Direct hardware processing of internal data structure fields
View Patent ↗In some embodiments, the execution of load and store instructions for internal fields of data structures is accelerated by using on-chip template registers and appropriate machine code instructions. A load/store machine code instruction comprises an identifier of a memory address offset of an internal field word relative to a base address of the data structure, an identifier of an intra-word start bit of the internal field, and an identifier of an intra-word length of the internal field. The three identifiers may coincide, for example if the three identifiers are represented by an identity of a template register storing a template entry including the memory address offset, the start position, and the field length. The three identifiers may also be provided as part of a machine code instruction itself. Further provided are compilers, compiler methods, and hardware systems for implementing accelerated internal-field load and store operations.
1. A data processing method comprising employing at least one processor to perform the steps of:
receiving high-level source code containing a reference to an internal field of a data structure; and
from the high-level source code, generating a low-level language memory access instruction for the internal field, the low-level language memory access instruction including an identifier of a memory address offset of the internal field relative to a base address of the data structure, and an identifier of a field bitspan within a word, wherein the field bitspan defines an intra-byte field start position and a field length, and wherein the identifier of the memory address offset and the identifier of the field bitspan are not necessarily distinct.
2. The method of claim 1 , further comprising generating a low level language template entry assignment instruction for assigning a template entry containing the memory address offset to a template register.
3. The method of claim 2 , further comprising determining whether to preserve the template entry in the template register across a function call.
4. The method of claim 1 further comprising determining a plurality of usage weights for a corresponding plurality of data structure internal fields, and assigning a plurality of template entries for data structure internal fields to a corresponding plurality of template registers according to the plurality of usage weights.
5. The method of claim 1 , wherein the low-level language memory access instruction is a load instruction.
6. The method of claim 5 , wherein the load instruction is a load-and-extract instruction.
7. The method of claim 5 , wherein the load instruction is a load-extract-and-sign instruction.
8. The method of claim 5 , wherein the load instruction is a masked-load instruction.
9. The method of claim 1 , wherein the low-level language memory access instruction is a store instruction.
10. The method of claim 9 , wherein the store instruction is a masked-store instruction.
11. A non-transitory computer-readable medium encoding instructions to perform a method comprising the steps of:
receiving high-level source code containing a reference to an internal field of a data structure; and
from the high-level source code, generating a low-level language memory access instruction for the internal field, the low-level language memory access instruction including an identifier of a memory address offset of the internal field relative to a base address of the data structure, and an identifier of a field bitspan within a word, wherein the field bitspan defines an intra-byte field start position and a field length, and wherein the identifier of the memory address offset and the identifier of the field bitspan are not necessarily distinct.
12. The computer-readable medium of claim 11 , wherein the method further comprises generating a low level language template entry assignment instruction assigning a template entry containing the memory address offset to a template register.
13. The computer-readable medium of claim 12 , wherein the method further comprises determining whether to preserve the template entry in the template register across a function call.
14. The computer-readable medium of claim 11 , wherein the method comprises determining a plurality of usage weights for a corresponding plurality of data structure internal fields, and assigning a plurality of template entries for data structure internal fields to a corresponding plurality of template registers according to the plurality of usage weights.
15. A data processing apparatus having a processor comprising:
means for receiving high-level source code containing a reference to an internal field of a data structure; and
means for generating from the high-level source code a low-level language memory access instruction for the internal field, the low-level language memory access instruction including an identifier of a memory address offset of the internal field relative to a base address of the data structure, and an identifier of a field bitspan within a word, wherein the field bitspan defines an intra-byte field start position and a field length, and wherein the identifier of the memory address offset and the identifier of the field bitspan are not necessarily distinct.
16. The data processing apparatus of claim 15 , further comprising means for generating a low level language template entry assignment instruction assigning a template entry containing the memory address offset to a template register.
17. The data processing apparatus of claim 16 , further comprising means for determining whether to preserve the template entry in the template register across a function call.
18. The data processing apparatus of claim 15 , further comprising
means for determining a plurality of usage weights for a corresponding plurality of data structure internal fields, and
means for assigning a plurality of template entries for data structure internal fields to a corresponding plurality of template registers according to the plurality of usage weights.