diff options
author | Radhika Jagtap <radhika.jagtap@arm.com> | 2016-09-15 18:01:16 +0100 |
---|---|---|
committer | Radhika Jagtap <radhika.jagtap@arm.com> | 2016-09-15 18:01:16 +0100 |
commit | d067327fc0c1a67ad2d456840e83db7b593b94a0 (patch) | |
tree | 693f455704b384e6215fde31ef15195466751936 /src/cpu/trace/trace_cpu.cc | |
parent | d7724d5f542b3ffcb1de05bc2960c820c7b96c85 (diff) | |
download | gem5-d067327fc0c1a67ad2d456840e83db7b593b94a0.tar.xz |
cpu: Adjust for trace offset and fix stats
This change subtracts the time offset present in the trace from
all the event times when nodes and request are sent so that the
replay starts immediately when the simulation starts. This makes
the stats accurate when the time offset in traces is large, for
example when traces are generated in the middle of a workload
execution. It also solves the problem of unnecessary DRAM
refresh events that would keep occuring during the large time
offset before even a single request is replayed into the system.
Change-Id: Ie0898842615def867ffd5c219948386d952af7f7
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Diffstat (limited to 'src/cpu/trace/trace_cpu.cc')
-rw-r--r-- | src/cpu/trace/trace_cpu.cc | 51 |
1 files changed, 32 insertions, 19 deletions
diff --git a/src/cpu/trace/trace_cpu.cc b/src/cpu/trace/trace_cpu.cc index 42620a177..ef0b252ed 100644 --- a/src/cpu/trace/trace_cpu.cc +++ b/src/cpu/trace/trace_cpu.cc @@ -60,7 +60,7 @@ TraceCPU::TraceCPU(TraceCPUParams *params) icacheNextEvent(this), dcacheNextEvent(this), oneTraceComplete(false), - firstFetchTick(0), + traceOffset(0), execCompleteEvent(nullptr) { // Increment static counter for number of Trace CPUs. @@ -116,22 +116,31 @@ TraceCPU::init() BaseCPU::init(); - // Get the send tick of the first instruction read request and schedule - // icacheNextEvent at that tick. + // Get the send tick of the first instruction read request Tick first_icache_tick = icacheGen.init(); - schedule(icacheNextEvent, first_icache_tick); - // Get the send tick of the first data read/write request and schedule - // dcacheNextEvent at that tick. + // Get the send tick of the first data read/write request Tick first_dcache_tick = dcacheGen.init(); - schedule(dcacheNextEvent, first_dcache_tick); + + // Set the trace offset as the minimum of that in both traces + traceOffset = std::min(first_icache_tick, first_dcache_tick); + inform("%s: Time offset (tick) found as min of both traces is %lli.\n", + name(), traceOffset); + + // Schedule next icache and dcache event by subtracting the offset + schedule(icacheNextEvent, first_icache_tick - traceOffset); + schedule(dcacheNextEvent, first_dcache_tick - traceOffset); + + // Adjust the trace offset for the dcache generator's ready nodes + // We don't need to do this for the icache generator as it will + // send its first request at the first event and schedule subsequent + // events using a relative tick delta + dcacheGen.adjustInitTraceOffset(traceOffset); // The static counter for number of Trace CPUs is correctly set at this // point so create an event and pass it. execCompleteEvent = new CountedExitEvent("end of all traces reached.", numTraceCPUs); - // Save the first fetch request tick to dump it as tickOffset - firstFetchTick = first_icache_tick; } void @@ -164,6 +173,9 @@ TraceCPU::schedDcacheNext() { DPRINTF(TraceCPUData, "DcacheGen event.\n"); + // Update stat for numCycles + numCycles = clockEdge() / clockPeriod(); + dcacheGen.execute(); if (dcacheGen.isExecComplete()) { checkAndSchedExitEvent(); @@ -179,11 +191,6 @@ TraceCPU::checkAndSchedExitEvent() // Schedule event to indicate execution is complete as both // instruction and data access traces have been played back. inform("%s: Execution complete.\n", name()); - - // Record stats which are computed at the end of simulation - tickOffset = firstFetchTick; - numCycles = (clockEdge() - firstFetchTick) / clockPeriod(); - numOps = dcacheGen.getMicroOpCount(); schedule(*execCompleteEvent, curTick()); } } @@ -216,11 +223,6 @@ TraceCPU::regStats() ; cpi = numCycles/numOps; - tickOffset - .name(name() + ".tickOffset") - .desc("The first execution tick for the root node of elastic traces") - ; - icacheGen.regStats(); dcacheGen.regStats(); } @@ -312,6 +314,13 @@ TraceCPU::ElasticDataGen::init() } void +TraceCPU::ElasticDataGen::adjustInitTraceOffset(Tick& offset) { + for (auto& free_node : readyList) { + free_node.execTick -= offset; + } +} + +void TraceCPU::ElasticDataGen::exit() { trace.reset(); @@ -534,6 +543,8 @@ TraceCPU::ElasticDataGen::execute() hwResource.release(node_ptr); // clear the dynamically allocated set of dependents (node_ptr->dependents).clear(); + // Update the stat for numOps simulated + owner.updateNumOps(node_ptr->robNum); // delete node delete node_ptr; // remove from graph @@ -736,6 +747,8 @@ TraceCPU::ElasticDataGen::completeMemAccess(PacketPtr pkt) // clear the dynamically allocated set of dependents (node_ptr->dependents).clear(); + // Update the stat for numOps completed + owner.updateNumOps(node_ptr->robNum); // delete node delete node_ptr; // remove from graph |