In Intel Intrinsics Guide, latency and throughput are given as mesurements of performance of instructions. According to Intel’s Measuring Instruction Latency and Throughput:

• Latency is the number of processor clocks it takes for an instruction to have its data available for use by another instruction. Therefore, an instruction which has a latency of 6 clocks will have its data available for another instruction that many clocks after it starts its execution.

• Throughput is the number of processor clocks it takes for an instruction to execute or perform its calculations. An instruction with a throughput of 2 clocks would tie up its execution unit for that many cycles which prevents an instruction needing that execution unit from being executed. Only after the instruction is done with the execution unit can the next instruction enter.

So both of them have the unit of cycles/instruction (some people use instructions/cycle when describing throughput, which is more similar to what throughput means in real life, i.e. some quantity per amount of time).

Usually an instruction’s latency is larger than its throughput so you can use latency as the wrost case measurement. However, also according to Intel’s Measuring Instruction Latency and Throughput:

• If a processor has multiple execution units for a certain type of operation then the throughput of an instruction using that execution unit will effectively be divided by the number of those execution units because multiple instructions using those units can be executing at once.

That means, multiple instructions may be issued at the same time so if you write your program properly you can get the maximum throughput out of Intel’s CPU. For instance, on Haswell, the latency of _mm256_blend_ps() is 1, so it takes 1 cycle for a sigle instruction to complete. However, the throughput of this instruction is only 0.33, which means there’re 3 execution units that can perform this instruction in parallel. Therefore, if you can write you code in a way that every three blend instrutions are issued together it would effectively take only 1 cycle for these three instructions to complete.