summaryrefslogtreecommitdiff
path: root/src/mesa/glapi/glX_XML.py
diff options
context:
space:
mode:
authorIan Romanick <idr@us.ibm.com>2004-12-03 20:31:59 +0000
committerIan Romanick <idr@us.ibm.com>2004-12-03 20:31:59 +0000
commit74764061facb8f518c3ac7ebdb844bcd27a36ee0 (patch)
treea3cabfedc78fb25babd788c307c4970c35f853ff /src/mesa/glapi/glX_XML.py
parent47719fda0c2fba13c81e84e33523d5489263182e (diff)
Move common GLX code to glX_XML.py. This will make adding glX_proto_recv.py
easier later on.
Diffstat (limited to 'src/mesa/glapi/glX_XML.py')
-rw-r--r--src/mesa/glapi/glX_XML.py466
1 files changed, 466 insertions, 0 deletions
diff --git a/src/mesa/glapi/glX_XML.py b/src/mesa/glapi/glX_XML.py
new file mode 100644
index 0000000000..44f945a9ee
--- /dev/null
+++ b/src/mesa/glapi/glX_XML.py
@@ -0,0 +1,466 @@
+#!/usr/bin/python2
+
+# (C) Copyright IBM Corporation 2004
+# All Rights Reserved.
+#
+# 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
+# on the rights to use, copy, modify, merge, publish, distribute, sub
+# license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+# IBM AND/OR ITS SUPPLIERS 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.
+#
+# Authors:
+# Ian Romanick <idr@us.ibm.com>
+
+from xml.sax import saxutils
+from xml.sax import make_parser
+from xml.sax.handler import feature_namespaces
+
+import gl_XML
+import license
+import sys, getopt
+
+
+def printPure():
+ print """# if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ >= 96)
+# define PURE __attribute__((pure))
+# else
+# define PURE
+# endif"""
+
+def printFastcall():
+ print """# if defined(__i386__) && defined(__GNUC__)
+# define FASTCALL __attribute__((fastcall))
+# else
+# define FASTCALL
+# endif"""
+
+def printVisibility(S, s):
+ print """# if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 3)
+# define %s __attribute__((visibility("%s")))
+# else
+# define %s
+# endif""" % (S, s, S)
+
+def printNoinline():
+ print """# if defined(__GNUC__)
+# define NOINLINE __attribute__((noinline))
+# else
+# define NOINLINE
+# endif"""
+
+
+class glXItemFactory(gl_XML.glItemFactory):
+ """Factory to create GLX protocol oriented objects derived from glItem."""
+
+ def create(self, context, name, attrs):
+ if name == "function":
+ return glXFunction(context, name, attrs)
+ elif name == "enum":
+ return glXEnum(context, name, attrs)
+ elif name == "param":
+ return glXParameter(context, name, attrs)
+ else:
+ return gl_XML.glItemFactory.create(self, context, name, attrs)
+
+class glXEnumFunction:
+ def __init__(self, name):
+ self.name = name
+
+ # "enums" is a set of lists. The element in the set is the
+ # value of the enum. The list is the list of names for that
+ # value. For example, [0x8126] = {"POINT_SIZE_MIN",
+ # "POINT_SIZE_MIN_ARB", "POINT_SIZE_MIN_EXT",
+ # "POINT_SIZE_MIN_SGIS"}.
+
+ self.enums = {}
+
+ # "count" is indexed by count values. Each element of count
+ # is a list of index to "enums" that have that number of
+ # associated data elements. For example, [4] =
+ # {GL_AMBIENT, GL_DIFFUSE, GL_SPECULAR, GL_EMISSION,
+ # GL_AMBIENT_AND_DIFFUSE} (the enum names are used here,
+ # but the actual hexadecimal values would be in the array).
+
+ self.count = {}
+
+
+ def append(self, count, value, name):
+ if self.enums.has_key( value ):
+ self.enums[value].append(name)
+ else:
+ if not self.count.has_key(count):
+ self.count[count] = []
+
+ self.enums[value] = []
+ self.enums[value].append(name)
+ self.count[count].append(value)
+
+
+ def signature( self ):
+ sig = ""
+ for i in self.count:
+ for e in self.count[i]:
+ sig += "%04x,%u," % (e, i)
+
+ return sig;
+
+
+ def PrintUsingTable(self):
+ """Emit the body of the __gl*_size function using a pair
+ of look-up tables and a mask. The mask is calculated such
+ that (e & mask) is unique for all the valid values of e for
+ this function. The result of (e & mask) is used as an index
+ into the first look-up table. If it matches e, then the
+ same entry of the second table is returned. Otherwise zero
+ is returned.
+
+ It seems like this should cause better code to be generated.
+ However, on x86 at least, the resulting .o file is about 20%
+ larger then the switch-statment version. I am leaving this
+ code in because the results may be different on other
+ platforms (e.g., PowerPC or x86-64)."""
+
+ return 0
+ count = 0
+ for a in self.enums:
+ count += 1
+
+ # Determine if there is some mask M, such that M = (2^N) - 1,
+ # that will generate unique values for all of the enums.
+
+ mask = 0
+ for i in [1, 2, 3, 4, 5, 6, 7, 8]:
+ mask = (1 << i) - 1
+
+ fail = 0;
+ for a in self.enums:
+ for b in self.enums:
+ if a != b:
+ if (a & mask) == (b & mask):
+ fail = 1;
+
+ if not fail:
+ break;
+ else:
+ mask = 0
+
+ if (mask != 0) and (mask < (2 * count)):
+ masked_enums = {}
+ masked_count = {}
+
+ for i in range(0, mask + 1):
+ masked_enums[i] = "0";
+ masked_count[i] = 0;
+
+ for c in self.count:
+ for e in self.count[c]:
+ i = e & mask
+ masked_enums[i] = '0x%04x /* %s */' % (e, self.enums[e][0])
+ masked_count[i] = c
+
+
+ print ' static const GLushort a[%u] = {' % (mask + 1)
+ for e in masked_enums:
+ print ' %s, ' % (masked_enums[e])
+ print ' };'
+
+ print ' static const GLubyte b[%u] = {' % (mask + 1)
+ for c in masked_count:
+ print ' %u, ' % (masked_count[c])
+ print ' };'
+
+ print ' const unsigned idx = (e & 0x%02xU);' % (mask)
+ print ''
+ print ' return (e == a[idx]) ? (GLint) b[idx] : 0;'
+ return 1;
+ else:
+ return 0;
+
+ def PrintUsingSwitch(self):
+ """Emit the body of the __gl*_size function using a
+ switch-statement."""
+
+ print ' switch( e ) {'
+
+ for c in self.count:
+ for e in self.count[c]:
+ first = 1
+
+ # There may be multiple enums with the same
+ # value. This happens has extensions are
+ # promoted from vendor-specific or EXT to
+ # ARB and to the core. Emit the first one as
+ # a case label, and emit the others as
+ # commented-out case labels.
+
+ for j in self.enums[e]:
+ if first:
+ print ' case %s:' % (j)
+ first = 0
+ else:
+ print '/* case %s:*/' % (j)
+
+ print ' return %u;' % (c)
+
+ print ' default: return 0;'
+ print ' }'
+
+
+ def Print(self, name):
+ print 'INTERNAL PURE FASTCALL GLint'
+ print '__gl%s_size( GLenum e )' % (name)
+ print '{'
+
+ if not self.PrintUsingTable():
+ self.PrintUsingSwitch()
+
+ print '}'
+ print ''
+
+
+
+class glXEnum(gl_XML.glEnum):
+ def __init__(self, context, name, attrs):
+ gl_XML.glEnum.__init__(self, context, name, attrs)
+ self.glx_functions = []
+
+ def startElement(self, name, attrs):
+ if name == "size":
+ n = attrs.get('name', None)
+ if not self.context.glx_enum_functions.has_key( n ):
+ f = glXEnumFunction( n )
+ self.context.glx_enum_functions[ f.name ] = f
+
+ temp = attrs.get('count', None)
+ try:
+ c = int(temp)
+ except Exception,e:
+ raise RuntimeError('Invalid count value "%s" for enum "%s" in function "%s" when an integer was expected.' % (temp, self.name, n))
+
+ self.context.glx_enum_functions[ n ].append( c, self.value, self.name )
+ else:
+ gl_XML.glEnum.startElement(self, context, name, attrs)
+ return
+
+
+class glXParameter(gl_XML.glParameter):
+ def __init__(self, context, name, attrs):
+ self.order = 1;
+ gl_XML.glParameter.__init__(self, context, name, attrs);
+
+
+class glXFunction(gl_XML.glFunction):
+ glx_rop = 0
+ glx_sop = 0
+ glx_vendorpriv = 0
+
+ # If this is set to true, it means that GLdouble parameters should be
+ # written to the GLX protocol packet in the order they appear in the
+ # prototype. This is different from the "classic" ordering. In the
+ # classic ordering GLdoubles are written to the protocol packet first,
+ # followed by non-doubles. NV_vertex_program was the first extension
+ # to break with this tradition.
+
+ glx_doubles_in_order = 0
+
+ vectorequiv = None
+ handcode = 0
+ ignore = 0
+ can_be_large = 0
+
+ def __init__(self, context, name, attrs):
+ self.vectorequiv = attrs.get('vectorequiv', None)
+ self.count_parameters = None
+ self.counter = None
+ self.output = None
+ self.can_be_large = 0
+ self.reply_always_array = 0
+
+ gl_XML.glFunction.__init__(self, context, name, attrs)
+ return
+
+ def startElement(self, name, attrs):
+ """Process elements within a function that are specific to GLX."""
+
+ if name == "glx":
+ self.glx_rop = int(attrs.get('rop', "0"))
+ self.glx_sop = int(attrs.get('sop', "0"))
+ self.glx_vendorpriv = int(attrs.get('vendorpriv', "0"))
+
+ if attrs.get('handcode', "false") == "true":
+ self.handcode = 1
+ else:
+ self.handcode = 0
+
+ if attrs.get('ignore', "false") == "true":
+ self.ignore = 1
+ else:
+ self.ignore = 0
+
+ if attrs.get('large', "false") == "true":
+ self.can_be_large = 1
+ else:
+ self.can_be_large = 0
+
+ if attrs.get('doubles_in_order', "false") == "true":
+ self.glx_doubles_in_order = 1
+ else:
+ self.glx_doubles_in_order = 0
+
+ if attrs.get('always_array', "false") == "true":
+ self.reply_always_array = 1
+ else:
+ self.reply_always_array = 0
+
+ else:
+ gl_XML.glFunction.startElement(self, name, attrs)
+
+
+ def append(self, tag_name, p):
+ gl_XML.glFunction.append(self, tag_name, p)
+
+ if p.is_variable_length_array():
+ p.order = 2;
+ elif not self.glx_doubles_in_order and p.p_type.size == 8:
+ p.order = 0;
+
+ if p.p_count_parameters != None:
+ self.count_parameters = p.p_count_parameters
+
+ if p.is_counter:
+ self.counter = p.name
+
+ if p.is_output:
+ self.output = p
+
+ return
+
+ def variable_length_parameter(self):
+ for param in self.fn_parameters:
+ if param.is_variable_length_array():
+ return param
+
+ return None
+
+
+ def command_payload_length(self):
+ size = 0
+ size_string = ""
+ for p in self:
+ if p.is_output: continue
+ temp = p.size_string()
+ try:
+ s = int(temp)
+ size += s
+ except Exception,e:
+ size_string = size_string + " + __GLX_PAD(%s)" % (temp)
+
+ return [size, size_string]
+
+ def command_length(self):
+ [size, size_string] = self.command_payload_length()
+
+ if self.glx_rop != 0:
+ size += 4
+
+ size = ((size + 3) & ~3)
+ return "%u%s" % (size, size_string)
+
+
+ def opcode_real_value(self):
+ if self.glx_vendorpriv != 0:
+ if self.needs_reply():
+ return 17
+ else:
+ return 16
+ else:
+ return self.opcode_value()
+
+ def opcode_value(self):
+ if self.glx_rop != 0:
+ return self.glx_rop
+ elif self.glx_sop != 0:
+ return self.glx_sop
+ elif self.glx_vendorpriv != 0:
+ return self.glx_vendorpriv
+ else:
+ return -1
+
+ def opcode_rop_basename(self):
+ if self.vectorequiv == None:
+ return self.name
+ else:
+ return self.vectorequiv
+
+ def opcode_name(self):
+ if self.glx_rop != 0:
+ return "X_GLrop_%s" % (self.opcode_rop_basename())
+ elif self.glx_sop != 0:
+ return "X_GLsop_%s" % (self.name)
+ elif self.glx_vendorpriv != 0:
+ return "X_GLvop_%s" % (self.name)
+ else:
+ return "ERROR"
+
+ def opcode_real_name(self):
+ if self.glx_vendorpriv != 0:
+ if self.needs_reply():
+ return "X_GLXVendorPrivateWithReply"
+ else:
+ return "X_GLXVendorPrivate"
+ else:
+ return self.opcode_name()
+
+
+ def return_string(self):
+ if self.fn_return_type != 'void':
+ return "return retval;"
+ else:
+ return "return;"
+
+
+ def needs_reply(self):
+ return self.fn_return_type != 'void' or self.output != None
+
+
+class GlxProto(gl_XML.FilterGLAPISpecBase):
+ name = "glX_proto_send.py (from Mesa)"
+
+ def __init__(self):
+ gl_XML.FilterGLAPISpecBase.__init__(self)
+ self.factory = glXItemFactory()
+ self.glx_enum_functions = {}
+
+
+ def endElement(self, name):
+ if name == 'OpenGLAPI':
+ # Once all the parsing is done, we have to go back and
+ # fix-up some cross references between different
+ # functions.
+
+ for k in self.functions:
+ f = self.functions[k]
+ if f.vectorequiv != None:
+ equiv = self.find_function(f.vectorequiv)
+ if equiv != None:
+ f.glx_doubles_in_order = equiv.glx_doubles_in_order
+ f.glx_rop = equiv.glx_rop
+ else:
+ raise RuntimeError("Could not find the vector equiv. function %s for %s!" % (f.name, f.vectorequiv))
+ else:
+ gl_XML.FilterGLAPISpecBase.endElement(self, name)
+ return