summaryrefslogtreecommitdiff
path: root/src/mem/protocol/MOESI_hammer-cache.sm
diff options
context:
space:
mode:
authorNilay Vaish ext:(%2C%20Malek%20Musleh%20%3Cmalek.musleh%40gmail.com%3E) <nilay@cs.wisc.edu>2013-05-21 11:31:31 -0500
committerNilay Vaish ext:(%2C%20Malek%20Musleh%20%3Cmalek.musleh%40gmail.com%3E) <nilay@cs.wisc.edu>2013-05-21 11:31:31 -0500
commit59a7abff29aa5a687e1693f003c20d7e2000c40a (patch)
treee1cf2cf822cf5b1002a6b72d8d613f65e0e1df8d /src/mem/protocol/MOESI_hammer-cache.sm
parentd3c33d91b68e917478dba48c03a674b21ebd2747 (diff)
downloadgem5-59a7abff29aa5a687e1693f003c20d7e2000c40a.tar.xz
ruby: add stats to .sm files, remove cache profiler
This patch changes the way cache statistics are collected in ruby. As of now, there is separate entity called CacheProfiler which holds statistical variables for caches. The CacheMemory class defines different functions for accessing the CacheProfiler. These functions are then invoked in the .sm files. I find this approach opaque and prone to error. Secondly, we probably should not be paying the cost of a function call for recording statistics. Instead, this patch allows for accessing statistical variables in the .sm files. The collection would become transparent. Secondly, it would happen in place, so no function calls. The patch also removes the CacheProfiler class. --HG-- rename : src/mem/slicc/ast/InfixOperatorExprAST.py => src/mem/slicc/ast/OperatorExprAST.py
Diffstat (limited to 'src/mem/protocol/MOESI_hammer-cache.sm')
-rw-r--r--src/mem/protocol/MOESI_hammer-cache.sm155
1 files changed, 114 insertions, 41 deletions
diff --git a/src/mem/protocol/MOESI_hammer-cache.sm b/src/mem/protocol/MOESI_hammer-cache.sm
index bc3b700d3..3680294de 100644
--- a/src/mem/protocol/MOESI_hammer-cache.sm
+++ b/src/mem/protocol/MOESI_hammer-cache.sm
@@ -1231,17 +1231,28 @@ machine(L1Cache, "AMD Hammer-like protocol")
}
}
- action(uu_profileMiss, "\u", desc="Profile the demand miss") {
- peek(mandatoryQueue_in, RubyRequest) {
- if (L1IcacheMemory.isTagPresent(address)) {
- L1IcacheMemory.profileMiss(in_msg);
- } else if (L1DcacheMemory.isTagPresent(address)) {
- L1DcacheMemory.profileMiss(in_msg);
- }
- if (L2cacheMemory.isTagPresent(address) == false) {
- L2cacheMemory.profileMiss(in_msg);
- }
- }
+ action(uu_profileL1DataMiss, "\udm", desc="Profile the demand miss") {
+ ++L1DcacheMemory.demand_misses;
+ }
+
+ action(uu_profileL1DataHit, "\udh", desc="Profile the demand hits") {
+ ++L1DcacheMemory.demand_hits;
+ }
+
+ action(uu_profileL1InstMiss, "\uim", desc="Profile the demand miss") {
+ ++L1IcacheMemory.demand_misses;
+ }
+
+ action(uu_profileL1InstHit, "\uih", desc="Profile the demand hits") {
+ ++L1IcacheMemory.demand_hits;
+ }
+
+ action(uu_profileL2Miss, "\um", desc="Profile the demand miss") {
+ ++L2cacheMemory.demand_misses;
+ }
+
+ action(uu_profileL2Hit, "\uh", desc="Profile the demand hits ") {
+ ++L2cacheMemory.demand_hits;
}
action(zz_stallAndWaitMandatoryQueue, "\z", desc="Send the head of the mandatory queue to the back of the queue.") {
@@ -1317,7 +1328,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
ii_allocateL1DCacheBlock;
nb_copyFromTBEToL1; // Not really needed for state I
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1328,7 +1338,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
ii_allocateL1DCacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1339,7 +1348,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
ii_allocateL1DCacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1350,7 +1358,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
ii_allocateL1DCacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1361,7 +1368,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
ii_allocateL1DCacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1372,7 +1378,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
jj_allocateL1ICacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1383,7 +1388,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
jj_allocateL1ICacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1394,7 +1398,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
jj_allocateL1ICacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1405,7 +1408,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
jj_allocateL1ICacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1416,7 +1418,6 @@ machine(L1Cache, "AMD Hammer-like protocol")
jj_allocateL1ICacheBlock;
nb_copyFromTBEToL1;
s_deallocateTBE;
- uu_profileMiss;
zz_stallAndWaitMandatoryQueue;
ll_L2toL1Transfer;
}
@@ -1447,34 +1448,36 @@ machine(L1Cache, "AMD Hammer-like protocol")
}
// Transitions from Idle
- transition({I, IR}, Load, IS) {
+ transition({I,IR}, Load, IS) {
ii_allocateL1DCacheBlock;
i_allocateTBE;
a_issueGETS;
- uu_profileMiss;
+ uu_profileL1DataMiss;
+ uu_profileL2Miss;
k_popMandatoryQueue;
}
- transition({I, IR}, Ifetch, IS) {
+ transition({I,IR}, Ifetch, IS) {
jj_allocateL1ICacheBlock;
i_allocateTBE;
a_issueGETS;
- uu_profileMiss;
+ uu_profileL1InstMiss;
+ uu_profileL2Miss;
k_popMandatoryQueue;
}
- transition({I, IR}, Store, IM) {
+ transition({I,IR}, Store, IM) {
ii_allocateL1DCacheBlock;
i_allocateTBE;
b_issueGETX;
- uu_profileMiss;
+ uu_profileL1DataMiss;
+ uu_profileL2Miss;
k_popMandatoryQueue;
}
transition({I, IR}, Flush_line, IM_F) {
it_allocateTBE;
bf_issueGETF;
- uu_profileMiss;
k_popMandatoryQueue;
}
@@ -1489,28 +1492,45 @@ machine(L1Cache, "AMD Hammer-like protocol")
}
// Transitions from Shared
- transition({S, SM, ISM}, {Load, Ifetch}) {
+ transition({S, SM, ISM}, Load) {
+ h_load_hit;
+ uu_profileL1DataHit;
+ k_popMandatoryQueue;
+ }
+
+ transition({S, SM, ISM}, Ifetch) {
h_load_hit;
+ uu_profileL1InstHit;
k_popMandatoryQueue;
}
- transition(SR, {Load, Ifetch}, S) {
+ transition(SR, Load, S) {
h_load_hit;
+ uu_profileL1DataMiss;
+ uu_profileL2Hit;
k_popMandatoryQueue;
ka_wakeUpAllDependents;
}
- transition({S, SR}, Store, SM) {
+ transition(SR, Ifetch, S) {
+ h_load_hit;
+ uu_profileL1InstMiss;
+ uu_profileL2Hit;
+ k_popMandatoryQueue;
+ ka_wakeUpAllDependents;
+ }
+
+ transition({S,SR}, Store, SM) {
i_allocateTBE;
b_issueGETX;
- uu_profileMiss;
+ uu_profileL1DataMiss;
+ uu_profileL2Miss;
k_popMandatoryQueue;
}
transition({S, SR}, Flush_line, SM_F) {
i_allocateTBE;
bf_issueGETF;
- uu_profileMiss;
forward_eviction_to_cpu;
gg_deallocateL1CacheBlock;
k_popMandatoryQueue;
@@ -1534,29 +1554,47 @@ machine(L1Cache, "AMD Hammer-like protocol")
}
// Transitions from Owned
- transition({O, OM, SS, MM_W, M_W}, {Load, Ifetch}) {
+ transition({O, OM, SS, MM_W, M_W}, {Load}) {
h_load_hit;
+ uu_profileL1DataHit;
k_popMandatoryQueue;
}
- transition(OR, {Load, Ifetch}, O) {
+ transition({O, OM, SS, MM_W, M_W}, {Ifetch}) {
h_load_hit;
+ uu_profileL1InstHit;
+ k_popMandatoryQueue;
+ }
+
+ transition(OR, Load, O) {
+ h_load_hit;
+ uu_profileL1DataMiss;
+ uu_profileL2Hit;
+ k_popMandatoryQueue;
+ ka_wakeUpAllDependents;
+ }
+
+ transition(OR, Ifetch, O) {
+ h_load_hit;
+ uu_profileL1InstMiss;
+ uu_profileL2Hit;
k_popMandatoryQueue;
ka_wakeUpAllDependents;
}
- transition({O, OR}, Store, OM) {
+ transition({O,OR}, Store, OM) {
i_allocateTBE;
b_issueGETX;
p_decrementNumberOfMessagesByOne;
- uu_profileMiss;
+ uu_profileL1DataMiss;
+ uu_profileL2Miss;
k_popMandatoryQueue;
}
+
transition({O, OR}, Flush_line, OM_F) {
i_allocateTBE;
bf_issueGETF;
p_decrementNumberOfMessagesByOne;
- uu_profileMiss;
forward_eviction_to_cpu;
gg_deallocateL1CacheBlock;
k_popMandatoryQueue;
@@ -1587,24 +1625,44 @@ machine(L1Cache, "AMD Hammer-like protocol")
}
// Transitions from Modified
- transition({MM, M}, {Load, Ifetch}) {
+ transition({MM, M}, {Ifetch}) {
h_load_hit;
+ uu_profileL1InstHit;
+ k_popMandatoryQueue;
+ }
+
+ transition({MM, M}, {Load}) {
+ h_load_hit;
+ uu_profileL1DataHit;
k_popMandatoryQueue;
}
transition(MM, Store) {
hh_store_hit;
+ uu_profileL1DataHit;
+ k_popMandatoryQueue;
+ }
+
+ transition(MMR, Load, MM) {
+ h_load_hit;
+ uu_profileL1DataMiss;
+ uu_profileL2Hit;
k_popMandatoryQueue;
+ ka_wakeUpAllDependents;
}
- transition(MMR, {Load, Ifetch}, MM) {
+ transition(MMR, Ifetch, MM) {
h_load_hit;
+ uu_profileL1InstMiss;
+ uu_profileL2Hit;
k_popMandatoryQueue;
ka_wakeUpAllDependents;
}
transition(MMR, Store, MM) {
hh_store_hit;
+ uu_profileL1DataMiss;
+ uu_profileL2Hit;
k_popMandatoryQueue;
ka_wakeUpAllDependents;
}
@@ -1662,17 +1720,30 @@ machine(L1Cache, "AMD Hammer-like protocol")
// Transitions from Dirty Exclusive
transition(M, Store, MM) {
hh_store_hit;
+ uu_profileL1DataHit;
k_popMandatoryQueue;
}
- transition(MR, {Load, Ifetch}, M) {
+ transition(MR, Load, M) {
+ h_load_hit;
+ uu_profileL1DataMiss;
+ uu_profileL2Hit;
+ k_popMandatoryQueue;
+ ka_wakeUpAllDependents;
+ }
+
+ transition(MR, Ifetch, M) {
h_load_hit;
+ uu_profileL1InstMiss;
+ uu_profileL2Hit;
k_popMandatoryQueue;
ka_wakeUpAllDependents;
}
transition(MR, Store, MM) {
hh_store_hit;
+ uu_profileL1DataMiss;
+ uu_profileL2Hit;
k_popMandatoryQueue;
ka_wakeUpAllDependents;
}
@@ -1947,6 +2018,7 @@ machine(L1Cache, "AMD Hammer-like protocol")
transition(MM_W, Store) {
hh_store_hit;
+ uu_profileL1DataHit;
k_popMandatoryQueue;
}
@@ -1972,6 +2044,7 @@ machine(L1Cache, "AMD Hammer-like protocol")
transition(M_W, Store, MM_W) {
hh_store_hit;
+ uu_profileL1DataHit;
k_popMandatoryQueue;
}