[Bug rtl-optimization/60969] [4.9/4.10 Regression] ICE in output_129 in MMXMOV of mode MODE_SF for march=pentium4

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60969

--- Comment #19 from H.J. Lu <hjl.tools at gmail dot com> ---
There is

#define VALID_MMX_REG_MODE_3DNOW(MODE) \
  ((MODE) == V2SFmode || (MODE) == SFmode)

and ix86_hard_regno_mode_ok returns TRUE for holding SFmode
in MMX registers.  It is wrong to tell RA that MMX registers
can have V2SFmode/SFmode when 3DNOW is disabled.  Even if 3DNOW
is enabled, we can't use MMX registers for floating point when
x87 registers are also used.