Path: utzoo!utgpu!water!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!mailrus!ames!joyce!sri-unix!garth!smryan From: smryan@garth.UUCP (Steven Ryan) Newsgroups: comp.arch Subject: Re: m88000 benchmarks (LONG) Message-ID: <886@garth.UUCP> Date: 6 Jul 88 21:04:38 GMT References: <1359@claude.oakhill.UUCP> <9442@lll-winken.llnl.gov> Reply-To: smryan@garth.UUCP (Steven Ryan) Organization: INTERGRAPH (APD) -- Palo Alto, CA Lines: 7 >For even better pipeline utilization from compiler B, the public domain >compiler, change the register allocation strategy so that when looking >for a scratch register is remembers the last one allocated and starts >upward folding around to R0 when it hits the top of the register file. >By doing this the false dependencies caused by register reuse will be Or do the scheduling before assigning registers (FTN200 EBBO).