[solarexplorer]

In the same section the part about the SiFive optimization is also misleading. The goal of the optimization is obviously to avoid interrupting the instruction fetch. But he makes it sound like the goal was to reduce instruction count by fusing two instructions to get a single monster op with five (!) register operands. That just doesn't make sense.

[brucehoult]

They don't fuse it. They have the branch and the following instruction both go down one integer pipeline just as they would if the branch was predicted "not taken. They are linked/tagged together such that if the branch actually happens then the result from the second instruction is simply not written back instead of taking a branch mispredict.

So the four input operands and one output operand are not a problem because that's just what two pipelines do all the time anyway.