Sequential Consistency

Result of the execution is as if all operations were executed in some order

Operations of each processor occured in the order of the prgram

Two requirements

Maintaining program order on individual processors
Ensuring write atomicity
- If any processor reads the result of the write, all subsequent reads will be the same

Limitations

Hardware optimizations

Complicates write buffers
Can’t reorder overlapping write operations
Complicates non-blocking reads
- Cannot prefetch data
Makes Cache Coherence more expensive
- Must delay write completion until invalidation and update of old write

Compiler optimizations

Code motion
- Moving code around to improve performance
- e.g. load-use hazard
Loop blocking
- Rearrange loops for better cache performance
  - keep inner loop completely in cache
Software pipelining
- Move instructions across iterations of a loop

Example

for (i = 0; i<N; i += 2) {
  f(i);
  f(i+1);
  g(i);
  g(i+1);

}