Revamp cache timing access mshr check to make stats sane again.

--HG--
extra : convert_revision : 37009b8ee536807073b5a5ca07ed1d097a496aea

diff --git a/src/mem/cache/blk.hh b/src/mem/cache/blk.hh
index bafb46a89051be7cbdc8a171b387a9b83e17be95..127c547ac82d25e514f4053c3de24830381cbd2e 100644 (file)
--- a/src/mem/cache/blk.hh
+++ b/src/mem/cache/blk.hh
@@ -51,8 +51,10 @@ enum CacheBlkStatusBits {
     BlkValid =         0x01,
     /** write permission */
     BlkWritable =      0x02,
+    /** read permission (yes, block can be valid but not readable) */
+    BlkReadable =      0x04,
     /** dirty (modified) */
-    BlkDirty =         0x04,
+    BlkDirty =         0x08,
     /** block was referenced */
     BlkReferenced =    0x10,
     /** block was a hardware prefetch yet unaccessed*/
@@ -162,7 +164,19 @@ class CacheBlk
     }
 
     /**
-     * Checks that a block is valid (readable).
+     * Checks the read permissions of this block.  Note that a block
+     * can be valid but not readable if there is an outstanding write
+     * upgrade miss.
+     * @return True if the block is readable.
+     */
+    bool isReadable() const
+    {
+        const int needed_bits = BlkReadable | BlkValid;
+        return (status & needed_bits) == needed_bits;
+    }
+
+    /**
+     * Checks that a block is valid.
      * @return True if the block is valid.
      */
     bool isValid() const

diff --git a/src/mem/cache/cache_impl.hh b/src/mem/cache/cache_impl.hh
index 15de76532f4ced30a9eca5e1a5b47460d7ca4de0..e546e2a9ae13ac6d81fc6568013efcd49397267e 100644 (file)
--- a/src/mem/cache/cache_impl.hh
+++ b/src/mem/cache/cache_impl.hh
@@ -292,7 +292,7 @@ Cache<TagStore>::access(PacketPtr pkt, BlkType *&blk, int &lat)
             }
         }
 
-        if (pkt->needsExclusive() ? blk->isWritable() : blk->isValid()) {
+        if (pkt->needsExclusive() ? blk->isWritable() : blk->isReadable()) {
             // OK to satisfy access
             hits[pkt->cmdToIndex()][0/*pkt->req->getThreadNum()*/]++;
             satisfyCpuSideRequest(pkt, blk);
@@ -318,7 +318,7 @@ Cache<TagStore>::access(PacketPtr pkt, BlkType *&blk, int &lat)
                 incMissCount(pkt);
                 return false;
             }
-            blk->status = BlkValid;
+            blk->status = BlkValid | BlkReadable;
         }
         std::memcpy(blk->data, pkt->getPtr<uint8_t>(), blkSize);
         blk->status |= BlkDirty;
@@ -422,21 +422,8 @@ Cache<TagStore>::timingAccess(PacketPtr pkt)
     }
 
     int lat = hitLatency;
-    bool satisfied = false;
-
-    Addr blk_addr = pkt->getAddr() & ~(Addr(blkSize-1));
-    MSHR *mshr = mshrQueue.findMatch(blk_addr);
-
-    if (!mshr) {
-        // no outstanding access to this block, look up in cache
-        // (otherwise if we allow reads while there's an outstanding
-        // write miss, the read could return stale data out of the
-        // cache block... a more aggressive system could detect the
-        // overlap (if any) and forward data out of the MSHRs, but we
-        // don't do that yet)
-        BlkType *blk = NULL;
-        satisfied = access(pkt, blk, lat);
-    }
+    BlkType *blk = NULL;
+    bool satisfied = access(pkt, blk, lat);
 
 #if 0
     /** @todo make the fast write alloc (wh64) work with coherence. */
@@ -483,6 +470,9 @@ Cache<TagStore>::timingAccess(PacketPtr pkt)
         if (prefetchMiss)
             prefetcher->handleMiss(pkt, time);
 
+        Addr blk_addr = pkt->getAddr() & ~(Addr(blkSize-1));
+        MSHR *mshr = mshrQueue.findMatch(blk_addr);
+
         if (mshr) {
             // MSHR hit
             //@todo remove hw_pf here
@@ -508,6 +498,26 @@ Cache<TagStore>::timingAccess(PacketPtr pkt)
             if (pkt->cmd == MemCmd::Writeback) {
                 allocateWriteBuffer(pkt, time, true);
             } else {
+                if (blk && blk->isValid()) {
+                    // If we have a write miss to a valid block, we
+                    // need to mark the block non-readable.  Otherwise
+                    // if we allow reads while there's an outstanding
+                    // write miss, the read could return stale data
+                    // out of the cache block... a more aggressive
+                    // system could detect the overlap (if any) and
+                    // forward data out of the MSHRs, but we don't do
+                    // that yet.  Note that we do need to leave the
+                    // block valid so that it stays in the cache, in
+                    // case we get an upgrade response (and hence no
+                    // new data) when the write miss completes.
+                    // As long as CPUs do proper store/load forwarding
+                    // internally, and have a sufficiently weak memory
+                    // model, this is probably unnecessary, but at some
+                    // point it must have seemed like we needed it...
+                    assert(pkt->needsExclusive() && !blk->isWritable());
+                    blk->status &= ~BlkReadable;
+                }
+
                 allocateMissBuffer(pkt, time, true);
             }
         }
@@ -934,10 +944,10 @@ Cache<TagStore>::handleFill(PacketPtr pkt, BlkType *blk,
     }
 
     if (!pkt->sharedAsserted()) {
-        blk->status = BlkValid | BlkWritable;
+        blk->status = BlkValid | BlkReadable | BlkWritable;
     } else {
         assert(!pkt->needsExclusive());
-        blk->status = BlkValid;
+        blk->status = BlkValid | BlkReadable;
     }
 
     DPRINTF(Cache, "Block addr %x moving from state %i to %i\n",