A10.5
CHI transactions
CHI transactions are sent to a specific node in the interconnect based on the following criteria:
• Type of access.
• Address of the access.
• Settings of the System Address Map.
Addresses that map to an HN-F node can be marked as cacheable memory in the page tables, and can
take part in the cache coherency protocol. Addresses that map to an HN-I or MN must be marked as
device or non-cacheable memory.
Table A10-5 CHI transaction IDs
Transaction ID Description
000nnxxx
Transaction from core nn. Can be a:
•
Read transaction.
•
Write transaction.
•
Cache maintenance transaction.
•
DVM transaction.
•
Barrier transaction.
001001xx
Transaction from the ACP interface. Can be a read or write.
00101110
Barrier generated in response to a DVM sync snoop from the interconnect.
0100xxxx
Eviction from L1 or L2 cache. The number of IDs used depends on the configuration.
Table A10-6 CHI transactions
Transaction
Operation
ReadNoSnp
Non-cacheable loads or instruction fetches. Linefills of non-shareable cache lines into L1 or L2.
ReadOnce
Cacheable loads that are not allocating into the cache, or cacheable instruction fetches when there is no L2 cache.
ReadClean
Not used.
ReadShared
L1 Data linefills started by a load instruction, or L2 linefills started by an instruction fetch.
ReadUnique
L1 Data linefills started by a store instruction.
CleanUnique
Store instructions that hit in the cache but the line is not in a unique coherence state.
MakeUnique
Store instructions of a full cache line of data, that miss in the caches, and are allocating into the L2 cache.
CleanShared
Cache maintenance instructions.
CleanInvalid
Cache maintenance instructions.
MakeInvalid
Cache maintenance instructions.
DVMOp
TLB and instruction cache maintenance instructions.
EOBarrier
DMB instructions.
ECBarrier
DSB instructions. DVM sync snoops received from the interconnect.
WriteNoSnpPtl
Non-cacheable store instructions.
WriteNoSnpFull Non-cacheable store instructions, or evictions of non-shareable cache lines from the L1 and L2 cache.
WriteUniqueFull Cacheable writes of a full cache line, that are not allocating into L1 or L2 caches, for example streaming writes.
WriteUniquePtl
Cacheable writes of less than a full cache line that are not allocating into L1 or L2.
A10 CHI Master Interface
A10.5 CHI transactions
100236_0100_00_en
Copyright © 2015–2017, 2019 Arm Limited or its affiliates. All rights
reserved.
A10-131
Non-Confidential
Summary of Contents for Cortex-A35
Page 4: ......
Page 18: ......
Page 26: ......
Page 27: ...Part A Functional Description ...
Page 28: ......
Page 145: ...Part B Register Descriptions ...
Page 146: ......
Page 573: ...Part C Debug ...
Page 574: ......
Page 845: ...Part D Appendices ...
Page 846: ......