added top element size computation information

doc/Makefile

@@ -49,7 +49,7 @@ occam.X:
 		cd occam; make "PIC="$(PIC) "TBL="$(TBL) "EQN="$(EQN) "TARGET="$(TARGET)
 ego.doc:	ego.X
 ego.X:
-		cd ego; make "REFER="$(REFER) "TARGET="$(TARGET)
+		cd ego; make "REFER="$(REFER) "TARGET="$(TARGET) "TBL="$(TBL)
 em.$(SUF):	em.X
 em.X:
 		cd em; make "TBL="$(TBL) "NROFF="$(NROFF) "SUF="$(SUF) "TARGET="$(TARGET)

doc/ansi_C.doc

@@ -13,7 +13,7 @@ Vrije Universiteit
 Amsterdam, The Netherlands
 .PP
 This document specifies the implementation-defined behaviour of the ANSI-C
-front end of the Amsterdam Compiler Kit as required by ANS X3.159-1989. Since
+front end of the Amsterdam Compiler Kit as required by ANS X3.159-1989.  Since
 the implementation-defined behaviour sometimes depends on the machine
 compiling on or for, some items will be left unspecified in this
 document\(dg.
@@ -24,7 +24,7 @@ compile-time behaviour
 The compiler assumes that it runs on a UNIX system.
 .NS A.6.3.1
 .IP -
-Diagnostics are placed on the standard error output. They have the
+Diagnostics are placed on the standard error output.  They have the
 following specification:
 .br
 "<file>", line <nr>: [(<class>)] <diagnostic>
@@ -33,21 +33,21 @@ There are three classes of diagnostics: "error", "strict" and "warning".
 When the class is "error", the <class> is absent.
 .br
 The class "strict" is used for violations of the standard which are
-not severe enough to stop compilation. An example is the the occurrence
+not severe enough to stop compilation.  An example is the the occurrence
 of non white-space after an '#else' or '#endif' pre-processing
-directive. The class "warning" is used for legal but dubious
+directive.  The class "warning" is used for legal but dubious
-constructions. An example is overflow of constant expressions.
+constructions.  An example is overflow of constant expressions.
 .NS A.6.3.2
 .IP -
-The function 'main' can have two arguments. The first argument is an
+The function 'main' can have two arguments.  The first argument is an
-integer specifying the number of arguments on the command line. The second
+integer specifying the number of arguments on the command line.  The second
 argument is a pointer to an array of pointers to the arguments (as
 strings).
 .IP -
 Interactive devices are terminals.
 .NS A.6.3.3
 .IP -
-The number of significant characters is an option. By default it is 64.
+The number of significant characters is an option.  By default it is 64.
 There is a distinction between upper and lower case.
 .NS A.6.3.4
 .IP -
@@ -70,13 +70,13 @@ A plain 'char' has the same range of values as 'signed char'.
 .NS A.6.3.5
 .IP -
 The compiler assumes that it works on and compiles for a
-2-complement binary-number system. Shorts will use 2 bytes and longs
+2-complement binary-number system.  Shorts will use 2 bytes and longs
-will use 4 bytes. The size of integers are machine dependent.
+will use 4 bytes.  The size of integers are machine dependent.
 .IP -
 Converting an integer to a shorter signed integer is implemented by
 ignoring the high-order byte(s) of the former.
 Converting a unsigned integer to a signed integer of the same type is
-only done in administration. This means that the bit-pattern remains
+only done in administration.  This means that the bit-pattern remains
 unchanged.
 .IP -
 The result of bitwise operations on signed integers are what can be
@@ -98,45 +98,45 @@ Truncation is always to the nearest floating-point number that can
 be represented.
 .NS A.6.3.7
 .IP -
-The type returned by the sizeof-operator is 'unsigned int'. This is done
+The type returned by the sizeof-operator (also known as size_t)
-for backward compatibility reasons.
+is 'unsigned int'.  This is done for backward compatibility reasons.
 .IP -
 Casting an integer to a pointer or vice versa has no effect in
-bit-pattern when the sizes are equal. Otherwise the value will be
+bit-pattern when the sizes are equal.  Otherwise the value will be
 truncated or zero-extended (depending on the direction of the
 conversion and the relative sizes).
 .IP -
 When a pointer is as large as an integer, the type of a 'ptrdiff_t' will
-be 'int'. Otherwise the type will be 'long'.
+be 'int'.  Otherwise the type will be 'long'.
 .NS A.6.3.8
 .IP -
 Since the front end has only limited control over the registers, it can
 only make it more likely that variables that are declared as
-registers also end up in registers. The only things that can possibly be
+registers also end up in registers.  The only things that can possibly be
 put into registers are : 'int', 'long', 'float', 'double', 'long double'
 and pointers.
 .NS A.6.3.9
 .IP -
 When a member of a union object is accessed using a member of a
-different type, the resulting value will usually be garbage. The
+different type, the resulting value will usually be garbage.  The
 compiler makes no effort to catch these errors.
 .IP -
-The alignment of types is a compile-time option. The alignment of
+The alignment of types is a compile-time option.  The alignment of
-a structure-member is the alignment of its type. Usually, the
+a structure-member is the alignment of its type.  Usually, the
-alignment is passed on to the compiler by the 'ack' program. When a
+alignment is passed on to the compiler by the 'ack' program.  When a
 user wants to do this manually, he/she should be prepared for trouble.
 .IP -
-A "plain" 'int' bit-field is taken as a 'signed int'. This means that
+A "plain" 'int' bit-field is taken as a 'signed int'.  This means that
 a field with a size 1 bit can only store the values 0 and -1.
 .IP -
-The order of allocation of bit-fields is a compile-time option. By
+The order of allocation of bit-fields is a compile-time option.  By
 default, high-order bits are allocated first.
 .IP -
 An enum has the same size as a "plain" 'int'.
 .NS A.6.3.10
 .IP -
 An access to a volatile declared variable is done by just mentioning
-the variable. E.g. the statement "x;" where x is declared volatile,
+the variable.  E.g. the statement "x;" where x is declared volatile,
 constitutes an access.
 .S A.6.3.11
 .IP -
@@ -153,10 +153,10 @@ Since both the pre-processor and the compiler assume ASCII-characters,
 a single character constant in a conditional-inclusion directive
 matches the same value in the execution character set.
 .IP -
-The pre-processor recognizes -I... command-line options. The
+The pre-processor recognizes -I... command-line options.  The
-directories thus specified are searched first. After that, depending on the
+directories thus specified are searched first.  After that, depending on the
 command that the preprocessor is called with, machine/system-dependant
-directories are searched. After that, ~em/include/_tail_ac and
+directories are searched.  After that, ~em/include/_tail_ac and
 /usr/include are visited.
 .IP -
 Quoted names are first looked for in the directory in which the file
@@ -171,7 +171,7 @@ Since the compiler runs on UNIX, __DATE__ and __TIME__ will always be
 defined.
 .NS A.6.3.14
 .IP -
-NULL is defined as ((void *)0). This in order to flag dubious
+NULL is defined as ((void *)0).  This in order to flag dubious
 constructions like "int x = NULL;".
 .IP -
 The diagnostic printed by 'assert' is as follows:
@@ -195,7 +195,7 @@ isprint()	<space>-~ (== \e040-\e176)
 .TE
 .KE
 As an addition, there is an isascii() macro, which tests whether a character
-is an ascii character. Characters in the range from \e000 to \e177 are ascii
+is an ascii character.  Characters in the range from \e000 to \e177 are ascii
 characters.
 .KS
 .IP -
@@ -221,9 +221,9 @@ The function fmod() returns 0.0 and sets errno to EDOM when the second
 argument is 0.0.
 .IP -
 The set of signals for the signal() function depends on the UNIX-system
-which the compiler is compiling for. The default handling, semantics
+which the compiler is compiling for.  The default handling, semantics
 and behaviour of these signals are those specified by the operating
-system vendor. The default handling is not reset when SIGILL is
+system vendor.  The default handling is not reset when SIGILL is
 received.
 .IP -
 A text-stream need not end in a new-line character.
@@ -249,11 +249,11 @@ the '/'.
 A file can be open multiple times.
 .IP -
 When a remove() is done on an open file, reading and writing behave
-just as can be expected from a non-removed file. When the associated
+just as can be expected from a non-removed file.  When the associated
 stream is closed, all written data will be lost.
 .IP -
 When a file exists prior to a call to rename(), the behaviour is that
-of the underlying UNIX system. Normally, the call would fail.
+of the underlying UNIX system.  Normally, the call would fail.
 .IP -
 The %p conversion in fprintf() has the same effect as %#x or %#lx,
 depending on the sizes of pointer and integer.
@@ -262,7 +262,7 @@ The %p conversion in fscanf() has the same effect as %x or %lx,
 depending on the sizes of pointer and integer.
 .IP -
 A - character that is neither the first nor the last character in the
-scanlist for %[ conversion is taken to be a range indicator. When the
+scanlist for %[ conversion is taken to be a range indicator.  When the
 first character has a higher ASCII-value than the second, the - will
 just be put into the scanlist.
 .IP -
@@ -284,7 +284,7 @@ The mapping of errors to strings is done by strerror().
 When the requested size is zero, malloc(), calloc() and realloc()
 return a null-pointer.
 .IP -
-When abort() is called, output buffers will be flushed. Temporary files
+When abort() is called, output buffers will be flushed.  Temporary files
 (made with the tmpfile() function) will have disappeared when SIGABRT
 is not caught or ignored.
 .IP -
@@ -300,7 +300,7 @@ putenv("TERM=a230");
 .br
 The argument to putenv() is stored in an internal table, so malloc'ed
 strings can not be freed until another call to putenv() (which sets the
-same environment variable) is made. The function returns 1 if it fails,
+same environment variable) is made.  The function returns 1 if it fails,
 0 otherwise.
 .LP
 .IP -
@@ -349,7 +349,7 @@ ERANGE	"Result too large"
 .TE
 everything else causes strerror() to return "unknown error"
 .IP -
-The local time zone is per default MET (GMT + 1:00:00). This can be
+The local time zone is per default MET (GMT + 1:00:00).  This can be
 changed through the TZ environment variable, or by some changes in the
 sources.
 .IP -

doc/ncg.doc

@@ -301,16 +301,17 @@ Here is a list of reserved words; all of these are unavailable as identifiers.
 .TS
 box;
 l l l l l.
-ADDR	STACK	from	proc	sfit
+ADDR	STACKINGRULES	gen	proc	test
-COERCIONS	STACKINGRULES	gen	reg_any	test
+COERCIONS	TESTS	highw	reg_any	to
-INSTRUCTIONS	TESTS	highw	reg_float	to
+INSTRUCTIONS	TIMEFACTOR	inreg	reg_float	topeltsize
-INT	TIMEFACTOR	inreg	reg_loop	ufit
+INT	TOKENS	is_rom	reg_loop	ufit
-MOVES	TOKENS	is_rom	reg_pointer	uses
+MOVES	call	kills	reg_pointer	uses
-PATTERNS	call	kills	regvar	with
+PATTERNS	cost	lab	regvar	with
-PROPERTIES	cost	leaving	return	yields
+PROPERTIES	defined	labeldef	return	yields
-REGISTERS	defined	loww	reusing
+REGISTERS	exact	leaving	reusing	
-SETS	exact	move	rom
+SETS	example	loww	rom	
-SIZEFACTOR	example	pat	samesign
+SIZEFACTOR	fallthrough	move	samesign	
+STACK	from	pat	sfit	
 .TE
 C style comments are accepted.
 .DS
@@ -928,6 +929,17 @@ If
 .I t
 is omitted reg_any is assumed.
 Undefined if inreg(\fIe\fP)<=0 .
+.LP
+The next two `functions' are only needed in a table that
+uses the top element size information.
+.IP topeltsize($a) 16
+Returns the size of the element on top of the EM-stack at the label
+identified by $a.  This can be used to put the top of the stack in a
+register at the moment of an unconditional jump.  At an unconditional jump,
+the size of the top-element will always look 0.
+.IP fallthrough($a)
+Returns 1 if the label identified by $a can be reached via fallthrough, 0
+otherwise.
 .NH 2
 Token descriptions
 .PP
@@ -1017,6 +1029,13 @@ This rule of three is an unfortunate implementation dependent restriction,
 but patterns longer than three EM instructions are luckily not needed
 too often.
 .PP
+The EM mnemonic may also be the pseudo-instruction \fBlab\fP, which matches
+a label.  Its argument can be used in testing on topeltsize and
+fallthrough.  When this pattern is specified, the label should be defined
+explicitly with a
+.I labeldef
+statement.
+.PP
 Following the EM-pattern there may be more than one code
 rule,
 .I cg
@@ -1308,6 +1327,16 @@ C-routine
 Explanation of this must wait for the description of the
 file mach.c below.
 .IP 5)
+The
+.I labeldef
+statement.  Its only argument should be that of the
+.I lab
+pseudo-instruction.  This is needed to generate local labels when the
+top element size information is used.  It takes the form
+.DS
+	labeldef $i
+.DE
+.IP 6)
 A temporary label of the form <digit>: may be placed here.
 Expressions of the form [0-9][bf] in this code rule
 generate the same string as is used for this label.
@@ -1552,6 +1581,35 @@ A simple jump.
 The stack pattern guarantees that everything will be stacked
 before the jump is taken.
 .DS
+pat lab	topeltsize($1)==2 && !fallthrough($1)
+gen labeldef $1					yields r0
+
+pat lab	topeltsize($1)==2 && fallthrough($1)
+with src2
+gen move %1,r0
+    labeldef $1					yields r0
+
+pat lab topeltsize($1)!=2
+with STACK
+kills all
+gen labeldef $1
+
+pat bra topeltsize($1)==2
+with src2 STACK
+    gen move %1,d0
+	jbr {label, $1}
+
+pat bra topeltsize($1)!=2
+with STACK
+    gen jbr {label, $1}
+.DE
+The combination of these patterns make sure that the top of the EM-stack will
+be in register r0 whenever necessary.  The top element size mechanism will
+also show a size of 0 whenever a conditional branch to a label
+occurs.  This saves a lot of patterns and hardly decreases performance.
+When the same register is used to return function results, this can save
+many moves to and from the stack.
+.DS
 .ta 7.5c
 pat cal
 with STACK
@@ -2738,6 +2796,11 @@ DO_RETURN
 Returns from this level of codegen().
 Is used at the end of coercions,
 move rules etc..
+.NH 4
+DO_LABDEF
+.PP
+This prints a label when the top element size mechanism is used.  Only done on
+toplevel.
 .NH 3
 compute.c
 .PP
@@ -2803,6 +2866,10 @@ This module maintains a list of global symbols that have a
 pseudo associated.
 There are functions to enter a symbol and to find a symbol.
 .NH 3
+label.c
+.PP
+This module contains routines to handle the top element size messages.
+.NH 3
 main.c
 .PP
 Main routine of the code generator.