Path: utzoo!attcan!uunet!mcsun!ukc!inmos!elberton!des From: des@elberton.inmos.co.uk (David Shepherd) Newsgroups: comp.arch Subject: Re: Compiler Costs Message-ID: <8352@ganymede.inmos.co.uk> Date: 12 Jul 90 15:56:00 GMT References: <23268@boulder.Colorado.EDU> <23285@boulder.Colorado.EDU> <9743@brazos.Rice.edu> Sender: news@inmos.co.uk Reply-To: des@elberton.inmos.co.uk (David Shepherd) Organization: INMOS architecture group Lines: 28 In article <9743@brazos.Rice.edu>, cliffc@sicilia.rice.edu (Cliff Click) writes: |> In article meissner@osf.org (Michael Meissner) writes: |> >I also wonder about the computational complexity of global register |> >analysis. |> |> Global register allocation is NP-complete. |> Some folks at Rice here experimenting with a nameless commerical compiler |> discovered a simple 2000-line fortran program that took 90+ HOURS to compile |> on a 10+ Mips, 16 Meg workstation. The program spent roughly 98% of it's time |> thrashing the disk. The algorithm the compiler used required exponentional |> memory space. The same compile took ~15 minutes on a 48 Meg workstation. When compiling one of the files to build the HOL theorem prover the C code that AKCL produces from the lisp "compilation" of the original ML code (ok so we're not talking about normal "user code" here) takes > 24 hours to compile on -O2. Without optimization it takes about 5 mins. I can't remember exactly but I don't think it was actually trashing the disk that much -- i.e. cpu seemed solid at 100%. -------------------------------------------------------------------------- David Shepherd: des@inmos.co.uk or des@inmos.com tel: 0454-616616 x 529 INMOS ltd, 1000 Aztec West, Almondsbury, Bristol, BS12 4SQ