[patch] ARM half-precision support
Nick Clifton
nickc@redhat.com
Mon Nov 17 16:53:00 GMT 2008
Hi Catherine,
> This patch adds the binutils support for arm half-precision
> instructions. Does this look okay to install?
It does, but - how was it tested ? Did you check that it does not break
an arm-vxworks target for example ?
> gas/
> * config/tc-arm.c (neon_type_mask): Renumber.
> (type_chk_of_el_type): Handle F_F16.
> (neon_cvt_flavour): Recognize half-precision conversions.
> (do_neon_cvt): New shapes NS_QD and NS_DQ. Encode half-precision
> conversions.
> (do_neon_cvtt): Encode the T bit.
> (asm_opcode_insns): vcvt, vcvtt support.
> (arm_option_cpu_value): Add neon-fp16 support.
>
> include/
> * opcode/arm.h (FPU_NEON_FP16): New.
> (FPU_ARCH_NEON_FP16): New.
> * elf/arm.h (Tag_ABI_FP_16bit_format): Define.
>
> opcodes/
> * arm-dis.c (coprocessor_opcodes): Add half-precision vcvt
> instructions.
> (neon_opcodes): Likewise.
> (print_insn_coprocessor): Print 't' or 'b' for
> vcvt instructions.
>
> binutils/
> * readelf.c (arm_attr_tag_ABI_FP_16bit_format): New.
> (arm_attr_tag_VFP_HP_extension): New.
> (arm_attr_public_tag arm_attr_public_tags): Support new
> attributes.
>
> bfd/
> * elf32-arm.c (elf32_arm_merge_eabi_attributes): Merge
> half-precision attributes.
> (elf32_arm_copy_one_eabi_other_attribute): New.
> (elf32_arm_copy_other_attribute_list): New.
Approved - assuming that the testing does not reveal any problems -
please apply.
Cheers
Nick
More information about the Binutils
mailing list