Hi Renlin, On Thu, Nov 01, 2018 at 06:50:22PM +0000, Renlin Li wrote: > I got the following dump from the test case. > > x1 is an early clobber operand in the inline assembly statement, > r92 should conflict with x1?
Yes, I think so. Or LRA should not pick something that was already a hard register to spill (this should work without the earlyclobbers, too... In this testcase that works anyhow, but that seems to be by accident). Segher