cpu: Add frequency scaling to the Trace CPU

This change adds a simple feature to scale the frequency of
the Trace CPU.

The compute delays in the input traces provide timing. This
change adds a freqency multiplier parameter to the Trace CPU
set to 1.0 by default. The compute delay is manipulated to
effectively achieve the  frequency at which the nodes become
ready and thus scale the frequency of the Trace CPU.

Change-Id: Iaabbd57806941ad56094fcddbeb38fcee1172431
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>

src/cpu/trace/TraceCPU.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2013 - 2015 ARM Limited
+# Copyright (c) 2013 - 2016 ARM Limited
 # All rights reserved.
 #
 # The license below extends only to copyright in the software and shall
@@ -69,3 +69,9 @@ class TraceCPU(BaseCPU):
     sizeLoadBuffer = Param.Unsigned(16, "Number of entries in the load buffer")
     sizeROB =  Param.Unsigned(40, "Number of entries in the re-order buffer")
 
+    # Frequency multiplier used to effectively scale the Trace CPU frequency
+    # either up or down. Note that the Trace CPU's clock domain must also be
+    # changed when frequency is scaled. A default value of 1.0 means the same
+    # frequency as was used for generating the traces.
+    freqMultiplier = Param.Float(1.0, "Multiplier scale the Trace CPU "\
+                                 "frequency up or down")

src/cpu/trace/trace_cpu.cc

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013 - 2015 ARM Limited
+ * Copyright (c) 2013 - 2016 ARM Limited
  * All rights reserved
  *
  * The license below extends only to copyright in the software and shall
@@ -56,8 +56,7 @@ TraceCPU::TraceCPU(TraceCPUParams *params)
         dataTraceFile(params->dataTraceFile),
         icacheGen(*this, ".iside", icachePort, instMasterID, instTraceFile),
         dcacheGen(*this, ".dside", dcachePort, dataMasterID, dataTraceFile,
-                    params->sizeROB, params->sizeStoreBuffer,
+                  params),
-                    params->sizeLoadBuffer),
         icacheNextEvent(this),
         dcacheNextEvent(this),
         oneTraceComplete(false),
@@ -1225,8 +1224,11 @@ TraceCPU::DcachePort::recvReqRetry()
     owner->dcacheRetryRecvd();
 }
 
-TraceCPU::ElasticDataGen::InputStream::InputStream(const std::string& filename)
+TraceCPU::ElasticDataGen::InputStream::InputStream(
+    const std::string& filename,
+    const double time_multiplier)
     : trace(filename),
+      timeMultiplier(time_multiplier),
       microOpCount(0)
 {
     // Create a protobuf message for the header and read it from the stream
@@ -1259,7 +1261,8 @@ TraceCPU::ElasticDataGen::InputStream::read(GraphNode* element)
         // Required fields
         element->seqNum = pkt_msg.seq_num();
         element->type = pkt_msg.type();
-        element->compDelay = pkt_msg.comp_delay();
+        // Scale the compute delay to effectively scale the Trace CPU frequency
+        element->compDelay = pkt_msg.comp_delay() * timeMultiplier;
 
         // Repeated field robDepList
         element->clearRobDep();

src/cpu/trace/trace_cpu.hh

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013 - 2015 ARM Limited
+ * Copyright (c) 2013 - 2016 ARM Limited
  * All rights reserved
  *
  * The license below extends only to copyright in the software and shall
@@ -795,6 +795,14 @@ class TraceCPU : public BaseCPU
             /** Input file stream for the protobuf trace */
             ProtoInputStream trace;
 
+            /**
+             * A multiplier for the compute delays in the trace to modulate
+             * the Trace CPU frequency either up or down. The Trace CPU's
+             * clock domain frequency must also be set to match the expected
+             * result of frequency scaling.
+             */
+            const double timeMultiplier;
+
             /** Count of committed ops read from trace plus the filtered ops */
             uint64_t microOpCount;
 
@@ -809,8 +817,10 @@ class TraceCPU : public BaseCPU
              * Create a trace input stream for a given file name.
              *
              * @param filename Path to the file to read from
+             * @param time_multiplier used to scale the compute delays
              */
-            InputStream(const std::string& filename);
+            InputStream(const std::string& filename,
+                        const double time_multiplier);
 
             /**
              * Reset the stream such that it can be played once
@@ -840,19 +850,19 @@ class TraceCPU : public BaseCPU
         /* Constructor */
         ElasticDataGen(TraceCPU& _owner, const std::string& _name,
                    MasterPort& _port, MasterID master_id,
-                   const std::string& trace_file, uint16_t max_rob,
+                   const std::string& trace_file, TraceCPUParams *params)
-                   uint16_t max_stores, uint16_t max_loads)
             : owner(_owner),
               port(_port),
               masterID(master_id),
-              trace(trace_file),
+              trace(trace_file, 1.0 / params->freqMultiplier),
               genName(owner.name() + ".elastic" + _name),
               retryPkt(nullptr),
               traceComplete(false),
               nextRead(false),
               execComplete(false),
               windowSize(trace.getWindowSize()),
-              hwResource(max_rob, max_stores, max_loads)
+              hwResource(params->sizeROB, params->sizeStoreBuffer,
+                         params->sizeLoadBuffer)
         {
             DPRINTF(TraceCPUData, "Window size in the trace is %d.\n",
                     windowSize);