On Jul 14, 2012, at 3:24 AM, David Chisnall wrote: > On 13 Jul 2012, at 17:40, Warner Losh wrote: > >> We shouldn't be gating the new math on an issue that only affects sparc64 machines > > Mostly agreed, but it's worth noting that the APCS for ARMv8[1] specifies that long double should be IEEE 754- 2008 quad precision. I think ARMv8 is going to be an important platform for us in the next few years, so ld128 is not just for SPARC (I'd also like us to switch to ld128 on PowerPC, since that's the ABI that everyone else uses, but I don't anticipate PowerPC becoming tier 1 any time soon, whereas I would very much like to see ARMv8 become tier 1 within a year of shipping silicon). > > That's not to say that we should hold things up waiting for ld128 versions, just that adding ld128 versions soon after adding ld80 ones would be very helpful. Agreed. My point was more "don't gate everything on the 128-bit versions" not "hey, nobody do this, it is stupid." If the 128-bit versions are ready at the same time as the 80-bit versions, so much the better. Warner > David > > [1] http://infocenter.arm.com/help/topic/com.arm.doc.ihi0055a/IHI0055A_aapcs64.pdfReceived on Sat Jul 14 2012 - 14:12:46 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:29 UTC