opcode-gen/opcode-gen.awk

# Copyright (C) 2007 The Android Open Source Project
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

#
# Awk helper script for opcode-gen.
#

#
# Initialization.
#

BEGIN {
    MAX_OPCODE = 65535;
    MAX_PACKED_OPCODE = 511;
    MAX_PACKED_OPCODE = 255; # TODO: Not for long!
    initIndexTypes();
    initFlags();
    if (readBytecodes()) exit 1;
    deriveOpcodeChains();
    createPackedTables();
    consumeUntil = "";
    emission = "";
}

#
# General control (must appear above directive handlers).
#

# Clear out the preexisting output within a directive section.
consumeUntil != "" {
    if (index($0, consumeUntil) != 0) {
        consumeUntil = "";
        print;
    }

    next;
}

# Detect directives.
/BEGIN\([a-z-]*\)/ {
    i = match($0, /BEGIN\([a-z-]*\)/);
    emission = substr($0, i + 6, RLENGTH - 7);
    consumeUntil = "END(" emission ")";
    emissionHandled = 0;
}

# Most lines just get copied from the source as-is, including the start
# comment for directives.
{
    print;
}

#
# Handlers for all of the directives.
#

emission == "opcodes" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i) || isOptimized(i)) continue;
        printf("    public static final int %s = 0x%s;\n",
               constName[i], hex[i]);
    }
}

emission == "first-opcodes" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i) || isOptimized(i)) continue;
        if (isFirst[i] == "true") {
            printf("    //     Opcodes.%s\n", constName[i]);
        }
    }
}

emission == "dops" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i) || isOptimized(i)) continue;

        nextOp = nextOpcode[i];
        nextOp = (nextOp == -1) ? "NO_NEXT" : constName[nextOp];

        printf("    public static final Dop %s =\n" \
               "        new Dop(Opcodes.%s, Opcodes.%s,\n" \
               "            Opcodes.%s, Form%s.THE_ONE, %s);\n\n",
               constName[i], constName[i], family[i], nextOp, format[i],
               hasResult[i]);
    }
}

emission == "opcode-info-defs" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i) || isOptimized(i)) continue;

        itype = toupper(indexType[i]);
        gsub(/-/, "_", itype);

        printf("    public static final Info %s =\n" \
               "        new Info(Opcodes.%s, \"%s\",\n" \
               "            InstructionCodec.FORMAT_%s, IndexType.%s);\n\n", \
               constName[i], constName[i], name[i], toupper(format[i]), itype);
    }
}

emission == "dops-init" || emission == "opcode-info-init" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i) || isOptimized(i)) continue;
        printf("        set(%s);\n", constName[i]);
    }
}

emission == "libcore-opcodes" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i) || isOptimized(i)) continue;
        printf("    int OP_%-28s = 0x%04x;\n", constName[i], i);
    }
}

emission == "libcore-maximum-values" {
    emissionHandled = 1;

    printf("        MAXIMUM_VALUE = %d;\n", MAX_OPCODE);
    printf("        MAXIMUM_PACKED_VALUE = %d;\n", MAX_PACKED_OPCODE);
}

emission == "libdex-maximum-values" {
    emissionHandled = 1;

    printf("#define kMaxOpcodeValue 0x%x\n", MAX_OPCODE);
    printf("#define kNumPackedOpcodes 0x%x\n", MAX_PACKED_OPCODE + 1);
}

emission == "libdex-opcode-enum" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        printf("    OP_%-28s = 0x%02x,\n", packedConstName[i], i);
    }
}

emission == "libdex-goto-table" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        content = sprintf("        H(OP_%s),", packedConstName[i]);
        printf("%-78s\\\n", content);
    }
}

emission == "libdex-opcode-names" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        printf("    \"%s\",\n", packedName[i]);
    }
}

emission == "libdex-widths" {
    emissionHandled = 1;

    col = 1;
    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        value = sprintf("%d,", packedWidth[i]);
        col = colPrint(value, (i == MAX_PACKED_OPCODE), col, 16, 2, "    ");
    }
}

emission == "libdex-flags" {
    emissionHandled = 1;

    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        value = flagsToC(packedFlags[i]);
        printf("    %s,\n", value);
    }
}

emission == "libdex-formats" {
    emissionHandled = 1;

    col = 1;
    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        value = sprintf("kFmt%s,", packedFormat[i]);
        col = colPrint(value, (i == MAX_PACKED_OPCODE), col, 7, 9, "    ");
    }
}

emission == "libdex-index-types" {
    emissionHandled = 1;

    col = 1;
    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        value = sprintf("%s,", indexTypeValues[packedIndexType[i]]);
        col = colPrint(value, (i == MAX_PACKED_OPCODE), col, 3, 19, "    ");
    }
}

# Handle the end of directive processing (must appear after the directive
# clauses).
emission != "" {
    if (!emissionHandled) {
        printf("WARNING: unknown tag \"%s\"\n", emission) >"/dev/stderr";
        consumeUntil = "";
    }

    emission = "";
}

#
# Helper functions.
#

# Helper to print out an element in a multi-column fashion. It returns
# the (one-based) column number that the next element will be printed
# in.
function colPrint(value, isLast, col, numCols, colWidth, linePrefix) {
    isLast = (isLast || (col == numCols));
    printf("%s%-*s%s",
        (col == 1) ? linePrefix : " ",
        isLast ? 1 : colWidth, value,
        isLast ? "\n" : "");

    return (col % numCols) + 1;
}

# Read the bytecode description file.
function readBytecodes(i, parts, line, cmd, status, count) {
    # locals: parts, line, cmd, status, count
    for (;;) {
        # Read a line.
        status = getline line <bytecodeFile;
        if (status == 0) break;
        if (status < 0) {
            print "trouble reading bytecode file";
            exit 1;
        }

        # Clean up the line and extract the command.
        gsub(/  */, " ", line);
        sub(/ *#.*$/, "", line);
        sub(/ $/, "", line);
        sub(/^ /, "", line);
        count = split(line, parts);
        if (count == 0) continue; # Blank or comment line.
        cmd = parts[1];
        sub(/^[a-z][a-z]* */, "", line); # Remove the command from line.

        if (cmd == "op") {
            status = defineOpcode(line);
        } else if (cmd == "format") {
            status = defineFormat(line);
        } else {
            status = -1;
        }

        if (status != 0) {
            printf("syntax error on line: %s\n", line) >"/dev/stderr";
            return 1;
        }
    }

    return 0;
}

# Define an opcode.
function defineOpcode(line, count, parts, idx) {
    # locals: count, parts, idx
    count = split(line, parts);
    if (count != 6)  return -1;
    idx = parseHex(parts[1]);
    if (idx < 0) return -1;

    # Extract directly specified values from the line.
    hex[idx] = parts[1];
    name[idx] = parts[2];
    format[idx] = parts[3];
    hasResult[idx] = (parts[4] == "n") ? "false" : "true";
    indexType[idx] = parts[5];
    flags[idx] = parts[6];

    # Calculate derived values.

    constName[idx] = toupper(name[idx]);
    gsub("[/-]", "_", constName[idx]);   # Dash and slash become underscore.
    gsub("[+^]", "", constName[idx]);    # Plus and caret are removed.
    split(name[idx], parts, "/");

    family[idx] = toupper(parts[1]);
    gsub("-", "_", family[idx]);         # Dash becomes underscore.
    gsub("[+^]", "", family[idx]);       # Plus and caret are removed.

    split(format[idx], parts, "");       # Width is the first format char.
    width[idx] = parts[1];

    # This association is used when computing "next" opcodes.
    familyFormat[family[idx],format[idx]] = idx;

    # Verify values.

    if (nextFormat[format[idx]] == "") {
        printf("unknown format: %s\n", format[idx]) >"/dev/stderr";
        return 1;
    }

    if (indexTypeValues[indexType[idx]] == "") {
        printf("unknown index type: %s\n", indexType[idx]) >"/dev/stderr";
        return 1;
    }

    if (flagsToC(flags[idx]) == "") {
        printf("bogus flags: %s\n", flags[idx]) >"/dev/stderr";
        return 1;
    }

    return 0;
}

# Define a format family.
function defineFormat(line, count, parts, i) {
    # locals: count, parts, i
    count = split(line, parts);
    if (count < 1)  return -1;
    formats[parts[1]] = line;

    parts[count + 1] = "none";
    for (i = 1; i <= count; i++) {
        nextFormat[parts[i]] = parts[i + 1];
    }

    return 0;
}

# Produce the nextOpcode and isFirst arrays. The former indicates, for
# each opcode, which one should be tried next when doing instruction
# fitting. The latter indicates which opcodes are at the head of an
# instruction fitting chain.
function deriveOpcodeChains(i, op) {
    # locals: i, op

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i)) continue;
        isFirst[i] = "true";
    }

    for (i = 0; i <= MAX_OPCODE; i++) {
        if (isUnused(i)) continue;
        op = findNextOpcode(i);
        nextOpcode[i] = op;
        if (op != -1) {
            isFirst[op] = "false";
        }
    }
}

# Given an opcode by index, find the next opcode in the same family
# (that is, with the same base name) to try when matching instructions
# to opcodes. This simply walks the nextFormat chain looking for a
# match. This returns the index of the matching opcode or -1 if there
# is none.
function findNextOpcode(idx, fam, fmt, result) {
    # locals: fam, fmt, result
    fam = family[idx];
    fmt = format[idx];

    # Not every opcode has a version with every possible format, so
    # we have to iterate down the chain until we find one or run out of
    # formats to try.
    for (fmt = nextFormat[format[idx]]; fmt != "none"; fmt = nextFormat[fmt]) {
        result = familyFormat[fam,fmt];
        if (result != "") {
            return result;
        }
    }

    return -1;
}

# Construct the tables of info indexed by packed opcode. The packed opcode
# values are in the range 0-0x1ff, whereas the unpacked opcodes sparsely
# span the range 0-0xffff.
function createPackedTables(i, op) {
    # locals: i, op
    for (i = 0; i <= MAX_PACKED_OPCODE; i++) {
        op = unpackOpcode(i);
        if (isUnused(op)) {
            packedName[i]      = unusedName(op);
            packedConstName[i] = unusedConstName(op);
            packedFormat[i]    = "00x";
            packedFlags[i]     = 0;
            packedWidth[i]     = 0;
            packedIndexType[i] = "unknown";
        } else {
            packedName[i]      = name[op];
            packedConstName[i] = constName[op];
            packedFormat[i]    = format[op];
            packedFlags[i]     = flags[op];
            packedWidth[i]     = width[op];
            packedIndexType[i] = indexType[op];
        }
    }
}

# Given a packed opcode, returns the raw (unpacked) opcode value.
function unpackOpcode(idx) {
    # Note: This must be the inverse of the corresponding code in
    # libdex/DexOpcodes.h.
    if (idx <= 255) {
        return idx;
    } else {
        idx -= 256;
        return (idx * 256) + 255;
    }
}

# Returns the "unused" name of the given opcode (by index).
# That is, this is the human-oriented name to use for an opcode
# definition in cases
# where the opcode isn't used.
function unusedName(idx) {
    if (idx <= 255) {
         return sprintf("unused-%02x", idx);
    } else {
         return sprintf("unused-%04x", idx);
    }
}

# Returns the "unused" constant name of the given opcode (by index).
# That is, this is the name to use for a constant definition in cases
# where the opcode isn't used.
function unusedConstName(idx) {
    if (idx <= 255) {
         return toupper(sprintf("UNUSED_%02x", idx));
    } else {
         return toupper(sprintf("UNUSED_%04x", idx));
    }
}

# Convert a hex value to an int.
function parseHex(hex, result, chars, count, c, i) {
    # locals: result, chars, count, c, i
    hex = tolower(hex);
    count = split(hex, chars, "");
    result = 0;
    for (i = 1; i <= count; i++) {
        c = index("0123456789abcdef", chars[i]);
        if (c == 0) {
            printf("bogus hex value: %s\n", hex) >"/dev/stderr";
            return -1;
        }
        result = (result * 16) + c - 1;
    }
    return result;
}

# Initialize the indexTypes data.
function initIndexTypes() {
    indexTypeValues["unknown"]              = "kIndexUnknown";
    indexTypeValues["none"]                 = "kIndexNone";
    indexTypeValues["varies"]               = "kIndexVaries";
    indexTypeValues["type-ref"]             = "kIndexTypeRef";
    indexTypeValues["string-ref"]           = "kIndexStringRef";
    indexTypeValues["method-ref"]           = "kIndexMethodRef";
    indexTypeValues["field-ref"]            = "kIndexFieldRef";
    indexTypeValues["inline-method"]        = "kIndexInlineMethod";
    indexTypeValues["vtable-offset"]        = "kIndexVtableOffset";
    indexTypeValues["field-offset"]         = "kIndexFieldOffset";
    indexTypeValues["method-and-proto-ref"] = "kIndexMethodAndProtoRef";
    indexTypeValues["call-site-ref"]        = "kIndexCallSiteRef";
    indexTypeValues["method-handle-ref"]    = "kIndexMethodHandleRef";
    indexTypeValues["proto-ref"]            = "kIndexProtoRef";
}

# Initialize the flags data.
function initFlags() {
    flagValues["branch"]        = "kInstrCanBranch";
    flagValues["continue"]      = "kInstrCanContinue";
    flagValues["switch"]        = "kInstrCanSwitch";
    flagValues["throw"]         = "kInstrCanThrow";
    flagValues["return"]        = "kInstrCanReturn";
    flagValues["invoke"]        = "kInstrInvoke";
    flagValues["optimized"]     = "0"; # Not represented in C output
    flagValues["0"]             = "0";
}

# Translate the given flags into the equivalent C expression. Returns
# "" on error.
function flagsToC(f, parts, result, i) {
    # locals: parts, result, i
    count = split(f, parts, /\|/); # Split input at pipe characters.
    result = "0";

    for (i = 1; i <= count; i++) {
        f = flagValues[parts[i]];
        if (f == "") {
            printf("bogus flag: %s\n", f) >"/dev/stderr";
            return ""; # Bogus flag name.
        } else if (f == "0") {
            # Nothing to append for this case.
        } else if (result == "0") {
            result = f;
        } else {
            result = result "|" f;
        }
    }

    return result;
}

# Returns true if the given opcode (by index) is an "optimized" opcode.
function isOptimized(idx, parts, f) {
    # locals: parts, f
    split(flags[idx], parts, /\|/); # Split flags[idx] at pipes.
    for (f in parts) {
        if (parts[f] == "optimized") return 1;
    }
    return 0;
}

# Returns true if there is no definition for the given opcode (by index).
function isUnused(idx) {
    return (name[idx] == "");
}