I witnessed some code sequence like the following in specjbb2005 and other benchmarks:
  61976 0x0000007f7e17ff44: mov w13, wzr
  61977 0x0000007f7e17ff48: cmp w12, w13
  61978 0x0000007f7e17ff4c: csel w11, w12, w13, gt ;*iinc {reexecute=0 rethrow=0 return_oop=0}
  61979 ; - jdk.internal.org.objectweb.asm.Type::getArgumentsAndReturnSizes@103 (line 451)
......
 123349 0x0000007f7e1db660: mov w16, wzr
 123350 0x0000007f7e1db664: sub w10, w19, w23 ;*iload {reexecute=0 rethrow=0 return_oop=0}
 123351 ; - java.math.MutableBigInteger::divideMagnitude@182 (line 1503)
 123352
 123353 0x0000007f7e1db668: cmp w15, w16
 123354 0x0000007f7e1db66c: csel w18, w15, w16, gt
 123355 0x0000007f7e1db670: mov w25, wzr ;*iastore {reexecute=0 rethrow=0 return_oop=0}
 123356 ; - java.math.MutableBigInteger::divideMagnitude@221 (line 1506)
......
 187167 0x0000007f7e23a6f0: mov w12, wzr
 187168 0x0000007f7e23a6f4: cmp w11, w12
 187169 0x0000007f7e23a6f8: csel w10, w11, w12, gt
......
  17114 0x0000007f7e124058: orr w14, wzr, #0x1
  17115
  17116 0x0000007f7e12405c: mov w10, wzr
  17117 0x0000007f7e124060: cmp w13, w14
  17118 0x0000007f7e124064: csel w12, w13, w14, gt
......
  61072 0x0000007f7e179450: orr w11, wzr, #0x1
  61073 0x0000007f7e179454: cmp w10, w11
  61074 0x0000007f7e179458: csel w12, w10, w11, gt

For the aarch64 port, we have csel/csinc/csinv instructions which can be applied for these cases eliminating one extra mov instruction here.

I mean something like this:
min(x, 1) which becomes

cmp x, 0
csinc x, x, zr, le

min(x, -1) which becomes

cmp x, 0
csinv x, x, zr, lt

max(x, 1)

cmp x, 0
csinc x, x, zr, gt

max(x, -1)

cmp x, 0
csinv x, x, zr, ge