bifrost/valhall/valhall.py

#encoding=utf-8

# Copyright (C) 2016 Intel Corporation
# Copyright (C) 2016 Broadcom
# Copyright (C) 2020 Collabora, Ltd.
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the "Software"),
# to deal in the Software without restriction, including without limitation
# the rights to use, copy, modify, merge, publish, distribute, sublicense,
# and/or sell copies of the Software, and to permit persons to whom the
# Software is furnished to do so, subject to the following conditions:
#
# The above copyright notice and this permission notice (including the next
# paragraph) shall be included in all copies or substantial portions of the
# Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
# IN THE SOFTWARE.

import os
import textwrap
import xml.etree.ElementTree as ET
import sys

tree = ET.parse(os.path.join(os.path.dirname(__file__), 'ISA.xml'))
root = tree.getroot()

# All instructions in the ISA
instructions = []

# All immediates in the ISA
ilut = root.findall('lut')[0]
assert(ilut.attrib['name'] == "Immediates")
immediates = [int(imm.text, base=0) for imm in ilut.findall('constant')]
enums = {}

def xmlbool(s):
    assert(s.lower() in ["false", "true"])
    return False if s.lower() == "false" else True

class EnumValue:
    def __init__(self, value, default):
        self.value = value
        self.default = default

class Enum:
    def __init__(self, name, values):
        self.name = name
        self.values = values
        self.bare_values = [x.value for x in values]

        defaults = [x.value for x in values if x.default]
        if len(defaults) > 0:
            assert(len(defaults) == 1)
            self.default = defaults[0]

def build_enum(el):
    values = []

    for child in el:
        if child.tag == 'value':
            is_default = child.attrib.get('default', False)
            values.append(EnumValue(child.text, is_default))
        elif child.tag == 'reserved':
            values.append(EnumValue("reserved", False))

    return Enum(el.attrib['name'], values)

class Modifier:
    def __init__(self, name, start, size, implied = False):
        self.name = name
        self.start = start
        self.size = size
        self.implied = implied

        if size == 1:
            self.bare_values = ['', name]
            self.default = 0
        else:
            enum = enums[name]
            self.bare_values = [x.value for x in enum.values]
            defaults = [x for x in enum.values if x.default]
            assert(len(defaults) <= 1)

            if len(defaults) > 0:
                self.default = self.bare_values.index(defaults[0].value)
            else:
                self.default = None

def Flag(name, start):
    return Modifier(name, start, 1)

# Model a single instruction
class Source:
    def __init__(self, index, size, is_float = False, swizzle = False, widen = False, lanes = False, lane = None, absneg = False, notted = False, name = ""):
        self.is_float = is_float or absneg
        self.size = size
        self.absneg = absneg
        self.notted = notted
        self.swizzle = swizzle
        self.widen = widen
        self.lanes = lanes
        self.lane = lane
        self.name = name

        self.offset = {}
        self.bits = {}
        if absneg:
            self.offset['neg'] = 32 + 2 + ((2 - index) * 2)
            self.offset['abs'] = 33 + 2 + ((2 - index) * 2)
            self.bits['neg'] = 1
            self.bits['abs'] = 1
        if notted:
            self.offset['not'] = 35
            self.bits['not'] = 1
        if widen or lanes:
            self.offset['widen'] = 26 if index == 1 else 36
            self.bits['widen'] = 4 # XXX: too much?
        if lane:
            self.offset['lane'] = self.lane
            self.bits['lane'] = 2 if size in (8, 32) else 1
        if swizzle:
            assert(size in [16, 32])
            self.offset['swizzle'] = 24 + ((2 - index) * 2)
            self.bits['swizzle'] = 2

class Dest:
    def __init__(self, name = ""):
        self.name = name

class Staging:
    def __init__(self, read = False, write = False, index = 0, count = 0, flags = True, name = ""):
        self.name = name
        self.read = read
        self.write = write
        self.count = count
        self.flags = flags

        # For compatibility
        self.absneg = False
        self.swizzle = False
        self.notted = False
        self.widen = False
        self.lanes = False
        self.lane = False
        self.size = 32

        assert(index < 2)
        self.start = 40 if index == 0 else 16

        if not flags:
            self.encoded_flags = 0
        elif index > 0:
            self.encoded_flags = 0xC0
        else:
            self.encoded_flags = (0x80 if write else 0) | (0x40 if read else 0)

class Immediate:
    def __init__(self, name, start, size, signed):
        self.name = name
        self.start = start
        self.size = size
        self.signed = signed

class Instruction:
    def __init__(self, name, opcode, opcode2, srcs = [], dests = [], immediates = [], modifiers = [], staging = None):
        self.name = name
        self.srcs = srcs
        self.dests = dests
        self.opcode = opcode
        self.opcode2 = opcode2 or 0
        self.immediates = immediates
        self.modifiers = modifiers
        self.staging = staging

        self.secondary_shift = max(len(self.srcs) * 8, 16)
        self.secondary_mask = 0xF if opcode2 is not None else 0x0
        if "left" in [x.name for x in self.modifiers]:
            self.secondary_mask |= 0x100
        if len(srcs) == 3 and (srcs[1].widen or srcs[1].lanes):
            self.secondary_mask &= ~0xC # conflicts
        if opcode == 0x90:
            # XXX: XMLify this, but disambiguates sign of conversions
            self.secondary_mask |= 0x10
        if name.startswith("LOAD.i") or name.startswith("STORE.i"):
            self.secondary_shift = 27 # Alias with memory_size
            self.secondary_mask = 0x7

        assert(len(dests) == 0 or not staging)
        assert(not opcode2 or (opcode2 & self.secondary_mask) == opcode2)

    def __str__(self):
        return self.name

# Build a single source from XML
def build_source(el, i, size):
    lane = el.get('lane', None)
    if lane == "true":
        lane = 38 if i == 0 else 36
    elif lane is not None:
        lane = int(lane)

    return Source(i, int(el.get('size', size)),
            absneg = el.get('absneg', False),
            is_float = el.get('float', False),
            swizzle = el.get('swizzle', False),
            widen = el.get('widen', False),
            lanes = el.get('lanes', False),
            lane = lane,
            notted = el.get('not', False),
            name = el.text or "")

def build_imm(el):
    return Immediate(el.attrib['name'], int(el.attrib['start']),
            int(el.attrib['size']), bool(el.attrib.get('signed', False)))

def build_staging(i, el):
    r = xmlbool(el.attrib.get('read', 'false'))
    w = xmlbool(el.attrib.get('write', 'false'))
    count = int(el.attrib.get('count', '0'))
    flags = xmlbool(el.attrib.get('flags', 'true'))

    return Staging(r, w, i, count, flags, el.text or '')

def build_modifier(el):
    name = el.attrib['name']
    start = int(el.attrib['start'])
    size = int(el.attrib['size'])
    implied = xmlbool(el.get('implied', 'false'))

    return Modifier(name, start, size, implied)

# Build a single instruction from XML and group based overrides
def build_instr(el, overrides = {}):
    # Get overridables
    name = overrides.get('name') or el.attrib.get('name')
    opcode = overrides.get('opcode') or el.attrib.get('opcode')
    opcode2 = overrides.get('opcode2') or el.attrib.get('opcode2')
    opcode = int(opcode, base=0)
    opcode2 = int(opcode2, base=0) if opcode2 else None

    # Get explicit sources/dests
    tsize = typesize(name)
    sources = [build_source(src, i, tsize) for i, src in enumerate(el.findall('src'))]
    dests = [Dest(dest.text or '') for dest in el.findall('dest')]

    # Get implicit ones
    sources = sources + ([Source(i, int(tsize)) for i in range(int(el.attrib.get('srcs', 0)))])
    dests = dests + ([Dest()] * int(el.attrib.get('dests', 0)))

    # Get staging registers
    staging = [build_staging(i, el) for i, el in enumerate(el.findall('sr'))]

    # Get immediates
    imms = [build_imm(imm) for imm in el.findall('imm')]

    modifiers = []
    for mod in el:
        if mod.tag in MODIFIERS:
            modifiers.append(MODIFIERS[mod.tag])
        elif mod.tag =='mod':
            modifiers.append(build_modifier(mod))

    instr = Instruction(name, opcode, opcode2, srcs = sources, dests = dests, immediates = imms, modifiers = modifiers, staging = staging)

    instructions.append(instr)

# Build all the instructions in a group by duplicating the group itself with
# overrides for each distinct instruction
def build_group(el):
    for ins in el.findall('ins'):
        build_instr(el, overrides = {
            'name': ins.attrib['name'],
            'opcode': ins.attrib.get('opcode'),
            'opcode2': ins.attrib.get('opcode2'),
        })

def to_alphanum(name):
    substitutions = {
        ' ': '_',
        '/': '_',
        '[': '',
        ']': '',
        '(': '',
        ')': '',
        '-': '_',
        ':': '',
        '.': '',
        ',': '',
        '=': '',
        '>': '',
        '#': '',
        '&': '',
        '*': '',
        '"': '',
        '+': '',
        '\'': '',
    }

    for i, j in substitutions.items():
        name = name.replace(i, j)

    return name

def safe_name(name):
    name = to_alphanum(name)
    if not name[0].isalpha():
        name = '_' + name

    return name.lower()

# Parses out the size part of an opocde name
def typesize(opcode):
    if opcode[-3:] == '128':
        return 128
    if opcode[-2:] == '48':
        return 48
    elif opcode[-1] == '8':
        return 8
    else:
        try:
            return int(opcode[-2:])
        except:
            return 32

for child in root.findall('enum'):
    enums[safe_name(child.attrib['name'])] = build_enum(child)

MODIFIERS = {
    "inactive_result": Modifier("inactive_result", 22, 4),
    "store_segment": Modifier("store_segment", 24, 2),
    "regfmt": Modifier("register_format", 24, 3),
    "vecsize": Modifier("vector_size", 28, 2),

    "slot": Modifier("slot", 30, 3),
    "roundmode": Modifier("round_mode", 30, 2),
    "result_type": Modifier("result_type", 30, 2),
    "saturate": Flag("saturate", 30),
    "not_result": Flag("not_result", 30),

    "lane_op": Modifier("lane_operation", 32, 2),
    "cmp": Modifier("condition", 32, 3),
    "clamp": Modifier("clamp", 32, 2),
    "sr_count": Modifier("staging_register_count", 33, 3, implied = True),

    "subgroup": Modifier("subgroup_size", 36, 2),
}

# Parse the ISA
for child in root:
    if child.tag == 'group':
        build_group(child)
    elif child.tag == 'ins':
        build_instr(child)

instruction_dict = { ins.name: ins for ins in instructions }

# Validate there are no duplicated instructions
if len(instruction_dict) != len(instructions):
    import collections
    counts = collections.Counter([i.name for i in instructions])
    for c in counts:
        if counts[c] != 1:
            print(f'{c} appeared {counts[c]} times.')

assert(len(instruction_dict) == len(instructions))