Sector cache for compression
In an example, an apparatus comprises a plurality of execution units, and a cache memory communicatively coupled to the plurality of execution units, wherein the cache memory is structured into a plurality of sectors, wherein each sector in the plurality of sectors comprises at least two cache lines. Other embodiments are also disclosed and claimed.
1. An apparatus comprising:
a general-purpose graphics processing compute block comprising a plurality of execution units;
a shared cache memory communicatively coupled to the plurality of execution units; and
a processing unit to:
structure the shared cache memory into a plurality of sectors, each sector in the plurality of sectors comprising at least a first cache line and a second cache line, wherein the first cache line and the second cache line in each of the plurality of sectors are indicated by a pairing bit field that associates the first cache line and the second cache line with a sector in the plurality of sectors and which is included with cache access operations and in a cache tag field;
receive a cache access request to the shared cache memory, the cache access request comprising the pairing bit field; and
use the pairing bit field to locate the at least a first cache line and a second cache line associated with the cache access request to the shared cache memory.
2. The apparatus of claim 1 , further comprising:
a data compression unit communicatively coupled to the cache memory, wherein the data compression unit operates at a sector-level granularity.
3. The apparatus of claim 1 , wherein the general-purpose graphics processing compute block includes multiple compute clusters, each compute cluster including multiple graphics multiprocessors.
4. The apparatus of claim 1 , further comprising a processor to implement a victim selection algorithm during a cache eviction.
5. The apparatus of claim 4 , wherein the victim selection algorithm is to assign a high priority to cache sectors which are invalid.
6. The apparatus of claim 5 , wherein the victim selection algorithm is to assign a high priority to cache sectors which are valid and include at least one invalid cacheline.
7. The apparatus of claim 5 , wherein the victim selection algorithm is to assign a low priority to cache sectors which are valid and include at least all invalid cachelines.
8. A method comprising:
structuring a shared cache memory communicatively coupled to the plurality of execution units of a general purpose graphics compute block into a plurality of sectors, each sector in the plurality of sectors comprising at least a first cache line and a second cache line, wherein the first cache line and the second cache line in each of the plurality of sectors are indicated by a pairing bit field that associates the first cache line and the second cache line with a sector in the plurality of sectors and which is included with cache access operations and in a cache tag field;
receiving, in a processing unit, a cache access request to the shared cache memory, the cache access request comprising the pairing bit field; and
using, in the processing unit, the pairing bit field to locate the at least a first cache line and a second cache line associated with the cache access request to the shared cache memory.
9. The method of claim 8 , further comprising:
compressing the data in the shared cache memory at a sector-level granularity.
10. The method of claim 9 , further comprising implementing a process to assign the one or more extra threads to one or more streaming multiprocessors which are assigned to a different context.
11. The method of claim 10 , wherein the general-purpose graphics processing compute block includes multiple compute clusters, each compute cluster including multiple graphics multiprocessors.
12. The method of claim 11 , further comprising:
implementing a victim selection algorithm during a cache eviction.
13. The method of claim 12 , wherein the victim selection algorithm is to assign a high priority to cache sectors which are invalid.
14. The method of claim 12 , wherein the victim selection algorithm is to assign a high priority to cache sectors which are valid and include at least one invalid cacheline.
15. The method of claim 12 , wherein the victim selection algorithm is to assign a low priority to cache sectors which are valid and include at least all invalid cachelines.
16. One or more non-transitory computer-readable medium comprising one or more instructions that when executed on a processing unit of a general purpose graphics processing, configure the processing unit to perform one or more operations to:
structure a shared cache memory communicatively coupled to a plurality of execution units of a general purpose graphics compute block into a plurality of sectors, each sector in the plurality of sectors comprising at least a first cache line and a second cache line, wherein the first cache line and the second cache line in each of the plurality of sectors are indicated by a pairing bit field that associates the first cache line and the second cache line with a sector in the plurality of sectors and which is included with cache access operations and in a cache tag field;
receive, in a processing unit, a cache access request to the shared cache memory, the cache access request comprising the pairing bit field; and
use, in the processing unit, the pairing bit field to locate the at least a first cache line and a second cache line associated with the cache access request to the shared cache memory.
17. The one or more non-transitory computer-readable medium of claim 16 , comprising one or more instructions that when executed on the at least one processor configure the at least one processor to:
compress the data in the shared cache memory at a sector-level granularity.
18. The one or more non-transitory computer-readable medium of claim 16 , comprising one or more instructions that when executed on the at least one processor configure the at least one processor to:
implement a process to assign the one or more extra threads to one or more streaming multiprocessors which are assigned to a different context.
19. The one or more non-transitory computer-readable medium of claim 16 , wherein the general-purpose graphics processing compute block includes multiple compute clusters, each compute cluster including multiple graphics multiprocessors.
20. The one or more non-transitory computer-readable medium of claim 19 , further comprising:
implementing a victim selection algorithm during a cache eviction.
21. The one or more non-transitory computer-readable medium of claim 20 , wherein the victim selection algorithm is to assign a high priority to cache sectors which are invalid.
22. The one or more non-transitory computer-readable medium of claim 20 , wherein the victim selection algorithm is to assign a high priority to cache sectors which are valid and include at least one invalid cacheline.
23. The one or more non-transitory computer-readable medium of claim 20 , wherein the victim selection algorithm is to assign a low priority to cache sectors which are valid and include at least all invalid cachelines.