Unified prefetching into instruction cache and branch target buffer

1. A method of coupling a content of a Branch Target Buffer (BTB) with an instruction cache content of an instruction cache comprising:

in a block-based BTB, tagging a plurality of target buffer entries that belong to branches within a same instruction block with a corresponding instruction block address and a branch bitmap to indicate individual branches in the block, whereby a predefined number of BTB entries tagged with an instruction block address and a bitmap constitute a bundle;

coupling an overflow buffer with the BTB to accommodate further BTB entries of instruction blocks, distinct from the plurality of BTB entries, which have more branches than the bundle is configured to accommodate in a corresponding instruction bundle in the BTB; and

predicting instructions or instruction blocks that are likely to be fetched by a core and fetching those instructions from lower levels of a memory hierarchy.

2. The method of claim 1 , wherein organization of the block-based BTB comprises a same number of sets and ways as the instruction cache to maintain only the bundles of the instruction-cache-resident blocks.

3. The method of claim 1 , further comprising

inserting an instruction block into the instruction cache;

pre-decoding the instructions in the instruction block to identify branch instructions and calculate target addresses of the branch instructions; and

inserting the branches identified within the block into the bundle of the corresponding instruction block.

4. The method of claim 1 , further comprising

evicting an instruction block from the instruction cache; and

evicting the corresponding bundle from the BTB.

5. The method of claim 1 , wherein a single common prefetcher performs insertions into the instruction cache and into the BTB in sync.

6. A computing system for coupling content of a Branch Target Buffer (BTB) with an instruction cache content of an instruction cache, comprising:

a system that tags a plurality of target buffer entries in a block-based BTB that belong to branches within a same instruction block with a corresponding instruction block address and a branch bitmap to indicate individual branches in the block, wherein a predefined number of BTB entries tagged with an instruction block address and a bitmap constitute a bundle;

a system that couples an overflow buffer with the BTB to accommodate further BTB entries of instruction blocks, distinct from the BTB entries, which have more branches than the bundle is configured to accommodate in a corresponding bundle in the BTB; and

a system that identifies predicted instructions or predicted instruction blocks that are likely to be fetched by a core and that fetches the predicted instructions or predicted instruction blocks from lower levels of a memory hierarchy.

7. The computing system of claim 6 , wherein the block-based BTB comprises a same number of sets and ways as the instruction cache to maintain only the bundles of the instruction-cache-resident blocks.

8. The computing system of claim 6 , further comprising:

a system that inserts an instruction block into the instruction cache;

a system that pre-decodes the instructions in the instruction block to identify branch instructions and calculates target addresses of the branch instructions; and

a system that inserts the branches identified within the block into the bundle of the corresponding instruction block.

9. The computing system of claim 6 , further comprising:

a system that evicts an instruction block from the instruction cache; and

a system that evicts the corresponding bundle from the BTB.

10. The computing system of claim 6 , wherein a system comprising a single common prefetcher performs insertions into the instruction cache and into the BTB in sync.

11. A computer program product stored on a non-transitory computer readable storage medium, which when executed by a computing system, couples content of a Branch Target Buffer (BTB) with an instruction cache content of an instruction cache, the computer program product comprising:

program instructions that tag a plurality of target buffer entries in a block-based BTB that belong to branches within a same instruction block with a corresponding instruction block address and a branch bitmap to indicate individual branches in the block, wherein a predefined number of BTB entries tagged with an instruction block address and a bitmap constitute a bundle;

program instructions that couple an overflow buffer with the BTB to accommodate further BTB entries of instruction blocks, distinct from the BTB entries, which have more branches than the bundle is configured to accommodate in a corresponding bundle in the BTB; and

program instructions that identify predicted instructions or predicted instruction blocks that are likely to be fetched by a core and that fetch the predicted instructions or predicted instruction blocks from lower levels of a memory hierarchy.

12. The computer program product of claim 11 , wherein the block-based BTB comprises a same number of sets and ways as the instruction cache to maintain only the bundles of the instruction-cache-resident blocks.

13. The computer program product of claim 11 , further comprising

program instructions that insert an instruction block into the instruction cache;

program instructions that pre-decode the instructions in the instruction block to identify branch instructions and calculate target addresses of the branch instructions; and

program instructions that insert the branches identified within the block into the bundle of the corresponding instruction block.

14. The computer program product of claim 11 , further comprising

program instructions that evict an instruction block from the instruction cache; and

program instruction that evict the corresponding bundle from the BTB.

15. The computer program of claim 11 , program instructions implementing a single common prefetcher that performs insertions into the instruction cache and into the BTB in sync.