[AArch64] Use unspecs for SVE conversions involving floats

[AArch64] Use unspecs for SVE conversions involving floats

This patch changes the SVE FP<->FP and FP<->INT patterns so that
they use unspecs rather than rtx codes, continuing the series
to make the patterns work with predicates that might not be all-true.

2019-08-14  Richard Sandiford  <richard.sandiford@arm.com>

gcc/
	* config/aarch64/aarch64.md (UNSPEC_FLOAT_CONVERT): Delete.
	* config/aarch64/iterators.md (UNSPEC_COND_FCVT, UNSPEC_COND_FCVTZS)
	(UNSPEC_COND_FCVTZU, UNSPEC_COND_SCVTF, UNSPEC_COND_UCVTF): New
	unspecs.
	(optab, su): Handle them.
	(SVE_COND_FCVTI, SVE_COND_ICVTF): New int iterators.
	* config/aarch64/aarch64-sve.md
	(<fix_trunc_optab><SVE_F:mode><v_int_equiv>2): Replace with...
	(<SVE_COND_FCVTI:optab><SVE_F:mode><v_int_equiv>2): ...this.
	(*<fix_trunc_optab>v16hsf<:SVE_HSDImode>2): Replace with...
	(*<SVE_COND_FCVTI:optab>v16hsf<SVE_F:mode>2): ...this.
	(*<fix_trunc_optab>vnx4sf<SVE_SDI:mode>2): Replace with...
	(*<SVE_COND_FCVTI:optab>vnx4sf<SVE_SDI:mode>2): ...this.
	(*<fix_trunc_optab>vnx2df<SVE_SDI:mode>2): Replace with...
	(*<SVE_COND_FCVTI:optab>vnx2df<SVE_SDI:mode>2): ...this.
	(vec_pack_<su>fix_trunc_vnx2df): Use SVE_COND_FCVTI instead of
	FIXUORS.
	(<FLOATUORS:optab><v_int_equiv><SVE_F:mode>2): Replace with...
	(<SVE_COND_ICVTF:optab><v_int_equiv><SVE_F:mode>2): ...this.
	(*<FLOATUORS:optab><SVE_HSDI:mode>vnx8hf2): Replace with...
	(*<SVE_COND_ICVTF:optab><SVE_HSDI:mode>vnx8hf2): ...this.
	(*<FLOATUORS:optab><SVE_SDI:mode>vnx4sf2): Replace with...
	(*<SVE_COND_ICVTF:optab><SVE_SDI:mode>vnx4sf2): ...this.
	(aarch64_sve_<FLOATUORS:optab><SVE_SDI:mode>vnx2df2): Replace with...
	(aarch64_sve_<SVE_COND_ICVTF:optab><SVE_SDI:mode>vnx2df2): ...this.
	(vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Pass a GP strictness
	operand to aarch64_sve_<SVE_COND_ICVTF:optab><SVE_SDI:mode>vnx2df2.
	(vec_pack_trunc_<SVE_HSF:Vwide>, *trunc<Vwide><SVE_HSF:mode>2)
	(aarch64_sve_extend<mode><Vwide>2): Use UNSPEC_COND_FCVT instead
	of UNSPEC_FLOAT_CONVERT.
	(vec_unpacks_<perm_hilo>_<mode>): Pass a GP strictness operand to
	aarch64_sve_extend<mode><Vwide>2.

From-SVN: r274423