refactor: remove excessive chunk_max_size calls

These calls are virtual and cause notable overhead

fesvr/htif.cc

@@ -277,8 +277,8 @@ void htif_t::clear_chunk(addr_t taddr, size_t len)
 {
   std::vector<uint8_t> zeros(chunk_max_size(), 0);
 
-  for (size_t pos = 0; pos < len; pos += chunk_max_size())
+  for (size_t pos = 0; pos < len; pos += zeros.size())
-    write_chunk(taddr + pos, std::min(len - pos, chunk_max_size()), &zeros[0]);
+    write_chunk(taddr + pos, std::min(len - pos, zeros.size()), &zeros[0]);
 }
 
 int htif_t::run()

fesvr/memif.cc

@@ -35,8 +35,9 @@ void memif_t::read(addr_t addr, size_t len, void* bytes)
   }
 
   // now we're aligned
-  for (size_t pos = 0; pos < len; pos += cmemif->chunk_max_size())
+  size_t max_chunk = cmemif->chunk_max_size();
-    cmemif->read_chunk(addr + pos, std::min(cmemif->chunk_max_size(), len - pos), (char*)bytes + pos);
+  for (size_t pos = 0; pos < len; pos += max_chunk)
+    cmemif->read_chunk(addr + pos, std::min(max_chunk, len - pos), (char*)bytes + pos);
 }
 
 void memif_t::write(addr_t addr, size_t len, const void* bytes)