Branch target buffer operation with auxiliary indirect cache
Disclosed embodiments provide techniques for branch prediction. A processor core is accessed. The processor core is coupled to memory and includes branch prediction circuitry. The branch prediction circuitry includes a branch target buffer (BTB) and an indirect branch target buffer (BTBI). A hashed program counter within the processor core is read. The BTB and BTBI are searched. The searching the BTB is accomplished with the hashed program counter and the searching the BTBI is accomplished with the hashed program counter and branch history information. A predicted branch target address within the BTBI or the BTB is matched. The matching within the BTBI is based on an indirect branch instruction, and the matching within the BTB is based on other branch instruction types. The predicted branch target address that was matched is predicted taken. The processor core is directed to fetch a next instruction from the predicted branch target address.
1. A processor-implemented method for branch prediction comprising:
accessing a processor core, wherein the processor core is coupled to memory, and wherein the processor core includes branch prediction circuitry, wherein the branch prediction circuitry includes a branch target buffer (BTB) and an indirect branch target buffer (BTBI);
reading a hashed program counter within the processor core;
searching the BTB and the BTBI, wherein the searching the BTB is accomplished with the hashed program counter and the searching the BTBI is accomplished with the hashed program counter and branch history information;
matching a predicted branch target address within the BTBI or the BTB, wherein the matching within the BTBI is based on an indirect branch instruction, and wherein the matching within the BTB is based on other branch instruction types;
deciding that the predicted branch target address that was matched is predicted taken; and
directing the processor core to fetch a next instruction from the predicted branch target address.
2. The method of claim 1 wherein the BTB comprises an n-way set associative cache, wherein each way of the n-ways comprises multiple banks.
3. The method of claim 2 wherein the number of multiple banks equals a number of instructions fetched by the processor core in a single cycle.
4. The method of claim 3 wherein each bank within the number of multiple banks can be accessed by the fetched instruction of its address offset matches.
5. The method of claim 2 wherein the BTB comprises a tag cache and a data cache.
6. The method of claim 5 wherein a data size of a tag offset is parameterized.
7. The method of claim 1 wherein the BTB stores compressed branch target addresses.
8. The method of claim 7 wherein multiple compressed instructions are stored in a single BTB entry.
9. The method of claim 8 wherein each entry of the BTB stores a portion of a branch target address.
10. The method of claim 9 wherein the portion of the branch target address is combined with the hashed program counter to obtain the branch target address.
11. The method of claim 10 wherein the hashed program counter includes information from a program counter and a global branch history table.
12. The method of claim 9 wherein the portion of the branch target address is 12 bits.
13. The method of claim 1 wherein the BTB supports unaligned instructions.
14. The method of claim 13 wherein each bank of the BTB is accessed either by a fetch group “N” or a mix of fetch group “N” and fetch group “N+1”.
15. The method of claim 1 wherein the BTBI includes a cache structure that comprises a different size, associativity, or number of banks than the BTB.
16. The method of claim 15 wherein the BTBI is direct mapped.
17. The method of claim 15 wherein the BTBI comprises 512 entries.
18. The method of claim 15 wherein each entry of the BTBI comprises a target address of the indirect branch instruction and a valid bit.
19. The method of claim 18 wherein the valid bit within the BTBI remains valid once set.
20. The method of claim 1 further comprising updating the BTB with branch mispredict information.
21. The method of claim 20 wherein the BTB is updated for a taken misprediction of all branch instructions, and wherein a branch target address within the BTB is updated only for direct branch instructions.
22. The method of claim 1 further comprising updating the BTBI with indirect branch mispredict information.
23. The method of claim 22 wherein the BTBI is updated for a taken misprediction of the indirect branch instruction.
24. The method of claim 1 wherein the processor core is included in a multiprocessor system.
25. A computer program product embodied in a non-transitory computer readable medium for branch prediction, 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 is coupled to memory, and wherein the processor core includes branch prediction circuitry, wherein the branch prediction circuitry includes a branch target buffer (BTB) and an indirect branch target buffer (BTBI);
reading a hashed program counter within the processor core;
searching the BTB and the BTBI, wherein the searching the BTB is accomplished with the hashed program counter and the searching the BTBI is accomplished with the hashed program counter and branch history information;
matching a predicted branch target address within the BTBI or the BTB, wherein the matching within the BTBI is based on an indirect branch instruction, and wherein the matching within the BTB is based on other branch instruction types;
deciding that the predicted branch target address that was matched is predicted taken; and
directing the processor core to fetch a next instruction from the predicted branch target address.
26. An apparatus for branch prediction comprising:
a processor core coupled to memory, wherein the processor core and the memory are used to perform operations comprising:
accessing the processor core, wherein the processor core is coupled to the memory, and wherein the processor core includes branch prediction circuitry, wherein the branch prediction circuitry includes a branch target buffer (BTB) and an indirect branch target buffer (BTBI);
reading a hashed program counter within the processor core;
searching the BTB and the BTBI, wherein the searching the BTB is accomplished with the hashed program counter and the searching the BTBI is accomplished with the hashed program counter and branch history information;
matching a predicted branch target address within the BTBI or the BTB, wherein the matching within the BTBI is based on an indirect branch instruction, and wherein the matching within the BTB is based on other branch instruction types;
deciding that the predicted branch target address that was matched is predicted taken; and
directing the processor core to fetch a next instruction from the predicted branch target address.