Factor out pre-DataPath pure pipelines into Spine module
laforest opened this issue · comments
Related to #31.
Once the I/O Predication is factored out, there is no point in having stage 1-3 empty pipelines in the DataPath, so factor out the pipelines of things that don't get processed, like the opcode, into a "spine" which runs parallel to the other sub-systems and just carries stuff along.
The addition of the AOM/BTM has functionally already done this, but it's not explicit in the code.
Having an explicit spine pipeline, as shown in fig. 6.4, pg 113 in the thesis, would make it clearer in the code from/to which pipeline stages information moves.