The cache in question is not something which stores the result of these speculatively executed instructions, but the normal L1-L2-L3 caches we are used to. The

The cache in question is not something which stores the result of these speculatively executed instructions, but the normal L1-L2-L3 caches we are used to. The result of these instructions is discarded, but as a side effect, they may load something from memory into the cache. The exploit detects whether or not a particular memory address was loaded into the cache (reading from something already in the cache is much faster than reading from the main memory).