cmp.geu p6,p7=r19,r17 // (syscall > 0 && syscall < 1024+NR_syscalls)?
;;
(p6) ld8 r18=[r18]
- mov r29=psr // read psr (12 cyc load latency)
+ mov r21=ar.fpsr
add r14=-8,r14 // r14 <- addr of fsys_bubble_down entry
;;
(p6) mov b7=r18
(p6) tbit.z p8,p0=r18,0
(p8) br.dptk.many b7
+(p6) rsm psr.i
mov r27=ar.rsc
- mov r21=ar.fpsr
mov r26=ar.pfs
+ ;;
+ mov r29=psr // read psr (12 cyc load latency)
/*
* brl.cond doesn't work as intended because the linker would convert this branch
* into a branch to a PLT. Perhaps there will be a way to avoid this with some