Dan Gohman 79f02e42dd Use movss/movsd rather than movd/movq for floating-point loads and stores. ...

While there may be CPUs that have a domain crossing penalty here,
this also helps the generated code look more like the code produced
by other compilers.

2018-03-27 11:53:59 -07:00

..

__init__.py

Add the very basics of Intel 32-bit instruction encodings.

2017-05-08 16:57:38 -07:00

defs.py

Rename I32 -> X86_32 and I64 -> X86_64 (#271)

2018-03-18 13:50:51 -07:00

encodings.py

Use movss/movsd rather than movd/movq for floating-point loads and stores.

2018-03-27 11:53:59 -07:00

instructions.py

Tidy up whitespace.

2018-03-05 06:55:27 -08:00

legalize.py

Adds support to transform integer div and rem by constants into cheaper equivalents.

2018-02-28 11:41:36 -08:00

recipes.py

Rename %eflags to %rflags.

2018-03-27 11:52:57 -07:00

registers.py

Rename %eflags to %rflags.

2018-03-27 11:52:57 -07:00

settings.py

Remove the "has_sse2" flag.

2018-03-12 12:38:01 -07:00