X-Authentication-Warning: delorie.com: mailnull set sender to djgpp-workers-bounces using -f Date: Sun, 9 Dec 2001 14:13:30 +0200 (IST) From: Eli Zaretskii X-Sender: eliz AT is To: Richard Dawe cc: djgpp-workers AT delorie DOT com Subject: Re: libemu and profiling (LONG) [Was: Re: Building a profiled version of libc] In-Reply-To: <3C134C9D.D057D932@phekda.freeserve.co.uk> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Reply-To: djgpp-workers AT delorie DOT com Errors-To: nobody AT delorie DOT com X-Mailing-List: djgpp-workers AT delorie DOT com X-Unsubscribes-To: listserv AT delorie DOT com Precedence: bulk On Sun, 9 Dec 2001, Richard Dawe wrote: > Below is a diff for the fixes for src/libemu. OK > to commit? Yes, thanks. > Using the math co-pro (387 not set or set to y) gives a result of > 0.122385; using a patched emu387.dxe gives a result of 0.1. Shouldn't the > results be the same or at least a little closer? They should be very close. Does the same happen with other argument values to atan? What about sin and cos? Also, could you please try the same on plain DOS? I don't trust the Windows DPMI server as far as FP emulation is considered. > Also if I run the test with 387=y, rerun the test with 387=n and finally > rerun the test with set 387=y, I get an 'Exiting due to signal SIGNOFP' > error. That's a known issue: when you run with 387=n, the startup code issues a DPMI call that turns the emulation on, and the emulation stays on until switched off. But our startup code doesn't switch the emulation off unless 387 is not set at all, so it stays on in the DPMI server, but our program thinks the emulation is off. The solution is to run the program without 387 being set at all, instead of setting it to y.