From 1a7d3f9fcb76a68540dd948f91413533a383bfde Mon Sep 17 00:00:00 2001 From: Tony Gutierrez Date: Tue, 19 Jan 2016 14:28:22 -0500 Subject: gpu-compute: AMD's baseline GPU model --- src/mem/protocol/MOESI_AMD_Base-CorePair.sm | 2904 +++++++++++++++++++++++++++ 1 file changed, 2904 insertions(+) create mode 100644 src/mem/protocol/MOESI_AMD_Base-CorePair.sm (limited to 'src/mem/protocol/MOESI_AMD_Base-CorePair.sm') diff --git a/src/mem/protocol/MOESI_AMD_Base-CorePair.sm b/src/mem/protocol/MOESI_AMD_Base-CorePair.sm new file mode 100644 index 000000000..76fe77230 --- /dev/null +++ b/src/mem/protocol/MOESI_AMD_Base-CorePair.sm @@ -0,0 +1,2904 @@ +/* + * Copyright (c) 2010-2015 Advanced Micro Devices, Inc. + * All rights reserved. + * + * For use for simulation and test purposes only + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * 2. Redistributions in binary form must reproduce the above copyright notice, + * this list of conditions and the following disclaimer in the documentation + * and/or other materials provided with the distribution. + * + * 3. Neither the name of the copyright holder nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + * + * Author: Lisa Hsu + */ + +machine(MachineType:CorePair, "CP-like Core Coherence") + : Sequencer * sequencer; + Sequencer * sequencer1; + CacheMemory * L1Icache; + CacheMemory * L1D0cache; + CacheMemory * L1D1cache; + CacheMemory * L2cache; // func mem logic looks in this CacheMemory + bool send_evictions := "False"; + Cycles issue_latency := 5; // time to send data down to NB + Cycles l2_hit_latency := 18; + + // BEGIN Core Buffers + + // To the Network + MessageBuffer * requestFromCore, network="To", virtual_network="0", vnet_type="request"; + MessageBuffer * responseFromCore, network="To", virtual_network="2", vnet_type="response"; + MessageBuffer * unblockFromCore, network="To", virtual_network="4", vnet_type="unblock"; + + // From the Network + MessageBuffer * probeToCore, network="From", virtual_network="0", vnet_type="request"; + MessageBuffer * responseToCore, network="From", virtual_network="2", vnet_type="response"; + + MessageBuffer * mandatoryQueue; + + MessageBuffer * triggerQueue, ordered="true"; + + // END Core Buffers + +{ + // BEGIN STATES + state_declaration(State, desc="Cache states", default="CorePair_State_I") { + + // Base States + I, AccessPermission:Invalid, desc="Invalid"; + S, AccessPermission:Read_Only, desc="Shared"; + E0, AccessPermission:Read_Write, desc="Exclusive with Cluster 0 ownership"; + E1, AccessPermission:Read_Write, desc="Exclusive with Cluster 1 ownership"; + Es, AccessPermission:Read_Write, desc="Exclusive in core"; + O, AccessPermission:Read_Only, desc="Owner state in core, both clusters and other cores may be sharing line"; + Ms, AccessPermission:Read_Write, desc="Modified in core, both clusters may be sharing line"; + M0, AccessPermission:Read_Write, desc="Modified with cluster ownership"; + M1, AccessPermission:Read_Write, desc="Modified with cluster ownership"; + + // Transient States + I_M0, AccessPermission:Busy, desc="Invalid, issued RdBlkM, have not seen response yet"; + I_M1, AccessPermission:Busy, desc="Invalid, issued RdBlkM, have not seen response yet"; + I_M0M1, AccessPermission:Busy, desc="Was in I_M0, got a store request from other cluster as well"; + I_M1M0, AccessPermission:Busy, desc="Was in I_M1, got a store request from other cluster as well"; + I_M0Ms, AccessPermission:Busy, desc="Was in I_M0, got a load request from other cluster as well"; + I_M1Ms, AccessPermission:Busy, desc="Was in I_M1, got a load request from other cluster as well"; + I_E0S, AccessPermission:Busy, desc="Invalid, issued RdBlk, have not seen response yet"; + I_E1S, AccessPermission:Busy, desc="Invalid, issued RdBlk, have not seen response yet"; + I_ES, AccessPermission:Busy, desc="S_F got hit by invalidating probe, RdBlk response needs to go to both clusters"; + + IF_E0S, AccessPermission:Busy, desc="something got hit with Probe Invalidate, now just I_E0S but expecting a L2_to_L1D0 trigger, just drop when receive"; + IF_E1S, AccessPermission:Busy, desc="something got hit with Probe Invalidate, now just I_E1S but expecting a L2_to_L1D1 trigger, just drop when receive"; + IF_ES, AccessPermission:Busy, desc="same, but waiting for two fills"; + IF0_ES, AccessPermission:Busy, desc="same, but waiting for two fills, got one"; + IF1_ES, AccessPermission:Busy, desc="same, but waiting for two fills, got one"; + F_S0, AccessPermission:Busy, desc="same, but going to S0 when trigger received"; + F_S1, AccessPermission:Busy, desc="same, but going to S1 when trigger received"; + + ES_I, AccessPermission:Read_Only, desc="L2 replacement, waiting for clean writeback ack"; + MO_I, AccessPermission:Read_Only, desc="L2 replacement, waiting for dirty writeback ack"; + MO_S0, AccessPermission:Read_Only, desc="M/O got Ifetch Miss, must write back first, then send RdBlkS"; + MO_S1, AccessPermission:Read_Only, desc="M/O got Ifetch Miss, must write back first, then send RdBlkS"; + S_F0, AccessPermission:Read_Only, desc="Shared, filling L1"; + S_F1, AccessPermission:Read_Only, desc="Shared, filling L1"; + S_F, AccessPermission:Read_Only, desc="Shared, filling L1"; + O_F0, AccessPermission:Read_Only, desc="Owned, filling L1"; + O_F1, AccessPermission:Read_Only, desc="Owned, filling L1"; + O_F, AccessPermission:Read_Only, desc="Owned, filling L1"; + Si_F0, AccessPermission:Read_Only, desc="Shared, filling icache"; + Si_F1, AccessPermission:Read_Only, desc="Shared, filling icache"; + S_M0, AccessPermission:Read_Only, desc="Shared, issued CtoD, have not seen response yet"; + S_M1, AccessPermission:Read_Only, desc="Shared, issued CtoD, have not seen response yet"; + O_M0, AccessPermission:Read_Only, desc="Shared, issued CtoD, have not seen response yet"; + O_M1, AccessPermission:Read_Only, desc="Shared, issued CtoD, have not seen response yet"; + S0, AccessPermission:Busy, desc="RdBlkS on behalf of cluster 0, waiting for response"; + S1, AccessPermission:Busy, desc="RdBlkS on behalf of cluster 1, waiting for response"; + + Es_F0, AccessPermission:Read_Write, desc="Es, Cluster read, filling"; + Es_F1, AccessPermission:Read_Write, desc="Es, Cluster read, filling"; + Es_F, AccessPermission:Read_Write, desc="Es, other cluster read, filling"; + E0_F, AccessPermission:Read_Write, desc="E0, cluster read, filling"; + E1_F, AccessPermission:Read_Write, desc="..."; + E0_Es, AccessPermission:Read_Write, desc="..."; + E1_Es, AccessPermission:Read_Write, desc="..."; + Ms_F0, AccessPermission:Read_Write, desc="..."; + Ms_F1, AccessPermission:Read_Write, desc="..."; + Ms_F, AccessPermission:Read_Write, desc="..."; + M0_F, AccessPermission:Read_Write, desc="..."; + M0_Ms, AccessPermission:Read_Write, desc="..."; + M1_F, AccessPermission:Read_Write, desc="..."; + M1_Ms, AccessPermission:Read_Write, desc="..."; + + I_C, AccessPermission:Invalid, desc="Invalid, but waiting for WBAck from NB from canceled writeback"; + S0_C, AccessPermission:Busy, desc="MO_S0 hit by invalidating probe, waiting for WBAck form NB for canceled WB"; + S1_C, AccessPermission:Busy, desc="MO_S1 hit by invalidating probe, waiting for WBAck form NB for canceled WB"; + S_C, AccessPermission:Busy, desc="S*_C got NB_AckS, still waiting for WBAck"; + + } // END STATES + + // BEGIN EVENTS + enumeration(Event, desc="CP Events") { + // CP Initiated events + C0_Load_L1miss, desc="Cluster 0 load, L1 missed"; + C0_Load_L1hit, desc="Cluster 0 load, L1 hit"; + C1_Load_L1miss, desc="Cluster 1 load L1 missed"; + C1_Load_L1hit, desc="Cluster 1 load L1 hit"; + Ifetch0_L1hit, desc="Instruction fetch, hit in the L1"; + Ifetch1_L1hit, desc="Instruction fetch, hit in the L1"; + Ifetch0_L1miss, desc="Instruction fetch, missed in the L1"; + Ifetch1_L1miss, desc="Instruction fetch, missed in the L1"; + C0_Store_L1miss, desc="Cluster 0 store missed in L1"; + C0_Store_L1hit, desc="Cluster 0 store hit in L1"; + C1_Store_L1miss, desc="Cluster 1 store missed in L1"; + C1_Store_L1hit, desc="Cluster 1 store hit in L1"; + // NB Initiated events + NB_AckS, desc="NB Ack to Core Request"; + NB_AckM, desc="NB Ack to Core Request"; + NB_AckE, desc="NB Ack to Core Request"; + + NB_AckWB, desc="NB Ack for writeback"; + + // Memory System initiatied events + L1I_Repl, desc="Replace address from L1I"; // Presumed clean + L1D0_Repl, desc="Replace address from L1D0"; // Presumed clean + L1D1_Repl, desc="Replace address from L1D1"; // Presumed clean + L2_Repl, desc="Replace address from L2"; + + L2_to_L1D0, desc="L1 fill from L2"; + L2_to_L1D1, desc="L1 fill from L2"; + L2_to_L1I, desc="L1 fill from L2"; + + // Probe Events + PrbInvData, desc="probe, return O or M data"; + PrbInv, desc="probe, no need for data"; + PrbShrData, desc="probe downgrade, return O or M data"; + + } // END EVENTS + + enumeration(RequestType, desc="To communicate stats from transitions to recordStats") { + L1D0DataArrayRead, desc="Read the data array"; + L1D0DataArrayWrite, desc="Write the data array"; + L1D0TagArrayRead, desc="Read the data array"; + L1D0TagArrayWrite, desc="Write the data array"; + L1D1DataArrayRead, desc="Read the data array"; + L1D1DataArrayWrite, desc="Write the data array"; + L1D1TagArrayRead, desc="Read the data array"; + L1D1TagArrayWrite, desc="Write the data array"; + L1IDataArrayRead, desc="Read the data array"; + L1IDataArrayWrite, desc="Write the data array"; + L1ITagArrayRead, desc="Read the data array"; + L1ITagArrayWrite, desc="Write the data array"; + L2DataArrayRead, desc="Read the data array"; + L2DataArrayWrite, desc="Write the data array"; + L2TagArrayRead, desc="Read the data array"; + L2TagArrayWrite, desc="Write the data array"; + } + + + // BEGIN STRUCTURE DEFINITIONS + + + // Cache Entry + structure(Entry, desc="...", interface="AbstractCacheEntry") { + State CacheState, desc="cache state"; + bool Dirty, desc="Is the data dirty (diff than memory)?"; + DataBlock DataBlk, desc="data for the block"; + bool FromL2, default="false", desc="block just moved from L2"; + } + + structure(TBE, desc="...") { + State TBEState, desc="Transient state"; + DataBlock DataBlk, desc="data for the block, required for concurrent writebacks"; + bool Dirty, desc="Is the data dirty (different than memory)?"; + int NumPendingMsgs, desc="Number of acks/data messages that this processor is waiting for"; + bool Shared, desc="Victim hit by shared probe"; + } + + structure(TBETable, external="yes") { + TBE lookup(Addr); + void allocate(Addr); + void deallocate(Addr); + bool isPresent(Addr); + } + + TBETable TBEs, template="", constructor="m_number_of_TBEs"; + + void set_cache_entry(AbstractCacheEntry b); + void unset_cache_entry(); + void set_tbe(TBE b); + void unset_tbe(); + void wakeUpAllBuffers(); + void wakeUpBuffers(Addr a); + Cycles curCycle(); + + // END STRUCTURE DEFINITIONS + + // BEGIN INTERNAL FUNCTIONS + + Tick clockEdge(); + Tick cyclesToTicks(Cycles c); + + bool addressInCore(Addr addr) { + return (L2cache.isTagPresent(addr) || L1Icache.isTagPresent(addr) || L1D0cache.isTagPresent(addr) || L1D1cache.isTagPresent(addr)); + } + + Entry getCacheEntry(Addr address), return_by_pointer="yes" { + Entry L2cache_entry := static_cast(Entry, "pointer", L2cache.lookup(address)); + return L2cache_entry; + } + + DataBlock getDataBlock(Addr addr), return_by_ref="yes" { + TBE tbe := TBEs.lookup(addr); + if(is_valid(tbe)) { + return tbe.DataBlk; + } else { + return getCacheEntry(addr).DataBlk; + } + } + + Entry getL1CacheEntry(Addr addr, int cluster), return_by_pointer="yes" { + if (cluster == 0) { + Entry L1D0_entry := static_cast(Entry, "pointer", L1D0cache.lookup(addr)); + return L1D0_entry; + } else { + Entry L1D1_entry := static_cast(Entry, "pointer", L1D1cache.lookup(addr)); + return L1D1_entry; + } + } + + Entry getICacheEntry(Addr addr), return_by_pointer="yes" { + Entry c_entry := static_cast(Entry, "pointer", L1Icache.lookup(addr)); + return c_entry; + } + + bool presentOrAvail2(Addr addr) { + return L2cache.isTagPresent(addr) || L2cache.cacheAvail(addr); + } + + bool presentOrAvailI(Addr addr) { + return L1Icache.isTagPresent(addr) || L1Icache.cacheAvail(addr); + } + + bool presentOrAvailD0(Addr addr) { + return L1D0cache.isTagPresent(addr) || L1D0cache.cacheAvail(addr); + } + + bool presentOrAvailD1(Addr addr) { + return L1D1cache.isTagPresent(addr) || L1D1cache.cacheAvail(addr); + } + + State getState(TBE tbe, Entry cache_entry, Addr addr) { + if(is_valid(tbe)) { + return tbe.TBEState; + } else if (is_valid(cache_entry)) { + return cache_entry.CacheState; + } + return State:I; + } + + void setState(TBE tbe, Entry cache_entry, Addr addr, State state) { + if (is_valid(tbe)) { + tbe.TBEState := state; + } + + if (is_valid(cache_entry)) { + cache_entry.CacheState := state; + } + } + + AccessPermission getAccessPermission(Addr addr) { + TBE tbe := TBEs.lookup(addr); + if(is_valid(tbe)) { + return CorePair_State_to_permission(tbe.TBEState); + } + + Entry cache_entry := getCacheEntry(addr); + if(is_valid(cache_entry)) { + return CorePair_State_to_permission(cache_entry.CacheState); + } + + return AccessPermission:NotPresent; + } + + void functionalRead(Addr addr, Packet *pkt) { + TBE tbe := TBEs.lookup(addr); + if(is_valid(tbe)) { + testAndRead(addr, tbe.DataBlk, pkt); + } else { + functionalMemoryRead(pkt); + } + } + + int functionalWrite(Addr addr, Packet *pkt) { + int num_functional_writes := 0; + + TBE tbe := TBEs.lookup(addr); + if(is_valid(tbe)) { + num_functional_writes := num_functional_writes + + testAndWrite(addr, tbe.DataBlk, pkt); + } + + num_functional_writes := num_functional_writes + functionalMemoryWrite(pkt); + return num_functional_writes; + } + + void setAccessPermission(Entry cache_entry, Addr addr, State state) { + if (is_valid(cache_entry)) { + cache_entry.changePermission(CorePair_State_to_permission(state)); + } + } + + MachineType testAndClearLocalHit(Entry cache_entry) { + assert(is_valid(cache_entry)); + if (cache_entry.FromL2) { + cache_entry.FromL2 := false; + return MachineType:L2Cache; + } else { + return MachineType:L1Cache; + } + } + + void recordRequestType(RequestType request_type, Addr addr) { + if (request_type == RequestType:L1D0DataArrayRead) { + L1D0cache.recordRequestType(CacheRequestType:DataArrayRead, addr); + } else if (request_type == RequestType:L1D0DataArrayWrite) { + L1D0cache.recordRequestType(CacheRequestType:DataArrayWrite, addr); + } else if (request_type == RequestType:L1D0TagArrayRead) { + L1D0cache.recordRequestType(CacheRequestType:TagArrayRead, addr); + } else if (request_type == RequestType:L1D0TagArrayWrite) { + L1D0cache.recordRequestType(CacheRequestType:TagArrayWrite, addr); + } else if (request_type == RequestType:L1D1DataArrayRead) { + L1D1cache.recordRequestType(CacheRequestType:DataArrayRead, addr); + } else if (request_type == RequestType:L1D1DataArrayWrite) { + L1D1cache.recordRequestType(CacheRequestType:DataArrayWrite, addr); + } else if (request_type == RequestType:L1D1TagArrayRead) { + L1D1cache.recordRequestType(CacheRequestType:TagArrayRead, addr); + } else if (request_type == RequestType:L1D1TagArrayWrite) { + L1D1cache.recordRequestType(CacheRequestType:TagArrayWrite, addr); + } else if (request_type == RequestType:L1IDataArrayRead) { + L1Icache.recordRequestType(CacheRequestType:DataArrayRead, addr); + } else if (request_type == RequestType:L1IDataArrayWrite) { + L1Icache.recordRequestType(CacheRequestType:DataArrayWrite, addr); + } else if (request_type == RequestType:L1ITagArrayRead) { + L1Icache.recordRequestType(CacheRequestType:TagArrayRead, addr); + } else if (request_type == RequestType:L1ITagArrayWrite) { + L1Icache.recordRequestType(CacheRequestType:TagArrayWrite, addr); + } else if (request_type == RequestType:L2DataArrayRead) { + L2cache.recordRequestType(CacheRequestType:DataArrayRead, addr); + } else if (request_type == RequestType:L2DataArrayWrite) { + L2cache.recordRequestType(CacheRequestType:DataArrayWrite, addr); + } else if (request_type == RequestType:L2TagArrayRead) { + L2cache.recordRequestType(CacheRequestType:TagArrayRead, addr); + } else if (request_type == RequestType:L2TagArrayWrite) { + L2cache.recordRequestType(CacheRequestType:TagArrayWrite, addr); + } + } + + bool checkResourceAvailable(RequestType request_type, Addr addr) { + if (request_type == RequestType:L2DataArrayRead) { + return L2cache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L2DataArrayWrite) { + return L2cache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L2TagArrayRead) { + return L2cache.checkResourceAvailable(CacheResourceType:TagArray, addr); + } else if (request_type == RequestType:L2TagArrayWrite) { + return L2cache.checkResourceAvailable(CacheResourceType:TagArray, addr); + } else if (request_type == RequestType:L1D0DataArrayRead) { + return L1D0cache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L1D0DataArrayWrite) { + return L1D0cache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L1D0TagArrayRead) { + return L1D0cache.checkResourceAvailable(CacheResourceType:TagArray, addr); + } else if (request_type == RequestType:L1D0TagArrayWrite) { + return L1D0cache.checkResourceAvailable(CacheResourceType:TagArray, addr); + } else if (request_type == RequestType:L1D1DataArrayRead) { + return L1D1cache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L1D1DataArrayWrite) { + return L1D1cache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L1D1TagArrayRead) { + return L1D1cache.checkResourceAvailable(CacheResourceType:TagArray, addr); + } else if (request_type == RequestType:L1D1TagArrayWrite) { + return L1D1cache.checkResourceAvailable(CacheResourceType:TagArray, addr); + } else if (request_type == RequestType:L1IDataArrayRead) { + return L1Icache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L1IDataArrayWrite) { + return L1Icache.checkResourceAvailable(CacheResourceType:DataArray, addr); + } else if (request_type == RequestType:L1ITagArrayRead) { + return L1Icache.checkResourceAvailable(CacheResourceType:TagArray, addr); + } else if (request_type == RequestType:L1ITagArrayWrite) { + return L1Icache.checkResourceAvailable(CacheResourceType:TagArray, addr); + + } else { + return true; + } + } + + // END INTERNAL FUNCTIONS + + // ** OUT_PORTS ** + + out_port(requestNetwork_out, CPURequestMsg, requestFromCore); + out_port(responseNetwork_out, ResponseMsg, responseFromCore); + out_port(triggerQueue_out, TriggerMsg, triggerQueue); + out_port(unblockNetwork_out, UnblockMsg, unblockFromCore); + + // ** IN_PORTS ** + + in_port(triggerQueue_in, TriggerMsg, triggerQueue, block_on="addr") { + if (triggerQueue_in.isReady(clockEdge())) { + peek(triggerQueue_in, TriggerMsg) { + Entry cache_entry := getCacheEntry(in_msg.addr); + TBE tbe := TBEs.lookup(in_msg.addr); + + if (in_msg.Type == TriggerType:L2_to_L1) { + if (in_msg.Dest == CacheId:L1I) { + trigger(Event:L2_to_L1I, in_msg.addr, cache_entry, tbe); + } else if (in_msg.Dest == CacheId:L1D0) { + trigger(Event:L2_to_L1D0, in_msg.addr, cache_entry, tbe); + } else if (in_msg.Dest == CacheId:L1D1) { + trigger(Event:L2_to_L1D1, in_msg.addr, cache_entry, tbe); + } else { + error("unexpected trigger dest"); + } + } + } + } + } + + + in_port(probeNetwork_in, NBProbeRequestMsg, probeToCore) { + if (probeNetwork_in.isReady(clockEdge())) { + peek(probeNetwork_in, NBProbeRequestMsg, block_on="addr") { + Entry cache_entry := getCacheEntry(in_msg.addr); + TBE tbe := TBEs.lookup(in_msg.addr); + + if (in_msg.Type == ProbeRequestType:PrbInv) { + if (in_msg.ReturnData) { + trigger(Event:PrbInvData, in_msg.addr, cache_entry, tbe); + } else { + trigger(Event:PrbInv, in_msg.addr, cache_entry, tbe); + } + } else if (in_msg.Type == ProbeRequestType:PrbDowngrade) { + assert(in_msg.ReturnData); + trigger(Event:PrbShrData, in_msg.addr, cache_entry, tbe); + } + } + } + } + + + // ResponseNetwork + in_port(responseToCore_in, ResponseMsg, responseToCore) { + if (responseToCore_in.isReady(clockEdge())) { + peek(responseToCore_in, ResponseMsg, block_on="addr") { + + Entry cache_entry := getCacheEntry(in_msg.addr); + TBE tbe := TBEs.lookup(in_msg.addr); + + if (in_msg.Type == CoherenceResponseType:NBSysResp) { + if (in_msg.State == CoherenceState:Modified) { + trigger(Event:NB_AckM, in_msg.addr, cache_entry, tbe); + } else if (in_msg.State == CoherenceState:Shared) { + trigger(Event:NB_AckS, in_msg.addr, cache_entry, tbe); + } else if (in_msg.State == CoherenceState:Exclusive) { + trigger(Event:NB_AckE, in_msg.addr, cache_entry, tbe); + } + } else if (in_msg.Type == CoherenceResponseType:NBSysWBAck) { + trigger(Event:NB_AckWB, in_msg.addr, cache_entry, tbe); + } else { + error("Unexpected Response Message to Core"); + } + } + } + } + + // Nothing from the Unblock Network + + // Mandatory Queue + in_port(mandatoryQueue_in, RubyRequest, mandatoryQueue, desc="...") { + if (mandatoryQueue_in.isReady(clockEdge())) { + peek(mandatoryQueue_in, RubyRequest, block_on="LineAddress") { + + Entry cache_entry := getCacheEntry(in_msg.LineAddress); + TBE tbe := TBEs.lookup(in_msg.LineAddress); + + if (in_msg.Type == RubyRequestType:IFETCH) { + // FETCH ACCESS + + if (L1Icache.isTagPresent(in_msg.LineAddress)) { + if (mod(in_msg.contextId, 2) == 0) { + trigger(Event:Ifetch0_L1hit, in_msg.LineAddress, cache_entry, tbe); + } else { + trigger(Event:Ifetch1_L1hit, in_msg.LineAddress, cache_entry, tbe); + } + } else { + if (presentOrAvail2(in_msg.LineAddress)) { + if (presentOrAvailI(in_msg.LineAddress)) { + if (mod(in_msg.contextId, 2) == 0) { + trigger(Event:Ifetch0_L1miss, in_msg.LineAddress, cache_entry, + tbe); + } else { + trigger(Event:Ifetch1_L1miss, in_msg.LineAddress, cache_entry, + tbe); + } + } else { + Addr victim := L1Icache.cacheProbe(in_msg.LineAddress); + trigger(Event:L1I_Repl, victim, + getCacheEntry(victim), TBEs.lookup(victim)); + } + } else { // Not present or avail in L2 + Addr victim := L2cache.cacheProbe(in_msg.LineAddress); + trigger(Event:L2_Repl, victim, getCacheEntry(victim), + TBEs.lookup(victim)); + } + } + } else { + // DATA ACCESS + if (mod(in_msg.contextId, 2) == 1) { + if (L1D1cache.isTagPresent(in_msg.LineAddress)) { + if (in_msg.Type == RubyRequestType:LD) { + trigger(Event:C1_Load_L1hit, in_msg.LineAddress, cache_entry, + tbe); + } else { + // Stores must write through, make sure L2 avail. + if (presentOrAvail2(in_msg.LineAddress)) { + trigger(Event:C1_Store_L1hit, in_msg.LineAddress, cache_entry, + tbe); + } else { + Addr victim := L2cache.cacheProbe(in_msg.LineAddress); + trigger(Event:L2_Repl, victim, getCacheEntry(victim), + TBEs.lookup(victim)); + } + } + } else { + if (presentOrAvail2(in_msg.LineAddress)) { + if (presentOrAvailD1(in_msg.LineAddress)) { + if (in_msg.Type == RubyRequestType:LD) { + trigger(Event:C1_Load_L1miss, in_msg.LineAddress, + cache_entry, tbe); + } else { + trigger(Event:C1_Store_L1miss, in_msg.LineAddress, + cache_entry, tbe); + } + } else { + Addr victim := L1D1cache.cacheProbe(in_msg.LineAddress); + trigger(Event:L1D1_Repl, victim, + getCacheEntry(victim), TBEs.lookup(victim)); + } + } else { // not present or avail in L2 + Addr victim := L2cache.cacheProbe(in_msg.LineAddress); + trigger(Event:L2_Repl, victim, getCacheEntry(victim), TBEs.lookup(victim)); + } + } + } else { + Entry L1D0cache_entry := getL1CacheEntry(in_msg.LineAddress, 0); + if (is_valid(L1D0cache_entry)) { + if (in_msg.Type == RubyRequestType:LD) { + trigger(Event:C0_Load_L1hit, in_msg.LineAddress, cache_entry, + tbe); + } else { + if (presentOrAvail2(in_msg.LineAddress)) { + trigger(Event:C0_Store_L1hit, in_msg.LineAddress, cache_entry, + tbe); + } else { + Addr victim := L2cache.cacheProbe(in_msg.LineAddress); + trigger(Event:L2_Repl, victim, getCacheEntry(victim), + TBEs.lookup(victim)); + } + } + } else { + if (presentOrAvail2(in_msg.LineAddress)) { + if (presentOrAvailD0(in_msg.LineAddress)) { + if (in_msg.Type == RubyRequestType:LD) { + trigger(Event:C0_Load_L1miss, in_msg.LineAddress, + cache_entry, tbe); + } else { + trigger(Event:C0_Store_L1miss, in_msg.LineAddress, + cache_entry, tbe); + } + } else { + Addr victim := L1D0cache.cacheProbe(in_msg.LineAddress); + trigger(Event:L1D0_Repl, victim, getCacheEntry(victim), + TBEs.lookup(victim)); + } + } else { + Addr victim := L2cache.cacheProbe(in_msg.LineAddress); + trigger(Event:L2_Repl, victim, getCacheEntry(victim), + TBEs.lookup(victim)); + } + } + } + } + } + } + } + + + // ACTIONS + action(ii_invIcache, "ii", desc="invalidate iCache") { + if (L1Icache.isTagPresent(address)) { + L1Icache.deallocate(address); + } + } + + action(i0_invCluster, "i0", desc="invalidate cluster 0") { + if (L1D0cache.isTagPresent(address)) { + L1D0cache.deallocate(address); + } + } + + action(i1_invCluster, "i1", desc="invalidate cluster 1") { + if (L1D1cache.isTagPresent(address)) { + L1D1cache.deallocate(address); + } + } + + action(ib_invBothClusters, "ib", desc="invalidate both clusters") { + if (L1D0cache.isTagPresent(address)) { + L1D0cache.deallocate(address); + } + if (L1D1cache.isTagPresent(address)) { + L1D1cache.deallocate(address); + } + } + + action(i2_invL2, "i2", desc="invalidate L2") { + if(is_valid(cache_entry)) { + L2cache.deallocate(address); + } + unset_cache_entry(); + } + + action(mru_setMRU, "mru", desc="Update LRU state") { + L2cache.setMRU(address); + } + + action(mruD1_setD1cacheMRU, "mruD1", desc="Update LRU state") { + L1D1cache.setMRU(address); + } + + action(mruD0_setD0cacheMRU, "mruD0", desc="Update LRU state") { + L1D0cache.setMRU(address); + } + + action(mruI_setIcacheMRU, "mruI", desc="Update LRU state") { + L1Icache.setMRU(address); + } + + action(n_issueRdBlk, "n", desc="Issue RdBlk") { + enqueue(requestNetwork_out, CPURequestMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceRequestType:RdBlk; + out_msg.Requestor := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); + DPRINTF(RubySlicc,"%s\n",out_msg.Destination); + out_msg.MessageSize := MessageSizeType:Request_Control; + out_msg.InitialRequestTime := curCycle(); + } + } + + action(nM_issueRdBlkM, "nM", desc="Issue RdBlkM") { + enqueue(requestNetwork_out, CPURequestMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceRequestType:RdBlkM; + out_msg.Requestor := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.MessageSize := MessageSizeType:Request_Control; + out_msg.InitialRequestTime := curCycle(); + } + } + + action(nS_issueRdBlkS, "nS", desc="Issue RdBlkS") { + enqueue(requestNetwork_out, CPURequestMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceRequestType:RdBlkS; + out_msg.Requestor := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.MessageSize := MessageSizeType:Request_Control; + out_msg.InitialRequestTime := curCycle(); + } + } + + action(vd_victim, "vd", desc="Victimize M/O L2 Data") { + enqueue(requestNetwork_out, CPURequestMsg, issue_latency) { + out_msg.addr := address; + out_msg.Requestor := machineID; + assert(is_valid(cache_entry)); + out_msg.DataBlk := cache_entry.DataBlk; + assert(cache_entry.Dirty); + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.MessageSize := MessageSizeType:Request_Control; + out_msg.Type := CoherenceRequestType:VicDirty; + out_msg.InitialRequestTime := curCycle(); + if (cache_entry.CacheState == State:O) { + out_msg.Shared := true; + } else { + out_msg.Shared := false; + } + } + } + + action(vc_victim, "vc", desc="Victimize E/S L2 Data") { + enqueue(requestNetwork_out, CPURequestMsg, issue_latency) { + out_msg.addr := address; + out_msg.Requestor := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.MessageSize := MessageSizeType:Request_Control; + out_msg.Type := CoherenceRequestType:VicClean; + out_msg.InitialRequestTime := curCycle(); + if (cache_entry.CacheState == State:S) { + out_msg.Shared := true; + } else { + out_msg.Shared := false; + } + } + } + + action(a0_allocateL1D, "a0", desc="Allocate L1D0 Block") { + if (L1D0cache.isTagPresent(address) == false) { + L1D0cache.allocateVoid(address, new Entry); + } + } + + action(a1_allocateL1D, "a1", desc="Allocate L1D1 Block") { + if (L1D1cache.isTagPresent(address) == false) { + L1D1cache.allocateVoid(address, new Entry); + } + } + + action(ai_allocateL1I, "ai", desc="Allocate L1I Block") { + if (L1Icache.isTagPresent(address) == false) { + L1Icache.allocateVoid(address, new Entry); + } + } + + action(a2_allocateL2, "a2", desc="Allocate L2 Block") { + if (is_invalid(cache_entry)) { + set_cache_entry(L2cache.allocate(address, new Entry)); + } + } + + action(t_allocateTBE, "t", desc="allocate TBE Entry") { + check_allocate(TBEs); + assert(is_valid(cache_entry)); + TBEs.allocate(address); + set_tbe(TBEs.lookup(address)); + tbe.DataBlk := cache_entry.DataBlk; // Data only used for WBs + tbe.Dirty := cache_entry.Dirty; + tbe.Shared := false; + } + + action(d_deallocateTBE, "d", desc="Deallocate TBE") { + TBEs.deallocate(address); + unset_tbe(); + } + + action(p_popMandatoryQueue, "pm", desc="Pop Mandatory Queue") { + mandatoryQueue_in.dequeue(clockEdge()); + } + + action(pr_popResponseQueue, "pr", desc="Pop Response Queue") { + responseToCore_in.dequeue(clockEdge()); + } + + action(pt_popTriggerQueue, "pt", desc="Pop Trigger Queue") { + triggerQueue_in.dequeue(clockEdge()); + } + + action(pp_popProbeQueue, "pp", desc="pop probe queue") { + probeNetwork_in.dequeue(clockEdge()); + } + + action(il0_loadDone, "il0", desc="Cluster 0 i load done") { + Entry entry := getICacheEntry(address); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + assert(is_valid(entry)); + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer.readCallback(address, + l2entry.DataBlk, + true, + testAndClearLocalHit(entry)); + } + + action(il1_loadDone, "il1", desc="Cluster 1 i load done") { + Entry entry := getICacheEntry(address); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + assert(is_valid(entry)); + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer1.readCallback(address, + l2entry.DataBlk, + true, + testAndClearLocalHit(entry)); + } + + action(l0_loadDone, "l0", desc="Cluster 0 load done") { + Entry entry := getL1CacheEntry(address, 0); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + assert(is_valid(entry)); + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer.readCallback(address, + l2entry.DataBlk, + true, + testAndClearLocalHit(entry)); + } + + action(l1_loadDone, "l1", desc="Cluster 1 load done") { + Entry entry := getL1CacheEntry(address, 1); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + assert(is_valid(entry)); + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer1.readCallback(address, + l2entry.DataBlk, + true, + testAndClearLocalHit(entry)); + } + + action(xl0_loadDone, "xl0", desc="Cluster 0 load done") { + peek(responseToCore_in, ResponseMsg) { + assert((machineIDToMachineType(in_msg.Sender) == MachineType:Directory) || + (machineIDToMachineType(in_msg.Sender) == MachineType:L3Cache)); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + DPRINTF(ProtocolTrace, "CP Load Done 0 -- address %s, data: %s\n", address, l2entry.DataBlk); + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer.readCallback(address, + l2entry.DataBlk, + false, + machineIDToMachineType(in_msg.Sender), + in_msg.InitialRequestTime, + in_msg.ForwardRequestTime, + in_msg.ProbeRequestStartTime); + } + } + + action(xl1_loadDone, "xl1", desc="Cluster 1 load done") { + peek(responseToCore_in, ResponseMsg) { + assert((machineIDToMachineType(in_msg.Sender) == MachineType:Directory) || + (machineIDToMachineType(in_msg.Sender) == MachineType:L3Cache)); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer1.readCallback(address, + l2entry.DataBlk, + false, + machineIDToMachineType(in_msg.Sender), + in_msg.InitialRequestTime, + in_msg.ForwardRequestTime, + in_msg.ProbeRequestStartTime); + } + } + + action(xi0_loadDone, "xi0", desc="Cluster 0 i-load done") { + peek(responseToCore_in, ResponseMsg) { + assert((machineIDToMachineType(in_msg.Sender) == MachineType:Directory) || + (machineIDToMachineType(in_msg.Sender) == MachineType:L3Cache)); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer.readCallback(address, + l2entry.DataBlk, + false, + machineIDToMachineType(in_msg.Sender), + in_msg.InitialRequestTime, + in_msg.ForwardRequestTime, + in_msg.ProbeRequestStartTime); + } + } + + action(xi1_loadDone, "xi1", desc="Cluster 1 i-load done") { + peek(responseToCore_in, ResponseMsg) { + assert((machineIDToMachineType(in_msg.Sender) == MachineType:Directory) || + (machineIDToMachineType(in_msg.Sender) == MachineType:L3Cache)); + Entry l2entry := getCacheEntry(address); // Used for functional accesses + // L2 supplies data (functional accesses only look in L2, ok because L1 + // writes through to L2) + sequencer1.readCallback(address, + l2entry.DataBlk, + false, + machineIDToMachineType(in_msg.Sender), + in_msg.InitialRequestTime, + in_msg.ForwardRequestTime, + in_msg.ProbeRequestStartTime); + } + } + + action(s0_storeDone, "s0", desc="Cluster 0 store done") { + Entry entry := getL1CacheEntry(address, 0); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + sequencer.writeCallback(address, + cache_entry.DataBlk, + true, + testAndClearLocalHit(entry)); + cache_entry.Dirty := true; + entry.DataBlk := cache_entry.DataBlk; + entry.Dirty := true; + DPRINTF(RubySlicc, "%s\n", cache_entry.DataBlk); + } + + action(s1_storeDone, "s1", desc="Cluster 1 store done") { + Entry entry := getL1CacheEntry(address, 1); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + sequencer1.writeCallback(address, + cache_entry.DataBlk, + true, + testAndClearLocalHit(entry)); + cache_entry.Dirty := true; + entry.Dirty := true; + entry.DataBlk := cache_entry.DataBlk; + DPRINTF(RubySlicc, "%s\n", cache_entry.DataBlk); + } + + action(xs0_storeDone, "xs0", desc="Cluster 0 store done") { + peek(responseToCore_in, ResponseMsg) { + Entry entry := getL1CacheEntry(address, 0); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + assert((machineIDToMachineType(in_msg.Sender) == MachineType:Directory) || + (machineIDToMachineType(in_msg.Sender) == MachineType:L3Cache)); + sequencer.writeCallback(address, + cache_entry.DataBlk, + false, + machineIDToMachineType(in_msg.Sender), + in_msg.InitialRequestTime, + in_msg.ForwardRequestTime, + in_msg.ProbeRequestStartTime); + cache_entry.Dirty := true; + entry.Dirty := true; + entry.DataBlk := cache_entry.DataBlk; + DPRINTF(RubySlicc, "%s\n", cache_entry.DataBlk); + } + } + + action(xs1_storeDone, "xs1", desc="Cluster 1 store done") { + peek(responseToCore_in, ResponseMsg) { + Entry entry := getL1CacheEntry(address, 1); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + assert((machineIDToMachineType(in_msg.Sender) == MachineType:Directory) || + (machineIDToMachineType(in_msg.Sender) == MachineType:L3Cache)); + sequencer1.writeCallback(address, + cache_entry.DataBlk, + false, + machineIDToMachineType(in_msg.Sender), + in_msg.InitialRequestTime, + in_msg.ForwardRequestTime, + in_msg.ProbeRequestStartTime); + cache_entry.Dirty := true; + entry.Dirty := true; + entry.DataBlk := cache_entry.DataBlk; + DPRINTF(RubySlicc, "%s\n", cache_entry.DataBlk); + } + } + + action(forward_eviction_to_cpu0, "fec0", desc="sends eviction information to processor0") { + if (send_evictions) { + DPRINTF(RubySlicc, "Sending invalidation for %s to the CPU\n", address); + sequencer.evictionCallback(address); + } + } + + action(forward_eviction_to_cpu1, "fec1", desc="sends eviction information to processor1") { + if (send_evictions) { + DPRINTF(RubySlicc, "Sending invalidation for %s to the CPU\n", address); + sequencer1.evictionCallback(address); + } + } + + action(ci_copyL2ToL1, "ci", desc="copy L2 data to L1") { + Entry entry := getICacheEntry(address); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + entry.Dirty := cache_entry.Dirty; + entry.DataBlk := cache_entry.DataBlk; + entry.FromL2 := true; + } + + action(c0_copyL2ToL1, "c0", desc="copy L2 data to L1") { + Entry entry := getL1CacheEntry(address, 0); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + entry.Dirty := cache_entry.Dirty; + entry.DataBlk := cache_entry.DataBlk; + entry.FromL2 := true; + } + + action(c1_copyL2ToL1, "c1", desc="copy L2 data to L1") { + Entry entry := getL1CacheEntry(address, 1); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + entry.Dirty := cache_entry.Dirty; + entry.DataBlk := cache_entry.DataBlk; + entry.FromL2 := true; + } + + action(fi_L2ToL1, "fi", desc="L2 to L1 inst fill") { + enqueue(triggerQueue_out, TriggerMsg, l2_hit_latency) { + out_msg.addr := address; + out_msg.Type := TriggerType:L2_to_L1; + out_msg.Dest := CacheId:L1I; + } + } + + action(f0_L2ToL1, "f0", desc="L2 to L1 data fill") { + enqueue(triggerQueue_out, TriggerMsg, l2_hit_latency) { + out_msg.addr := address; + out_msg.Type := TriggerType:L2_to_L1; + out_msg.Dest := CacheId:L1D0; + } + } + + action(f1_L2ToL1, "f1", desc="L2 to L1 data fill") { + enqueue(triggerQueue_out, TriggerMsg, l2_hit_latency) { + out_msg.addr := address; + out_msg.Type := TriggerType:L2_to_L1; + out_msg.Dest := CacheId:L1D1; + } + } + + action(wi_writeIcache, "wi", desc="write data to icache (and l2)") { + peek(responseToCore_in, ResponseMsg) { + Entry entry := getICacheEntry(address); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + entry.DataBlk := in_msg.DataBlk; + entry.Dirty := in_msg.Dirty; + cache_entry.DataBlk := in_msg.DataBlk; + cache_entry.Dirty := in_msg.Dirty; + } + } + + action(w0_writeDcache, "w0", desc="write data to dcache 0 (and l2)") { + peek(responseToCore_in, ResponseMsg) { + Entry entry := getL1CacheEntry(address, 0); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + DPRINTF(ProtocolTrace, "CP writeD0: address %s, data: %s\n", address, in_msg.DataBlk); + entry.DataBlk := in_msg.DataBlk; + entry.Dirty := in_msg.Dirty; + cache_entry.DataBlk := in_msg.DataBlk; + cache_entry.Dirty := in_msg.Dirty; + } + } + + action(w1_writeDcache, "w1", desc="write data to dcache 1 (and l2)") { + peek(responseToCore_in, ResponseMsg) { + Entry entry := getL1CacheEntry(address, 1); + assert(is_valid(entry)); + assert(is_valid(cache_entry)); + entry.DataBlk := in_msg.DataBlk; + entry.Dirty := in_msg.Dirty; + cache_entry.DataBlk := in_msg.DataBlk; + cache_entry.Dirty := in_msg.Dirty; + } + } + + action(ss_sendStaleNotification, "ss", desc="stale data; nothing to writeback") { + peek(responseToCore_in, ResponseMsg) { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:StaleNotif; + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.MessageSize := MessageSizeType:Response_Control; + DPRINTF(RubySlicc, "%s\n", out_msg); + } + } + } + + action(wb_data, "wb", desc="write back data") { + peek(responseToCore_in, ResponseMsg) { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUData; + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.DataBlk := tbe.DataBlk; + out_msg.Dirty := tbe.Dirty; + if (tbe.Shared) { + out_msg.NbReqShared := true; + } else { + out_msg.NbReqShared := false; + } + out_msg.State := CoherenceState:Shared; // faux info + out_msg.MessageSize := MessageSizeType:Writeback_Data; + DPRINTF(RubySlicc, "%s\n", out_msg); + } + } + } + + action(pi_sendProbeResponseInv, "pi", desc="send probe ack inv, no data") { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUPrbResp; // L3 and CPUs respond in same way to probes + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); // will this always be ok? probably not for multisocket + out_msg.Dirty := false; + out_msg.Hit := false; + out_msg.Ntsl := true; + out_msg.State := CoherenceState:NA; + out_msg.MessageSize := MessageSizeType:Response_Control; + } + } + + action(pim_sendProbeResponseInvMs, "pim", desc="send probe ack inv, no data") { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUPrbResp; // L3 and CPUs respond in same way to probes + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); // will this always be ok? probably not for multisocket + out_msg.Dirty := false; + out_msg.Ntsl := true; + out_msg.Hit := false; + out_msg.State := CoherenceState:NA; + out_msg.MessageSize := MessageSizeType:Response_Control; + } + } + + action(ph_sendProbeResponseHit, "ph", desc="send probe ack PrbShrData, no data") { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUPrbResp; // L3 and CPUs respond in same way to probes + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); // will this always be ok? probably not for multisocket + assert(addressInCore(address) || is_valid(tbe)); + out_msg.Dirty := false; // only true if sending back data i think + out_msg.Hit := true; + out_msg.Ntsl := false; + out_msg.State := CoherenceState:NA; + out_msg.MessageSize := MessageSizeType:Response_Control; + } + } + + action(pb_sendProbeResponseBackprobe, "pb", desc="send probe ack PrbShrData, no data, check for L1 residence") { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUPrbResp; // L3 and CPUs respond in same way to probes + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); // will this always be ok? probably not for multisocket + if (addressInCore(address)) { + out_msg.Hit := true; + } else { + out_msg.Hit := false; + } + out_msg.Dirty := false; // not sending back data, so def. not dirty + out_msg.Ntsl := false; + out_msg.State := CoherenceState:NA; + out_msg.MessageSize := MessageSizeType:Response_Control; + } + } + + action(pd_sendProbeResponseData, "pd", desc="send probe ack, with data") { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + assert(is_valid(cache_entry)); + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUPrbResp; + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); // will this always be ok? probably not for multisocket + out_msg.DataBlk := cache_entry.DataBlk; + assert(cache_entry.Dirty); + out_msg.Dirty := true; + out_msg.Hit := true; + out_msg.State := CoherenceState:NA; + out_msg.MessageSize := MessageSizeType:Response_Data; + } + } + + action(pdm_sendProbeResponseDataMs, "pdm", desc="send probe ack, with data") { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + assert(is_valid(cache_entry)); + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUPrbResp; + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); // will this always be ok? probably not for multisocket + out_msg.DataBlk := cache_entry.DataBlk; + assert(cache_entry.Dirty); + out_msg.Dirty := true; + out_msg.Hit := true; + out_msg.State := CoherenceState:NA; + out_msg.MessageSize := MessageSizeType:Response_Data; + } + } + + action(pdt_sendProbeResponseDataFromTBE, "pdt", desc="send probe ack with data") { + enqueue(responseNetwork_out, ResponseMsg, issue_latency) { + assert(is_valid(tbe)); + out_msg.addr := address; + out_msg.Type := CoherenceResponseType:CPUPrbResp; + out_msg.Sender := machineID; + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.DataBlk := tbe.DataBlk; + assert(tbe.Dirty); + out_msg.Dirty := true; + out_msg.Hit := true; + out_msg.State := CoherenceState:NA; + out_msg.MessageSize := MessageSizeType:Response_Data; + } + } + + action(s_setSharedFlip, "s", desc="hit by shared probe, status may be different") { + assert(is_valid(tbe)); + tbe.Shared := true; + } + + action(uu_sendUnblock, "uu", desc="state changed, unblock") { + enqueue(unblockNetwork_out, UnblockMsg, issue_latency) { + out_msg.addr := address; + out_msg.Destination.add(map_Address_to_Directory(address)); + out_msg.MessageSize := MessageSizeType:Unblock_Control; + DPRINTF(RubySlicc, "%s\n", out_msg); + } + } + + action(l2m_profileMiss, "l2m", desc="l2m miss profile") { + ++L2cache.demand_misses; + } + + action(l10m_profileMiss, "l10m", desc="l10m miss profile") { + ++L1D0cache.demand_misses; + } + + action(l11m_profileMiss, "l11m", desc="l11m miss profile") { + ++L1D1cache.demand_misses; + } + + action(l1im_profileMiss, "l1lm", desc="l1im miss profile") { + ++L1Icache.demand_misses; + } + + action(yy_recycleProbeQueue, "yy", desc="recycle probe queue") { + probeNetwork_in.recycle(clockEdge(), cyclesToTicks(recycle_latency)); + } + + action(xx_recycleResponseQueue, "xx", desc="recycle response queue") { + responseToCore_in.recycle(clockEdge(), cyclesToTicks(recycle_latency)); + } + + action(zz_recycleMandatoryQueue, "\z", desc="recycle mandatory queue") { + mandatoryQueue_in.recycle(clockEdge(), cyclesToTicks(recycle_latency)); + } + + // END ACTIONS + + // BEGIN TRANSITIONS + + // transitions from base + transition(I, C0_Load_L1miss, I_E0S) {L1D0TagArrayRead, L2TagArrayRead} { + // track misses, if implemented + // since in I state, L2 miss as well + l2m_profileMiss; + l10m_profileMiss; + a0_allocateL1D; + a2_allocateL2; + i1_invCluster; + ii_invIcache; + n_issueRdBlk; + p_popMandatoryQueue; + } + + transition(I, C1_Load_L1miss, I_E1S) {L1D1TagArrayRead, L2TagArrayRead} { + // track misses, if implemented + // since in I state, L2 miss as well + l2m_profileMiss; + l11m_profileMiss; + a1_allocateL1D; + a2_allocateL2; + i0_invCluster; + ii_invIcache; + n_issueRdBlk; + p_popMandatoryQueue; + } + + transition(I, Ifetch0_L1miss, S0) {L1ITagArrayRead,L2TagArrayRead} { + // track misses, if implemented + // L2 miss as well + l2m_profileMiss; + l1im_profileMiss; + ai_allocateL1I; + a2_allocateL2; + ib_invBothClusters; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + transition(I, Ifetch1_L1miss, S1) {L1ITagArrayRead, L2TagArrayRead} { + // track misses, if implemented + // L2 miss as well + l2m_profileMiss; + l1im_profileMiss; + ai_allocateL1I; + a2_allocateL2; + ib_invBothClusters; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + transition(I, C0_Store_L1miss, I_M0) {L1D0TagArrayRead, L2TagArrayRead} { + l2m_profileMiss; + l10m_profileMiss; + a0_allocateL1D; + a2_allocateL2; + i1_invCluster; + ii_invIcache; + nM_issueRdBlkM; + p_popMandatoryQueue; + } + + transition(I, C1_Store_L1miss, I_M1) {L1D0TagArrayRead, L2TagArrayRead} { + l2m_profileMiss; + l11m_profileMiss; + a1_allocateL1D; + a2_allocateL2; + i0_invCluster; + ii_invIcache; + nM_issueRdBlkM; + p_popMandatoryQueue; + } + + transition(S, C0_Load_L1miss, S_F0) {L1D0TagArrayRead, L2TagArrayRead, L2DataArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(S, C1_Load_L1miss, S_F1) {L1D1TagArrayRead,L2TagArrayRead, L2DataArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(S, Ifetch0_L1miss, Si_F0) {L1ITagArrayRead, L2TagArrayRead, L2DataArrayRead} { + l1im_profileMiss; + ai_allocateL1I; + fi_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(S, Ifetch1_L1miss, Si_F1) {L1ITagArrayRead,L2TagArrayRead, L2DataArrayRead} { + l1im_profileMiss; + ai_allocateL1I; + fi_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition({S}, {C0_Store_L1hit, C0_Store_L1miss}, S_M0) {L1D0TagArrayRead, L2TagArrayRead} { + l2m_profileMiss; + l10m_profileMiss; + a0_allocateL1D; + mruD0_setD0cacheMRU; + i1_invCluster; + ii_invIcache; + nM_issueRdBlkM; + p_popMandatoryQueue; + } + + transition({S}, {C1_Store_L1hit, C1_Store_L1miss}, S_M1) {L1D1TagArrayRead, L2TagArrayRead} { + l2m_profileMiss; + l11m_profileMiss; + a1_allocateL1D; + mruD1_setD1cacheMRU; + i0_invCluster; + ii_invIcache; + nM_issueRdBlkM; + p_popMandatoryQueue; + } + + transition(Es, C0_Load_L1miss, Es_F0) {L1D0TagArrayRead, L2TagArrayRead, L2DataArrayRead} { // can this be folded with S_F? + a0_allocateL1D; + l10m_profileMiss; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(Es, C1_Load_L1miss, Es_F1) {L1D1TagArrayRead, L2TagArrayRead, L2DataArrayRead} { // can this be folded with S_F? + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(Es, Ifetch0_L1miss, S0) {L1ITagArrayRead, L1ITagArrayWrite, L2TagArrayRead, L2TagArrayWrite} { + l1im_profileMiss; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + ib_invBothClusters; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + transition(Es, Ifetch1_L1miss, S1) {L1ITagArrayRead, L2TagArrayRead} { + l1im_profileMiss; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + ib_invBothClusters; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + // THES SHOULD NOT BE INSTANTANEOUS BUT OH WELL FOR NOW + transition(Es, {C0_Store_L1hit, C0_Store_L1miss}, M0) {L1D0TagArrayRead, L1D0TagArrayWrite, L1D0DataArrayWrite, L2TagArrayRead, L2TagArrayWrite, L2DataArrayWrite} { + a0_allocateL1D; + i1_invCluster; + s0_storeDone; // instantaneous L1/L2 dirty - no writethrough delay + mruD0_setD0cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(Es, {C1_Store_L1hit, C1_Store_L1miss}, M1) {L1D1TagArrayRead, L1D1TagArrayWrite, L1D1DataArrayWrite, L2TagArrayRead, L2TagArrayWrite, L2DataArrayWrite} { + a1_allocateL1D; + i0_invCluster; + s1_storeDone; + mruD1_setD1cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E0, C0_Load_L1miss, E0_F) {L1D0TagArrayRead,L2TagArrayRead, L2DataArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E0, C1_Load_L1miss, E0_Es) {L1D1TagArrayRead, L2TagArrayRead, L2DataArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E0, Ifetch0_L1miss, S0) {L2TagArrayRead, L1ITagArrayRead} { + l2m_profileMiss; // permissions miss, still issue RdBlkS + l1im_profileMiss; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + i0_invCluster; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + transition(E0, Ifetch1_L1miss, S1) {L2TagArrayRead, L1ITagArrayRead} { + l2m_profileMiss; // permissions miss, still issue RdBlkS + l1im_profileMiss; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + i0_invCluster; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + transition(E0, {C0_Store_L1hit, C0_Store_L1miss}, M0) {L1D0TagArrayRead, L1D0DataArrayWrite, L1D0TagArrayWrite, L2TagArrayRead, L2DataArrayWrite, L2TagArrayWrite} { + a0_allocateL1D; + s0_storeDone; + mruD0_setD0cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E0, C1_Store_L1miss, M1) {L1D1TagArrayRead, L1D1TagArrayWrite, L1D1TagArrayWrite, L2TagArrayRead, L2TagArrayWrite, L2DataArrayWrite} { + l11m_profileMiss; + a1_allocateL1D; + i0_invCluster; + s1_storeDone; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E1, C1_Load_L1miss, E1_F) {L1D1TagArrayRead, L2TagArrayRead, L2DataArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E1, C0_Load_L1miss, E1_Es) {L1D0TagArrayRead, L2TagArrayRead, L2DataArrayRead} { + l11m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E1, Ifetch1_L1miss, S1) {L2TagArrayRead, L1ITagArrayRead} { + l2m_profileMiss; // permissions miss, still issue RdBlkS + l1im_profileMiss; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + i1_invCluster; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + transition(E1, Ifetch0_L1miss, S0) {L2TagArrayRead, L1ITagArrayRead} { + l2m_profileMiss; // permissions miss, still issue RdBlkS + l1im_profileMiss; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + i1_invCluster; + nS_issueRdBlkS; + p_popMandatoryQueue; + } + + transition(E1, {C1_Store_L1hit, C1_Store_L1miss}, M1) {L1D1TagArrayRead, L2TagArrayRead, L2DataArrayWrite, L1D1TagArrayWrite, L2TagArrayWrite} { + a1_allocateL1D; + s1_storeDone; + mruD1_setD1cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(E1, C0_Store_L1miss, M0) {L1D0TagArrayRead, L2TagArrayRead, L2TagArrayWrite, L1D0TagArrayWrite, L1D0DataArrayWrite, L2DataArrayWrite} { + l10m_profileMiss; + a0_allocateL1D; + i1_invCluster; + s0_storeDone; + mru_setMRU; + p_popMandatoryQueue; + } + + transition({O}, {C0_Store_L1hit, C0_Store_L1miss}, O_M0) {L1D0TagArrayRead,L2TagArrayRead} { + l2m_profileMiss; // permissions miss, still issue CtoD + l10m_profileMiss; + a0_allocateL1D; + mruD0_setD0cacheMRU; + i1_invCluster; + ii_invIcache; + nM_issueRdBlkM; + p_popMandatoryQueue; + } + + transition({O}, {C1_Store_L1hit, C1_Store_L1miss}, O_M1) {L1D1TagArrayRead, L2TagArrayRead} { + l2m_profileMiss; // permissions miss, still issue RdBlkS + l11m_profileMiss; + a1_allocateL1D; + mruD1_setD1cacheMRU; + i0_invCluster; + ii_invIcache; + nM_issueRdBlkM; + p_popMandatoryQueue; + } + + transition(O, C0_Load_L1miss, O_F0) {L2TagArrayRead, L2DataArrayRead, L1D0TagArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(O, C1_Load_L1miss, O_F1) {L2TagArrayRead, L2DataArrayRead, L1D1TagArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(Ms, C0_Load_L1miss, Ms_F0) {L2TagArrayRead, L2DataArrayRead, L1D0TagArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(Ms, C1_Load_L1miss, Ms_F1) {L2TagArrayRead, L2DataArrayRead, L1D1TagArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition({Ms, M0, M1, O}, Ifetch0_L1miss, MO_S0) {L1ITagArrayRead, L2DataArrayRead, L2TagArrayRead} { + l2m_profileMiss; // permissions miss + l1im_profileMiss; + ai_allocateL1I; + t_allocateTBE; + ib_invBothClusters; + vd_victim; +// i2_invL2; + p_popMandatoryQueue; + } + + transition({Ms, M0, M1, O}, Ifetch1_L1miss, MO_S1) {L1ITagArrayRead, L2TagArrayRead, L2DataArrayRead } { + l2m_profileMiss; // permissions miss + l1im_profileMiss; + ai_allocateL1I; + t_allocateTBE; + ib_invBothClusters; + vd_victim; +// i2_invL2; + p_popMandatoryQueue; + } + + transition(Ms, {C0_Store_L1hit, C0_Store_L1miss}, M0) {L1D0TagArrayRead, L1D0TagArrayWrite, L1D0DataArrayWrite, L2TagArrayRead, L2DataArrayWrite, L2TagArrayWrite} { + a0_allocateL1D; + i1_invCluster; + s0_storeDone; + mruD0_setD0cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(Ms, {C1_Store_L1hit, C1_Store_L1miss}, M1) {L1D1TagArrayRead, L1D1TagArrayWrite, L1D1DataArrayWrite, L2TagArrayRead, L2DataArrayWrite, L2TagArrayWrite} { + a1_allocateL1D; + i0_invCluster; + s1_storeDone; + mruD1_setD1cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M0, C0_Load_L1miss, M0_F) {L1D0TagArrayRead, L2TagArrayRead, L2DataArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M0, C1_Load_L1miss, M0_Ms) {L2TagArrayRead, L2DataArrayRead,L1D0TagArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M0, {C0_Store_L1hit, C0_Store_L1miss}) {L1D0TagArrayRead,L1D0DataArrayWrite, L2DataArrayWrite, L2TagArrayRead} { + a0_allocateL1D; + s0_storeDone; + mruD0_setD0cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M0, {C1_Store_L1hit, C1_Store_L1miss}, M1) {L1D1TagArrayRead, L1D1TagArrayWrite, L1D0DataArrayWrite, L2DataArrayWrite, L2TagArrayRead, L2TagArrayWrite} { + a1_allocateL1D; + i0_invCluster; + s1_storeDone; + mruD1_setD1cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M1, C0_Load_L1miss, M1_Ms) {L2TagArrayRead, L2DataArrayRead, L1D0TagArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M1, C1_Load_L1miss, M1_F) {L1D1TagArrayRead,L2TagArrayRead, L2DataArrayRead} { + a1_allocateL1D; + f1_L2ToL1; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M1, {C0_Store_L1hit, C0_Store_L1miss}, M0) {L1D0TagArrayRead, L1D0TagArrayWrite, L1D0DataArrayWrite, L2TagArrayRead, L2DataArrayWrite, L2TagArrayWrite} { + a0_allocateL1D; + i1_invCluster; + s0_storeDone; + mruD0_setD0cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(M1, {C1_Store_L1hit, C1_Store_L1miss}) {L1D1TagArrayRead, L2TagArrayRead, L2DataArrayWrite} { + a1_allocateL1D; + s1_storeDone; + mruD1_setD1cacheMRU; + mru_setMRU; + p_popMandatoryQueue; + } + + // end transitions from base + + // Begin simple hit transitions + transition({S, Es, E0, O, Ms, M0, O_F1, S_F1, Si_F0, Si_F1, Es_F1, E0_Es, + Ms_F1, M0_Ms}, C0_Load_L1hit) {L1D0TagArrayRead, L1D0DataArrayRead} { + // track hits, if implemented + l0_loadDone; + mruD0_setD0cacheMRU; + p_popMandatoryQueue; + } + + transition({S, Es, E1, O, Ms, M1, O_F0, S_F0, Si_F0, Si_F1, Es_F0, E1_Es, + Ms_F0, M1_Ms}, C1_Load_L1hit) {L1D1TagArrayRead, L1D1DataArrayRead} { + // track hits, if implemented + l1_loadDone; + mruD1_setD1cacheMRU; + p_popMandatoryQueue; + } + + transition({S, S_C, S_F0, S_F1, S_F}, Ifetch0_L1hit) {L1ITagArrayRead, L1IDataArrayRead} { + // track hits, if implemented + il0_loadDone; + mruI_setIcacheMRU; + p_popMandatoryQueue; + } + + transition({S, S_C, S_F0, S_F1, S_F}, Ifetch1_L1hit) {L1ITagArrayRead, L1IDataArrayWrite} { + // track hits, if implemented + il1_loadDone; + mruI_setIcacheMRU; + p_popMandatoryQueue; + } + + // end simple hit transitions + + // Transitions from transient states + + // recycles + transition({I_M0, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E0S, I_ES, IF_E0S, IF_ES, + IF0_ES, IF1_ES, S_F0, S_F, O_F0, O_F, S_M0, O_M0, Es_F0, Es_F, E0_F, + E1_Es, Ms_F0, Ms_F, M0_F, M1_Ms}, C0_Load_L1hit) {} { + zz_recycleMandatoryQueue; + } + + transition({IF_E1S, F_S0, F_S1, ES_I, MO_I, MO_S0, MO_S1, Si_F0, Si_F1, S_M1, + O_M1, S0, S1, I_C, S0_C, S1_C, S_C}, C0_Load_L1miss) {} { + zz_recycleMandatoryQueue; + } + + transition({I_M1, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E1S, I_ES, IF_E1S, IF_ES, + IF0_ES, IF1_ES, S_F1, S_F, O_F1, O_F, S_M1, O_M1, Es_F1, Es_F, E1_F, + E0_Es, Ms_F1, Ms_F, M0_Ms, M1_F}, C1_Load_L1hit) {} { + zz_recycleMandatoryQueue; + } + + transition({IF_E0S, F_S0, F_S1, ES_I, MO_I, MO_S0, MO_S1, Si_F0, Si_F1, S_M0, + O_M0, S0, S1, I_C, S0_C, S1_C, S_C}, C1_Load_L1miss) {} { + zz_recycleMandatoryQueue; + } + + transition({F_S0, F_S1, MO_S0, MO_S1, Si_F0, Si_F1, S0, S1, S0_C, S1_C}, {Ifetch0_L1hit, Ifetch1_L1hit}) {} { + zz_recycleMandatoryQueue; + } + + transition({I_M0, I_M1, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E0S, I_E1S, I_ES, + IF_E0S, IF_E1S, IF_ES, IF0_ES, IF1_ES, ES_I, MO_I, S_F0, S_F1, S_F, + O_F0, O_F1, O_F, S_M0, S_M1, O_M0, O_M1, Es_F0, Es_F1, Es_F, E0_F, + E1_F, E0_Es, E1_Es, Ms_F0, Ms_F1, Ms_F, M0_F, M0_Ms, M1_F, M1_Ms, I_C, + S_C}, {Ifetch0_L1miss, Ifetch1_L1miss}) {} { + zz_recycleMandatoryQueue; + } + + transition({I_E1S, IF_E1S, F_S0, F_S1, ES_I, MO_I, MO_S0, MO_S1, S_F1, O_F1, + Si_F0, Si_F1, S_M1, O_M1, S0, S1, Es_F1, E1_F, E0_Es, Ms_F1, M0_Ms, + M1_F, I_C, S0_C, S1_C, S_C}, {C0_Store_L1miss}) {} { + zz_recycleMandatoryQueue; + } + + transition({I_E0S, IF_E0S, F_S0, F_S1, ES_I, MO_I, MO_S0, MO_S1 S_F0, O_F0, + Si_F0, Si_F1, S_M0, O_M0, S0, S1, Es_F0, E0_F, E1_Es, Ms_F0, M0_F, + M1_Ms, I_C, S0_C, S1_C, S_C}, {C1_Store_L1miss}) {} { + zz_recycleMandatoryQueue; + } + + transition({I_M0, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E0S, I_ES, IF_E0S, IF_ES, + IF0_ES, IF1_ES, S_F0, S_F1, S_F, O_F0, O_F1, O_F, Si_F0, Si_F1, S_M0, O_M0, Es_F0, Es_F1, Es_F, E0_F, E0_Es, E1_Es, Ms_F0, Ms_F1, Ms_F, M0_F, M0_Ms, M1_Ms}, {C0_Store_L1hit}) {} { + zz_recycleMandatoryQueue; + } + + transition({I_M1, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E1S, I_ES, IF_E1S, IF_ES, + IF0_ES, IF1_ES, S_F0, S_F1, S_F, O_F0, O_F1, O_F, Si_F0, Si_F1, S_M1, + O_M1, Es_F0, Es_F1, Es_F, E1_F, E0_Es, E1_Es, Ms_F0, Ms_F1, Ms_F, + M0_Ms, M1_F, M1_Ms}, {C1_Store_L1hit}) {} { + zz_recycleMandatoryQueue; + } + + transition({I_M0, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E0S, I_ES, IF_E0S, IF_ES, + IF0_ES, IF1_ES, S_F0, S_F, O_F0, O_F, S_M0, O_M0, Es_F0, Es_F, E0_F, + E1_Es, Ms_F0, Ms_F, M0_F, M1_Ms}, L1D0_Repl) {} { + zz_recycleMandatoryQueue; + } + + transition({I_M1, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E1S, I_ES, IF_E1S, IF_ES, + IF0_ES, IF1_ES, S_F1, S_F, O_F1, O_F, S_M1, O_M1, Es_F1, Es_F, E1_F, + E0_Es, Ms_F1, Ms_F, M0_Ms, M1_F}, L1D1_Repl) {} { + zz_recycleMandatoryQueue; + } + + transition({F_S0, F_S1, MO_S0, MO_S1, Si_F0, Si_F1, S0, S1, S0_C, S1_C}, L1I_Repl) {} { + zz_recycleMandatoryQueue; + } + + transition({S_C, S0_C, S1_C, S0, S1, Si_F0, Si_F1, I_M0, I_M1, I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_E0S, I_E1S, I_ES, S_F0, S_F1, S_F, O_F0, O_F1, O_F, S_M0, O_M0, S_M1, O_M1, Es_F0, Es_F1, Es_F, E0_F, E1_F, E0_Es, E1_Es, Ms_F0, Ms_F1, Ms_F, M0_F, M0_Ms, M1_F, M1_Ms, MO_S0, MO_S1, IF_E0S, IF_E1S, IF_ES, IF0_ES, IF1_ES, F_S0, F_S1}, L2_Repl) {} { + zz_recycleMandatoryQueue; + } + + transition({IF_E0S, IF_E1S, IF_ES, IF0_ES, IF1_ES, F_S0, F_S1}, {NB_AckS, + PrbInvData, PrbInv, PrbShrData}) {} { + yy_recycleProbeQueue; // these should be resolved soon, but I didn't want to add more states, though technically they could be solved now, and probes really could be solved but i don't think it's really necessary. + } + + transition({IF_E0S, IF_E1S, IF_ES, IF0_ES, IF1_ES}, NB_AckE) {} { + xx_recycleResponseQueue; // these should be resolved soon, but I didn't want to add more states, though technically they could be solved now, and probes really could be solved but i don't think it's really necessary. + } + + transition({E0_Es, E1_F, Es_F1}, C0_Load_L1miss, Es_F) {L2DataArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + p_popMandatoryQueue; + } + + transition(S_F1, C0_Load_L1miss, S_F) {L2DataArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + p_popMandatoryQueue; + } + + transition(O_F1, C0_Load_L1miss, O_F) {L2DataArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + p_popMandatoryQueue; + } + + transition({Ms_F1, M0_Ms, M1_F}, C0_Load_L1miss, Ms_F) {L2DataArrayRead} { + l10m_profileMiss; + a0_allocateL1D; + f0_L2ToL1; + p_popMandatoryQueue; + } + + transition(I_M0, C1_Load_L1miss, I_M0Ms) {} { + l2m_profileMiss; + l11m_profileMiss; + a1_allocateL1D; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(I_M1, C0_Load_L1miss, I_M1Ms) {} { + l2m_profileMiss; + l10m_profileMiss; + a0_allocateL1D; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(I_M0, C1_Store_L1miss, I_M0M1) {} { + l2m_profileMiss; + l11m_profileMiss; + a1_allocateL1D; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(I_M1, C0_Store_L1miss, I_M1M0) {} { + l2m_profileMiss; + l10m_profileMiss; + a0_allocateL1D; + mru_setMRU; + p_popMandatoryQueue; + } + + transition(I_E0S, C1_Load_L1miss, I_ES) {} { + l2m_profileMiss; + l11m_profileMiss; + a1_allocateL1D; + p_popMandatoryQueue; + } + + transition(I_E1S, C0_Load_L1miss, I_ES) {} { + l2m_profileMiss; + l10m_profileMiss; + a0_allocateL1D; + p_popMandatoryQueue; + } + + transition({E1_Es, E0_F, Es_F0}, C1_Load_L1miss, Es_F) {L2DataArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + p_popMandatoryQueue; + } + + transition(S_F0, C1_Load_L1miss, S_F) {L2DataArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + p_popMandatoryQueue; + } + + transition(O_F0, C1_Load_L1miss, O_F) {L2DataArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + p_popMandatoryQueue; + } + + transition({Ms_F0, M1_Ms, M0_F}, C1_Load_L1miss, Ms_F) { L2DataArrayRead} { + l11m_profileMiss; + a1_allocateL1D; + f1_L2ToL1; + p_popMandatoryQueue; + } + + transition({S, Es, E0, O, Ms, M0, O_F1, S_F1, Si_F0, Si_F1, Es_F1, E0_Es, Ms_F1, M0_Ms}, L1D0_Repl) {L1D0TagArrayRead} { + i0_invCluster; + } + + transition({S, Es, E1, O, Ms, M1, O_F0, S_F0, Si_F0, Si_F1, Es_F0, E1_Es, Ms_F0, M1_Ms}, L1D1_Repl) {L1D1TagArrayRead} { + i1_invCluster; + } + + transition({S, S_C, S_F0, S_F1}, L1I_Repl) {L1ITagArrayRead} { + ii_invIcache; + } + + transition({S, E0, E1, Es}, L2_Repl, ES_I) {L2TagArrayRead, L2DataArrayRead, L1D0TagArrayRead, L1D1TagArrayRead} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + t_allocateTBE; + vc_victim; + ib_invBothClusters; + i2_invL2; + ii_invIcache; + } + + transition({Ms, M0, M1, O}, L2_Repl, MO_I) {L2TagArrayRead, L2DataArrayRead, L1D0TagArrayRead, L1D1TagArrayRead} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + t_allocateTBE; + vd_victim; + i2_invL2; + ib_invBothClusters; // nothing will happen for D0 on M1, vice versa + } + + transition(S0, NB_AckS, S) {L1D0DataArrayWrite, L1D0TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + wi_writeIcache; + xi0_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(S1, NB_AckS, S) {L1D1DataArrayWrite, L1D1TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + wi_writeIcache; + xi1_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(S0_C, NB_AckS, S_C) {L1D0DataArrayWrite,L2DataArrayWrite} { + wi_writeIcache; + xi0_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(S1_C, NB_AckS, S_C) {L1D1DataArrayWrite, L2DataArrayWrite} { + wi_writeIcache; + xi1_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(I_M0, NB_AckM, M0) {L1D0DataArrayWrite, L1D0TagArrayWrite,L2DataArrayWrite, L2TagArrayWrite} { + w0_writeDcache; + xs0_storeDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(I_M1, NB_AckM, M1) {L1D1DataArrayWrite, L1D1TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + w1_writeDcache; + xs1_storeDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + // THESE MO->M1 should not be instantaneous but oh well for now. + transition(I_M0M1, NB_AckM, M1) {L1D1DataArrayWrite, L1D1TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + w0_writeDcache; + xs0_storeDone; + uu_sendUnblock; + i0_invCluster; + s1_storeDone; + pr_popResponseQueue; + } + + transition(I_M1M0, NB_AckM, M0) {L1D0DataArrayWrite, L1D0TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + w1_writeDcache; + xs1_storeDone; + uu_sendUnblock; + i1_invCluster; + s0_storeDone; + pr_popResponseQueue; + } + + // Above shoudl be more like this, which has some latency to xfer to L1 + transition(I_M0Ms, NB_AckM, M0_Ms) {L1D0DataArrayWrite,L2DataArrayWrite} { + w0_writeDcache; + xs0_storeDone; + uu_sendUnblock; + f1_L2ToL1; + pr_popResponseQueue; + } + + transition(I_M1Ms, NB_AckM, M1_Ms) {L1D1DataArrayWrite, L2DataArrayWrite} { + w1_writeDcache; + xs1_storeDone; + uu_sendUnblock; + f0_L2ToL1; + pr_popResponseQueue; + } + + transition(I_E0S, NB_AckE, E0) {L1D0DataArrayWrite, L1D0TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + w0_writeDcache; + xl0_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(I_E1S, NB_AckE, E1) {L1D1DataArrayWrite, L1D1TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + w1_writeDcache; + xl1_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(I_ES, NB_AckE, Es) {L1D1DataArrayWrite, L1D1TagArrayWrite, L1D0DataArrayWrite, L1D0TagArrayWrite, L2DataArrayWrite, L2TagArrayWrite } { + w0_writeDcache; + xl0_loadDone; + w1_writeDcache; + xl1_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(I_E0S, NB_AckS, S) {L1D0DataArrayWrite, L1D0TagArrayWrite,L2DataArrayWrite, L2TagArrayWrite} { + w0_writeDcache; + xl0_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(I_E1S, NB_AckS, S) {L1D1TagArrayWrite, L1D1DataArrayWrite, L2TagArrayWrite, L2DataArrayWrite} { + w1_writeDcache; + xl1_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(I_ES, NB_AckS, S) {L1D0TagArrayWrite, L1D0DataArrayWrite, L2TagArrayWrite, L2DataArrayWrite} { + w0_writeDcache; + xl0_loadDone; + w1_writeDcache; + xl1_loadDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(S_F0, L2_to_L1D0, S) {L1D0TagArrayWrite, L1D0DataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(S_F1, L2_to_L1D1, S) {L1D1TagArrayWrite, L1D1DataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(Si_F0, L2_to_L1I, S) {L1ITagArrayWrite, L1IDataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + ci_copyL2ToL1; + mru_setMRU; + il0_loadDone; + pt_popTriggerQueue; + } + + transition(Si_F1, L2_to_L1I, S) {L1ITagArrayWrite, L1IDataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + ci_copyL2ToL1; + mru_setMRU; + il1_loadDone; + pt_popTriggerQueue; + } + + transition(S_F, L2_to_L1D0, S_F1) { L1D0DataArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(S_F, L2_to_L1D1, S_F0) { L1D1DataArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(O_F0, L2_to_L1D0, O) { L1D0DataArrayWrite, L1D0TagArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(O_F1, L2_to_L1D1, O) {L1D1DataArrayWrite, L1D1TagArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(O_F, L2_to_L1D0, O_F1) { L1D0DataArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(O_F, L2_to_L1D1, O_F0) { L1D1DataArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(M1_F, L2_to_L1D1, M1) {L1D1DataArrayWrite, L1D1TagArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(M0_F, L2_to_L1D0, M0) {L1D0DataArrayWrite, L1D0TagArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(Ms_F0, L2_to_L1D0, Ms) {L1D0DataArrayWrite, L1D0TagArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(Ms_F1, L2_to_L1D1, Ms) {L1D1DataArrayWrite, L1D1TagArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(Ms_F, L2_to_L1D0, Ms_F1) {L1D0DataArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(Ms_F, L2_to_L1D1, Ms_F0) {L1IDataArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(M1_Ms, L2_to_L1D0, Ms) {L1D0TagArrayWrite, L1D0DataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(M0_Ms, L2_to_L1D1, Ms) {L1D1TagArrayWrite, L1D1DataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(Es_F0, L2_to_L1D0, Es) {L1D0TagArrayWrite, L1D0DataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(Es_F1, L2_to_L1D1, Es) {L1D1TagArrayWrite, L1D1DataArrayWrite, L2TagArrayWrite, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(Es_F, L2_to_L1D0, Es_F1) {L2TagArrayRead, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(Es_F, L2_to_L1D1, Es_F0) {L2TagArrayRead, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(E0_F, L2_to_L1D0, E0) {L2TagArrayRead, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(E1_F, L2_to_L1D1, E1) {L2TagArrayRead, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(E1_Es, L2_to_L1D0, Es) {L2TagArrayRead, L2DataArrayRead} { + c0_copyL2ToL1; + mru_setMRU; + l0_loadDone; + pt_popTriggerQueue; + } + + transition(E0_Es, L2_to_L1D1, Es) {L2TagArrayRead, L2DataArrayRead} { + c1_copyL2ToL1; + mru_setMRU; + l1_loadDone; + pt_popTriggerQueue; + } + + transition(IF_E0S, L2_to_L1D0, I_E0S) {} { + pt_popTriggerQueue; + } + + transition(IF_E1S, L2_to_L1D1, I_E1S) {} { + pt_popTriggerQueue; + } + + transition(IF_ES, L2_to_L1D0, IF1_ES) {} { + pt_popTriggerQueue; + } + + transition(IF_ES, L2_to_L1D1, IF0_ES) {} { + pt_popTriggerQueue; + } + + transition(IF0_ES, L2_to_L1D0, I_ES) {} { + pt_popTriggerQueue; + } + + transition(IF1_ES, L2_to_L1D1, I_ES) {} { + pt_popTriggerQueue; + } + + transition(F_S0, L2_to_L1I, S0) {} { + pt_popTriggerQueue; + } + + transition(F_S1, L2_to_L1I, S1) {} { + pt_popTriggerQueue; + } + + transition({S_M0, O_M0}, NB_AckM, M0) {L1D0TagArrayWrite, L1D0DataArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + mru_setMRU; + xs0_storeDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition({S_M1, O_M1}, NB_AckM, M1) {L1D1TagArrayWrite, L1D1DataArrayWrite, L2DataArrayWrite, L2TagArrayWrite} { + mru_setMRU; + xs1_storeDone; + uu_sendUnblock; + pr_popResponseQueue; + } + + transition(MO_I, NB_AckWB, I) {L2TagArrayWrite} { + wb_data; + d_deallocateTBE; + pr_popResponseQueue; + } + + transition(ES_I, NB_AckWB, I) {L2TagArrayWrite} { + wb_data; + d_deallocateTBE; + pr_popResponseQueue; + } + + transition(MO_S0, NB_AckWB, S0) {L2TagArrayWrite} { + wb_data; + i2_invL2; + a2_allocateL2; + d_deallocateTBE; // FOO + nS_issueRdBlkS; + pr_popResponseQueue; + } + + transition(MO_S1, NB_AckWB, S1) {L2TagArrayWrite} { + wb_data; + i2_invL2; + a2_allocateL2; + d_deallocateTBE; // FOO + nS_issueRdBlkS; + pr_popResponseQueue; + } + + // Writeback cancel "ack" + transition(I_C, NB_AckWB, I) {L2TagArrayWrite} { + ss_sendStaleNotification; + d_deallocateTBE; + pr_popResponseQueue; + } + + transition(S0_C, NB_AckWB, S0) {L2TagArrayWrite} { + ss_sendStaleNotification; + pr_popResponseQueue; + } + + transition(S1_C, NB_AckWB, S1) {L2TagArrayWrite} { + ss_sendStaleNotification; + pr_popResponseQueue; + } + + transition(S_C, NB_AckWB, S) {L2TagArrayWrite} { + ss_sendStaleNotification; + pr_popResponseQueue; + } + + // Begin Probe Transitions + + transition({Ms, M0, M1, O}, PrbInvData, I) {L2TagArrayRead, L2TagArrayWrite, L2DataArrayRead} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pd_sendProbeResponseData; + i2_invL2; + ib_invBothClusters; + pp_popProbeQueue; + } + + transition({Es, E0, E1, S, I}, PrbInvData, I) {L2TagArrayRead, L2TagArrayWrite} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + i2_invL2; + ib_invBothClusters; + ii_invIcache; // only relevant for S + pp_popProbeQueue; + } + + transition(S_C, PrbInvData, I_C) {L2TagArrayWrite} { + t_allocateTBE; + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + i2_invL2; + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition(I_C, PrbInvData, I_C) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; + pp_popProbeQueue; + } + + transition({Ms, M0, M1, O, Es, E0, E1, S, I}, PrbInv, I) {L2TagArrayRead, L2TagArrayWrite} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + i2_invL2; // nothing will happen in I + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition(S_C, PrbInv, I_C) {L2TagArrayWrite} { + t_allocateTBE; + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + i2_invL2; + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition(I_C, PrbInv, I_C) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition({Ms, M0, M1, O}, PrbShrData, O) {L2TagArrayRead, L2TagArrayWrite, L2DataArrayRead} { + pd_sendProbeResponseData; + pp_popProbeQueue; + } + + transition({Es, E0, E1, S}, PrbShrData, S) {L2TagArrayRead, L2TagArrayWrite} { + ph_sendProbeResponseHit; + pp_popProbeQueue; + } + + transition(S_C, PrbShrData) {} { + ph_sendProbeResponseHit; + pp_popProbeQueue; + } + + transition({I, I_C}, PrbShrData) {L2TagArrayRead} { + pb_sendProbeResponseBackprobe; + pp_popProbeQueue; + } + + transition({I_M0, I_E0S}, {PrbInv, PrbInvData}) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; // must invalidate current data (only relevant for I_M0) + a0_allocateL1D; // but make sure there is room for incoming data when it arrives + pp_popProbeQueue; + } + + transition({I_M1, I_E1S}, {PrbInv, PrbInvData}) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; // must invalidate current data (only relevant for I_M1) + a1_allocateL1D; // but make sure there is room for incoming data when it arrives + pp_popProbeQueue; + } + + transition({I_M0M1, I_M1M0, I_M0Ms, I_M1Ms, I_ES}, {PrbInv, PrbInvData, PrbShrData}) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; + a0_allocateL1D; + a1_allocateL1D; + pp_popProbeQueue; + } + + transition({I_M0, I_E0S, I_M1, I_E1S}, PrbShrData) {} { + pb_sendProbeResponseBackprobe; + pp_popProbeQueue; + } + + transition(ES_I, PrbInvData, I_C) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition(MO_I, PrbInvData, I_C) {} { + pdt_sendProbeResponseDataFromTBE; + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition(MO_I, PrbInv, I_C) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition(ES_I, PrbInv, I_C) {} { + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + pp_popProbeQueue; + } + + transition(ES_I, PrbShrData, ES_I) {} { + ph_sendProbeResponseHit; + s_setSharedFlip; + pp_popProbeQueue; + } + + transition(MO_I, PrbShrData, MO_I) {} { + pdt_sendProbeResponseDataFromTBE; + s_setSharedFlip; + pp_popProbeQueue; + } + + transition(MO_S0, PrbInvData, S0_C) {L2TagArrayWrite} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pdt_sendProbeResponseDataFromTBE; + i2_invL2; + a2_allocateL2; + d_deallocateTBE; + nS_issueRdBlkS; + pp_popProbeQueue; + } + + transition(MO_S1, PrbInvData, S1_C) {L2TagArrayWrite} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pdt_sendProbeResponseDataFromTBE; + i2_invL2; + a2_allocateL2; + d_deallocateTBE; + nS_issueRdBlkS; + pp_popProbeQueue; + } + + transition(MO_S0, PrbInv, S0_C) {L2TagArrayWrite} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + i2_invL2; + a2_allocateL2; + d_deallocateTBE; + nS_issueRdBlkS; + pp_popProbeQueue; + } + + transition(MO_S1, PrbInv, S1_C) {L2TagArrayWrite} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + i2_invL2; + a2_allocateL2; + d_deallocateTBE; + nS_issueRdBlkS; + pp_popProbeQueue; + } + + transition({MO_S0, MO_S1}, PrbShrData) {} { + pdt_sendProbeResponseDataFromTBE; + s_setSharedFlip; + pp_popProbeQueue; + } + + transition({S_F0, Es_F0, E0_F, E1_Es}, {PrbInvData, PrbInv}, IF_E0S) {}{ + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + // invalidate everything you've got + ib_invBothClusters; + ii_invIcache; + i2_invL2; + // but make sure you have room for what you need from the fill + a0_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({S_F1, Es_F1, E1_F, E0_Es}, {PrbInvData, PrbInv}, IF_E1S) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + // invalidate everything you've got + ib_invBothClusters; + ii_invIcache; + i2_invL2; + // but make sure you have room for what you need from the fill + a1_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({S_F, Es_F}, {PrbInvData, PrbInv}, IF_ES) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + // invalidate everything you've got + ib_invBothClusters; + ii_invIcache; + i2_invL2; + // but make sure you have room for what you need from the fill + a0_allocateL1D; + a1_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition(Si_F0, {PrbInvData, PrbInv}, F_S0) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + nS_issueRdBlkS; + pp_popProbeQueue; + } + + transition(Si_F1, {PrbInvData, PrbInv}, F_S1) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + nS_issueRdBlkS; + pp_popProbeQueue; + } + + transition({Es_F0, E0_F, E1_Es}, PrbShrData, S_F0) {} { + ph_sendProbeResponseHit; + pp_popProbeQueue; + } + + transition({Es_F1, E1_F, E0_Es}, PrbShrData, S_F1) {} { + ph_sendProbeResponseHit; + pp_popProbeQueue; + } + + transition(Es_F, PrbShrData, S_F) {} { + ph_sendProbeResponseHit; + pp_popProbeQueue; + } + + transition({S_F0, S_F1, S_F, Si_F0, Si_F1}, PrbShrData) {} { + ph_sendProbeResponseHit; + pp_popProbeQueue; + } + + transition(S_M0, PrbInvData, I_M0) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pim_sendProbeResponseInvMs; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + a0_allocateL1D; + a2_allocateL2; + pp_popProbeQueue; + } + + transition(O_M0, PrbInvData, I_M0) {L2DataArrayRead} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pdm_sendProbeResponseDataMs; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + a0_allocateL1D; + a2_allocateL2; + pp_popProbeQueue; + } + + transition({S_M0, O_M0}, {PrbInv}, I_M0) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pim_sendProbeResponseInvMs; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + a0_allocateL1D; + a2_allocateL2; + pp_popProbeQueue; + } + + transition(S_M1, PrbInvData, I_M1) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pim_sendProbeResponseInvMs; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + a1_allocateL1D; + a2_allocateL2; + pp_popProbeQueue; + } + + transition(O_M1, PrbInvData, I_M1) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pdm_sendProbeResponseDataMs; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + a1_allocateL1D; + a2_allocateL2; + pp_popProbeQueue; + } + + transition({S_M1, O_M1}, {PrbInv}, I_M1) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pim_sendProbeResponseInvMs; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + a1_allocateL1D; + a2_allocateL2; + pp_popProbeQueue; + } + + transition({S0, S0_C}, {PrbInvData, PrbInv}) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + pp_popProbeQueue; + } + + transition({S1, S1_C}, {PrbInvData, PrbInv}) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + ib_invBothClusters; + ii_invIcache; + i2_invL2; + ai_allocateL1I; + a2_allocateL2; + pp_popProbeQueue; + } + + transition({S_M0, S_M1}, PrbShrData) {} { + ph_sendProbeResponseHit; + pp_popProbeQueue; + } + + transition({O_M0, O_M1}, PrbShrData) {L2DataArrayRead} { + pd_sendProbeResponseData; + pp_popProbeQueue; + } + + transition({S0, S1, S0_C, S1_C}, PrbShrData) {} { + pb_sendProbeResponseBackprobe; + pp_popProbeQueue; + } + + transition({Ms_F0, M0_F, M1_Ms, O_F0}, PrbInvData, IF_E0S) { L2DataArrayRead} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pd_sendProbeResponseData; + ib_invBothClusters; + i2_invL2; + a0_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({Ms_F1, M1_F, M0_Ms, O_F1}, PrbInvData, IF_E1S) {L2DataArrayRead} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pd_sendProbeResponseData; + ib_invBothClusters; + i2_invL2; + a1_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({Ms_F, O_F}, PrbInvData, IF_ES) {L2DataArrayRead} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pd_sendProbeResponseData; + ib_invBothClusters; + i2_invL2; + a0_allocateL1D; + a1_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({Ms_F0, M0_F, M1_Ms, O_F0}, PrbInv, IF_E0S) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + ib_invBothClusters; + i2_invL2; + a0_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({Ms_F1, M1_F, M0_Ms, O_F1}, PrbInv, IF_E1S) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + ib_invBothClusters; + i2_invL2; + a1_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({Ms_F, O_F}, PrbInv, IF_ES) {} { + forward_eviction_to_cpu0; + forward_eviction_to_cpu1; + pi_sendProbeResponseInv; + ib_invBothClusters; + i2_invL2; + a0_allocateL1D; + a1_allocateL1D; + a2_allocateL2; + n_issueRdBlk; + pp_popProbeQueue; + } + + transition({Ms_F0, M0_F, M1_Ms}, PrbShrData, O_F0) {L2DataArrayRead} { + pd_sendProbeResponseData; + pp_popProbeQueue; + } + + transition({Ms_F1, M1_F, M0_Ms}, PrbShrData, O_F1) {} { + } + + transition({Ms_F}, PrbShrData, O_F) {L2DataArrayRead} { + pd_sendProbeResponseData; + pp_popProbeQueue; + } + + transition({O_F0, O_F1, O_F}, PrbShrData) {L2DataArrayRead} { + pd_sendProbeResponseData; + pp_popProbeQueue; + } + + // END TRANSITIONS +} + + -- cgit v1.2.3