[PATCH 4/14][AArch64] Use new reduc_plus_scal optabs, inc. for __builtins

Alan Lawrence alan.lawrence@arm.com
Thu Sep 18 11:59:00 GMT 2014


This migrates AArch64 over to the new optab for 'plus' reductions, i.e. so the 
define_expands produce scalars by generating a MOV to a GPR. Effectively, this 
moves the vget_lane inside every arm_neon.h intrinsic, into the inside of the 
define_expand.

Tested: aarch64.exp vect.exp on aarch64-none-elf and aarch64_be-none-elf (full 
check-gcc on next patch for reduc_min/max)

gcc/ChangeLog:

	* config/aarch64/aarch64-simd-builtins.def
	(reduc_splus_<mode>/VDQF, reduc_uplus_<mode>/VDQF, reduc_splus_v4sf):
	Remove.
	(reduc_plus_scal_<mode>, reduc_plus_scal_v4sf): New.

	* config/aarch64/aarch64-simd.md (reduc_<sur>plus_mode): Remove.
	(reduc_splus_<mode>, reduc_uplus_<mode>, reduc_plus_scal_<mode>): New.

	(reduc_<sur>plus_mode): Change SUADDV -> UNSPEC_ADDV, rename to...
	(aarch64_reduc_plus_internal<mode>): ...this.

	(reduc_<sur>plus_v2si): Change SUADDV -> UNSPEC_ADDV, rename to...
	(aarch64_reduc_plus_internalv2si): ...this.

	(reduc_splus_<mode>/V2F): Rename to...
	(aarch64_reduc_plus_internal<mode>): ...this.

	* config/aarch64/iterators.md
	(UNSPEC_SADDV, UNSPEC_UADDV, SUADDV): Remove.
	(UNSPEC_ADDV): New.
	(sur): Remove elements for UNSPEC_SADDV and UNSPEC_UADDV.

	* config/aarch64/arm_neon.h (vaddv_s8, vaddv_s16, vaddv_s32, vaddv_u8,
	vaddv_u16, vaddv_u32, vaddvq_s8, vaddvq_s16, vaddvq_s32, vaddvq_s64,
	vaddvq_u8, vaddvq_u16, vaddvq_u32, vaddvq_u64, vaddv_f32, vaddvq_f32,
	vaddvq_f64): Change __builtin_aarch64_reduc_[us]plus_... to
	__builtin_aarch64_reduc_plus_scal, remove vget_lane wrapper.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 4_aarch64_reduc_plus.patch
Type: text/x-patch
Size: 10855 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20140918/88c316fa/attachment.bin>


More information about the Gcc-patches mailing list