[openpower-isa.git] / src / openpower / sv / trans / svp64.py

# SPDX-License-Identifier: LGPLv3+
# Copyright (C) 2021 Luke Kenneth Casson Leighton <lkcl@lkcl.net>
# Funded by NLnet http://nlnet.nl

"""SVP64 OpenPOWER v3.0B assembly translator

This class takes raw svp64 assembly mnemonics (aliases excluded) and creates
an EXT001-encoded "svp64 prefix" (as a .long) followed by a v3.0B opcode.

It is very simple and straightforward, the only weirdness being the
extraction of the register information and conversion to v3.0B numbering.

Encoding format of svp64: https://libre-soc.org/openpower/sv/svp64/
Encoding format of arithmetic: https://libre-soc.org/openpower/sv/normal/
Encoding format of LDST: https://libre-soc.org/openpower/sv/ldst/
**TODO format of branches: https://libre-soc.org/openpower/sv/branches/**
**TODO format of CRs: https://libre-soc.org/openpower/sv/cr_ops/**
Bugtracker: https://bugs.libre-soc.org/show_bug.cgi?id=578
"""

import functools
import os
import sys
from collections import OrderedDict
import inspect

from openpower.decoder.pseudo.pagereader import ISA
from openpower.decoder.power_svp64 import SVP64RM, get_regtype, decode_extra
from openpower.decoder.selectable_int import SelectableInt
from openpower.consts import SVP64MODE
from openpower.decoder.power_insn import SVP64Instruction
from openpower.decoder.power_insn import Database
from openpower.decoder.power_enums import find_wiki_dir

# for debug logging
from openpower.util import log


def instruction(*fields):
    def instruction(insn, desc):
        (value, start, end) = desc
        bits = ((1,) * ((end + 1) - start))
        mask = 0
        for bit in bits:
            mask = ((mask << 1) | bit)
        return (insn | ((value & mask) << (31 - end)))

    return functools.reduce(instruction, fields, 0)


CUSTOM_INSNS = {}


def _insn(name, **kwargs):
    return name, kwargs


def _custom_insns(*insns):
    """ a decorator that adds the function to `CUSTOM_INSNS` """

    def decorator(fn):
        FIELDS_ARG = object()
        if len(insns) == 0:
            insns_ = (fn.__name__, {}),
        else:
            insns_ = insns
        for name, kwargs in insns_:
            if not isinstance(name, str):
                raise TypeError("instruction name must be a str: {name!r}")
            if name in CUSTOM_INSNS:
                raise ValueError(f"duplicate instruction mnemonic: {name!r}")
            # use getcallargs to check that arguments work:
            inspect.getcallargs(fn, FIELDS_ARG, **kwargs)
            CUSTOM_INSNS[name] = functools.partial(fn, **kwargs)
        return fn
    return decorator


@_custom_insns(
    _insn("setvl", Rc=0),
    _insn("setvl.", Rc=1),
)
def setvl(fields, Rc):
    """
    setvl is a *32-bit-only* instruction. It controls SVSTATE.
    It is *not* a 64-bit-prefixed Vector instruction (no sv.setvl, yet),
    it is a Vector *control* instruction.

    * setvl RT,RA,SVi,vf,vs,ms

    1.6.28 SVL-FORM - from fields.txt
    |0     |6    |11    |16   |23 |24 |25 |26    |31 |
    | PO   |  RT |   RA | SVi |ms |vs |vf |   XO |Rc |
    """
    PO = 22
    XO = 0b11011
    # ARRRGH these are in a non-obvious order in openpower/isa/simplev.mdwn
    # compared to the SVL-Form above. sigh
    # setvl RT,RA,SVi,vf,vs,ms
    (RT, RA, SVi, vf, vs, ms) = fields
    SVi -= 1
    return instruction(
        (PO, 0, 5),
        (RT, 6, 10),
        (RA, 11, 15),
        (SVi, 16, 22),
        (ms, 23, 23),
        (vs, 24, 24),
        (vf, 25, 25),
        (XO, 26, 30),
        (Rc, 31, 31),
    )


@_custom_insns(
    _insn("svstep", Rc=0),
    _insn("svstep.", Rc=1),
)
def svstep(fields, Rc):
    """
    svstep is a 32-bit instruction. It updates SVSTATE.
    It *can* be SVP64-prefixed, to indicate that its registers
    are Vectorised.

    * svstep RT,SVi,vf

    # 1.6.28 SVL-FORM - from fields.txt
    # |0     |6    |11    |16   |23 |24 |25 |26    |31 |
    # | PO   |  RT | /    | SVi |/  |/  |vf |   XO |Rc |

    """
    PO = 22
    XO = 0b10011
    (RT, SVi, vf) = fields
    SVi -= 1
    return instruction(
        (PO, 0, 5),
        (RT, 6, 10),
        (0, 11, 15),
        (SVi, 16, 22),
        (0, 23, 23),
        (0, 24, 24),
        (vf, 25, 25),
        (XO, 26, 30),
        (Rc, 31, 31),
    )


@_custom_insns()
def svshape(fields):
    """
    svshape is a *32-bit-only* instruction. It updates SVSHAPE and SVSTATE.
    It is *not* a 64-bit-prefixed Vector instruction (no sv.svshape, yet),
    it is a Vector *control* instruction.

    https://libre-soc.org/openpower/sv/remap/#svshape

    * svshape SVxd,SVyd,SVzd,SVrm,vf

    # 1.6.33 SVM-FORM from fields.txt
    # |0     |6        |11      |16    |21    |25 |26    |31  |
    # | PO   |  SVxd   |   SVyd | SVzd | SVrm |vf |   XO      |

    note that SVrm is not permitted to be 0b0111, 0b1000 or 0b1001.
    0b0111 is reserved and 0b100- is for svshape2

    """
    PO = 22
    XO = 0b011001
    (SVxd, SVyd, SVzd, SVrm, vf) = fields
    SVxd -= 1
    SVyd -= 1
    SVzd -= 1

    # check SVrm for reserved (and svshape2) values
    assert SVrm not in [0b0111, 0b1000, 0b1001], \
            "svshape reserved SVrm value %s" % bin(SVrm)

    return instruction(
        (PO, 0, 5),
        (SVxd, 6, 10),
        (SVyd, 11, 15),
        (SVzd, 16, 20),
        (SVrm, 21, 24),
        (vf, 25, 25),
        (XO, 26, 31),
    )


@_custom_insns()
def svshape2(fields):
    """
    svshape2 is a *32-bit-only* instruction. It updates SVSHAPE and SVSTATE.
    It is *not* a 64-bit-prefixed Vector instruction (no sv.svshape2, yet),
    it is a Vector *control* instruction, and is a sort-of hybrid of
    svshape and svindex, with the key important feature being the "offset".

    https://libre-soc.org/openpower/sv/remap/discussion

    * svshape2 offs,yx,rmm,SVd,sk,mm

    # 1.6.35.1 SVM2-FORM from fields.txt
    # |0     |6     |10|11      |16    |21 |24|25 |26    |31  |
    # | PO   | offs |yx|   rmm  | SVd  |XO |mm|sk |   XO      |

    note that this fits into the space of svshape and that XO is
    split across 2 areas.

    """
    PO = 22
    XO = 0b011001
    XO2 = 0b100 # not really XO2 but hey
    (offs, yx, rmm, SVd, sk, mm) = fields
    SVd -= 1 # offset by one

    return instruction(
        (PO, 0, 5),
        (offs, 6, 9),  # offset (the whole point of adding svshape2)
        (yx, 10, 10),  # like svindex
        (rmm, 11, 15), # ditto svindex
        (SVd, 16, 20), # ditto svindex
        (XO2, 21, 23), # actually XO split across 2 places...
        (mm, 24, 24),  # ditto svindex
        (sk, 25, 25),  # ditto svindex
        (XO, 26, 31),
    )


@_custom_insns()
def svindex(fields):
    """
    svindex is a *32-bit-only* instruction. It is a convenience
    instruction that reduces instruction count for Indexed REMAP
    Mode.
    It is *not* a 64-bit-prefixed Vector instruction (no sv.svindex, yet),
    it is a Vector *control* instruction.

    1.6.28 SVI-FORM
      |0     |6    |11    |16   |21 |23|24|25|26    31|
      | PO   |  SVG|rmm   | SVd |ew |yx|mm|sk|   XO   |
    """
    # note that the dimension field one subtracted
    PO = 22
    XO = 0b101001
    (SVG, rmm, SVd, ew, yx, mm, sk) = fields
    SVd -= 1
    return instruction(
        (PO, 0, 5),
        (SVG, 6, 10),
        (rmm, 11, 15),
        (SVd, 16, 20),
        (ew, 21, 22),
        (yx, 23, 23),
        (mm, 24, 24),
        (sk, 25, 25),
        (XO, 26, 31),
    )


@_custom_insns()
def svremap(fields):
    """
    this is a *32-bit-only* instruction. It updates the SVSHAPE SPR
    it is *not* a 64-bit-prefixed Vector instruction (no sv.svremap),
    it is a Vector *control* instruction.

    * svremap SVme,mi0,mi1,mi2,mo0,mo1,pst

    # 1.6.34 SVRM-FORM
       |0     |6     |11  |13   |15   |17   |19   |21  |22   |26     |31 |
       | PO   | SVme |mi0 | mi1 | mi2 | mo0 | mo1 |pst |///  | XO        |

    """
    PO = 22
    XO = 0b111001
    (SVme, mi0, mi1, mi2, mo0, mo1, pst) = fields
    return instruction(
        (PO, 0, 5),
        (SVme, 6, 10),
        (mi0, 11, 12),
        (mi1, 13, 14),
        (mi2, 15, 16),
        (mo0, 17, 18),
        (mo1, 19, 20),
        (pst, 21, 21),
        (0, 22, 25),
        (XO, 26, 31),
    )


# ok from here-on down these are added as 32-bit instructions
# and are here only because binutils (at present) doesn't have
# them (that's being fixed!)
# they can - if implementations then choose - be Vectorised
# because they are general-purpose scalar instructions
@_custom_insns()
def bmask(fields):
    """
    1.6.2.2 BM2-FORM
    |0     |6    |11    |16    |21   |26 |27    31|
    | PO   |  RT |   RA |   RB |bm   |L  |   XO   |
    """
    PO = 22
    XO = 0b010001
    (RT, RA, RB, bm, L) = fields
    return instruction(
        (PO, 0, 5),
        (RT, 6, 10),
        (RA, 11, 15),
        (RB, 16, 20),
        (bm, 21, 25),
        (L, 26, 26),
        (XO, 27, 31),
    )


@_custom_insns(
    _insn("fsins", Rc=0),
    _insn("fsins.", Rc=1),
)
def fsins(fields, Rc):
    # XXX WARNING THESE ARE NOT APPROVED BY OPF ISA WG
    # however we are out of space with opcode 22
    # 1.6.7 X-FORM
    # |0     |6 |7|8|9  |10  |11|12|13  |15|16|17     |20|21    |31  |
    # | PO   |   FRT         |     ///     |   FRB       |   XO |Rc  |
    PO = 59
    XO = 0b1000001110
    (FRT, FRB) = fields
    return instruction(
        (PO, 0, 5),
        (FRT, 6, 10),
        (0, 11, 15),
        (FRB, 16, 20),
        (XO, 21, 30),
        (Rc, 31, 31),
    )


@_custom_insns(
    _insn("fcoss", Rc=0),
    _insn("fcoss.", Rc=1),
)
def fcoss(fields, Rc):
    # XXX WARNING THESE ARE NOT APPROVED BY OPF ISA WG
    # however we are out of space with opcode 22
    # 1.6.7 X-FORM
    # |0     |6 |7|8|9  |10  |11|12|13  |15|16|17     |20|21    |31  |
    # | PO   |   FRT         |     ///     |   FRB       |   XO |Rc  |
    PO = 59
    XO = 0b1000101110
    (FRT, FRB) = fields
    return instruction(
        (PO, 0, 5),
        (FRT, 6, 10),
        (0, 11, 15),
        (FRB, 16, 20),
        (XO, 21, 30),
        (Rc, 31, 31),
    )


@_custom_insns(
    _insn("ternlogi", Rc=0),
    _insn("ternlogi.", Rc=1),
)
def ternlogi(fields, Rc):
    # XXX WARNING THESE ARE NOT APPROVED BY OPF ISA WG
    # however we are out of space with opcode 22
    # 1.6.34 TLI-FORM
    #   |0   |6   |11   |16   |21   |29  |31 |
    #   | PO | RT |  RA |  RB | TLI | XO |Rc |
    PO = 5
    XO = 0
    (RT, RA, RB, TLI) = fields
    return instruction(
        (PO, 0, 5),
        (RT, 6, 10),
        (RA, 11, 15),
        (RB, 16, 20),
        (TLI, 21, 28),
        (XO, 29, 30),
        (Rc, 31, 31),
    )


@_custom_insns(
    _insn("grev", Rc=0, imm=0, word=0),
    _insn("grevw", Rc=0, imm=0, word=1),
    _insn("grevi", Rc=0, imm=1, word=0),
    _insn("grevwi", Rc=0, imm=1, word=1),
    _insn("grev.", Rc=1, imm=0, word=0),
    _insn("grevw.", Rc=1, imm=0, word=1),
    _insn("grevi.", Rc=1, imm=1, word=0),
    _insn("grevwi.", Rc=1, imm=1, word=1),
)
def grev(fields, Rc, imm, word):
    # XXX WARNING THESE ARE NOT APPROVED BY OPF ISA WG
    # however we are out of space with opcode 22
    insn = PO = 5
    # _ matches fields in table at:
    # https://libre-soc.org/openpower/sv/bitmanip/
    XO = 0b1_0010_110
    if word:
        XO |= 0b100_000
    if imm:
        XO |= 0b1000_000
    (RT, RA, XBI) = fields
    insn = (insn << 5) | RT
    insn = (insn << 5) | RA
    if imm and not word:
        assert 0 <= XBI < 64
        insn = (insn << 6) | XBI
        insn = (insn << 9) | XO
    else:
        assert 0 <= XBI < 32
        insn = (insn << 5) | XBI
        insn = (insn << 10) | XO
    insn = (insn << 1) | Rc
    return insn


@_custom_insns(
    _insn("maxs", XO=0b0111001110, Rc=0),
    _insn("maxs.", XO=0b0111001110, Rc=1),
    _insn("maxu", XO=0b0011001110, Rc=0),
    _insn("maxu.", XO=0b0011001110, Rc=1),
    _insn("minu", XO=0b0001001110, Rc=0),
    _insn("minu.", XO=0b0001001110, Rc=1),
    _insn("mins", XO=0b0101001110, Rc=0),
    _insn("mins.", XO=0b0101001110, Rc=1),
    _insn("absdu", XO=0b1011110110, Rc=0),
    _insn("absdu.", XO=0b1011110110, Rc=1),
    _insn("absds", XO=0b1001110110, Rc=0),
    _insn("absds.", XO=0b1001110110, Rc=1),
    _insn("avgadd", XO=0b1101001110, Rc=0),
    _insn("avgadd.", XO=0b1101001110, Rc=1),
    _insn("absdacu", XO=0b1111110110, Rc=0),
    _insn("absdacu.", XO=0b1111110110, Rc=1),
    _insn("absdacs", XO=0b0111110110, Rc=0),
    _insn("absdacs.", XO=0b0111110110, Rc=1),
    _insn("cprop", XO=0b0110001110, Rc=0),
    _insn("cprop.", XO=0b0110001110, Rc=1),
)
def av(fields, XO, Rc):
    # 1.6.7 X-FORM
    #   |0     |6 |7|8|9  |10  |11|12|13  |15|16|17     |20|21    |31  |
    #   | PO   |       RT      |    RA       |    RB       |   XO |Rc  |
    PO = 22
    (RT, RA, RB) = fields
    return instruction(
        (PO, 0, 5),
        (RT, 6, 10),
        (RA, 11, 15),
        (RB, 16, 20),
        (XO, 21, 30),
        (Rc, 31, 31),
    )


@_custom_insns()
def fmvis(fields):
    # XXX WARNING THESE ARE NOT APPROVED BY OPF ISA WG
    # V3.0B 1.6.6 DX-FORM
    # |0     |6 |7|8|9  |10  |11|12|13  |15|16|17     |26|27    |31  |
    # | PO   |   FRS         |     d1      |      d0  |      XO |d2  |
    PO = 22
    XO = 0b00011
    (FRS, imm) = fields
    # first split imm into d1, d0 and d2. sigh
    d2 = (imm & 1)  # LSB (0)
    d1 = (imm >> 1) & 0b11111  # bits 1-5
    d0 = (imm >> 6)  # MSBs 6-15
    return instruction(
        (PO, 0, 5),
        (FRS, 6, 10),
        (d1,  11, 15),
        (d0,  16, 25),
        (XO, 26, 30),
        (d2, 31, 31),
    )


@_custom_insns()
def fishmv(fields):
    # XXX WARNING THESE ARE NOT APPROVED BY OPF ISA WG
    # V3.0B 1.6.6 DX-FORM
    # |0     |6 |7|8|9  |10  |11|12|13  |15|16|17     |26|27   |31  |
    # | PO   |   FRS         |     d1      |      d0  |     XO |d2  |
    PO = 22
    XO = 0b01011
    (FRS, imm) = fields
    # first split imm into d1, d0 and d2. sigh
    d2 = (imm & 1)  # LSB (0)
    d1 = (imm >> 1) & 0b11111  # bits 1-5
    d0 = (imm >> 6)  # MSBs 6-15
    return instruction(
        (PO, 0, 5),
        (FRS, 6, 10),
        (d1,  11, 15),
        (d0,  16, 25),
        (XO, 26, 30),
        (d2, 31, 31),
    )


# decode GPR into sv extra
def get_extra_gpr(etype, regmode, field):
    if regmode == 'scalar':
        # cut into 2-bits 5-bits SS FFFFF
        sv_extra = field >> 5
        field = field & 0b11111
    else:
        # cut into 5-bits 2-bits FFFFF SS
        sv_extra = field & 0b11
        field = field >> 2
    return sv_extra, field


# decode 3-bit CR into sv extra
def get_extra_cr_3bit(etype, regmode, field):
    if regmode == 'scalar':
        # cut into 2-bits 3-bits SS FFF
        sv_extra = field >> 3
        field = field & 0b111
    else:
        # cut into 3-bits 4-bits FFF SSSS but will cut 2 zeros off later
        sv_extra = field & 0b1111
        field = field >> 4
    return sv_extra, field


# decodes SUBVL
def decode_subvl(encoding):
    pmap = {'2': 0b01, '3': 0b10, '4': 0b11}
    assert encoding in pmap, \
        "encoding %s for SUBVL not recognised" % encoding
    return pmap[encoding]


# decodes elwidth
def decode_elwidth(encoding):
    pmap = {'8': 0b11, '16': 0b10, '32': 0b01}
    assert encoding in pmap, \
        "encoding %s for elwidth not recognised" % encoding
    return pmap[encoding]


# decodes predicate register encoding
def decode_predicate(encoding):
    pmap = {  # integer
        '1<<r3': (0, 0b001),
        'r3': (0, 0b010),
        '~r3': (0, 0b011),
        'r10': (0, 0b100),
        '~r10': (0, 0b101),
        'r30': (0, 0b110),
        '~r30': (0, 0b111),
        # CR
        'lt': (1, 0b000),
        'nl': (1, 0b001), 'ge': (1, 0b001),  # same value
        'gt': (1, 0b010),
        'ng': (1, 0b011), 'le': (1, 0b011),  # same value
        'eq': (1, 0b100),
        'ne': (1, 0b101),
        'so': (1, 0b110), 'un': (1, 0b110),  # same value
        'ns': (1, 0b111), 'nu': (1, 0b111),  # same value
    }
    assert encoding in pmap, \
        "encoding %s for predicate not recognised" % encoding
    return pmap[encoding]


# decodes "Mode" in similar way to BO field (supposed to, anyway)
def decode_bo(encoding):
    pmap = {  # TODO: double-check that these are the same as Branch BO
        'lt': 0b000,
        'nl': 0b001, 'ge': 0b001,  # same value
        'gt': 0b010,
        'ng': 0b011, 'le': 0b011,  # same value
        'eq': 0b100,
        'ne': 0b101,
        'so': 0b110, 'un': 0b110,  # same value
        'ns': 0b111, 'nu': 0b111,  # same value
    }
    assert encoding in pmap, \
        "encoding %s for BO Mode not recognised" % encoding
    return pmap[encoding]


# partial-decode fail-first mode
def decode_ffirst(encoding):
    if encoding in ['RC1', '~RC1']:
        return encoding
    return decode_bo(encoding)


def decode_reg(field, macros=None):
    if macros is None:
        macros = {}
    # decode the field number. "5.v" or "3.s" or "9"
    # and now also "*0", and "*%0".  note: *NOT* to add "*%rNNN" etc.
    # https://bugs.libre-soc.org/show_bug.cgi?id=884#c0
    if field.startswith(("*%", "*")):
        if field.startswith("*%"):
            field = field[2:]
        else:
            field = field[1:]
        while field in macros:
            field = macros[field]
        return int(field), "vector"  # actual register number

    # try old convention (to be retired)
    field = field.split(".")
    regmode = 'scalar'  # default
    if len(field) == 2:
        if field[1] == 's':
            regmode = 'scalar'
        elif field[1] == 'v':
            regmode = 'vector'
    field = int(field[0])  # actual register number
    return field, regmode


def decode_imm(field):
    ldst_imm = "(" in field and field[-1] == ')'
    if ldst_imm:
        return field[:-1].split("(")
    else:
        return None, field


def crf_extra(etype, regmode, field, extras):
    """takes a CR Field number (CR0-CR127), splits into EXTRA2/3 and v3.0
    the scalar/vector mode (crNN.v or crNN.s) changes both the format
    of the EXTRA2/3 encoding as well as what range of registers is possible.
    this function can be used for both BF/BFA and BA/BB/BT by first removing
    the bottom 2 bits of BA/BB/BT then re-instating them after encoding.
    see https://libre-soc.org/openpower/sv/svp64/appendix/#cr_extra
    for specification
    """
    sv_extra, field = get_extra_cr_3bit(etype, regmode, field)
    # now sanity-check (and shrink afterwards)
    if etype == 'EXTRA2':
        # 3-bit CR Field (BF, BFA) EXTRA2 encoding
        if regmode == 'scalar':
            # range is CR0-CR15 in increments of 1
            assert (sv_extra >> 1) == 0, \
                "scalar CR %s cannot fit into EXTRA2 %s" % \
                (rname, str(extras[extra_idx]))
            # all good: encode as scalar
            sv_extra = sv_extra & 0b01
        else:  # vector
            # range is CR0-CR127 in increments of 16
            assert sv_extra & 0b111 == 0, \
                "vector CR %s cannot fit into EXTRA2 %s" % \
                (rname, str(extras[extra_idx]))
            # all good: encode as vector (bit 2 set)
            sv_extra = 0b10 | (sv_extra >> 3)
    else:
        # 3-bit CR Field (BF, BFA) EXTRA3 encoding
        if regmode == 'scalar':
            # range is CR0-CR31 in increments of 1
            assert (sv_extra >> 2) == 0, \
                "scalar CR %s cannot fit into EXTRA3 %s" % \
                (rname, str(extras[extra_idx]))
            # all good: encode as scalar
            sv_extra = sv_extra & 0b11
        else:  # vector
            # range is CR0-CR127 in increments of 8
            assert sv_extra & 0b11 == 0, \
                "vector CR %s cannot fit into EXTRA3 %s" % \
                (rname, str(extras[extra_idx]))
            # all good: encode as vector (bit 3 set)
            sv_extra = 0b100 | (sv_extra >> 2)
    return sv_extra, field


def to_number(field):
    if field.startswith("0x"):
        return eval(field)
    if field.startswith("0b"):
        return eval(field)
    return int(field)


# decodes svp64 assembly listings and creates EXT001 svp64 prefixes
class SVP64Asm:
    def __init__(self, lst, bigendian=False, macros=None):
        if macros is None:
            macros = {}
        self.macros = macros
        self.lst = lst
        self.trans = self.translate(lst)
        self.isa = ISA()  # reads the v3.0B pseudo-code markdown files
        self.svp64 = SVP64RM()  # reads the svp64 Remap entries for registers
        assert bigendian == False, "error, bigendian not supported yet"

    def __iter__(self):
        yield from self.trans

    def translate_one(self, insn, macros=None):
        if macros is None:
            macros = {}
        macros.update(self.macros)
        isa = self.isa
        svp64 = self.svp64
        insn_no_comments = insn.partition('#')[0]
        # find first space, to get opcode
        ls = insn_no_comments.split(' ')
        opcode = ls[0]
        # now find opcode fields
        fields = ''.join(ls[1:]).split(',')
        mfields = list(map(str.strip, fields))
        log("opcode, fields", ls, opcode, mfields)
        fields = []
        # macro substitution
        for field in mfields:
            fields.append(macro_subst(macros, field))
        log("opcode, fields substed", ls, opcode, fields)

        # identify if it is a special instruction
        custom_insn_hook = CUSTOM_INSNS.get(opcode)
        if custom_insn_hook is not None:
            fields = tuple(map(to_number, fields))
            insn_num = custom_insn_hook(fields)
            log(opcode, bin(insn_num))
            yield ".long 0x%X # %s" % (insn_num, insn)
            return

        # identify if is a svp64 mnemonic
        if not opcode.startswith('sv.'):
            yield insn  # unaltered
            return
        opcode = opcode[3:]  # strip leading "sv"

        # start working on decoding the svp64 op: sv.basev30Bop/vec2/mode
        opmodes = opcode.split("/")  # split at "/"
        v30b_op_orig = opmodes.pop(0)    # first is the v3.0B
        # check instruction ends with dot
        rc_mode = v30b_op_orig.endswith('.')
        if rc_mode:
            v30b_op = v30b_op_orig[:-1]
        else:
            v30b_op = v30b_op_orig

        # look up the 32-bit op (original, with "." if it has it)
        if v30b_op_orig in isa.instr:
            isa_instr = isa.instr[v30b_op_orig]
        else:
            raise Exception("opcode %s of '%s' not supported" %
                            (v30b_op_orig, insn))

        # look up the svp64 op, first the original (with "." if it has it)
        if v30b_op_orig in svp64.instrs:
            rm = svp64.instrs[v30b_op_orig]  # one row of the svp64 RM CSV
        # then without the "." (if there was one)
        elif v30b_op in svp64.instrs:
            rm = svp64.instrs[v30b_op]  # one row of the svp64 RM CSV
        else:
            raise Exception(f"opcode {v30b_op_orig!r} of "
                            f"{insn!r} not an svp64 instruction")

        # get regs info e.g. "RT,RA,RB"
        v30b_regs = isa_instr.regs[0]
        log("v3.0B op", v30b_op, "Rc=1" if rc_mode else '')
        log("v3.0B regs", opcode, v30b_regs)
        log("RM", rm)

        # right.  the first thing to do is identify the ordering of
        # the registers, by name.  the EXTRA2/3 ordering is in
        # rm['0']..rm['3'] but those fields contain the names RA, BB
        # etc.  we have to read the pseudocode to understand which
        # reg is which in our instruction. sigh.

        # first turn the svp64 rm into a "by name" dict, recording
        # which position in the RM EXTRA it goes into
        # also: record if the src or dest was a CR, for sanity-checking
        # (elwidth overrides on CRs are banned)
        decode = decode_extra(rm)
        dest_reg_cr, src_reg_cr, svp64_src, svp64_dest = decode

        log("EXTRA field index, src", svp64_src)
        log("EXTRA field index, dest", svp64_dest)

        # okaaay now we identify the field value (opcode N,N,N) with
        # the pseudo-code info (opcode RT, RA, RB)
        assert len(fields) == len(v30b_regs), \
            "length of fields %s must match insn `%s` fields %s" % \
            (str(v30b_regs), insn, str(fields))
        opregfields = zip(fields, v30b_regs)  # err that was easy

        # now for each of those find its place in the EXTRA encoding
        # note there is the possibility (for LD/ST-with-update) of
        # RA occurring **TWICE**.  to avoid it getting added to the
        # v3.0B suffix twice, we spot it as a duplicate, here
        extras = OrderedDict()
        for idx, (field, regname) in enumerate(opregfields):
            imm, regname = decode_imm(regname)
            rtype = get_regtype(regname)
            log("    idx find", rtype, idx, field, regname, imm)
            if rtype is None:
                # probably an immediate field, append it straight
                extras[('imm', idx, False)] = (idx, field, None, None, None)
                continue
            extra = svp64_src.get(regname, None)
            if extra is not None:
                extra = ('s', extra, False)  # not a duplicate
                extras[extra] = (idx, field, regname, rtype, imm)
                log("    idx src", idx, extra, extras[extra])
            dextra = svp64_dest.get(regname, None)
            log("regname in", regname, dextra)
            if dextra is not None:
                is_a_duplicate = extra is not None  # duplicate spotted
                dextra = ('d', dextra, is_a_duplicate)
                extras[dextra] = (idx, field, regname, rtype, imm)
                log("    idx dst", idx, extra, extras[dextra])

        # great! got the extra fields in their associated positions:
        # also we know the register type. now to create the EXTRA encodings
        etype = rm['Etype']  # Extra type: EXTRA3/EXTRA2
        ptype = rm['Ptype']  # Predication type: Twin / Single
        extra_bits = 0
        v30b_newfields = []
        for extra_idx, (idx, field, rname, rtype, iname) in extras.items():
            # is it a field we don't alter/examine?  if so just put it
            # into newfields
            if rtype is None:
                v30b_newfields.append(field)
                continue

            # identify if this is a ld/st immediate(reg) thing
            ldst_imm = "(" in field and field[-1] == ')'
            if ldst_imm:
                immed, field = field[:-1].split("(")

            field, regmode = decode_reg(field, macros=macros)
            log("    ", extra_idx, rname, rtype,
                regmode, iname, field, end=" ")

            # see Mode field https://libre-soc.org/openpower/sv/svp64/
            # XXX TODO: the following is a bit of a laborious repeated
            # mess, which could (and should) easily be parameterised.
            # XXX also TODO: the LD/ST modes which are different
            # https://libre-soc.org/openpower/sv/ldst/

            # rright.  SVP64 register numbering is from 0 to 127
            # for GPRs, FPRs *and* CR Fields, where for v3.0 the GPRs and RPFs
            # are 0-31 and CR Fields are only 0-7.  the SVP64 RM "Extra"
            # area is used to extend the numbering from the 32-bit
            # instruction, and also to record whether the register
            # is scalar or vector. on a per-operand basis.  this
            # results in a slightly finnicky encoding: here we go...

            # encode SV-GPR and SV-FPR field into extra, v3.0field
            if rtype in ['GPR', 'FPR']:
                sv_extra, field = get_extra_gpr(etype, regmode, field)
                # now sanity-check. EXTRA3 is ok, EXTRA2 has limits
                # (and shrink to a single bit if ok)
                if etype == 'EXTRA2':
                    if regmode == 'scalar':
                        # range is r0-r63 in increments of 1
                        assert (sv_extra >> 1) == 0, \
                            "scalar GPR %s cannot fit into EXTRA2 %s" % \
                            (rname, str(extras[extra_idx]))
                        # all good: encode as scalar
                        sv_extra = sv_extra & 0b01
                    else:
                        # range is r0-r127 in increments of 2 (r0 r2 ... r126)
                        assert sv_extra & 0b01 == 0, \
                            "%s: vector field %s cannot fit " \
                            "into EXTRA2 %s" % \
                            (insn, rname, str(extras[extra_idx]))
                        # all good: encode as vector (bit 2 set)
                        sv_extra = 0b10 | (sv_extra >> 1)
                elif regmode == 'vector':
                    # EXTRA3 vector bit needs marking
                    sv_extra |= 0b100

            # encode SV-CR 3-bit field into extra, v3.0field.
            # 3-bit is for things like BF and BFA
            elif rtype == 'CR_3bit':
                sv_extra, field = crf_extra(etype, regmode, field, extras)

            # encode SV-CR 5-bit field into extra, v3.0field
            # 5-bit is for things like BA BB BC BT etc.
            # *sigh* this is the same as 3-bit except the 2 LSBs of the
            # 5-bit field are passed through unaltered.
            elif rtype == 'CR_5bit':
                cr_subfield = field & 0b11  # record bottom 2 bits for later
                field = field >> 2         # strip bottom 2 bits
                # use the exact same 3-bit function for the top 3 bits
                sv_extra, field = crf_extra(etype, regmode, field, extras)
                # reconstruct the actual 5-bit CR field (preserving the
                # bottom 2 bits, unaltered)
                field = (field << 2) | cr_subfield

            else:
                raise Exception("no type match: %s" % rtype)

            # capture the extra field info
            log("=>", "%5s" % bin(sv_extra), field)
            extras[extra_idx] = sv_extra

            # append altered field value to v3.0b, differs for LDST
            # note that duplicates are skipped e.g. EXTRA2 contains
            # *BOTH* s:RA *AND* d:RA which happens on LD/ST-with-update
            srcdest, idx, duplicate = extra_idx
            if duplicate:  # skip adding to v3.0b fields, already added
                continue
            if ldst_imm:
                v30b_newfields.append(("%s(%s)" % (immed, str(field))))
            else:
                v30b_newfields.append(str(field))

        log("new v3.0B fields", v30b_op, v30b_newfields)
        log("extras", extras)

        # rright. now we have all the info. start creating SVP64 instruction.
        db = Database(find_wiki_dir())
        svp64_insn = SVP64Instruction.pair(prefix=0, suffix=0)
        svp64_prefix = svp64_insn.prefix
        svp64_rm = svp64_insn.prefix.rm

        # begin with EXTRA fields
        for idx, sv_extra in extras.items():
            log(idx)
            if idx is None:
                continue
            if idx[0] == 'imm':
                continue
            srcdest, idx, duplicate = idx
            if etype == 'EXTRA2':
                svp64_rm.extra2[idx] = sv_extra
            else:
                svp64_rm.extra3[idx] = sv_extra

        # identify if the op is a LD/ST. the "blegh" way. copied
        # from power_enums.  TODO, split the list _insns down.
        is_ld = v30b_op in [
            "lbarx", "lbz", "lbzu", "lbzux", "lbzx",            # load byte
            "ld", "ldarx", "ldbrx", "ldu", "ldux", "ldx",       # load double
            "lfs", "lfsx", "lfsu", "lfsux",                     # FP load single
            "lfd", "lfdx", "lfdu", "lfdux", "lfiwzx", "lfiwax",  # FP load dbl
            "lha", "lharx", "lhau", "lhaux", "lhax",            # load half
            "lhbrx", "lhz", "lhzu", "lhzux", "lhzx",            # more load half
            "lwa", "lwarx", "lwaux", "lwax", "lwbrx",           # load word
            "lwz", "lwzcix", "lwzu", "lwzux", "lwzx",           # more load word
        ]
        is_st = v30b_op in [
            "stb", "stbcix", "stbcx", "stbu", "stbux", "stbx",
            "std", "stdbrx", "stdcx", "stdu", "stdux", "stdx",
            "stfs", "stfsx", "stfsu", "stfux",                  # FP store sgl
            "stfd", "stfdx", "stfdu", "stfdux", "stfiwx",       # FP store dbl
            "sth", "sthbrx", "sthcx", "sthu", "sthux", "sthx",
            "stw", "stwbrx", "stwcx", "stwu", "stwux", "stwx",
        ]
        # use this to determine if the SVP64 RM format is different.
        # see https://libre-soc.org/openpower/sv/ldst/
        is_ldst = is_ld or is_st

        # branch-conditional detection
        is_bc = v30b_op in [
            "bc", "bclr",
        ]

        # parts of svp64_rm
        mmode = 0  # bit 0
        pmask = 0  # bits 1-3
        destwid = 0  # bits 4-5
        srcwid = 0  # bits 6-7
        subvl = 0   # bits 8-9
        smask = 0  # bits 16-18 but only for twin-predication
        mode = 0  # bits 19-23

        mask_m_specified = False
        has_pmask = False
        has_smask = False

        saturation = None
        src_zero = 0
        dst_zero = 0
        sv_mode = None

        mapreduce = False
        reverse_gear = False
        mapreduce_crm = False
        mapreduce_svm = False

        predresult = False
        failfirst = False
        ldst_elstride = 0

        # branch-conditional bits
        bc_all = 0
        bc_lru = 0
        bc_brc = 0
        bc_svstep = 0
        bc_vsb = 0
        bc_vlset = 0
        bc_vli = 0
        bc_snz = 0

        # ok let's start identifying opcode augmentation fields
        for encmode in opmodes:
            # predicate mask (src and dest)
            if encmode.startswith("m="):
                pme = encmode
                pmmode, pmask = decode_predicate(encmode[2:])
                smmode, smask = pmmode, pmask
                mmode = pmmode
                mask_m_specified = True
            # predicate mask (dest)
            elif encmode.startswith("dm="):
                pme = encmode
                pmmode, pmask = decode_predicate(encmode[3:])
                mmode = pmmode
                has_pmask = True
            # predicate mask (src, twin-pred)
            elif encmode.startswith("sm="):
                sme = encmode
                smmode, smask = decode_predicate(encmode[3:])
                mmode = smmode
                has_smask = True
            # vec2/3/4
            elif encmode.startswith("vec"):
                subvl = decode_subvl(encmode[3:])
            # elwidth
            elif encmode.startswith("ew="):
                destwid = decode_elwidth(encmode[3:])
            elif encmode.startswith("sw="):
                srcwid = decode_elwidth(encmode[3:])
            # element-strided LD/ST
            elif encmode == 'els':
                ldst_elstride = 1
            # saturation
            elif encmode == 'sats':
                assert sv_mode is None
                saturation = 1
                sv_mode = 0b10
            elif encmode == 'satu':
                assert sv_mode is None
                sv_mode = 0b10
                saturation = 0
            # predicate zeroing
            elif encmode == 'sz':
                src_zero = 1
            elif encmode == 'dz':
                dst_zero = 1
            # failfirst
            elif encmode.startswith("ff="):
                assert sv_mode is None
                sv_mode = 0b01
                failfirst = decode_ffirst(encmode[3:])
            # predicate-result, interestingly same as fail-first
            elif encmode.startswith("pr="):
                assert sv_mode is None
                sv_mode = 0b11
                predresult = decode_ffirst(encmode[3:])
            # map-reduce mode, reverse-gear
            elif encmode == 'mrr':
                assert sv_mode is None
                sv_mode = 0b00
                mapreduce = True
                reverse_gear = True
            # map-reduce mode
            elif encmode == 'mr':
                assert sv_mode is None
                sv_mode = 0b00
                mapreduce = True
            elif encmode == 'crm':  # CR on map-reduce
                assert sv_mode is None
                sv_mode = 0b00
                mapreduce_crm = True
            elif encmode == 'svm':  # sub-vector mode
                mapreduce_svm = True
            elif is_bc:
                if encmode == 'all':
                    bc_all = 1
                elif encmode == 'st':  # svstep mode
                    bc_step = 1
                elif encmode == 'sr':  # svstep BRc mode
                    bc_step = 1
                    bc_brc = 1
                elif encmode == 'vs':  # VLSET mode
                    bc_vlset = 1
                elif encmode == 'vsi':  # VLSET mode with VLI (VL inclusives)
                    bc_vlset = 1
                    bc_vli = 1
                elif encmode == 'vsb':  # VLSET mode with VSb
                    bc_vlset = 1
                    bc_vsb = 1
                elif encmode == 'vsbi':  # VLSET mode with VLI and VSb
                    bc_vlset = 1
                    bc_vli = 1
                    bc_vsb = 1
                elif encmode == 'snz':  # sz (only) already set above
                    src_zero = 1
                    bc_snz = 1
                elif encmode == 'lu':  # LR update mode
                    bc_lru = 1
                else:
                    raise AssertionError("unknown encmode %s" % encmode)
            else:
                raise AssertionError("unknown encmode %s" % encmode)

        if ptype == '2P':
            # since m=xx takes precedence (overrides) sm=xx and dm=xx,
            # treat them as mutually exclusive
            if mask_m_specified:
                assert not has_smask,\
                    "cannot have both source-mask and predicate mask"
                assert not has_pmask,\
                    "cannot have both dest-mask and predicate mask"
            # since the default is INT predication (ALWAYS), if you
            # specify one CR mask, you must specify both, to avoid
            # mixing INT and CR reg types
            if has_pmask and pmmode == 1:
                assert has_smask, \
                    "need explicit source-mask in CR twin predication"
            if has_smask and smmode == 1:
                assert has_pmask, \
                    "need explicit dest-mask in CR twin predication"
            # sanity-check that 2Pred mask is same mode
            if has_pmask and has_smask:
                assert smmode == pmmode, \
                    "predicate masks %s and %s must be same reg type" % \
                    (pme, sme)

        # sanity-check that twin-predication mask only specified in 2P mode
        if ptype == '1P':
            assert not has_smask, \
                "source-mask can only be specified on Twin-predicate ops"
            assert not has_pmask, \
                "dest-mask can only be specified on Twin-predicate ops"

        # construct the mode field, doing sanity-checking along the way
        if mapreduce_svm:
            assert sv_mode == 0b00, "sub-vector mode in mapreduce only"
            assert subvl != 0, "sub-vector mode not possible on SUBVL=1"

        if src_zero:
            assert has_smask or mask_m_specified, \
                "src zeroing requires a source predicate"
        if dst_zero:
            assert has_pmask or mask_m_specified, \
                "dest zeroing requires a dest predicate"

        # okaaay, so there are 4 different modes, here, which will be
        # partly-merged-in: is_ldst is merged in with "normal", but
        # is_bc is so different it's done separately.  likewise is_cr
        # (when it is done).  here are the maps:

        # for "normal" arithmetic: https://libre-soc.org/openpower/sv/normal/
        """
            | 0-1 |  2  |  3   4  |  description              |
            | --- | --- |---------|-------------------------- |
            | 00  |   0 |  dz  sz | normal mode                      |
            | 00  |   1 | 0  RG   | scalar reduce mode (mapreduce), SUBVL=1 |
            | 00  |   1 | 1  /    | parallel reduce mode (mapreduce), SUBVL=1 |
            | 00  |   1 | SVM RG  | subvector reduce mode, SUBVL>1   |
            | 01  | inv | CR-bit  | Rc=1: ffirst CR sel              |
            | 01  | inv | VLi RC1 |  Rc=0: ffirst z/nonz |
            | 10  |   N | dz   sz |  sat mode: N=0/1 u/s |
            | 11  | inv | CR-bit  |  Rc=1: pred-result CR sel |
            | 11  | inv | dz  RC1 |  Rc=0: pred-result z/nonz |
        """

        # https://libre-soc.org/openpower/sv/ldst/
        # for LD/ST-immediate:
        """
            | 0-1 |  2  |  3   4  |  description               |
            | --- | --- |---------|--------------------------- |
            | 00  | 0   |  dz els | normal mode                |
            | 00  | 1   |  dz shf | shift mode                 |
            | 01  | inv | CR-bit  | Rc=1: ffirst CR sel        |
            | 01  | inv | els RC1 |  Rc=0: ffirst z/nonz       |
            | 10  |   N | dz  els |  sat mode: N=0/1 u/s       |
            | 11  | inv | CR-bit  |  Rc=1: pred-result CR sel  |
            | 11  | inv | els RC1 |  Rc=0: pred-result z/nonz  |
        """

        # for LD/ST-indexed (RA+RB):
        """
            | 0-1 |  2  |  3   4  |  description              |
            | --- | --- |---------|-------------------------- |
            | 00  | SEA |  dz  sz | normal mode        |
            | 01  | SEA | dz sz  | Strided (scalar only source)   |
            | 10  |   N | dz   sz |  sat mode: N=0/1 u/s |
            | 11  | inv | CR-bit  |  Rc=1: pred-result CR sel |
            | 11  | inv | dz  RC1 |  Rc=0: pred-result z/nonz |
        """

        # and leaving out branches and cr_ops for now because they're
        # under development
        """ TODO branches and cr_ops
        """

        # now create mode and (overridden) src/dst widths
        # XXX TODO: sanity-check bc modes
        if is_bc:
            sv_mode = ((bc_svstep << SVP64MODE.MOD2_MSB) |
                       (bc_vlset << SVP64MODE.MOD2_LSB) |
                       (bc_snz << SVP64MODE.BC_SNZ))
            srcwid = (bc_vsb << 1) | bc_lru
            destwid = (bc_lru << 1) | bc_all

        else:

            ######################################
            # "normal" mode
            if sv_mode is None:
                mode |= src_zero << SVP64MODE.SZ  # predicate zeroing
                mode |= dst_zero << SVP64MODE.DZ  # predicate zeroing
                if is_ldst:
                    # TODO: for now, LD/ST-indexed is ignored.
                    mode |= ldst_elstride << SVP64MODE.ELS_NORMAL  # el-strided
                else:
                    # TODO, reduce and subvector mode
                    # 00  1   dz CRM  reduce mode (mapreduce), SUBVL=1
                    # 00  1   SVM CRM subvector reduce mode, SUBVL>1
                    pass
                sv_mode = 0b00

            ######################################
            # "mapreduce" modes
            elif sv_mode == 0b00:
                mode |= (0b1 << SVP64MODE.REDUCE)  # sets mapreduce
                assert dst_zero == 0, "dest-zero not allowed in mapreduce mode"
                if reverse_gear:
                    mode |= (0b1 << SVP64MODE.RG)  # sets Reverse-gear mode
                if mapreduce_crm:
                    mode |= (0b1 << SVP64MODE.CRM)  # sets CRM mode
                    assert rc_mode, "CRM only allowed when Rc=1"
                # bit of weird encoding to jam zero-pred or SVM mode in.
                # SVM mode can be enabled only when SUBVL=2/3/4 (vec2/3/4)
                if subvl == 0:
                    mode |= dst_zero << SVP64MODE.DZ  # predicate zeroing
                elif mapreduce_svm:
                    mode |= (0b1 << SVP64MODE.SVM)  # sets SVM mode

            ######################################
            # "failfirst" modes
            elif sv_mode == 0b01:
                assert src_zero == 0, "dest-zero not allowed in failfirst mode"
                if failfirst == 'RC1':
                    mode |= (0b1 << SVP64MODE.RC1)  # sets RC1 mode
                    mode |= (dst_zero << SVP64MODE.DZ)  # predicate dst-zeroing
                    assert rc_mode == False, "ffirst RC1 only ok when Rc=0"
                elif failfirst == '~RC1':
                    mode |= (0b1 << SVP64MODE.RC1)  # sets RC1 mode
                    mode |= (dst_zero << SVP64MODE.DZ)  # predicate dst-zeroing
                    mode |= (0b1 << SVP64MODE.INV)  # ... with inversion
                    assert rc_mode == False, "ffirst RC1 only ok when Rc=0"
                else:
                    assert dst_zero == 0, "dst-zero not allowed in ffirst BO"
                    assert rc_mode, "ffirst BO only possible when Rc=1"
                    mode |= (failfirst << SVP64MODE.BO_LSB)  # set BO

            ######################################
            # "saturation" modes
            elif sv_mode == 0b10:
                mode |= src_zero << SVP64MODE.SZ  # predicate zeroing
                mode |= dst_zero << SVP64MODE.DZ  # predicate zeroing
                mode |= (saturation << SVP64MODE.N)  # signed/us saturation

            ######################################
            # "predicate-result" modes.  err... code-duplication from ffirst
            elif sv_mode == 0b11:
                assert src_zero == 0, "dest-zero not allowed in predresult mode"
                if predresult == 'RC1':
                    mode |= (0b1 << SVP64MODE.RC1)  # sets RC1 mode
                    mode |= (dst_zero << SVP64MODE.DZ)  # predicate dst-zeroing
                    assert rc_mode == False, "pr-mode RC1 only ok when Rc=0"
                elif predresult == '~RC1':
                    mode |= (0b1 << SVP64MODE.RC1)  # sets RC1 mode
                    mode |= (dst_zero << SVP64MODE.DZ)  # predicate dst-zeroing
                    mode |= (0b1 << SVP64MODE.INV)  # ... with inversion
                    assert rc_mode == False, "pr-mode RC1 only ok when Rc=0"
                else:
                    assert dst_zero == 0, "dst-zero not allowed in pr-mode BO"
                    assert rc_mode, "pr-mode BO only possible when Rc=1"
                    mode |= (predresult << SVP64MODE.BO_LSB)  # set BO

        # whewww.... modes all done :)
        # now put into svp64_rm
        mode |= sv_mode
        # mode: bits 19-23
        svp64_rm.mode = mode

        # put in predicate masks into svp64_rm
        if ptype == '2P':
            # source pred: bits 16-18
            svp64_rm.smask = smask
        # mask mode: bit 0
        svp64_rm.mmode = mmode
        # 1-pred: bits 1-3
        svp64_rm.mask = pmask

        # and subvl: bits 8-9
        svp64_rm.subvl = subvl

        # put in elwidths
        # srcwid: bits 6-7
        svp64_rm.ewsrc = srcwid
        # destwid: bits 4-5
        svp64_rm.elwidth = destwid

        # nice debug printout. (and now for something completely different)
        # https://youtu.be/u0WOIwlXE9g?t=146
        svp64_rm_value = int(svp64_rm)
        log("svp64_rm", hex(svp64_rm_value), bin(svp64_rm_value))
        log("    mmode  0    :", bin(mmode))
        log("    pmask  1-3  :", bin(pmask))
        log("    dstwid 4-5  :", bin(destwid))
        log("    srcwid 6-7  :", bin(srcwid))
        log("    subvl  8-9  :", bin(subvl))
        log("    mode   19-23:", bin(mode))
        offs = 2 if etype == 'EXTRA2' else 3  # 2 or 3 bits
        for idx, sv_extra in extras.items():
            if idx is None:
                continue
            if idx[0] == 'imm':
                continue
            srcdest, idx, duplicate = idx
            start = (10+idx*offs)
            end = start + offs-1
            log("    extra%d %2d-%2d:" % (idx, start, end),
                bin(sv_extra))
        if ptype == '2P':
            log("    smask  16-17:", bin(smask))
        log()

        # update prefix PO and ID (aka PID)
        svp64_prefix.po = 0x1
        svp64_prefix.id = 0b11

        # fiinally yield the svp64 prefix and the thingy.  v3.0b opcode
        rc = '.' if rc_mode else ''
        yield ".long 0x%08x" % int(svp64_prefix)
        log(v30b_op, v30b_newfields)

        v30b_op_rc = v30b_op
        if not v30b_op.endswith('.'):
            v30b_op_rc += rc

        # svstep is weird
        # FIXME(lkcl): should sv.svstep be like svstep?
        if v30b_op_rc in ("svstep", "svstep."):
            # compensate for `SVi -= 1` in svstep()
            v30b_newfields[1] = str(int(v30b_newfields[1]) + 1)

        custom_insn_hook = CUSTOM_INSNS.get(v30b_op_rc)
        if custom_insn_hook is not None:
            fields = tuple(map(to_number, v30b_newfields))
            insn_num = custom_insn_hook(fields)
            log(opcode, bin(insn_num))
            yield ".long 0x%X # %s" % (insn_num, insn)
            return
        # argh, sv.fmadds etc. need to be done manually
        elif v30b_op == 'ffmadds':
            opcode = 59 << (32-6)    # bits 0..6 (MSB0)
            opcode |= int(v30b_newfields[0]) << (32-11)  # FRT
            opcode |= int(v30b_newfields[1]) << (32-16)  # FRA
            opcode |= int(v30b_newfields[2]) << (32-21)  # FRB
            opcode |= int(v30b_newfields[3]) << (32-26)  # FRC
            opcode |= 0b00101 << (32-31)   # bits 26-30
            if rc:
                opcode |= 1  # Rc, bit 31.
            yield ".long 0x%x" % opcode
        # argh, sv.fdmadds need to be done manually
        elif v30b_op == 'fdmadds':
            opcode = 59 << (32-6)    # bits 0..6 (MSB0)
            opcode |= int(v30b_newfields[0]) << (32-11)  # FRT
            opcode |= int(v30b_newfields[1]) << (32-16)  # FRA
            opcode |= int(v30b_newfields[2]) << (32-21)  # FRB
            opcode |= int(v30b_newfields[3]) << (32-26)  # FRC
            opcode |= 0b01111 << (32-31)   # bits 26-30
            if rc:
                opcode |= 1  # Rc, bit 31.
            yield ".long 0x%x" % opcode
        # argh, sv.ffadds etc. need to be done manually
        elif v30b_op == 'ffadds':
            opcode = 59 << (32-6)    # bits 0..6 (MSB0)
            opcode |= int(v30b_newfields[0]) << (32-11)  # FRT
            opcode |= int(v30b_newfields[1]) << (32-16)  # FRA
            opcode |= int(v30b_newfields[2]) << (32-21)  # FRB
            opcode |= 0b01101 << (32-31)   # bits 26-30
            if rc:
                opcode |= 1  # Rc, bit 31.
            yield ".long 0x%x" % opcode
        else:
            if not v30b_op.endswith('.'):
                v30b_op += rc
            yield "%s %s" % (v30b_op, ", ".join(v30b_newfields))
        log("new v3.0B fields", v30b_op, v30b_newfields)

    def translate(self, lst):
        for insn in lst:
            yield from self.translate_one(insn)


def macro_subst(macros, txt):
    again = True
    log("subst", txt, macros)
    while again:
        again = False
        for macro, value in macros.items():
            if macro == txt:
                again = True
                replaced = txt.replace(macro, value)
                log("macro", txt, "replaced", replaced, macro, value)
                txt = replaced
                continue
            toreplace = '%s.s' % macro
            if toreplace == txt:
                again = True
                replaced = txt.replace(toreplace, "%s.s" % value)
                log("macro", txt, "replaced", replaced, toreplace, value)
                txt = replaced
                continue
            toreplace = '%s.v' % macro
            if toreplace == txt:
                again = True
                replaced = txt.replace(toreplace, "%s.v" % value)
                log("macro", txt, "replaced", replaced, toreplace, value)
                txt = replaced
                continue
            toreplace = '(%s)' % macro
            if toreplace in txt:
                again = True
                replaced = txt.replace(toreplace, '(%s)' % value)
                log("macro", txt, "replaced", replaced, toreplace, value)
                txt = replaced
                continue
    log("    processed", txt)
    return txt


def get_ws(line):
    # find whitespace
    ws = ''
    while line:
        if not line[0].isspace():
            break
        ws += line[0]
        line = line[1:]
    return ws, line


def asm_process():
    # get an input file and an output file
    args = sys.argv[1:]
    if len(args) == 0:
        infile = sys.stdin
        outfile = sys.stdout
        # read the whole lot in advance in case of in-place
        lines = list(infile.readlines())
    elif len(args) != 2:
        print("pysvp64asm [infile | -] [outfile | -]", file=sys.stderr)
        exit(0)
    else:
        if args[0] == '--':
            infile = sys.stdin
        else:
            infile = open(args[0], "r")
        # read the whole lot in advance in case of in-place overwrite
        lines = list(infile.readlines())

        if args[1] == '--':
            outfile = sys.stdout
        else:
            outfile = open(args[1], "w")

    # read the line, look for custom insn, process it
    macros = {}  # macros which start ".set"
    isa = SVP64Asm([])
    for line in lines:
        op = line.split("#")[0].strip()
        # identify macros
        if op.startswith(".set"):
            macro = op[4:].split(",")
            (macro, value) = map(str.strip, macro)
            macros[macro] = value
        if not op.startswith('sv.') and not op.startswith(tuple(CUSTOM_INSNS)):
            outfile.write(line)
            continue

        (ws, line) = get_ws(line)
        lst = isa.translate_one(op, macros)
        lst = '; '.join(lst)
        outfile.write("%s%s # %s\n" % (ws, lst, op))


if __name__ == '__main__':
    lst = ['slw 3, 1, 4',
           'extsw 5, 3',
           'sv.extsw 5, 3',
           'sv.cmpi 5, 1, 3, 2',
           'sv.setb 5, 31',
           'sv.isel 64.v, 3, 2, 65.v',
           'sv.setb/dm=r3/sm=1<<r3 5, 31',
           'sv.setb/m=r3 5, 31',
           'sv.setb/vec2 5, 31',
           'sv.setb/sw=8/ew=16 5, 31',
           'sv.extsw./ff=eq 5, 31',
           'sv.extsw./satu/sz/dz/sm=r3/dm=r3 5, 31',
           'sv.extsw./pr=eq 5.v, 31',
           'sv.add. 5.v, 2.v, 1.v',
           'sv.add./m=r3 5.v, 2.v, 1.v',
           ]
    lst += [
        'sv.stw 5.v, 4(1.v)',
        'sv.ld 5.v, 4(1.v)',
        'setvl. 2, 3, 4, 0, 1, 1',
        'sv.setvl. 2, 3, 4, 0, 1, 1',
    ]
    lst = [
        "sv.stfsu 0.v, 16(4.v)",
    ]
    lst = [
        "sv.stfsu/els 0.v, 16(4)",
    ]
    lst = [
        'sv.add./mr 5.v, 2.v, 1.v',
    ]
    macros = {'win2': '50', 'win': '60'}
    lst = [
        'sv.addi win2.v, win.v, -1',
        'sv.add./mrr 5.v, 2.v, 1.v',
        #'sv.lhzsh 5.v, 11(9.v), 15',
        #'sv.lwzsh 5.v, 11(9.v), 15',
        'sv.ffmadds 6.v, 2.v, 4.v, 6.v',
    ]
    lst = [
        #'sv.fmadds 0.v, 8.v, 16.v, 4.v',
        #'sv.ffadds 0.v, 8.v, 4.v',
        'svremap 11, 0, 1, 2, 3, 2, 1',
        'svshape 8, 1, 1, 1, 0',
        'svshape 8, 1, 1, 1, 1',
    ]
    lst = [
        #'sv.lfssh 4.v, 11(8.v), 15',
        #'sv.lwzsh 4.v, 11(8.v), 15',
        #'sv.svstep. 2.v, 4, 0',
        #'sv.fcfids. 48.v, 64.v',
        'sv.fcoss. 80.v, 0.v',
        'sv.fcoss. 20.v, 0.v',
    ]
    lst = [
        'sv.bc/all 3,12,192',
        'sv.bclr/vsbi 3,81.v,192',
        'sv.ld 5.v, 4(1.v)',
        'sv.svstep. 2.v, 4, 0',
    ]
    lst = [
        'maxs 3,12,5',
        'maxs. 3,12,5',
        'avgadd 3,12,5',
        'absdu 3,12,5',
        'absds 3,12,5',
        'absdacu 3,12,5',
        'absdacs 3,12,5',
        'cprop 3,12,5',
        'svindex 0,0,1,0,0,0,0',
    ]
    lst = [
        'sv.svstep./m=r3 2.v, 4, 0',
        'ternlogi 0,0,0,0x5',
        'fmvis 5,65535',
        'fmvis 5,1',
        'fmvis 5,2',
        'fmvis 5,4',
        'fmvis 5,8',
        'fmvis 5,16',
        'fmvis 5,32',
        'fmvis 5,64',
        'fmvis 5,32768',
    ]
    lst = [
        'sv.andi. *80, *80, 1',
        'sv.ffmadds. 6.v, 2.v, 4.v, 6.v',  # incorrectly inserted 32-bit op
        'sv.ffmadds 6.v, 2.v, 4.v, 6.v',  # correctly converted to .long
        'svshape2 8, 1, 31, 7, 1, 1',
    ]
    isa = SVP64Asm(lst, macros=macros)
    log("list:\n", "\n\t".join(list(isa)))
    # running svp64.py is designed to test hard-coded lists
    # (above) - which strictly speaking should all be unit tests.
    # if you need to actually do assembler translation at the
    # commandline use "pysvp64asm" - see setup.py
    # XXX NO. asm_process()