Add conditional move opcodes
Add jit_movnr(r0, r1, r2) and jit_movzr(r0, r1, r2).
jit_movnr() corresponds to the following operation:
r0 = r2 ? r1 : r0
jit_movzr() corresponds to the following operation:
r0 = r2 ? r0 : r1
Add implementations for x86/x86_64, MIPS, PowerPC, ARM and Aarch64.
The x86 implementation uses the CMOV feature, which was added in i686.
I don't think anybody will try to run Lightning on anything more
ancient, so this should be fine. Otherwise, a simple implementation
based on branches can be added later. This could be used as well for
architectures that lack conditional moves, like it's currently done on
PowerPC for instance.
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Diffstat (limited to 'lib/jit_alpha-cpu.c')
0 files changed, 0 insertions, 0 deletions