On Wed, Feb 03, 2016 at 10:41:03AM -0800, Bobby Holley wrote: > On Wed, Feb 3, 2016 at 10:32 AM, Ehsan Akhgari <[email protected]> > wrote: > > > On 2016-02-03 12:50 PM, Martin Thomson wrote: > > > >> On Thu, Feb 4, 2016 at 2:21 AM, Milan Sreckovic <[email protected]> > >> wrote: > >> > >>> 99.77% of the users on all channels have SSE2 support; > >>> 51.7% of all users are on 32-bit Windows; > >>> 0.44% of all users on 32-bit Windows do not have SSE2 support. > >>> > >> > >> Those numbers wouldn't justify a change to me. When we make decisions > >> about what we break with TLS by disabling something that is maybe > >> dangerous, we try to avoid changes that break any more than 0.1% of > >> our population. It looks like we are almost there, but unlike some of > >> the security changes, we can't just provide motivation to change [1] > >> because requesting a hardware change is a pretty high bar to clear. > >> > > > > As I said elsewhere in the thread, we can just pass the correct flag to > > rustc to select the correct target features. Dropping support for old > > processors seems to be orthogonal to what Henri wants to do in rust. > > > Except for the fact that the code he wants to replace uses dynamic SSE > switching for the hot code, which seems difficult to do when dropping in a > rust replacement, right?
GCC has a target __attribute__ that allows to selectively enable e.g. SSE per function (which, btw, we could use for C++ instead of using separate sources built with different flags, although I don't know if clang or MSVC support the same thing or similar). Rust could (should) grow something similar. Mike _______________________________________________ dev-platform mailing list [email protected] https://lists.mozilla.org/listinfo/dev-platform
