hardware graphics commands are ready to be submitted to a graphics processing unit based on a received notification message. In one embodiment, the received notification message may signify that a graphics processing unit is idle after finishing executing pending hardware graphics commands. Process 700 may submit a next batch of hardware graphics commands and a next generated hardware synchronization command to the graphics processing unit for execution at block 713.
At block 807, in one embodiment, process 800 may receive a notification message after the hardware graphics commands corresponding to a retrieved graphics command have been executed in a graphics processing unit. Process 800 may determine an association between a notification message and a retrieved graphics command. In one embodiment, process 800 may track submitted graphics command to determine the association. At block 809, process 800 may update a graphics resource status associated with a graphics resource subsequent to receiving a notification message. In one embodiment, process 800 may increment the value of a complete counter in a graphics resource status by one at block 809. At block 811, process 800 may determine if a graphic resource can be freed or released based on an associated graphics resource status. In one embodiment, process 800 may compare the values of a submit counter and a complete counter included in a graphics resource status to determine whether all graphics commands sent to the hardware have been executed or completed. Process 800 may determine an associated graphics resource can be released when the values of the submit counter and the complete counter are equal in the associated graphics resource status. When the values of the submit counter and complete counter are equal, that means all of the commands that have been sent to the hardware have been executed by the hardware. As a result, all graphics resources associated with those commands are no longer in use and these graphics resources can be released for use by other subsequent commands.
At time instance Time2903, a batch of hardware graphics commands are submitted to the hardware command buffer including 2D′b, 2D′a and 2D′a 923 corresponding to a stream of 2D graphics commands 2Db, 2Da and 2Da 909 from the command buffer. Additionally, a hardware synchronization command Syn 939 is submitted to the hardware command buffer following the last hardware graphics command 2D′b in the batch. The graphics resource status for graphics resource a 919 associated with two graphics commands 2Das′ is updated by incrementing its submit count CountS by 2 to be of a value 2 according to the submission. The graphics resource status for graphics resource b 921 associated with one graphics command 2Db is updated by incrementing its submit count CountS by 1 to be of a value 8 according to the submission.
At time instance Time3905, the graphics processing unit have executed the batch of hardware graphics commands and the hardware synchronization command 923 from the hardware command buffer. Executing the hardware synchronization command Syn 939 has resulted in a notification message signifying a completion of processing the stream of 2D graphics commands 2Db, 2Da and 2Da for updating associated graphics resource statuses. In one embodiment, the graphics resource status for graphics resource a 925 associated with two graphics commands 2Das′ is updated by incrementing its complete count CountC by 2 to be of a value 2 according to the notification. The graphics resource status for graphics resource b 927 associated with one graphics command 2Db is updated by incrementing its complete count CountC by 1 to be of a value 8 according to the notification.
At time instance Time4907, a next batch of hardware graphics command is submitted to the hardware command buffer including a single hardware graphic command 3D′c 937 corresponding to one 3D graphics command 3Dc 931 from the command buffer. Additionally, a newly generated hardware synchronization command Syn 935 is submitted to the hardware command buffer following hardware graphics command 3D′c. The graphics resource status for graphics resource c 941 associated with the graphics command 3Dc is updated by incrementing its submit count CountS by 1 to be of a value 1 according to the submission.
As described above, referring back to
Once the graphics commands have been sent to the hardware, graphics driver 117 still maintains the command stream buffer 101 that stores the graphics commands that have been sent to the hardware and stamped by the graphics driver 17. Subsequently, a notification is received by the graphics driver from the hardware, for example, via an interrupt via a host processor, indicating that the hardware has completed executing the graphics commands. In response to the notification, the graphics driver may update the complete counter to indicate the number of graphics commands that have been executed. In addition, the graphics driver revisits the graphics commands stored in the command stream buffer 101, where these graphics commands by now have been executed by the hardware. For each of those graphics commands that have been executed, the graphics resources associated with those commands may be released using the information maintained by graphics resource handler module 615. In this way, the graphics resources would not be released until the associated graphics commands have been processed. This is typically useful when an application or a graphics API requests for releasing certain graphics resources while the associated graphics commands are still being processed by the hardware, which will be described in details further below.
Asynchronous Notifications for Concurrent Graphics Processing
In one embodiment, requests to swap graphics data to one or more display devices may be constructed to ensure segments of graphics data are ready for display prior to swapping the graphics data. Each segment of graphics data may be rendered by separate media processors running substantially concurrently and/or asynchronously. A synchronization point may be provided to ensure all segments of the graphics data are displayed simultaneously in a graphics display (e.g. refresh). Content completion notifications that arrive asynchronously to a swap request may allow maximum parallelism among media processors and/or display devices which operate asynchronously in parallel for display the graphics data.
As described above, a video frame may include multiple components which may require different graphics resources and processed by different graphics rendering engines or processing units. In one embodiment, each component may be processed by a different graphics processing unit. Thus, multiple graphics processing units may process multiple components of a video frame substantially concurrently and asynchronously. The output of the graphics processing units are stored in graphics memory 1001. However, the processing speeds of different graphics processing units may be different dependent upon the specific graphics commands and/or graphics resources. As a result, the output data generated from different graphics processing units may arrive at the graphics memory 1001 at different time. According to one embodiment of the invention, a synchronization signal, such as a swap ID, is transmitted to the display controller 125 to allow the display controller 125 to collect all of the output data of all components of a video frame, based on the synchronization signal such as a swap ID, from all graphics processing units before display the video frame having all of the components ready in a display. In this embodiment, a swap ID is used to identify a particular video frame and each graphics components of the video frame include certain information linked with the swap ID. When display controller 125 is notified by each of the graphics processing units indicating that a respective component of the video frame identified by the swap ID has been completed, the display controller 125 can examine each component based on the swap ID to determine whether all components are ready to be displayed. If so, the video frame is displayed by the display controller 125; otherwise, the display controller 125 may wait until all of the components of the video frame have been completed.
At block 1105, process 1100 may generate a swap command via a swap API call to display a graphics data in a display device. The swap command may include a swap ID and one or more notification IDs′ in association with one or more notification commands at block 1101. In one embodiment, a swap command may be associated with a graphics display unit, such as the display device 127 of
At block 1207, according to one embodiment, process 1200 may wait for a notification message caused by a graphics processing unit executing a hardware synchronization command. In one embodiment, a notification message may be caused by a hardware interrupt according to a hardware synchronization command. Subsequently, at block 1209, process 1200 may generate a driver notification message including a swap ID and a notification ID associated with a notification command. In one embodiment, a notification command may be associated with a target graphics driver, such as display driver module 125 of
Graphics commands G 1403, G 1409 and notification command N21411 may be sent to the driver for a graphics processing unit, graphics gpu 1457, to draw GUI graphics data during time period 1429. Graphics gpu 1457 may be the 2D/3D graphics processing unit 1003 of
Notification commands N11407 and N21411 may be associated with two of the notification IDs associated with swap command S 1401. In one embodiment, the driver for special GPU 1459 may send a driver notification messages corresponding to notification commands N11407 subsequent to completing the rotation operation 1431. The driver for graphics GPU 1457 may send a driver notification message corresponding to notification command N21411 subsequent to completing drawing GUI graphics data according to graphics commands G 1403 and G 1409 at 1465. In one embodiment, LCD driver 1461 may determine a graphics data, e.g. in a frame buffer, is ready to display at 1465 because each notification ID associated with the swap ID for swap command S 1401 has a corresponding driver notification message received. The graphics data may include the GUI components drawn during time period 1429 and a video frame rated during time period 1467. When receiving video synchronization signal Vsync at time instance 1453, LCD driver 1461 may display the corresponding graphics data. In one embodiment, LCD driver 1461 may not display a graphics data corresponding to the swap ID of swap command S 1401 when receiving Vsync signal at time instance 1451 because the driver notification command associated with notification command N21411 has not been received.
In one embodiment, graphics command S 1413 may be a swap command associated with a display driver, such as TV display driver 1463. Swap command S 1413 may be associated with at least two notification IDs. TV display driver 1463 may receive swap command S 1413 to wait for at least two driver notification messages associated with the notification IDs included in the swap command S 1413 during time period 1441. Graphics commands G 1415 and notification command N21421 may be sent to the driver for graphics GPU 1457 to draw GUI graphics data during time period 1435. Graphics command Z 1417 and notification command N11419 may be sent to a driver for special GPU 1459 to perform zooming operation on a video frame during time period 1437. Notification commands N11419 and N21421 may be associated with two of the notification IDs associated with swap command S 1413. In one embodiment, TV display driver 1463 may determine a graphics data is ready to display at 1449 after receiving both driver notification messages associated with notification IDs of notification commands N11419 and N21421. TV display driver 1463 may display a graphics data corresponding to swap command 1413 when receiving the next video synchronization signal Vsync at 1455 after 1449.
Graphics Resources Management
In one embodiment, asynchronous deletion operations may be performed based on detecting if graphics processing units are idle or not to allow a host processor to immediately return to take on additional tasks after making a deleting request. Graphics processing units may be ensured to be idle when graphics resources are deleted while allowing parallel operations between the graphics processing units and a coupled host processor.
A graphics driver module 117 may include a resource management module 1509 for managing graphics resources allocated according to graphics commands retrieved from a command stream buffer 101. A resource management module may notify a memory management module to allocate required graphics resourced associated with a graphics command. In one embodiment, a graphics resource may be associated with a graphics resource data 613 including time stamps 617 and flags 619, using certain techniques described above. Time stamps 617 for a graphics resource may include a complete counter and submit counter as in
According to certain embodiments, the submit counter and complete counter may be used to identify which of the graphics commands have been sent to the hardware and which of the graphics commands have been completed by the hardware. The finish flag and delete flag may be used to indicate whether a specific graphics command has been completed and whether the graphics resources associated with the commands can be released or deleted. This typically useful when a request is received from an application or a graphics API registering a notification when certain graphics commands have been completed (e.g., finished) and/or requesting certain graphics resources can be released (e.g., delete).
For example, according to certain embodiments, when a request is received from an application or a graphics API for registering a notification call back when a specific graphics command has been completed, the graphics driver may check based on the finish flag of the requested command to determine whether the command has been completed. If the finish flag has been set which indicates that the associated command has been completed, the graphics driver may immediately return the request indicating that the command has been completed.
However, if the finish flag has not be set which indicates that the command has not been completed or finished by the hardware, the graphics driver may delay returning the request, for example, by marking the request as “pending”. Subsequently, when the graphics driver receives a notification from the hardware indicating that the previous command or previous batch of commands have been completed, the graphics driver will “walk through”, for example, in the command stream buffer, those commands that have been sent to the hardware and mark those commands as “finished” by setting the corresponding finish mark. Thereafter, a maintenance routine or a garbage collection routine maintained by the graphics driver may “clean up” the pending request for finish notifications by checking the finish flag of the commands and return the pending requests to indicate the caller (e.g., application or graphics API) that the requested command has been “finished.”
In another scenarios, according to one embodiment, when an application or a graphics API sends down a request for deleting or releasing certain graphics resources, the graphics driver may check whether the requested resources can be deleted or released using the graphics resource status data. If there is no graphics command pending to be finished that is associated with the requested graphics resources, the requested resources can be immediately deleted or released. However, if there is at least one graphics command associated with the requested graphics resources and is still pending (e.g., the corresponding “finish” flag has not been set), the requested graphics resources cannot be deleted or released. In this situation, the graphics driver may delay returning the request by marking the request as “pending” and set the “delete” flag of the requesting graphics resources to indicate that the associated graphics resources have been requested to be deleted or released.
Subsequently, when the graphics driver receives a notification from the hardware indicating that a previous command or a previous batch of commands has been completed. The graphics driver marks those commands as “finished” by setting the corresponding “finished” flag. In addition, graphics driver may check whether there are any outstanding graphics resources that require to be cleaned up by examining the delete pool 1511. For each graphics resource that has been marked in a delete flag, if the associated graphics command or commands have been executed indicated via the corresponding finish flag, the marked graphics resources are deleted or released. Other configurations apparent to those with ordinary skills in the art may exist.
At block 1605, in one embodiment, process 1600 may determine if there are graphics processing units busy processing graphics data after determining there are no associated outstanding graphics commands. In one embodiment, one or more graphics processing units may be coupled with a graphics memory storing the graphics resource to be deleted, such as the graphics memory 1001 of
At block 1603, in one embodiment, process 1600 may set a Finish flag associated with a graphics resource associated with a management request received at block 1601, if the management request is to wait for finishing processing the graphics resource instead of deleting the graphic resource as determined at block 1615. A Finish flag, when set, may signify a request has been made for a process, such as process 1600, to be notified when all outstanding graphics commands associated with a graphics resource have been processed. In one embodiment, a Finish flag associated with a graphics resource may include an identification of a process waiting to be notified. At block 1613, process 1600 may wait for a wake up notification message when all outstanding graphics commands associated with a graphics resource have been processed.
In one embodiment, the wake up notification message may be based on hardware interrupts generated from graphics processing units finishing processing outstanding graphics commands associated with the graphics resource. A graphics processing unit may notify a host processor which performs process 1600 when a hardware interrupt occurs. When waken up, process 1600 may return from the management request to the calling application at block 1615.
At block 1705, process 1700 may update time stamps associated with a graphics resource, such as time stamps 617 of
In one embodiment, if there are no more outstanding graphics commands for a graphics resource, for example when both the associated complete counter and submit counter have the same value, process 1700 may send a wake up notification message to a graphics application waiting for finishing graphics operations on the graphics resource at block 1713. One or more graphics processing units may perform graphics operations on a command graphics resource. In one embodiment, the status data associated with the graphics resource, such as graphics resource status data 613 of
Subsequent to checking the Finish flag, in one embodiment, process 1700 may retrieve a value of a Delete flag associated with the graphics resource, such as from Flags 619 of the graphics resource status data 613 of
Asymmetric Two-Pass Graphics Data Scaling
In one embodiment, a two-pass scaling approach utilizing a graphics operation memory and an extra scaling operation may provide an actual scaling effect more desirable than a single scaling operation based on a desired scaling factor which is not representable because of limited bit-precision in a graphics scaling unit. A larger number of discrete values may be explored when scaling up with an up scale factor greater than 1 to arrive at an effective scale factor closer to the desired scale factor than a single scaling operation based on a limited representation of the desired scale factor in the graphics scaling unit.
A graphics scaling unit 1801 may generate a scaled graphics data 1811 from an input graphics data 1809. A graphics memory, such as graphics memory 1001 of
At block 1909, in one embodiment, process 1900 may determine an up scale factor and a down scale factor which can be precisely represented under the limited number of fractional precisions in the graphics scaling unit to better represent the desired scale factor at block 1901, which may not be precisely represented based on the same number of fractional precisions in the graphics scaling unit. In one embodiment, process 1900 may compare a difference between the desired scale factor and a value by multiplying the up scale factor and the down scale factor against a scaling error determined at block 1907. Process 1900 may determine an up scale factor and a down scale factor such that the difference of block 1907 is smaller than the scaling error. In another embodiment, process 1900 may determine a maximum memory requirement based on an up scale factor, a down scale factor, and a size of the source texture. Process 1900 may determine an up scale factor and a down scale factor such that the maximum memory requirement does not exceed the size of the graphics operating memory.
In one embodiment, process 1900 may employ a binary search scheme among on a plurality of possible up scale factors precisely represented according to a limited fractional precision of arithmetic representation. Process 1900 may determine, based on each selected up scale factor, among all possible numbers precisely represented under the same limited fractional precision of arithmetic representation, a corresponding down scale factor such that multiplication of the selected up scale factor and the corresponding down scale factor has a value closest to the desired scale factor of block 1901. In one embodiment, process 1900 may select an up scale factor such that a combined scale factor with the corresponding down scale factor most closely (or more precisely) representing the desired scale factor among the plurality of possible up scale factors searched.
An up scale factor may be applied in a first pass graphics scaling operation. The corresponding down scale factor may be applied in a second pass graphics scaling operation. A combined scale factor may be based on multiplying an up scale factor and its corresponding down scale factor. In one embodiment, process 1900 may perform a binary search subject to a constraint that the maximum operating memory requirement based on an up scale factor, a corresponding down scale factor and a size of the source texture does not exceed the size of a graphics operating memory associated with the graphics scaling unit. In one embodiment, process 1900 may determine a size requirement for a graphics operating memory based on a maximum scaling error required to scale a source texture according to a limited fractional precision of arithmetic representations for an up scale factor and the corresponding down scale factor.
In one embodiment, at block 1911, process 1900 may generate a first graphics command to scale the source texture to a temporary texture according to an up scale factor. A temporary texture and a source texture may be stored in a graphics memory coupled to a graphics scaling unit, such as graphics memory 1001 of
In a single pass scaling, a desired scale factor 0.8 2009, which is binary represented with infinite fractional precisions as 0.1100110011001100 . . . 2001, has an actual value of 0.78125 2013 with 5 binary fractional precisions with a binary representation as 0.11001 2015. Thus, the single pass scaling for desired scale factor 0.8 may have a scaling error of 0.01875. An up scale factor 0.5 2017 for a first pass scaling may be precisely represented with 5 fractional precisions as 0.10000 for binary representation. Thus, the actual up scale factor 2021 with a binary representation 2023 are the same as the up scale factor 2017, 2019. A down scale factor 1.59375 2029 may be selected by representing a desired down scale factor 1.6 2025 for the second pass scaling based on 5 binary fractional precision representation 1.10011 2031. The desired down scale factor 1.6 2025 may be binary represented with infinite fractional precisions as 1.1001100110011001 . . . 2027. The actual combined scale factor based on the up scale factor 2021 and a down scale factor 2029 may be obtained as 0.796875, which corresponds to a smaller scaling error of 0.003125 than 0.01875.
Example of Data Processing System
This digital interface may include a graphical user interface which is similar to that shown on a typical computer, such as, for example, a Macintosh computer when running OS X operating system software. The system 2100 also includes a communication interface (e.g., wired or wireless communication interface) 2103, such as, for example, one or more wireless transceivers to communicate with another system or device. A wireless transceiver may be a WiFi transceiver, an infrared (IR) transceiver, a Bluetooth transceiver, and/or a wireless cellular telephony transceiver. It will be appreciated that additional components, not shown, may also be part of the system 2100 in certain embodiments, and in certain embodiments fewer components than shown in
The data processing system 2100 also includes one or more input devices 2113 which are provided to allow a user to provide input to the system. These input devices may be a keypad or a keyboard or a touch panel or a multi touch panel. Alternatively, input devices 2113 may include a voice interactive interface that can receive and interact with a voice command. The data processing system 2100 also includes an optional input/output device 2115 which may be a connector for a dock. It will be appreciated that one or more buses, not shown, may be used to interconnect the various components as is well known in the art. The data processing system shown in
At least certain embodiments of the inventions may be part of a digital media player, such as a portable music and/or video media player, which may include a media processing system to present the media, a storage device to store the media and may further include a radio frequency (RF) transceiver (e.g., an RF transceiver for a cellular telephone) coupled with an antenna system and the media processing system. In certain embodiments, media stored on a remote storage device may be transmitted to the media player through the RF transceiver. The media may be, for example, one or more of music or other audio, still pictures, or motion pictures.
The portable media player may include a media selection device, such as a click wheel input device on an iPod® or iPod Nano® media player from Apple Computer, Inc. of Cupertino, Calif., a touch screen input device, pushbutton device, movable pointing input device or other input device. The media selection device may be used to select the media stored on the storage device and/or the remote storage device. The portable media player may, in at least certain embodiments, include a display device which is coupled to the media processing system to display titles or other indicators of media being selected through the input device and being presented, either through a speaker or earphone(s), or on the display device, or on both display device and a speaker or earphone(s). Examples of a portable media player are described in published U.S. patent application numbers 2003/0095096 and 2004/0224638, both of which are incorporated herein by reference. Other configurations may exist.
Portions of what was described above may be implemented with logic circuitry such as a dedicated logic circuit or with a microcontroller or other form of processing core that executes program code instructions. Thus processes taught by the discussion above may be performed with program code such as machine-executable instructions that cause a machine that executes these instructions to perform certain functions. In this context, a “machine” may be a machine that converts intermediate form (or “abstract”) instructions into processor specific instructions (e.g., an abstract execution environment such as a “virtual machine” (e.g., a Java Virtual Machine), an interpreter, a Common Language Runtime, a high-level language virtual machine, etc.), and/or, electronic circuitry disposed on a semiconductor chip (e.g., “logic circuitry” implemented with transistors) designed to execute instructions such as a general-purpose processor and/or a special-purpose processor. Processes taught by the discussion above may also be performed by (in the alternative to a machine or in combination with a machine) electronic circuitry designed to perform the processes (or a portion thereof) without the execution of program code.
An article of manufacture may be used to store program code. An article of manufacture that stores program code may be embodied as, but is not limited to, one or more memories (e.g., one or more flash memories, random access memories (static, dynamic or other)), optical disks, CD-ROMs, DVD ROMs, EPROMs, EEPROMs, magnetic or optical cards or other type of machine-readable media suitable for storing electronic instructions. Program code may also be downloaded from a remote computer (e.g., a server) to a requesting computer (e.g., a client) by way of data signals embodied in a propagation medium (e.g., via a communication link (e.g., a network connection)).
The preceding detailed descriptions are presented in terms of algorithms and symbolic representations of operations on data bits within a computer memory. These algorithmic descriptions and representations are the tools used by those skilled in the data processing arts to most effectively convey the substance of their work to others skilled in the art. An algorithm is here, and generally, conceived to be a self-consistent sequence of operations leading to a desired result. The operations are those requiring physical manipulations of physical quantities. Usually, though not necessarily, these quantities take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, and otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to these signals as bits, values, elements, symbols, characters, terms, numbers, or the like.
It should be kept in mind, however, that all of these and similar terms are to be associated with the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless specifically stated otherwise as apparent from the above discussion, it is appreciated that throughout the description, discussions utilizing terms such as “processing” or “computing” or “calculating” or “determining” or “displaying” or the like, refer to the action and processes of a computer system, or similar electronic computing device, that manipulates and transforms data represented as physical (electronic) quantities within the computer system's registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices.
The present invention also relates to an apparatus for performing the operations described herein. This apparatus may be specially constructed for the required purpose, or it may comprise a general-purpose computer selectively activated or reconfigured by a computer program stored in the computer. Such a computer program may be stored in a computer readable storage medium, such as, but is not limited to, any type of disk including floppy disks, optical disks, CD-ROMs, and magnetic-optical disks, read-only memories (ROMs), RAMs, EPROMs, EEPROMs, magnetic or optical cards, or any type of media suitable for storing electronic instructions, and each coupled to a computer system bus.
The processes and displays presented herein are not inherently related to any particular computer or other apparatus. Various general-purpose systems may be used with programs in accordance with the teachings herein, or it may prove convenient to construct a more specialized apparatus to perform the operations described. The required structure for a variety of these systems will be evident from the description below. In addition, the present invention is not described with reference to any particular programming language. It will be appreciated that a variety of programming languages may be used to implement the teachings of the invention as described herein.
The foregoing discussion merely describes some exemplary embodiments of the present invention. One skilled in the art will readily recognize from such discussion, the accompanying drawings and the claims that various modifications can be made without departing from the spirit and scope of the invention.
Number | Name | Date | Kind |
---|---|---|---|
7151863 | Bradley et al. | Dec 2006 | B1 |
7619639 | Azar et al. | Nov 2009 | B1 |
20040183817 | Kaasila | Sep 2004 | A1 |
20070230827 | Haukijarvi et al. | Oct 2007 | A1 |
Number | Date | Country | |
---|---|---|---|
20080303846 A1 | Dec 2008 | US |