-
Notifications
You must be signed in to change notification settings - Fork 0
M8: MIR and below #17
Copy link
Copy link
Open
Labels
area/codegenSelectionDAG, GlobalISel, regalloc, MC, the back endSelectionDAG, GlobalISel, regalloc, MC, the back endkind/milestoneTracking issue for a whole milestoneTracking issue for a whole milestonepriority/p1Needed this milestoneNeeded this milestone
Milestone
Description
Activity
Metadata
Metadata
Assignees
Labels
area/codegenSelectionDAG, GlobalISel, regalloc, MC, the back endSelectionDAG, GlobalISel, regalloc, MC, the back endkind/milestoneTracking issue for a whole milestoneTracking issue for a whole milestonepriority/p1Needed this milestoneNeeded this milestone
Seven weeks. Part VII, from MIR down to the MC layer. Ten lessons, five blueprints, and the register allocation material that most people never get a clear picture of.
Exit criterion. M06's boss fight, predict the allocation and then force a spill and explain it, is completed by testers who had never read a MIR dump before Part VII.
Tasks
BP-CG-MIR,BP-CG-REGALLOC,BP-CG-FRAME,BP-CG-CC,BP-CG-MCirx.mirandirx.mcaMIRWalker,IntervalChartandMCATimelinewidgetsGates
llvm-exegesisavailability at E0 is settled one way or the other and the affected lessons say what they sayllvm-exegesisneeds performance counters, which a Colab VM may or may not expose. Settle it, do not leave it as a maybe, and if the answer is no then the lessons that wanted it say so plainly and use a recording instead of pretending.