Processor pipeline for interlocked data transfer operations with variable latency
Disclosed embodiments provide techniques for instruction execution with a processor pipeline for data transfer operations. A processor core is accessed. The processor core executes one or more instructions out of order. The processor core supports integer operations and floating-point operations. An instruction in the processor core is decoded. The instruction is a data transfer operation. The data transfer operation necessitates a floating-point operation and an integer operation. The floating-point operation and the integer operation are dispatched to one or more issue queues. The floating-point operation and the integer operation are interlocked. The interlocking is accomplished using at least one entry in the one or more issue queues. A first operation of the floating-point operation and the integer operation is executed. A second operation of the floating-point operation and the integer operation is executed. The execution of the second operation is based on the interlocking.
1 . A processor-implemented method for instruction execution comprising:
accessing a processor core, wherein the processor core executes one or more instructions out of order, and wherein the processor core supports integer operations and floating-point operations, wherein the integer operations and the floating-point operations include various latency operations, wherein the various latency operations include variable latency operations;
decoding an instruction in the processor core, wherein the instruction comprises a data transfer operation, and wherein the data transfer operation necessitates a floating-point operation and an integer operation;
dispatching the floating-point operation and the integer operation to one or more issue queues;
interlocking the floating-point operation and the integer operation, wherein the interlocking is accomplished using at least one entry in the one or more issue queues, wherein a request/grant protocol is employed for interlocking the variable latency operations;
executing a first operation of the floating-point operation and the integer operation; and
executing a second operation of the floating-point operation and the integer operation, based on the interlocking.
2 . The method of claim 1 wherein the one or more issue queues comprise a floating-point issue queue and an integer issue queue.
3 . The method of claim 1 wherein a latency of the various latency operations determines one or more fields in corresponding entries of a floating-point issue queue and an integer issue queue.
4 . The method of claim 1 wherein the at least one entry in the one or more issue queues includes one or more of a companion instruction queue identification, a dependency bit, and an operation type.
5 . The method of claim 3 wherein the companion instruction queue identification, the dependency bit, and/or the operation type enable the interlocking at an instruction level.
6 . The method of claim 1 wherein the at least one entry in the one or more issue queues includes a reorder buffer identification and/or an operand validity indication.
7 . The method of claim 6 wherein the reorder buffer identification and/or the operand validity indication enable interlocking at an out-of-order instruction level.
8 . The method of claim 1 wherein the interlocking enables marking the second operation eligible for issuing.
9 . The method of claim 8 wherein the interlocking enables issuing an operation from an instruction queue to an execution unit.
10 . The method of claim 9 wherein the interlocking enables a source operand for the second operation to be obtained from the first operation.
11 . The method of claim 1 wherein the first operation comprises a floating-point operation and the second operation comprises an integer operation.
12 . The method of claim 11 wherein the first operation is executed in a floating-point unit and the second operation is executed in an arithmetic logic unit.
13 . The method of claim 12 wherein data resulting from the first operation is transferred directly from the floating-point unit to the arithmetic logic unit.
14 . The method of claim 12 wherein data resulting from the first operation is transferred from the floating-point unit to the arithmetic logic unit using temporary register file storage.
15 . The method of claim 1 wherein the first operation comprises an integer operation and the second operation comprises a floating-point element operation.
16 . The method of claim 15 wherein the first operation is executed in an arithmetic logic unit and the second operation is executed in a floating-point unit.
17 . The method of claim 16 wherein data resulting from the first operation is transferred directly from the arithmetic logic unit to the floating-point unit.
18 . The method of claim 16 wherein data resulting from the first operation is transferred from the arithmetic logic unit to the floating-point unit using temporary register file storage.
19 . The method of claim 1 wherein the data transfer operations include floating-point convert operations, floating-point move operations, floating-point compare operations, or floating-point class operations.
20 . The method of claim 19 wherein the data transfer operations include single-precision floating-point operations and double-precision floating-point operations.
21 . The method of claim 1 wherein the one or more instructions include vector floating-point instructions.
22 . The method of claim 1 wherein the interlocking is managed by a decode unit.
23 . A computer program product embodied in a non-transitory computer readable medium for instruction execution, the computer program product comprising code which causes one or more processors to generate semiconductor logic for:
accessing a processor core, wherein the processor core executes one or more instructions out of order, and wherein the processor core supports integer operations and floating-point operations, wherein the integer operations and the floating-point operations include various latency operations, wherein the various latency operations include variable latency operations;
decoding an instruction in the processor core, wherein the instruction comprises a data transfer operation, and wherein the data transfer operation necessitates a floating-point operation and an integer operation;
dispatching the floating-point operation and the integer operation to one or more issue queues;
interlocking the floating-point operation and the integer operation, wherein the interlocking is accomplished using at least one entry in the one or more issue queues, wherein a request/grant protocol is employed for interlocking the variable latency operations;
executing a first operation of the floating-point operation and the integer operation; and
executing a second operation of the floating-point operation and the integer operation, based on the interlocking.
24 . A computer system for instruction execution comprising:
a memory which stores instructions;
one or more processors coupled to the memory wherein the one or more processors, when executing the instructions which are stored, are configured to:
access a processor core, wherein the processor core executes one or more instructions out of order, and wherein the processor core supports integer operations and floating-point operations, wherein the integer operations and the floating-point operations include various latency operations, wherein the various latency operations include variable latency operations;
decode an instruction in the processor core, wherein the instruction comprises a data transfer operation, and wherein the data transfer operation necessitates a floating-point operation and an integer operation;
dispatch the floating-point operation and the integer operation to one or more issue queues;
interlock the floating-point operation and the integer operation, wherein the interlocking is accomplished using at least one entry in the one or more issue queues, wherein a request/grant protocol is employed for interlocking the variable latency operations;
execute a first operation of the floating-point operation and the integer operation; and
execute a second operation of the floating-point operation and the integer operation, based on the interlocking.