O3PCU: Split loads and stores that cross cache line boundaries.

When each load or store is sent to the LSQ, we check whether it will cross a
cache line boundary and, if so, split it in two. This creates two TLB
translations and two memory requests. Care has to be taken if the first
packet of a split load is sent but the second blocks the cache. Similarly,
for a store, if the first packet cannot be sent, we must store the second
one somewhere to retry later.

This modifies the LSQSenderState class to record both packets in a split
load or store.

Finally, a new const variable, HasUnalignedMemAcc, is added to each ISA
to indicate whether unaligned memory accesses are allowed. This is used
throughout the changed code so that compiler can optimise away code dealing
with split requests for ISAs that don't need them.

1 files changed, 11 insertions, 4 deletions

diff --git a/src/cpu/o3/cpu.hh b/src/cpu/o3/cpu.hh
index 2ea918983..82d4ca25b 100644
--- a/src/cpu/o3/cpu.hh
+++ b/src/cpu/o3/cpu.hh
@@ -703,18 +703,25 @@ class FullO3CPU : public BaseO3CPU
 
     /** CPU read function, forwards read to LSQ. */
     template <class T>
-    Fault read(RequestPtr &req, T &data, int load_idx)
+    Fault read(RequestPtr &req, RequestPtr &sreqLow, RequestPtr &sreqHigh,
+               T &data, int load_idx)
     {
-        return this->iew.ldstQueue.read(req, data, load_idx);
+        return this->iew.ldstQueue.read(req, sreqLow, sreqHigh,
+                                        data, load_idx);
     }
 
     /** CPU write function, forwards write to LSQ. */
     template <class T>
-    Fault write(RequestPtr &req, T &data, int store_idx)
+    Fault write(RequestPtr &req, RequestPtr &sreqLow, RequestPtr &sreqHigh,
+                T &data, int store_idx)
     {
-        return this->iew.ldstQueue.write(req, data, store_idx);
+        return this->iew.ldstQueue.write(req, sreqLow, sreqHigh,
+                                         data, store_idx);
     }
 
+    /** Get the dcache port (used to find block size for translations). */
+    Port *getDcachePort() { return this->iew.ldstQueue.getDcachePort(); }
+
     Addr lockAddr;
 
     /** Temporary fix for the lock flag, works in the UP case. */