2010/5/12 Richard Henderson <r...@twiddle.net>: > On 05/11/2010 02:31 PM, Artyom Tarasenko wrote: >> Nack. It looks like you reverted carry generation to the previous >> (broken) behavior. > > Oh? I suppose I should go back and look at the logs, but the way > it's written there sure seems to match 5.1.5.1 of the sparcv9 manual: > You'll only get carry into the high bit on X - Y if X < Y. > > In any case, if you meant to fix carry computation for subtraction, > you missed changing the 64-bit carry computation too.
May very well be the case. I cared only about sparcv8. Was sort of expecting that someone would stop me telling I broke v9... > It is still > written the same way before and after my patch, and matches both > the expectation I have from the v9 manual and the post-patch code > along the 32-bit path. Probably v9 differs from v8? > Perhaps you could point out the change I'm reverting? I don't > see any change to the actual computation of the flags since > f0f26a06d51b7e7764f8951cdbf67ac9ad507f6d, last year. It is last year, but 3e6ba503400c34cbe0f9ad6e289921688bf303a3 -- Regards, Artyom Tarasenko solaris/sparc under qemu blog: http://tyom.blogspot.com/