Abstract: Systems, apparatuses, and methods for implementing register compaction with early release are disclosed. A processor includes at least a command processor, a plurality of compute units, a plurality of registers, and a control unit. Registers are statically allocated to wavefronts by the control unit when wavefronts are launched by the command processor on the compute units. In response to determining that a first set of registers, previously allocated to a first wavefront, are no longer needed, the first wavefront executes an instruction to release the first set of registers. The control unit detects the executed instruction and releases the first set of registers to the available pool of registers to potentially be used by other wavefronts. Then, the control unit can allocate the first set of registers to a second wavefront for use by threads of the second wavefront while the first wavefront is still active.

Abstract: An apparatus and method for efficiently processing multiplication and accumulate operations for matrices in applications. In various implementations, a computing system includes a parallel data processing circuit and a memory. The memory stores the instructions (or translated commands) of a parallel data application. The circuitry of the parallel data processing circuit performs a matrix multiplication operation using source operands accessed only once from a vector register file and multiple instantiations of a vector processing circuit capable of performing multiple matrix multiplication operations corresponding to multiple different types of instructions. The multiplier circuit and the adder circuit of the vector processing circuit perform each of the fused multiply add (FMA) operation and the dot product (inner product) operation without independent, dedicated execution pipelines with one execution pipeline for the FMA operation and the other separate execution pipeline for the dot product operation.

Abstract: Systems, apparatuses, and methods for implementing register compaction with early release are disclosed. A processor includes at least a command processor, a plurality of compute units, a plurality of registers, and a control unit. Registers are statically allocated to wavefronts by the control unit when wavefronts are launched by the command processor on the compute units. In response to determining that a first set of registers, previously allocated to a first wavefront, are no longer needed, the first wavefront executes an instruction to release the first set of registers. The control unit detects the executed instruction and releases the first set of registers to the available pool of registers to potentially be used by other wavefronts. Then, the control unit can allocate the first set of registers to a second wavefront for use by threads of the second wavefront while the first wavefront is still active.

Abstract: Systems, apparatuses, and methods for efficiently processing arithmetic operations are disclosed. A computing system includes a processor capable of executing single precision mathematical instructions on data sizes of M bits and half precision mathematical instructions on data sizes of N bits, which is less than M bits. At least two source operands with M bits indicated by a received instruction are read from a register file. If the instruction is a packed math instruction, at least a first source operand with a size of N bits less than M bits is selected from either a high portion or a low portion of one of the at least two source operands read from the register file. The instruction includes fields storing bits, each bit indicating the high portion or the low portion of a given source operand associated with a register identifier specified elsewhere in the instruction.

Abstract: Systems, apparatuses, and methods for implementing register compaction with early release are disclosed. A processor includes at least a command processor, a plurality of compute units, a plurality of registers, and a control unit. Registers are statically allocated to wavefronts by the control unit when wavefronts are launched by the command processor on the compute units. In response to determining that a first set of registers, previously allocated to a first wavefront, are no longer needed, the first wavefront executes an instruction to release the first set of registers. The control unit detects the executed instruction and releases the first set of registers to the available pool of registers to potentially be used by other wavefronts. Then, the control unit can allocate the first set of registers to a second wavefront for use by threads of the second wavefront while the first wavefront is still active.

Abstract: Systems, apparatuses, and methods for maintaining separate pending load and store counters are disclosed herein. In one embodiment, a system includes at least one execution unit, a memory subsystem, and a pair of counters for each thread of execution. In one embodiment, the system implements a software based approach for managing dependencies between instructions. In one embodiment, the execution unit(s) maintains counters to support the software-based approach for managing dependencies between instructions. The execution unit(s) are configured to execute instructions that are used to manage the dependencies during run-time. In one embodiment, the execution unit(s) execute wait instructions to wait until a given counter is equal to a specified value before continuing to execute the instruction sequence.

Abstract: Systems, apparatuses, and methods for processing variable wavefront sizes on a processor are disclosed. In one embodiment, a processor includes at least a scheduler, cache, and multiple execution units. When operating in a first mode, the processor executes the same instruction on multiple portions of a wavefront before proceeding to the next instruction of the shader program. When operating in a second mode, the processor executes a set of instructions on a first portion of a wavefront. In the second mode, when the processor finishes executing the set of instructions on the first portion of the wavefront, the processor executes the set of instructions on a second portion of the wavefront, and so on until all portions of the wavefront have been processed. The processor determines the operating mode based on one or more conditions.

Abstract: Systems, apparatuses, and methods for routing traffic between clients and system memory are disclosed. A computing system includes a processor capable of executing single precision mathematical instructions on data sizes of M bits and half precision mathematical instructions on data sizes of N bits, which is less than M bits. At least two source operands with M bits indicated by a received instruction are read from a register file. If the instruction is a packed math instruction, at least a first source operand with a size of N bits less than M bits is selected from either a high portion or a low portion of one of the at least two source operands read from the register file. The instruction includes fields storing bits, each bit indicating the high portion or the low portion of a given source operand associated with a register identifier specified elsewhere in the instruction.

Abstract: Systems, apparatuses, and methods for implementing a stream processor with overlapping execution are disclosed. In one embodiment, a system includes at least a parallel processing unit with a plurality of execution pipelines. The processing throughput of the parallel processing unit is increased by overlapping execution of multi-pass instructions with single pass instructions without increasing the instruction issue rate. A first plurality of operands of a first vector instruction are read from a shared vector register file in a single clock cycle and stored in temporary storage. The first plurality of operands are accessed and utilized to initiate multiple instructions on individual vector elements on a first execution pipeline in subsequent clock cycles. A second plurality of operands are read from the shared vector register file during the subsequent clock cycles to initiate execution of one or more second vector instructions on the second execution pipeline.

Abstract: Systems, apparatuses, and methods for maintaining separate pending load and store counters are disclosed herein. In one embodiment, a system includes at least one execution unit, a memory subsystem, and a pair of counters for each thread of execution. In one embodiment, the system implements a software based approach for managing dependencies between instructions. In one embodiment, the execution unit(s) maintains counters to support the software-based approach for managing dependencies between instructions. The execution unit(s) are configured to execute instructions that are used to manage the dependencies during run-time. In one embodiment, the execution unit(s) execute wait instructions to wait until a given counter is equal to a specified value before continuing to execute the instruction sequence.

Abstract: Systems, apparatuses, and methods for processing variable wavefront sizes on a processor are disclosed. In one embodiment, a processor includes at least a scheduler, cache, and multiple execution units. When operating in a first mode, the processor executes the same instruction on multiple portions of a wavefront before proceeding to the next instruction of the shader program. When operating in a second mode, the processor executes a set of instructions on a first portion of a wavefront. In the second mode, when the processor finishes executing the set of instructions on the first portion of the wavefront, the processor executes the set of instructions on a second portion of the wavefront, and so on until all portions of the wavefront have been processed. The processor determines the operating mode based on one or more conditions.

Abstract: An apparatus and methods for hardware-based performance monitoring of a computer system are presented. The apparatus includes: processing units; a memory; a connector device connecting the processing units and the memory; probes inserted the processing units, and the probes generating probe signals when selected processing events are detected; and a thread trace device connected to the connector device. The thread trace device includes an event interface to receive probe signals, and an event memory controller to send probe event messages to the memory, where probe event messages are based on probe signals. The probe event messages transferred to memory can be subsequently analyzed using a software program to determine, for example, thread-to-thread interactions.

Abstract: Embodiments of the present invention provide systems, methods, and computer program products for improving divergent conditional branches in code being executed by a processor. For example, in an embodiment, a method comprises detecting a conditional statement of a program being simultaneously executed by a plurality of threads, determining which threads evaluate a condition of the conditional statement as true and which threads evaluate the condition as false, pushing an identifier associated with the larger set of the threads onto a stack, executing code associated with a smaller set of the threads, and executing code associated with the larger set of the threads.

Abstract: Methods and apparatuses are provided for power control in a processor. The apparatus comprises a plurality of operational units arranged as a group of operational units. A power consumption monitor determines when cumulative power consumption of the group of operational units exceeds a threshold (e.g., either or both of the cumulative power threshold and the cumulative power rate threshold) during a time interval, after which a filter for issuing instructions to the group of operational units suspends instruction issuance to the group of operational units for the remainder of the time interval. The method comprises monitoring cumulative power consumption by a group of operational units within a processor over a time interval. If the cumulative power consumption of the group of operational units exceeds the threshold, instruction issuance to the group of operational units is suspended for the remainder of the time interval.

Abstract: Method, system, and computer program product embodiments for synchronizing workitems on one or more processors are disclosed. The embodiments include executing a barrier skip instruction by a first workitem from the group, and responsive to the executed barrier skip instruction, reconfiguring a barrier to synchronize other workitems from the group in a plurality of points in a sequence without requiring the first workitem to reach the barrier in any of the plurality of points.

Abstract: The present invention provides a system for handling extra contexts for shader constants, and applications thereof. In an embodiment there is provided a computer-based method for executing a series of compute packets in an execution pipeline. The execution pipeline includes a first plurality of registers configured to store state-updates of a first type and a second plurality of registers configured to store state-updates of a second type. A first number of state-updates of the first type and a second number of state-updates of the second type are respectively identified and stored in the first and second plurality of registers. A compute packet is sent to the execution pipeline responsive to the first number and the second number. Then, the compute packet is executed by the execution pipeline.

Abstract: Methods and apparatuses are provided for power control in a processor. The apparatus comprises a plurality of operational units arranged as a group of operational units. A power consumption monitor determines when cumulative power consumption of the group of operational units exceeds a threshold (e.g., either or both of the cumulative power threshold and the cumulative power rate threshold) during a time interval, after which a filter for issuing instructions to the group of operational units suspends instruction issuance to the group of operational units for the remainder of the time interval. The method comprises monitoring cumulative power consumption by a group of operational units within a processor over a time interval. If the cumulative power consumption of the group of operational units exceeds the threshold, instruction issuance to the group of operational units is suspended for the remainder of the time interval.

Abstract: Method, system, and computer program product embodiments for synchronizing workitems on one or more processors are disclosed. The embodiments include executing a barrier skip instruction by a first workitem from the group, and responsive to the executed barrier skip instruction, reconfiguring a barrier to synchronize other workitems from the group in a plurality of points in a sequence without requiring the first workitem to reach the barrier in any of the plurality of points.

Abstract: An apparatus and methods for hardware-based performance monitoring of a computer system are presented. The apparatus includes: processing units; a memory; a connector device connecting the processing units and the memory; probes inserted the processing units, and the probes generating probe signals when selected processing events are detected; and a thread trace device connected to the connector device. The thread trace device includes an event interface to receive probe signals, and an event memory controller to send probe event messages to the memory, where probe event messages are based on probe signals. The probe event messages transferred to memory can be subsequently analyzed using a software program to determine, for example, thread-to-thread interactions.

Abstract: Embodiments of the present invention provide systems, methods, and computer program products for improving divergent conditional branches in code being executed by a processor. For example, in an embodiment, a method comprises detecting a conditional statement of a program being simultaneously executed by a plurality of threads, determining which threads evaluate a condition of the conditional statement as true and which threads evaluate the condition as false, pushing an identifier associated with the larger set of the threads onto a stack, executing code associated with a smaller set of the threads, and executing code associated with the larger set of the threads.