A method, system, and computer program product synchronize a group of workitems executing an instruction stream on a processor. The processor is yielded by a first workitem responsive to a synchronization instruction in the instruction stream. A first one of a plurality of program counters is updated to point to a next instruction following the synchronization instruction in the instruction stream to be executed by the first workitem. A second workitem is run on the processor after the yielding.

An apparatus to facilitate thread scheduling is disclosed. The apparatus includes logic to store barrier usage data based on a magnitude of barrier messages in an application kernel and a scheduler to schedule execution of threads across a plurality of multiprocessors based on the barrier usage data.

A data processing apparatus and method are provided for executing a plurality of threads. Processing circuitry performs processing operations required by the plurality of threads, the processing operations including a lock-protected processing operation with which a lock is associated, where the lock needs to be acquired before the processing circuitry performs the lock-protected processing operation. Baton maintenance circuitry is used to maintain a baton in association with the plurality of threads, the baton forming a proxy for the lock, and the baton maintenance circuitry being configured to allocate the baton between the threads. Via communication between the processing circuitry and the baton maintenance circuitry, once the lock has been acquired for one of the threads, the processing circuitry performs the lock-protected processing operation for multiple threads before the lock is released, with the baton maintenance circuitry identifying a current thread amongst the multiple threads for which the lock-protected processing operation is to be performed by allocating the baton to that current thread. The baton can hence be passed from one thread to the next, without needing to release and re-acquire the lock. This provides a significant performance improvement when performing lock-protected processing operations across multiple threads.

Disclosed herein is an apparatus which comprises a plurality of execution units, and a first general register file (GRF) communicatively couple to the plurality of execution units, wherein the first GRF is shared by the plurality of execution units.

A method for allocating registers in a compute unit of a vector processor includes determining a maximum number of registers that are to be used concurrently by a plurality of threads of a kernel at the compute unit. The method further includes setting a mode of register allocation at the compute unit based on a comparison of the determined maximum number of registers and a total number of physical registers implemented at the compute unit.

One embodiment provides for a parallel processor comprising a processing array within the parallel processor, the processing array including multiple compute blocks, each compute block including multiple processing clusters configured for parallel operation, wherein each of the multiple compute blocks is independently preemptable. In one embodiment a preemption hint can be generated for source code during compilation to enable a compute unit to determine an efficient point for preemption.

Techniques to dispatch threads of a graphics kernel for execution to increase the interval between dependent threads and the associated are disclosed. The dispatch interval may be increased by dispatching associated threads, followed by threads without any dependencies, followed by threads dependent on the earlier dispatched associated threads. As such, the interval between dependent threads and their associated threads can be increased, leading to increased parallelism.

An apparatus to facilitate thread scheduling is disclosed. The apparatus includes logic to store barrier usage data based on a magnitude of barrier messages in an application kernel and a scheduler to schedule execution of threads across a plurality of multiprocessors based on the barrier usage data.

Embodiments described herein provide a technique to decompose 64-bit per-lane virtual addresses to access a plurality of data elements on behalf of a multi-lane parallel processing execution resource of a graphics or compute accelerator. The 64-bit per-lane addresses are decomposed into a base address and a plurality of per-lane offsets for transmission to memory access circuitry. The memory access circuitry then combines the base address and the per-lane offsets to reconstruct the per-lane addresses.

Various examples are directed to systems and methods for executing a loop in a reconfigurable compute fabric. A first flow controller may initiate a first thread at a first synchronous flow to execute a first portion of a first iteration of the loop. A second flow controller may receive a first asynchronous message instructing the second flow controller to initiate a first thread at a second synchronous flow to execute a second portion of the first iteration. The second flow controller may determine that the first iteration of the loop is the last iteration of the loop to be executed and initiate the first thread at the second synchronous flow with a last iteration flag set.