Not-taken path instruction for selectively generating a forwarded result from a previous instruction based on branch outcome
View Patent ↗A scalable processing system includes a memory device having a plurality of executable program instructions, wherein each of the executable program instructions includes a timetag data field indicative of the nominal sequential order of the associated executable program instructions. The system also includes a plurality of processing elements, which are configured and arranged to receive executable program instructions from the memory device, wherein each of the processing elements executes executable instructions having the highest priority as indicated by the state of the timetag data field.
1. In a processing system, a method for executing one or more program instructions, the method comprising:
executing the following instructions resident within an instruction storage:
a first instruction prior to a branch instruction, to thereby generate a first instruction output having a first register address;
the branch instruction establishing a branch; and
a second instruction between said branch instruction and a target location identified by a branch target address associated with said branch instruction, to thereby generate a second instruction output having a second register address;
forwarding the first instruction output to the second instruction in instruction storage;
storing in association with the second instruction, when the first and second register addresses are equal, the first instruction output; and
generating by the second instruction, when a branch-taken condition occurs, the stored first instruction output as the second instruction output;
wherein said branch-taken condition arises when a predicate evaluation indicates that execution of said second instruction should not be performed.
2. The method of claim 1 , wherein said first instruction output is a data output or a predicate output and said second instruction output is a data output or a predicate output.
3. The method of claim 1 , wherein said predicate evaluation is associated with said branch instruction, and further comprising broadcasting a predicate output of the branch instruction, indicating the branch-taken condition.
4. The method of claim 1 , further comprising:
initiating, in a common clock period, execution of a third instruction, including or subsequent to the target location, along with the first instruction and/or the second instruction.
5. The method of claim 4 , further comprising selectively executing the third instruction based on whether or not the branch is taken.
6. The method of claim 4 , further comprising restoring at least one data value and/or at least one predicate value to a previous value based on whether or not the branch is taken.
7. The method of claim 4 , wherein, when the branch is initially predicted to be taken, the executing the second instruction comprises broadcasting at least one pre-existing data value.
8. The method of claim 1 , further comprising:
executing a third instruction at or after the branch target address;
wherein, when an input operand address of the third instruction matches a relay address of the second instruction, the third instruction snarls relay data forwarded by the second instruction before the third instruction executes.
9. A system, comprising:
means for executing the following instructions resident within an instruction storage:
a first instruction prior to a branch instruction, to thereby generate a first instruction output having a first register address;
the branch instruction establishing a branch; and
a second instruction between said branch instruction and a target location identified by a branch target address associated with said branch instruction, to thereby generate a second instruction output having a second register address;
means for forwarding the first instruction output to the second instruction in instruction storage;
means for storing in association with the second instruction, when the first and second register addresses are equal, the first instruction output; and
means for generating by the second instruction, when a branch-taken condition occurs, the stored first instruction output as the second instruction output;
wherein said branch-taken condition arises when a predicate evaluation indicates that execution of said second instruction should not be performed.
10. The system of claim 9 , wherein said first instruction output is a data output or a predicate output and said second instruction output is a data output or a predicate output.
11. The system of claim 9 , wherein said predicate evaluation is associated with said branch instruction, said system further comprising means for broadcasting a predicate output of the branch instruction, indicating the branch-taken condition.
12. The system of claim 9 , wherein the means for executing is further configured for initiating, during a common clock period, execution of a third instruction, including or subsequent to the target location, along with the first instruction and/or the second instruction.
13. The system of claim 12 , wherein the means for executing is further configured for selectively executing the third instruction based on whether or not the branch is taken.
14. The system of claim 12 , wherein the system is further configured to restore at least one data value and/or at least one predicate value to a previous value based on whether or not the branch is taken.
15. The system of claim 12 , wherein the means for executing is further configured such that, if the branch is initially predicted to be taken, the executing the second instruction comprises broadcasting at least one pre-existing data value.
16. The system of claim 9 , further comprising:
a third instruction at or after the branch target address;
wherein, when an input operand address of the third instruction matches a relay address of the second instruction, the third instruction snarls relay data forwarded by the second instruction before the third instruction executes.