75 lines · plain
1Date: Sat, 18 Nov 2000 09:19:35 -0600 (CST)2From: Vikram Adve <vadve@cs.uiuc.edu>3To: Chris Lattner <lattner@cs.uiuc.edu>4Subject: a few thoughts5 6I've been mulling over the virtual machine problem and I had some7thoughts about some things for us to think about discuss:8 91. We need to be clear on our goals for the VM. Do we want to emphasize10 portability and safety like the Java VM? Or shall we focus on the11 architecture interface first (i.e., consider the code generation and12 processor issues), since the architecture interface question is also13 important for portable Java-type VMs?14 15 This is important because the audiences for these two goals are very16 different. Architects and many compiler people care much more about17 the second question. The Java compiler and OS community care much more18 about the first one.19 20 Also, while the architecture interface question is important for21 Java-type VMs, the design constraints are very different.22 23 242. Design issues to consider (an initial list that we should continue25 to modify). Note that I'm not trying to suggest actual solutions here,26 but just various directions we can pursue:27 28 a. A single-assignment VM, which we've both already been thinking about.29 30 b. A strongly-typed VM. One question is do we need the types to be31 explicitly declared or should they be inferred by the dynamic compiler?32 33 c. How do we get more high-level information into the VM while keeping34 to a low-level VM design?35 36 o Explicit array references as operands? An alternative is37 to have just an array type, and let the index computations be38 separate 3-operand instructions.39 40 o Explicit instructions to handle aliasing, e.g.s:41 -- an instruction to say "I speculate that these two values are not42 aliased, but check at runtime", like speculative execution in43 EPIC?44 -- or an instruction to check whether two values are aliased and45 execute different code depending on the answer, somewhat like46 predicated code in EPIC47 48 o (This one is a difficult but powerful idea.)49 A "thread-id" field on every instruction that allows the static50 compiler to generate a set of parallel threads, and then have51 the runtime compiler and hardware do what they please with it.52 This has very powerful uses, but thread-id on every instruction53 is expensive in terms of instruction size and code size.54 We would need to compactly encode it somehow.55 56 Also, this will require some reading on at least two other57 projects:58 -- Multiscalar architecture from Wisconsin59 -- Simultaneous multithreading architecture from Washington60 61 o Or forget all this and stick to a traditional instruction set?62 63 64BTW, on an unrelated note, after the meeting yesterday, I did remember65that you had suggested doing instruction scheduling on SSA form instead66of a dependence DAG earlier in the semester. When we talked about67it yesterday, I didn't remember where the idea had come from but I68remembered later. Just giving credit where its due...69 70Perhaps you can save the above as a file under RCS so you and I can71continue to expand on this.72 73--Vikram74 75