1Date: Fri, 6 Jul 2001 16:56:56 -0500 2From: Vikram S. Adve <vadve@cs.uiuc.edu> 3To: Chris Lattner <lattner@cs.uiuc.edu> 4Subject: lowering the IR 5 6BTW, I do think that we should consider lowering the IR as you said. I 7didn't get time to raise it today, but it comes up with the SPARC 8move-conditional instruction. I don't think we want to put that in the core 9VM -- it is a little too specialized. But without a corresponding 10conditional move instruction in the VM, it is pretty difficult to maintain a 11close mapping between VM and machine code. Other architectures may have 12other such instructions. 13 14What I was going to suggest was that for a particular processor, we define 15additional VM instructions that match some of the unusual opcodes on the 16processor but have VM semantics otherwise, i.e., all operands are in SSA 17form and typed. This means that we can re-generate core VM code from the 18more specialized code any time we want (so that portability is not lost). 19 20Typically, a static compiler like gcc would generate just the core VM, which 21is relatively portable. Anyone (an offline tool, the linker, etc., or even 22the static compiler itself if it chooses) can transform that into more 23specialized target-specific VM code for a particular architecture. If the 24linker does it, it can do it after all machine-independent optimizations. 25This would be the most convenient, but not necessary. 26 27The main benefit of lowering will be that we will be able to retain a close 28mapping between VM and machine code. 29 30--Vikram 31 32