asahi: Add a GenXML fork
authorAlyssa Rosenzweig <alyssa@rosenzweig.io>
Sat, 24 Apr 2021 23:10:10 +0000 (19:10 -0400)
committerAlyssa Rosenzweig <none>
Sun, 2 May 2021 21:41:21 +0000 (17:41 -0400)
Via Panfrost via v3d via Intel. Sour dough!

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10582>

src/asahi/lib/gen_pack.py [new file with mode: 0644]
src/asahi/lib/meson.build

diff --git a/src/asahi/lib/gen_pack.py b/src/asahi/lib/gen_pack.py
new file mode 100644 (file)
index 0000000..0cca73f
--- /dev/null
@@ -0,0 +1,691 @@
+#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 xml.parsers.expat
+import sys
+import operator
+from functools import reduce
+
+global_prefix = "agx"
+
+pack_header = """
+/* Generated code, see midgard.xml and gen_pack_header.py
+ *
+ * Packets, enums and structures for Panfrost.
+ *
+ * This file has been generated, do not hand edit.
+ */
+
+#ifndef AGX_PACK_H
+#define AGX_PACK_H
+
+#include <stdio.h>
+#include <stdint.h>
+#include <stdbool.h>
+#include <assert.h>
+#include <math.h>
+#include <inttypes.h>
+#include "util/macros.h"
+#include "util/u_math.h"
+
+#define __gen_unpack_float(x, y, z) uif(__gen_unpack_uint(x, y, z))
+
+static inline uint64_t
+__gen_uint(uint64_t v, uint32_t start, uint32_t end)
+{
+#ifndef NDEBUG
+   const int width = end - start + 1;
+   if (width < 64) {
+      const uint64_t max = (1ull << width) - 1;
+      assert(v <= max);
+   }
+#endif
+
+   return v << start;
+}
+
+static inline uint32_t
+__gen_sint(int32_t v, uint32_t start, uint32_t end)
+{
+#ifndef NDEBUG
+   const int width = end - start + 1;
+   if (width < 64) {
+      const int64_t max = (1ll << (width - 1)) - 1;
+      const int64_t min = -(1ll << (width - 1));
+      assert(min <= v && v <= max);
+   }
+#endif
+
+   return (((uint32_t) v) << start) & ((2ll << end) - 1);
+}
+
+static inline uint64_t
+__gen_unpack_uint(const uint8_t *restrict cl, uint32_t start, uint32_t end)
+{
+   uint64_t val = 0;
+   const int width = end - start + 1;
+   const uint64_t mask = (width == 64 ? ~0 : (1ull << width) - 1 );
+
+   for (unsigned byte = start / 8; byte <= end / 8; byte++) {
+      val |= ((uint64_t) cl[byte]) << ((byte - start / 8) * 8);
+   }
+
+   return (val >> (start % 8)) & mask;
+}
+
+static inline uint64_t
+__gen_unpack_sint(const uint8_t *restrict cl, uint32_t start, uint32_t end)
+{
+   int size = end - start + 1;
+   int64_t val = __gen_unpack_uint(cl, start, end);
+
+   /* Get the sign bit extended. */
+   return (val << (64 - size)) >> (64 - size);
+}
+
+#define agx_prepare(dst, T)                                 \\
+   *(dst) = (struct AGX_ ## T){ AGX_ ## T ## _header }
+
+#define agx_pack(dst, T, name)                              \\
+   for (struct AGX_ ## T name = { AGX_ ## T ## _header }, \\
+        *_loop_terminate = (void *) (dst);                  \\
+        __builtin_expect(_loop_terminate != NULL, 1);       \\
+        ({ AGX_ ## T ## _pack((uint32_t *) (dst), &name);  \\
+           _loop_terminate = NULL; }))
+
+#define agx_unpack(fp, src, T, name)                        \\
+        struct AGX_ ## T name;                         \\
+        AGX_ ## T ## _unpack(fp, (uint8_t *)(src), &name)
+
+#define agx_print(fp, T, var, indent)                   \\
+        AGX_ ## T ## _print(fp, &(var), indent)
+
+"""
+
+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
+
+def prefixed_upper_name(prefix, name):
+    if prefix:
+        name = prefix + "_" + name
+    return safe_name(name).upper()
+
+def enum_name(name):
+    return "{}_{}".format(global_prefix, safe_name(name)).lower()
+
+def num_from_str(num_str):
+    if num_str.lower().startswith('0x'):
+        return int(num_str, base=16)
+    else:
+        assert(not num_str.startswith('0') and 'octals numbers not allowed')
+        return int(num_str)
+
+MODIFIERS = ["shr", "minus", "align", "log2"]
+
+def parse_modifier(modifier):
+    if modifier is None:
+        return None
+
+    for mod in MODIFIERS:
+        if modifier[0:len(mod)] == mod:
+            if mod == "log2":
+                assert(len(mod) == len(modifier))
+                return [mod]
+
+            if modifier[len(mod)] == '(' and modifier[-1] == ')':
+                ret = [mod, int(modifier[(len(mod) + 1):-1])]
+                if ret[0] == 'align':
+                    align = ret[1]
+                    # Make sure the alignment is a power of 2
+                    assert(align > 0 and not(align & (align - 1)));
+
+                return ret
+
+    print("Invalid modifier")
+    assert(False)
+
+class Field(object):
+    def __init__(self, parser, attrs):
+        self.parser = parser
+        if "name" in attrs:
+            self.name = safe_name(attrs["name"]).lower()
+            self.human_name = attrs["name"]
+
+        if ":" in str(attrs["start"]):
+            (word, bit) = attrs["start"].split(":")
+            self.start = (int(word) * 32) + int(bit)
+        else:
+            self.start = int(attrs["start"])
+
+        self.end = self.start + int(attrs["size"]) - 1
+        self.type = attrs["type"]
+
+        if self.type == 'bool' and self.start != self.end:
+            print("#error Field {} has bool type but more than one bit of size".format(self.name));
+
+        if "prefix" in attrs:
+            self.prefix = safe_name(attrs["prefix"]).upper()
+        else:
+            self.prefix = None
+
+        if "exact" in attrs:
+            self.exact = int(attrs["exact"])
+        else:
+            self.exact = None
+
+        self.default = attrs.get("default")
+
+        # Map enum values
+        if self.type in self.parser.enums and self.default is not None:
+            self.default = safe_name('{}_{}_{}'.format(global_prefix, self.type, self.default)).upper()
+
+        self.modifier  = parse_modifier(attrs.get("modifier"))
+
+    def emit_template_struct(self, dim):
+        if self.type == 'address':
+            type = 'uint64_t'
+        elif self.type == 'bool':
+            type = 'bool'
+        elif self.type == 'float':
+            type = 'float'
+        elif self.type in ['uint', 'hex'] and self.end - self.start > 32:
+            type = 'uint64_t'
+        elif self.type == 'int':
+            type = 'int32_t'
+        elif self.type in ['uint', 'uint/float', 'hex']:
+            type = 'uint32_t'
+        elif self.type in self.parser.structs:
+            type = 'struct ' + self.parser.gen_prefix(safe_name(self.type.upper()))
+        elif self.type in self.parser.enums:
+            type = 'enum ' + enum_name(self.type)
+        else:
+            print("#error unhandled type: %s" % self.type)
+            type = "uint32_t"
+
+        print("   %-36s %s%s;" % (type, self.name, dim))
+
+        for value in self.values:
+            name = prefixed_upper_name(self.prefix, value.name)
+            print("#define %-40s %d" % (name, value.value))
+
+    def overlaps(self, field):
+        return self != field and max(self.start, field.start) <= min(self.end, field.end)
+
+class Group(object):
+    def __init__(self, parser, parent, start, count, label):
+        self.parser = parser
+        self.parent = parent
+        self.start = start
+        self.count = count
+        self.label = label
+        self.size = 0
+        self.length = 0
+        self.fields = []
+
+    def get_length(self):
+        # Determine number of bytes in this group.
+        calculated = max(field.end // 8 for field in self.fields) + 1 if len(self.fields) > 0 else 0
+        if self.length > 0:
+            assert(self.length >= calculated)
+        else:
+            self.length = calculated
+        return self.length
+
+
+    def emit_template_struct(self, dim):
+        if self.count == 0:
+            print("   /* variable length fields follow */")
+        else:
+            if self.count > 1:
+                dim = "%s[%d]" % (dim, self.count)
+
+            if len(self.fields) == 0:
+                print("   int dummy;")
+
+            for field in self.fields:
+                if field.exact is not None:
+                    continue
+
+                field.emit_template_struct(dim)
+
+    class Word:
+        def __init__(self):
+            self.size = 32
+            self.contributors = []
+
+    class FieldRef:
+        def __init__(self, field, path, start, end):
+            self.field = field
+            self.path = path
+            self.start = start
+            self.end = end
+
+    def collect_fields(self, fields, offset, path, all_fields):
+        for field in fields:
+            field_path = '{}{}'.format(path, field.name)
+            field_offset = offset + field.start
+
+            if field.type in self.parser.structs:
+                sub_struct = self.parser.structs[field.type]
+                self.collect_fields(sub_struct.fields, field_offset, field_path + '.', all_fields)
+                continue
+
+            start = field_offset
+            end = offset + field.end
+            all_fields.append(self.FieldRef(field, field_path, start, end))
+
+    def collect_words(self, fields, offset, path, words):
+        for field in fields:
+            field_path = '{}{}'.format(path, field.name)
+            start = offset + field.start
+
+            if field.type in self.parser.structs:
+                sub_fields = self.parser.structs[field.type].fields
+                self.collect_words(sub_fields, start, field_path + '.', words)
+                continue
+
+            end = offset + field.end
+            contributor = self.FieldRef(field, field_path, start, end)
+            first_word = contributor.start // 32
+            last_word = contributor.end // 32
+            for b in range(first_word, last_word + 1):
+                if not b in words:
+                    words[b] = self.Word()
+                words[b].contributors.append(contributor)
+
+    def emit_pack_function(self):
+        self.get_length()
+
+        words = {}
+        self.collect_words(self.fields, 0, '', words)
+
+        # Validate the modifier is lossless
+        for field in self.fields:
+            if field.modifier is None:
+                continue
+
+            assert(field.exact is None)
+
+            if field.modifier[0] == "shr":
+                shift = field.modifier[1]
+                mask = hex((1 << shift) - 1)
+                print("   assert((values->{} & {}) == 0);".format(field.name, mask))
+            elif field.modifier[0] == "minus":
+                print("   assert(values->{} >= {});".format(field.name, field.modifier[1]))
+            elif field.modifier[0] == "log2":
+                print("   assert(util_is_power_of_two_nonzero(values->{}));".format(field.name))
+
+        for index in range(self.length // 4):
+            # Handle MBZ words
+            if not index in words:
+                print("   cl[%2d] = 0;" % index)
+                continue
+
+            word = words[index]
+
+            word_start = index * 32
+
+            v = None
+            prefix = "   cl[%2d] =" % index
+
+            for contributor in word.contributors:
+                field = contributor.field
+                name = field.name
+                start = contributor.start
+                end = contributor.end
+                contrib_word_start = (start // 32) * 32
+                start -= contrib_word_start
+                end -= contrib_word_start
+
+                value = str(field.exact) if field.exact is not None else "values->{}".format(contributor.path)
+                if field.modifier is not None:
+                    if field.modifier[0] == "shr":
+                        value = "{} >> {}".format(value, field.modifier[1])
+                    elif field.modifier[0] == "minus":
+                        value = "{} - {}".format(value, field.modifier[1])
+                    elif field.modifier[0] == "align":
+                        value = "ALIGN_POT({}, {})".format(value, field.modifier[1])
+                    elif field.modifier[0] == "log2":
+                        value = "util_logbase2({})".format(value)
+
+                if field.type in ["uint", "hex", "address"]:
+                    s = "__gen_uint(%s, %d, %d)" % \
+                        (value, start, end)
+                elif field.type in self.parser.enums:
+                    s = "__gen_uint(%s, %d, %d)" % \
+                        (value, start, end)
+                elif field.type == "int":
+                    s = "__gen_sint(%s, %d, %d)" % \
+                        (value, start, end)
+                elif field.type == "bool":
+                    s = "__gen_uint(%s, %d, %d)" % \
+                        (value, start, end)
+                elif field.type == "float":
+                    assert(start == 0 and end == 31)
+                    s = "__gen_uint(fui({}), 0, 32)".format(value)
+                else:
+                    s = "#error unhandled field {}, type {}".format(contributor.path, field.type)
+
+                if not s == None:
+                    shift = word_start - contrib_word_start
+                    if shift:
+                        s = "%s >> %d" % (s, shift)
+
+                    if contributor == word.contributors[-1]:
+                        print("%s %s;" % (prefix, s))
+                    else:
+                        print("%s %s |" % (prefix, s))
+                    prefix = "           "
+
+            continue
+
+    # Given a field (start, end) contained in word `index`, generate the 32-bit
+    # mask of present bits relative to the word
+    def mask_for_word(self, index, start, end):
+        field_word_start = index * 32
+        start -= field_word_start
+        end -= field_word_start
+        # Cap multiword at one word
+        start = max(start, 0)
+        end = min(end, 32 - 1)
+        count = (end - start + 1)
+        return (((1 << count) - 1) << start)
+
+    def emit_unpack_function(self):
+        # First, verify there is no garbage in unused bits
+        words = {}
+        self.collect_words(self.fields, 0, '', words)
+
+        for index in range(self.length // 4):
+            base = index * 32
+            word = words.get(index, self.Word())
+            masks = [self.mask_for_word(index, c.start, c.end) for c in word.contributors]
+            mask = reduce(lambda x,y: x | y, masks, 0)
+
+            ALL_ONES = 0xffffffff
+
+            if mask != ALL_ONES:
+                TMPL = '   if (((const uint32_t *) cl)[{}] & {}) fprintf(fp, "XXX: Unknown field of {} unpacked at word {}: got %X, bad mask %X\\n", ((const uint32_t *) cl)[{}], ((const uint32_t *) cl)[{}] & {});'
+                print(TMPL.format(index, hex(mask ^ ALL_ONES), self.label, index, index, index, hex(mask ^ ALL_ONES)))
+
+        fieldrefs = []
+        self.collect_fields(self.fields, 0, '', fieldrefs)
+        for fieldref in fieldrefs:
+            field = fieldref.field
+            convert = None
+
+            args = []
+            args.append('cl')
+            args.append(str(fieldref.start))
+            args.append(str(fieldref.end))
+
+            if field.type in set(["uint", "uint/float", "address", "hex"]) | self.parser.enums:
+                convert = "__gen_unpack_uint"
+            elif field.type == "int":
+                convert = "__gen_unpack_sint"
+            elif field.type == "bool":
+                convert = "__gen_unpack_uint"
+            elif field.type == "float":
+                convert = "__gen_unpack_float"
+            else:
+                s = "/* unhandled field %s, type %s */\n" % (field.name, field.type)
+
+            suffix = ""
+            prefix = ""
+            if field.modifier:
+                if field.modifier[0] == "minus":
+                    suffix = " + {}".format(field.modifier[1])
+                elif field.modifier[0] == "shr":
+                    suffix = " << {}".format(field.modifier[1])
+                if field.modifier[0] == "log2":
+                    prefix = "1 << "
+
+            decoded = '{}{}({}){}'.format(prefix, convert, ', '.join(args), suffix)
+
+            print('   values->{} = {};'.format(fieldref.path, decoded))
+            if field.modifier and field.modifier[0] == "align":
+                mask = hex(field.modifier[1] - 1)
+                print('   assert(!(values->{} & {}));'.format(fieldref.path, mask))
+
+    def emit_print_function(self):
+        for field in self.fields:
+            convert = None
+            name, val = field.human_name, 'values->{}'.format(field.name)
+
+            if field.type in self.parser.structs:
+                pack_name = self.parser.gen_prefix(safe_name(field.type)).upper()
+                print('   fprintf(fp, "%*s{}:\\n", indent, "");'.format(field.human_name))
+                print("   {}_print(fp, &values->{}, indent + 2);".format(pack_name, field.name))
+            elif field.type == "address":
+                # TODO resolve to name
+                print('   fprintf(fp, "%*s{}: 0x%" PRIx64 "\\n", indent, "", {});'.format(name, val))
+            elif field.type in self.parser.enums:
+                print('   fprintf(fp, "%*s{}: %s\\n", indent, "", {}_as_str({}));'.format(name, enum_name(field.type), val))
+            elif field.type == "int":
+                print('   fprintf(fp, "%*s{}: %d\\n", indent, "", {});'.format(name, val))
+            elif field.type == "bool":
+                print('   fprintf(fp, "%*s{}: %s\\n", indent, "", {} ? "true" : "false");'.format(name, val))
+            elif field.type == "float":
+                print('   fprintf(fp, "%*s{}: %f\\n", indent, "", {});'.format(name, val))
+            elif field.type in ["uint", "hex"] and (field.end - field.start) >= 32:
+                print('   fprintf(fp, "%*s{}: 0x%" PRIx64 "\\n", indent, "", {});'.format(name, val))
+            elif field.type == "hex":
+                print('   fprintf(fp, "%*s{}: 0x%" PRIx32 "\\n", indent, "", {});'.format(name, val))
+            elif field.type == "uint/float":
+                print('   fprintf(fp, "%*s{}: 0x%X (%f)\\n", indent, "", {}, uif({}));'.format(name, val, val))
+            else:
+                print('   fprintf(fp, "%*s{}: %u\\n", indent, "", {});'.format(name, val))
+
+class Value(object):
+    def __init__(self, attrs):
+        self.name = attrs["name"]
+        self.value = int(attrs["value"], 0)
+
+class Parser(object):
+    def __init__(self):
+        self.parser = xml.parsers.expat.ParserCreate()
+        self.parser.StartElementHandler = self.start_element
+        self.parser.EndElementHandler = self.end_element
+
+        self.struct = None
+        self.structs = {}
+        # Set of enum names we've seen.
+        self.enums = set()
+
+    def gen_prefix(self, name):
+        return '{}_{}'.format(global_prefix.upper(), name)
+
+    def start_element(self, name, attrs):
+        if name == "agxml":
+            print(pack_header)
+        elif name == "struct":
+            name = attrs["name"]
+            self.no_direct_packing = attrs.get("no-direct-packing", False)
+            object_name = self.gen_prefix(safe_name(name.upper()))
+            self.struct = object_name
+
+            self.group = Group(self, None, 0, 1, name)
+            if "size" in attrs:
+                self.group.length = int(attrs["size"])
+            self.group.align = int(attrs["align"]) if "align" in attrs else None
+            self.structs[attrs["name"]] = self.group
+        elif name == "field":
+            self.group.fields.append(Field(self, attrs))
+            self.values = []
+        elif name == "enum":
+            self.values = []
+            self.enum = safe_name(attrs["name"])
+            self.enums.add(attrs["name"])
+            if "prefix" in attrs:
+                self.prefix = attrs["prefix"]
+            else:
+                self.prefix= None
+        elif name == "value":
+            self.values.append(Value(attrs))
+
+    def end_element(self, name):
+        if name == "struct":
+            self.emit_struct()
+            self.struct = None
+            self.group = None
+        elif name  == "field":
+            self.group.fields[-1].values = self.values
+        elif name  == "enum":
+            self.emit_enum()
+            self.enum = None
+        elif name == "agxml":
+            print('#endif')
+
+    def emit_header(self, name):
+        default_fields = []
+        for field in self.group.fields:
+            if not type(field) is Field:
+                continue
+            if field.default is not None:
+                default_fields.append("   .{} = {}".format(field.name, field.default))
+            elif field.type in self.structs:
+                default_fields.append("   .{} = {{ {}_header }}".format(field.name, self.gen_prefix(safe_name(field.type.upper()))))
+
+        print('#define %-40s\\' % (name + '_header'))
+        if default_fields:
+            print(",  \\\n".join(default_fields))
+        else:
+            print('   0')
+        print('')
+
+    def emit_template_struct(self, name, group):
+        print("struct %s {" % name)
+        group.emit_template_struct("")
+        print("};\n")
+
+    def emit_pack_function(self, name, group):
+        print("static inline void\n%s_pack(uint32_t * restrict cl,\n%sconst struct %s * restrict values)\n{" %
+              (name, ' ' * (len(name) + 6), name))
+
+        group.emit_pack_function()
+
+        print("}\n\n")
+
+        print('#define {} {}'.format (name + "_LENGTH", self.group.length))
+        if self.group.align != None:
+            print('#define {} {}'.format (name + "_ALIGN", self.group.align))
+        print('struct {}_packed {{ uint32_t opaque[{}]; }};'.format(name.lower(), self.group.length // 4))
+
+    def emit_unpack_function(self, name, group):
+        print("static inline void")
+        print("%s_unpack(FILE *fp, const uint8_t * restrict cl,\n%sstruct %s * restrict values)\n{" %
+              (name.upper(), ' ' * (len(name) + 8), name))
+
+        group.emit_unpack_function()
+
+        print("}\n")
+
+    def emit_print_function(self, name, group):
+        print("static inline void")
+        print("{}_print(FILE *fp, const struct {} * values, unsigned indent)\n{{".format(name.upper(), name))
+
+        group.emit_print_function()
+
+        print("}\n")
+
+    def emit_struct(self):
+        name = self.struct
+
+        self.emit_template_struct(self.struct, self.group)
+        self.emit_header(name)
+        if self.no_direct_packing == False:
+            self.emit_pack_function(self.struct, self.group)
+            self.emit_unpack_function(self.struct, self.group)
+        self.emit_print_function(self.struct, self.group)
+
+    def enum_prefix(self, name):
+        return 
+
+    def emit_enum(self):
+        e_name = enum_name(self.enum)
+        prefix = e_name if self.enum != 'Format' else global_prefix
+        print('enum {} {{'.format(e_name))
+
+        for value in self.values:
+            name = '{}_{}'.format(prefix, value.name)
+            name = safe_name(name).upper()
+            print('        % -36s = %6d,' % (name, value.value))
+        print('};\n')
+
+        print("static inline const char *")
+        print("{}_as_str(enum {} imm)\n{{".format(e_name.lower(), e_name))
+        print("    switch (imm) {")
+        for value in self.values:
+            name = '{}_{}'.format(prefix, value.name)
+            name = safe_name(name).upper()
+            print('    case {}: return "{}";'.format(name, value.name))
+        print('    default: return "XXX: INVALID";')
+        print("    }")
+        print("}\n")
+
+    def parse(self, filename):
+        file = open(filename, "rb")
+        self.parser.ParseFile(file)
+        file.close()
+
+if len(sys.argv) < 2:
+    print("No input xml file specified")
+    sys.exit(1)
+
+input_file = sys.argv[1]
+
+p = Parser()
+p.parse(input_file)
index a2677d7..7bde6af 100644 (file)
 # 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.
+
+agx_pack = custom_target(
+  'agx_pack.h',
+  input : ['gen_pack.py', 'cmdbuf.xml'],
+  output : 'agx_pack.h',
+  command : [prog_python, '@INPUT@'],
+  capture : true,
+)
+
+idep_agx_pack = declare_dependency(
+  sources : [agx_pack],
+  include_directories : include_directories('.'),
+)