Skip to content

Commit

Permalink
Refactor: remove micro-optimizations from ACS95 token reading function
Browse files Browse the repository at this point in the history
  • Loading branch information
@positively-charged
positively-charged committed Apr 23, 2017
1 parent 8082d76 commit 4388a16
Showing 1 changed file with 67 additions and 147 deletions.
214 changes: 67 additions & 147 deletions src/parse/token/source.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1125,8 +1125,7 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
int line = 0;
int column = 0;
enum tk tk = TK_END;
char* text = NULL;
int length = 0;
struct str* text = NULL;

whitespace:
// -----------------------------------------------------------------------
Expand Down Expand Up @@ -1427,62 +1426,28 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
{ "while", TK_WHILE },
{ "world", TK_WORLD },
};
text = temp_text( parse );
while ( isalnum( ch ) || ch == '_' ) {
append_ch( text, tolower( ch ) );
ch = read_ch( parse );
}
enum { MAX_IDENTIFIER_LENGTH = 31 };
struct text_buffer* text_buffer = t_get_text_buffer( parse->task,
MAX_IDENTIFIER_LENGTH + 1 );
text = text_buffer->left;
char* copied_text = text;
char* end = copied_text + MAX_IDENTIFIER_LENGTH;
char* source_text = parse->source->buffer +
parse->source->buffer_pos - 1;
while ( true ) {
if ( isalnum( *source_text ) || *source_text == '_' ) {
if ( copied_text == end ) {
struct pos pos;
t_init_pos( &pos,
parse->source->file_entry_id,
line, column );
p_diag( parse, DIAG_POS_ERR, &pos,
"identifier too long (maximum length is %d)",
MAX_IDENTIFIER_LENGTH );
p_bail( parse );
}
*copied_text = tolower( *source_text );
++copied_text;
++source_text;
}
// Read new data from the source file.
else if ( *source_text == '\n' && source_text[ 1 ] == '\0' ) {
size_t count = fread( parse->source->buffer,
sizeof( parse->source->buffer[ 0 ] ), SOURCE_BUFFER_SIZE,
parse->source->fh );
parse->source->buffer[ count ] = '\n';
parse->source->buffer[ count + 1 ] = '\0';
parse->source->buffer_pos = 0;
source_text = parse->source->buffer;
if ( count == 0 ) {
break;
}
}
else {
break;
}
if ( text->length > MAX_IDENTIFIER_LENGTH ) {
struct pos pos;
t_init_pos( &pos,
parse->source->file_entry_id,
line, column );
p_diag( parse, DIAG_POS_ERR, &pos,
"identifier too long (maximum length is %d)",
MAX_IDENTIFIER_LENGTH );
p_bail( parse );
}
*copied_text = '\0';
length = copied_text - text;
// Update source buffer. The 1 added to `buffer_pos` is for the character
// after the identifier, since we assume it is now read.
parse->source->buffer_pos = source_text - parse->source->buffer + 1;
parse->source->ch = *source_text;
parse->source->column += length;
// Update text buffer.
text_buffer->left = copied_text + 1;
// Reserved identifier. Uses binary search.
int left = 0;
int right = ARRAY_SIZE( table ) - 1;
while ( left <= right ) {
int middle = ( left + right ) / 2;
int result = strcmp( text, table[ middle ].name );
int result = strcmp( text->value, table[ middle ].name );
if ( result > 0 ) {
left = middle + 1;
}
Expand All @@ -1491,6 +1456,7 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
}
else {
tk = table[ middle ].tk;
text = NULL;
goto finish;
}
}
Expand All @@ -1508,9 +1474,9 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
goto hexadecimal;
}
else if ( ch == '.' ) {
str_clear( &parse->temp_text );
append_ch( &parse->temp_text, '0' );
append_ch( &parse->temp_text, '.' );
text = temp_text( parse );
append_ch( text, '0' );
append_ch( text, '.' );
ch = read_ch( parse );
goto fixedpoint;
}
Expand All @@ -1524,26 +1490,24 @@ void read_token_acs95( struct parse* parse, struct token* token ) {

hexadecimal:
// -----------------------------------------------------------------------
str_clear( &parse->temp_text );
text = temp_text( parse );
while ( true ) {
if ( isxdigit( ch ) ) {
append_ch( &parse->temp_text, ch );
append_ch( text, ch );
ch = read_ch( parse );
}
else {
if ( parse->temp_text.length == 0 ) {
if ( text->length == 0 ) {
struct pos pos;
t_init_pos( &pos,
parse->source->file_entry_id,
parse->source->line,
column );
p_diag( parse, DIAG_POS | DIAG_WARN, &pos,
"hexadecimal literal has no digits, will interpret it as 0x0" );
append_ch( &parse->temp_text, '0' );
append_ch( text, '0' );
}
tk = TK_LIT_HEX;
text = parse->temp_text.value;
length = parse->temp_text.length;
goto finish;
}
}
Expand All @@ -1557,48 +1521,46 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
goto decimal;
}
else if ( ch == '.' ) {
str_clear( &parse->temp_text );
append_ch( &parse->temp_text, '0' );
append_ch( &parse->temp_text, '.' );
text = temp_text( parse );
append_ch( text, '0' );
append_ch( text, '.' );
ch = read_ch( parse );
goto fixedpoint;
}
else if ( ch == '_' ) {
str_clear( &parse->temp_text );
append_ch( &parse->temp_text, '0' );
append_ch( &parse->temp_text, ch );
text = temp_text( parse );
append_ch( text, '0' );
append_ch( text, ch );
ch = read_ch( parse );
goto radix;
}
else {
text = "0";
length = 1;
text = temp_text( parse );
append_ch( text, '0' );
tk = TK_LIT_DECIMAL;
goto finish;
}

decimal:
// -----------------------------------------------------------------------
str_clear( &parse->temp_text );
text = temp_text( parse );
while ( true ) {
if ( isdigit( ch ) ) {
append_ch( &parse->temp_text, ch );
append_ch( text, ch );
ch = read_ch( parse );
}
else if ( ch == '.' ) {
append_ch( &parse->temp_text, ch );
append_ch( text, ch );
ch = read_ch( parse );
goto fixedpoint;
}
else if ( ch == '_' ) {
append_ch( &parse->temp_text, ch );
append_ch( text, ch );
ch = read_ch( parse );
goto radix;
}
else {
tk = TK_LIT_DECIMAL;
text = parse->temp_text.value;
length = parse->temp_text.length;
goto finish;
}
}
Expand All @@ -1607,24 +1569,22 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
// -----------------------------------------------------------------------
while ( true ) {
if ( isdigit( ch ) ) {
append_ch( &parse->temp_text, ch );
append_ch( text, ch );
ch = read_ch( parse );
}
else {
if ( parse->temp_text.value[ parse->temp_text.length - 1 ] == '.' ) {
if ( text->value[ text->length - 1 ] == '.' ) {
struct pos pos;
t_init_pos( &pos,
parse->source->file_entry_id,
parse->source->line,
column );
p_diag( parse, DIAG_POS | DIAG_WARN, &pos,
"fixed-point literal has no digits after point, will interpret "
"it as %s0", parse->temp_text.value );
append_ch( &parse->temp_text, '0' );
"it as %s0", text->value );
append_ch( text, '0' );
}
tk = TK_LIT_FIXED;
text = parse->temp_text.value;
length = parse->temp_text.length;
goto finish;
}
}
Expand All @@ -1633,87 +1593,47 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
// -----------------------------------------------------------------------
while ( true ) {
if ( isalnum( ch ) ) {
append_ch( &parse->temp_text, tolower( ch ) );
append_ch( text, tolower( ch ) );
ch = read_ch( parse );
}
else {
if ( parse->temp_text.value[ parse->temp_text.length - 1 ] == '_' ) {
if ( text->value[ text->length - 1 ] == '_' ) {
struct pos pos;
t_init_pos( &pos,
parse->source->file_entry_id,
parse->source->line,
column );
p_diag( parse, DIAG_POS | DIAG_WARN, &pos,
"radix literal has no digits after underscore, "
"will interpret it as %s0", parse->temp_text.value );
append_ch( &parse->temp_text, '0' );
"will interpret it as %s0", text->value );
append_ch( text, '0' );
}
text = parse->temp_text.value;
length = parse->temp_text.length;
tk = TK_LIT_RADIX;
goto finish;
}
}

string:
// -----------------------------------------------------------------------
{
// Most strings will be small, so copy the characters directly into the
// text buffer. For long strings, use an intermediate buffer.
enum { SEGMENTLENGTH = 255 };
enum { CUSHIONLENGTH = 1 };
enum { SAFELENGTH = SEGMENTLENGTH - CUSHIONLENGTH };
struct text_buffer* text_buffer = t_get_text_buffer( parse->task,
SEGMENTLENGTH + 1 );
text = text_buffer->left;
char* copied_text = text;
char* end = copied_text + SAFELENGTH;
struct str* temp_text = NULL;
while ( true ) {
if ( copied_text >= end ) {
if ( ! temp_text ) {
temp_text = &parse->temp_text;
str_clear( temp_text );
str_append_sub( temp_text, text, copied_text - text );
}
else {
temp_text->length = copied_text - temp_text->value;
str_grow( temp_text, temp_text->buffer_length * 2 );
}
copied_text = temp_text->value + temp_text->length;
end = temp_text->value + temp_text->buffer_length -
( CUSHIONLENGTH + 1 );
}
else if ( ! ch ) {
struct pos pos;
t_init_pos( &pos,
parse->source->file_entry_id,
line, column );
p_diag( parse, DIAG_POS_ERR, &pos,
"unterminated string" );
p_bail( parse );
}
else if ( ch == '"' ) {
read_ch( parse );
tk = TK_LIT_STRING;
*copied_text = '\0';
if ( temp_text ) {
temp_text->length = copied_text - temp_text->value;
text = t_intern_text( parse->task, temp_text->value,
temp_text->length );
length = temp_text->length;
}
else {
length = copied_text - text;
text_buffer->left = copied_text + 1;
}
goto finish;
}
else {
*copied_text = ch;
++copied_text;
ch = read_ch( parse );
}
text = temp_text( parse );
while ( true ) {
if ( ! ch ) {
struct pos pos;
t_init_pos( &pos,
parse->source->file_entry_id,
line, column );
p_diag( parse, DIAG_POS_ERR, &pos,
"unterminated string" );
p_bail( parse );
}
else if ( ch == '"' ) {
ch = read_ch( parse );
tk = TK_LIT_STRING;
goto finish;
}
else {
append_ch( text, ch );
ch = read_ch( parse );
}
}

Expand Down Expand Up @@ -1751,15 +1671,15 @@ void read_token_acs95( struct parse* parse, struct token* token ) {
finish:
// -----------------------------------------------------------------------
token->type = tk;
token->modifiable_text = NULL;
if ( text ) {
token->text = text;
token->length = length;
token->text = t_intern_text( parse->task, text->value, text->length );
token->length = text->length;
}
else {
const struct token_info* info = p_get_token_info( tk );
token->text = info->shared_text;
token->length = ( length > 0 ) ?
length : info->length;
token->length = info->length;
}
token->pos.line = line;
token->pos.column = column;
Expand Down

0 comments on commit 4388a16

Please sign in to comment.