Abstract: Systems, apparatuses, and methods for efficiently allocating data in a cache are described. In various embodiments, a processor decodes an indication in a software application identifying a temporal data set. The data set is flagged with a data set identifier (DSID) indicating temporal data to drop after consumption. When the data set is allocated in a cache, the data set is stored with a non-replaceable attribute to prevent a cache replacement policy from evicting the data set before it is dropped. A drop command with an indication of the DSID of the data set is later issued after the data set is read (consumed). A copy of the data set is not written back to the lower-level memory although the data set is removed from the cache. An interrupt is generated to notify firmware or other software of the completion of the drop command.

Abstract: Techniques are disclosed relating to circuitry configured to perform reciprocal-based floating-point division. In some embodiments, floating-point circuitry includes reciprocal circuitry configured to generate a reciprocal of a divisor, multiplication circuitry configured to multiply the reciprocal results with a dividend, and circuitry configured to clear a least significant bit of an integer representation of the multiplication output to generate a modified multiplication output. The floating-point circuitry may be configured to convert the modified multiplication output to a representation using the first precision to generate a division output. In some embodiments, the refinement using the integer representation may provide correctly-rounded subnormal division results. The disclosed techniques may improve accuracy, reduce processing time, and/or reduce instructions needed for floating-point division, with little to no increase in chip area.

Abstract: Systems, apparatuses, and methods for efficiently allocating data in a cache are described. In various embodiments, a processor decodes an indication in a software application identifying a temporal data set. The data set is flagged with a data set identifier (DSID) indicating temporal data to drop after consumption. When the data set is allocated in a cache, the data set is stored with a non-replaceable attribute to prevent a cache replacement policy from evicting the data set before it is dropped. A drop command with an indication of the DSID of the data set is later issued after the data set is read (consumed). A copy of the data set is not written back to the lower-level memory although the data set is removed from the cache. An interrupt is generated to notify firmware or other software of the completion of the drop command.

Abstract: Systems, apparatuses, and methods for efficiently allocating data in a cache are described. In various embodiments, a processor decodes an indication in a software application identifying a temporal data set. The data set is flagged with a data set identifier (DSID) indicating temporal data to drop after consumption. When the data set is allocated in a cache, the data set is stored with a non-replaceable attribute to prevent a cache replacement policy from evicting the data set before it is dropped. A drop command with an indication of the DSID of the data set is later issued after the data set is read (consumed). A copy of the data set is not written back to the lower-level memory although the data set is removed from the cache. An interrupt is generated to notify firmware or other software of the completion of the drop command.

Abstract: Systems, apparatuses, and methods for efficiently allocating data in a cache are described. In various embodiments, a processor decodes an indication in a software application identifying a temporal data set. The data set is flagged with a data set identifier (DSID) indicating temporal data to drop after consumption. When the data set is allocated in a cache, the data set is stored with a non-replaceable attribute to prevent a cache replacement policy from evicting the data set before it is dropped. A drop command with an indication of the DSID of the data set is later issued after the data set is read (consumed). A copy of the data set is not written back to the lower-level memory although the data set is removed from the cache. An interrupt is generated to notify firmware or other software of the completion of the drop command.

Abstract: Techniques are disclosed relating to flushing cache lines. In some embodiments, a graphics processing unit includes a cache and one or more storage elements configured to store a plurality of command buffers that include instructions executable to manipulate data stored in the cache. In some embodiments, ones of the cache lines in the cache are configured to store data to be operated on by instructions in the command buffers and a first tag portion that identifies a command buffer that has stored data in the cache line. In some embodiments, the graphics processing unit is configured to receive a request to flush cache lines that store data of a particular command buffer, and to flush ones of the cache lines having first tag portions indicating the particular command buffer as having data stored in the cache lines while maintaining data stored in other ones of the cache lines as valid.

Abstract: In one embodiment, a system controls a duty cycle of a processor (e.g. a graphics processing unit (GPU)) to manage power. The GPU may include a snoop indicator that may control whether or not the GPU captures and/or responds to communications from another processor (e.g. a central processing unit (CPU) that executes a driver associate with the GPU). A snooze indicator may control whether or not the GPU is automatically repowered at the start of the next frame, or is repowered only if the communication indicating additional work has been received. In an embodiment, the GPU mode controls discussed above may permit the GPU firmware executed with the GPU itself to control duty cycle power down, independent of the driver executing on the CPU.

Abstract: In one embodiment, a system controls a duty cycle of a processor (e.g. a graphics processing unit (GPU)) to manage power. The GPU may include a snoop indicator that may control whether or not the GPU captures and/or responds to communications from another processor (e.g. a central processing unit (CPU) that executes a driver associate with the GPU). A snooze indicator may control whether or not the GPU is automatically repowered at the start of the next frame, or is repowered only if the communication indicating additional work has been received. In an embodiment, the GPU mode controls discussed above may permit the GPU firmware executed with the GPU itself to control duty cycle power down, independent of the driver executing on the CPU.

Abstract: In an embodiment, a processor that includes multiple cores may implement a power/performance-efficient stop mechanism for power gating. One or more first cores of the multiple cores may have a higher latency stop than one or more second cores of the multiple cores. The power control mechanism may permit continued dispatching of work to the second cores until the first cores have stopped. The power control mechanism may prevent dispatch of additional work once the first cores have stopped, and may power gate the processing in response to the stopping of the second cores. Stopping a core may include one or more of: requesting a context switch from the core or preventing additional work from being dispatched to the core and permitting current work to complete normally. In an embodiment, the processor may be a graphics processing unit (GPU).