mem: per-thread cache occupancy and per-block ages

This patch enables tracking of cache occupancy per thread along with
ages (in buckets) per cache blocks.  Cache occupancy stats are
recalculated on each stat dump.

2 files changed, 8 insertions, 0 deletions

diff --git a/src/cpu/simple/atomic.cc b/src/cpu/simple/atomic.cc
index 13c4b9bd3..617e845a5 100644
--- a/src/cpu/simple/atomic.cc
+++ b/src/cpu/simple/atomic.cc
@@ -301,6 +301,7 @@ AtomicSimpleCPU::readMem(Addr addr, uint8_t * data,
 
     dcache_latency = 0;
 
+    req->taskId(taskId());
     while (1) {
         req->setVirt(0, addr, size, flags, dataMasterId(), thread->pcState().instAddr());
 
@@ -387,6 +388,7 @@ AtomicSimpleCPU::writeMem(uint8_t *data, unsigned size,
 
     dcache_latency = 0;
 
+    req->taskId(taskId());
     while(1) {
         req->setVirt(0, addr, size, flags, dataMasterId(), thread->pcState().instAddr());
 
@@ -492,6 +494,7 @@ AtomicSimpleCPU::tick()
         bool needToFetch = !isRomMicroPC(pcState.microPC()) &&
                            !curMacroStaticInst;
         if (needToFetch) {
+            ifetch_req.taskId(taskId());
             setupFetchRequest(&ifetch_req);
             fault = thread->itb->translateAtomic(&ifetch_req, tc,
                                                  BaseTLB::Execute);
diff --git a/src/cpu/simple/timing.cc b/src/cpu/simple/timing.cc
index 9253d8005..7996a6ddd 100644
--- a/src/cpu/simple/timing.cc
+++ b/src/cpu/simple/timing.cc
@@ -415,6 +415,8 @@ TimingSimpleCPU::readMem(Addr addr, uint8_t *data,
     RequestPtr req  = new Request(asid, addr, size,
                                   flags, dataMasterId(), pc, _cpuId, tid);
 
+    req->taskId(taskId());
+
     Addr split_addr = roundDown(addr + size - 1, block_size);
     assert(split_addr <= addr || split_addr - addr < block_size);
 
@@ -484,6 +486,8 @@ TimingSimpleCPU::writeMem(uint8_t *data, unsigned size,
     RequestPtr req = new Request(asid, addr, size,
                                  flags, dataMasterId(), pc, _cpuId, tid);
 
+    req->taskId(taskId());
+
     Addr split_addr = roundDown(addr + size - 1, block_size);
     assert(split_addr <= addr || split_addr - addr < block_size);
 
@@ -561,6 +565,7 @@ TimingSimpleCPU::fetch()
     if (needToFetch) {
         _status = BaseSimpleCPU::Running;
         Request *ifetch_req = new Request();
+        ifetch_req->taskId(taskId());
         ifetch_req->setThreadContext(_cpuId, /* thread ID */ 0);
         setupFetchRequest(ifetch_req);
         DPRINTF(SimpleCPU, "Translating address %#x\n", ifetch_req->getVaddr());