ACP slave interface 学习

ACPAccelerator Coherency PortACP是一个可选的slave interface(接口)接口协议规范是ACE-Lite协议的子集。只有访问cacheable memory空间时ACP slave接口允许外部master通过DSU的main memory interface(Master口)访问memory空间。ACP读和写的data bus是128bit为了提高效率每次访问以cache line长度为边界。为了保持cache的一致性对于ACP口的访问需要check cluster中的所有cache locations即L3 cache和每个core的data caches。By default(默认情况下)ACP write访问(acccess)隐含stash requests to the L3 cache即其他master想要把数据存在(stash)L3 cache中。或者隐含的stash request can target the L2 cache of a selected core即将数据存在指定的core的L2 cache中。ACP interface properties for the DSUACP propertySupported by the DSUPort_TypeAcceleratorContinuous_Cache_Line_Read_DataYesMulti_Copy_AtomicityYesCache_Stash_TransactionYesLow_Power_SignalsYesOrdered_Write_ObservationNoothers......NoACP slave口的实现是ACE-Lite协议的子集ACP ACE-Lite子集在AMBA AXI and ACE Protocol Specification AXI3, AXI4, and AXI4-Lite, ACE and ACE-Lite中有有描述。在DSU中实现的ACP Slave port还有如下额外的限制ARCACHES和AWCACHES的值仅限于——0b0111——0b1011——0b1111所有的传输(transactions)可以是Secure或Non-secure。Exclusive accesses不支持。ARLOCK和AWLOCK信号不存在。通过AWDOMAINS和ARDOMAINS信号所有的操作可以指定是Inner Shareable、Outer Shareable和Non-shareable。不支持Barriers操作。任何write transaction的BRESP响应表明该transaction可以被全局观察(global observability)到。ARSIZE和AWSIZE信号不存在默认认为是4(16 bytes)。ARLENS和AWLENS限制如下0 One beat3 Four beatsARBURST和AWBURST信号不存在默认认为是0b01即INCR类型。ARSNOOP信号不存在默认认为是0b0000。ARQOS和AWQOS信号不存在。ACP支持的read传输类型有如下transfer size和length组合16-byte INCR read transaction——ARLENS is 0 (one beat)。——Address aligned to 16-byte boundary (ARADDRS[3:0] is 0b0000)。64-byte INCR read transaction——ARLENS is 3 (four beats)——Address aligned to 64-byte boundary (ARADDRS[5:0] is 0b000000)。ACP支持的写传输类型有如下transfer size和length组合16-byte INCR write transaction——AWSTRBSany combination of bytesincluding no bytesare valid.——AWLENS is 0 (one beat).——AWSNOOPS is WriteUniquePtl.——Address aligned to 16-bytes boundary (AWADDRS[3:0] is 0b0000)64-byte INCR write transaction——AWSTRBSany combination of bytesincluding no bytesare valid. When AWSNOOPS is WriteUniqueFull, all bytes must be valid.——AWLENS is 3 (four beats).——AWSNOOPS is WriteUniquePtl or WriteUniqueFull.——Address aligned to 64-bytes boundary (AWADDRS[5:0] is 0b000000)ACP支持如下的Cache stash传输类型的transfer size和length组合64-byte INCR write stash transaction——AWSTRBSall bytes are valid.——AWLENS is 3 (four beats).——AWSNOOPS is WriteUniqueFullStash.——Address aligned to 64-bytes boundary (AWADDRS[5:0] is 0b000000)Dataless 64-byte INCR write stash transaction——No W-Channel transfer.——AWLENS is 3 (four beats).——AWSNOOPS is StashOnceShared or StashOnceUnique.——Address aligned to 64-bytes boundary (AWADDRS[5:0] is 0b000000)在AWSTASHLPIDENS信号有效时AWSTASHLPIDS[3:1]信号指示选定的core numberStash requests可以指向选定core的L2 cache。AWSTASHLPIDS[0]信号用于thread number但不会影响stash request。如果requests不符合这些约束(restrictions)将会在RRESPS和BRESPS通道上产生SLVERR response。下表列出了支持的ACP transactionsACP supported transactionsTransactionNotesReadOnce-WriteUniqueFull-WriteUniquePtl-WriteUniqueFullStash-StashOnceUnique-StashOnceShared-ACP performance为了提高ACP性能ACP transaction遵循如下指导(guidelines)为了避免second transaction stalling the ACP interface直到first transaction结束同一个master必须避免发出多个同AXI ID的outstanding transactions。如果有两笔transaction必须要保序Arm建议second transaction等first transaction回完response后再发送出去。当writes包含full cacheline data时会达到更高的性能。因为write小雨full cacheline data时通常会导致read-modify write sequence即需要先读后写(merge)L3的一些资源在ACP interface和cores之间是共享的因此在一些场景下ACP interface拥堵的访问会降低cores的性能。下表列出了ACP的可以接受outstanding能力ACP acceptance capabilitiesAttributeValueDescriptionWrite acceptance capability33The ACP can accept up to 33 write transactions.Read acceptance capability33The ACP can accept up to 33 read transactions.Combined acceptance capability34The ACP can accept up to 34 transactions. There is no performance benefit above 32 outstanding transactions.