Kenneth Zadeck <zad...@naturalbridge.com> writes: > On 11/01/2013 04:46 AM, Richard Sandiford wrote: >> I'm building one target for each supported CPU and comparing the wide-int >> assembly output of gcc.c-torture, gcc.dg and g++.dg with the corresponding >> output from the merge point. This patch removes all the differences I saw >> for alpha-linux-gnu in gcc.c-torture. >> >> Hunk 1: Preserve the current trunk behaviour in which the shift count >> is truncated if SHIFT_COUNT_TRUNCATED and not otherwise. This was by >> inspection after hunk 5. > i used to do this inside of wide-int so that i would get consistent > behavior for all clients, but i got beat up on it.
Right :-) SHIFT_COUNT_TRUNCATED is a bad enough interface as it is. Putting into a utility class like wide-int just makes it worse. This is something that clients _should_ think about if they want to emulate target behaviour (rather than doing "natural" maths). Not sure -- is the patch OK, or are you deferring judgement for now? Thanks, Richard
