Changeset 2160

Timestamp:

05/24/12 20:12:42 (13 months ago)

Author:

ksherdy

Message:

Refactored pablo template.

Location:

proto/parabix2

Files:

• pablo_template.cpp (modified) (5 diffs)
• src/TagMatcher.hpp (modified) (16 diffs)

proto/parabix2/pablo_template.cpp

@@ -13 +13 @@
 #include "../lib/bitblock_iterator.hpp"
 #include "../lib/s2p.hpp"
-
-#define SEGMENT_BLOCKS 12
-#define BUFFER_SIZE (BLOCK_SIZE * SEGMENT_BLOCKS)
-#define OVERLAP_BUFSIZE (sizeof(BitBlock))
+#include "../lib/perflib/perfsec.h"
 
 #include "xmldecl.h"
 #include "namechars.h"
-#include "../lib/perflib/perfsec.h"
-
 #include "TagMatcher.hpp"
 #include "LineColTracker.hpp"
@@ -29 +24 @@
 
 #ifdef BUFFER_PROFILING
-        BOM_Table * parser_timer;
-
+    BOM_Table * parser_timer;
 #elif CODE_CLOCKER
-        #define NUM_EVENTS 1
-        int Events[NUM_EVENTS] = {PAPI_TOT_CYC};
-        //int Events[NUM_EVENTS] = {PAPI_L2_DCM};
-        //int Events[NUM_EVENTS] = {PAPI_TOT_CYC, PAPI_BR_MSP};
-        int cal_size = 20;
-        CC * parser_timer = new CC(Events,NUM_EVENTS,cal_size);
+    //#define NUM_EVENTS 1
+    //int Events[NUM_EVENTS] = {PAPI_TOT_CYC};
+    //int Events[NUM_EVENTS] = {PAPI_L2_DCM};
+    #define NUM_EVENTS 2
+    int Events[NUM_EVENTS] = {PAPI_TOT_CYC, PAPI_BR_MSP};
+    int cal_size = 20;
+    CC * parser_timer = new CC(Events,NUM_EVENTS,cal_size);
 #else
-        void * parser_timer;
+    void * parser_timer;
 #endif
 
-int block_base=0;
-int buffer_base=0;
-char * source;
-
-LineColTracker tracker;
-TagMatcher matcher;
 ErrorTracker error_tracker;
 BitBlock EOF_mask = simd<1>::constant<1>();
 
-static inline int NameStrt_check(int pos);
-static inline int Name_check(int pos);
-static inline int PIName_check(int pos);
-static inline int CD_check(int pos);
-static inline int GenRef_check(int pos);
-static inline int HexRef_check(int pos);
-static inline int DecRef_check(int pos);
-static inline int AttRef_check(int pos);
-
+//////////////////////////////////////////////////////////////////////////////////////////
+// Buffer Management
+//////////////////////////////////////////////////////////////////////////////////////////
+#include "../lib/buffer.hpp"
+
+#define OVERLAP_BUFSIZE PADDING_SIZE //sizeof(BitBlock)
+
+//////////////////////////////////////////////////////////////////////////////////////////
+// @ global depends on 'error_tracker' and 'EOF_mask' definitions.
+//////////////////////////////////////////////////////////////////////////////////////////
 @global
 
-static inline void s2p_do_block(BytePack U8[], Basis_bits & basis_bits);
-static inline void s2p_do_final_block(BytePack U8[], Basis_bits & basis_bits, BitBlock EOF_mask);
-static inline void postprocess_do_block(Lex & lex, CtCDPI_Callouts & ctCDPI_Callouts, Ref_Callouts & ref_Callouts, Check_streams & check_streams, int chars_avail);
-
-void do_process(FILE *infile, FILE *outfile);
-
-static inline void validate_block(BitBlockForwardIterator & start, int block_base, int is_valid(int));
-static inline void validate_block(BitBlockForwardIterator & start, int block_base, int is_valid(int,int));
+//////////////////////////////////////////////////////////////////////////////////////////
+// Headers that depend @ global stream struct types.
+//////////////////////////////////////////////////////////////////////////////////////////
+#include "../lib/transpose.hpp"
+#include "post_process.hpp"
+
+static void do_process(FILE *infile, FILE *outfile);
 
 int main(int argc, char * argv[]) {
@@ -100 +89 @@
         }
 
-//      PERF_SEC_BIND(1);
+        PERF_SEC_BIND(1);
 
         PERF_SEC_INIT(parser_timer);
@@ -116 +105 @@
 }
 
-/* s2p Definitions */
-static inline void s2p_do_block(BytePack U8[], Basis_bits & basis_bits) {
-  s2p(U8[0], U8[1], U8[2], U8[3], U8[4], U8[5], U8[6], U8[7],
-        basis_bits.bit_0, basis_bits.bit_1, basis_bits.bit_2, basis_bits.bit_3, basis_bits.bit_4, basis_bits.bit_5, basis_bits.bit_6, basis_bits.bit_7);
-}
-
-static inline void s2p_do_final_block(BytePack U8[], Basis_bits & basis_bits, BitBlock EOF_mask) {
-  s2p_do_block(U8, basis_bits);
-  basis_bits.bit_0 = simd_and(basis_bits.bit_0, EOF_mask);
-  basis_bits.bit_1 = simd_and(basis_bits.bit_1, EOF_mask);
-  basis_bits.bit_2 = simd_and(basis_bits.bit_2, EOF_mask);
-  basis_bits.bit_3 = simd_and(basis_bits.bit_3, EOF_mask);
-  basis_bits.bit_4 = simd_and(basis_bits.bit_4, EOF_mask);
-  basis_bits.bit_5 = simd_and(basis_bits.bit_5, EOF_mask);
-  basis_bits.bit_6 = simd_and(basis_bits.bit_6, EOF_mask);
-  basis_bits.bit_7 = simd_and(basis_bits.bit_7, EOF_mask);
-}
-
-
-static inline int NameStrt_check(int pos) {
-        if(XML_10_UTF8_NameStrt_bytes((unsigned char*)&source[pos]) == 0){
-              return XMLTestSuiteError::NAME_START;
-        }
-        return 0;
-}
-
-static inline int Name_check(int pos) {
-        if(XML_10_UTF8_NameChar_bytes((unsigned char*)&source[pos]) == 0){
-                  return XMLTestSuiteError::NAME;
-        }
-        return 0;
-}
-
-static inline int PIName_check(int pos, int file_pos) {
-        if (at_XxMmLll<ASCII>((unsigned char*)&source[pos]) && (source[pos+3]=='?' || source[pos+3]<= ' ')) {
-              // "<?xml" legal at start of file.
-              if (!((file_pos == 2) && at_XmlDecl_start<ASCII>((unsigned char*)&source[0]))) {
-                  return XMLTestSuiteError::XMLPINAME;
-              }
-        }
-        return 0;
-}
-
-static inline int CD_check(int pos) {
-        if (!at_CDATA1<ASCII>((unsigned char*)&source[pos])){
-                  return XMLTestSuiteError::CDATA;
-        }
-        return 0;
-}
-
-static inline int GenRef_check(int pos) {
-        unsigned char* s = (unsigned char*)&source[pos];
-        if (!(at_Ref_gt<ASCII>(s)||at_Ref_lt<ASCII>(s)||at_Ref_amp<ASCII>(s)||at_Ref_quot<ASCII>(s)||at_Ref_apos<ASCII>(s))){
-              return XMLTestSuiteError::UNDEFREF;
-        }
-        return 0;
-}
-
-static inline int HexRef_check(int pos) {
-        unsigned char* s = (unsigned char*)&source[pos];
-        int ch_val = 0;
-        while(at_HexDigit<ASCII>(s)){
-          ch_val = HexVal<ASCII>(s[0]) + (ch_val<<4);
-          if (ch_val> 0x10FFFF ){
-                return XMLTestSuiteError::CHARREF;
-          }
-          s++;
-        }
-        if ((ch_val == 0x0) || ((ch_val | 0x7FF) == 0xDFFF)|| ((ch_val | 0x1) == 0xFFFF)){
-          return XMLTestSuiteError::CHARREF;
-        }
-        else if (((ch_val < 0x20) && (ch_val != 0x9) && (ch_val != 0xD) && (ch_val != 0xA))){
-          return XMLTestSuiteError::XML10CHARREF;
-        }
-        return 0;
-}
-
-static inline int DecRef_check(int pos) {
-        unsigned char* s = (unsigned char*)&source[pos];
-        int ch_val = 0;
-        while(at_HexDigit<ASCII>(s)){
-          ch_val = DigitVal<ASCII>(s[0]) + ch_val*10;
-          if (ch_val> 0x10FFFF ){
-                        return XMLTestSuiteError::CHARREF;
-          }
-          s++;
-        }
-        if ((ch_val == 0x0) || ((ch_val | 0x7FF) == 0xDFFF)|| ((ch_val | 0x1) == 0xFFFF)){
-                  return XMLTestSuiteError::CHARREF;
-        }
-        else if (((ch_val < 0x20) && (ch_val != 0x9) && (ch_val != 0xD) && (ch_val != 0xA))){
-                  return XMLTestSuiteError::XML10CHARREF;
-        }
-        return 0;
-}
-
-static inline int AttRef_check(int pos) {
-        unsigned char* s = (unsigned char*)&source[pos];
-        int ch_val = 0;
-        if(s[0]=='#'){
-          s++;
-          if(s[0]=='x' || s[0]=='X'){
-            s++;
-            while(at_HexDigit<ASCII>(s)){
-              ch_val = HexVal<ASCII>(s[0]) + (ch_val<<4);
-              s++;
-            }
-          }
-          else{
-            while(at_HexDigit<ASCII>(s)){
-              ch_val = DigitVal<ASCII>(s[0]) + ch_val*10;
-              s++;
-            }
-          }
-          if (ch_val==60){
-            return XMLTestSuiteError::ATTREF;
-          }
-        }
-        else if(at_Ref_lt<ASCII>(s)){
-          return XMLTestSuiteError::ATTREF;
-        }
-        return 0;
-}
-
-static inline void validate_block(BitBlockForwardIterator & start, int block_base, int is_valid(int)) {
-
-        int pos, block_pos;
-        BitBlockForwardIterator end;
-        while(start != end) {
-
-                block_pos = block_base + *start;
-                int rv = is_valid(block_pos);
-
-                if (rv) {
-                        int error_line, error_column;
-                        tracker.get_Line_and_Column(block_pos, error_line, error_column);
-                        ReportError(XMLTestSuiteError::get_msg(rv), error_line, error_column);
-                        exit(-1);
-                }
-                start++;
-        }
-}
-
-static inline void validate_block(BitBlockForwardIterator & start, int block_base, int buffer_base, int is_valid(int,int)) {
-
-        int pos, block_pos, file_pos;
-        BitBlockForwardIterator end;
-        while(start != end) {
-
-                block_pos = block_base + *start;
-                file_pos = block_pos+buffer_base;
-
-
-                int rv = is_valid(block_pos, file_pos);
-
-                if (rv) {
-                        int error_line, error_column;
-                        tracker.get_Line_and_Column(block_pos, error_line, error_column);
-                        ReportError(XMLTestSuiteError::get_msg(rv), error_line, error_column);
-                        exit(-1);
-                }
-                start++;
-        }
-}
-
-static inline void postprocess_do_block(Lex & lex, CtCDPI_Callouts & ctCDPI_Callouts, Ref_Callouts & ref_Callouts, Check_streams & check_streams, int chars_avail){
-BitBlockForwardIterator iter;
-
-tracker.StoreNewlines(lex.LF);
-
-if (bitblock::any(simd_or(check_streams.non_ascii_name_starts, check_streams.non_ascii_names))) {
-        iter.init(&check_streams.non_ascii_name_starts);
-        validate_block(iter, block_base, NameStrt_check);
-        iter.init(&check_streams.non_ascii_names);
-        validate_block(iter, block_base, Name_check);
-}
-if (bitblock::any(ctCDPI_Callouts.PI_name_starts)){
-        iter.init(&(ctCDPI_Callouts.PI_name_starts));
-        validate_block(iter, block_base, buffer_base, PIName_check);
-}
-if (bitblock::any(ctCDPI_Callouts.CD_starts)){
-        iter.init(&ctCDPI_Callouts.CD_starts);
-        validate_block(iter, block_base, CD_check);
-}
-if(bitblock::any(ref_Callouts.GenRef_starts)){
-        iter.init(&ref_Callouts.GenRef_starts);
-        validate_block(iter, block_base, GenRef_check);
-}
-if(bitblock::any(ref_Callouts.DecRef_starts)){
-        iter.init(&ref_Callouts.DecRef_starts);
-        validate_block(iter, block_base, DecRef_check);
-}
-if(bitblock::any(ref_Callouts.HexRef_starts)){
-        iter.init(&ref_Callouts.HexRef_starts);
-        validate_block(iter, block_base, HexRef_check);
-}
-if(bitblock::any(check_streams.att_refs)){
-        iter.init(&check_streams.att_refs);
-        validate_block(iter, block_base, AttRef_check);
-}
-
-if(error_tracker.Has_Noted_Error()){
-        int error_line, error_column;
-        tracker.get_Line_and_Column(error_tracker.Noted_Pos_In_Block(), error_line, error_column);
-        ReportError(error_tracker.Noted_Error_Msg(), error_line, error_column);
-        exit(-1);
-}
-
-matcher.store_streams(check_streams.tag_marks, check_streams.name_follows, check_streams.misc_mask, chars_avail);
-tracker.AdvanceBlock();
-
-}
-
 void do_process(FILE *infile, FILE *outfile) {
 
-@decl
-
-  int buf_pos = 0;
-  int block_pos = 0;
-  int chars_avail = 0;
-  int check_pos = 0;
-  int chars_read = 0;
-  BytePack buf[(BUFFER_SIZE+BLOCK_SIZE+OVERLAP_BUFSIZE*2)/sizeof(BitBlock)];
-
-  char * srcbuf = ((char *) buf) + OVERLAP_BUFSIZE;
-  buffer_base = buf_pos;
-  source = srcbuf;
-
-  chars_read = fread((void *)srcbuf, 1, BUFFER_SIZE + OVERLAP_BUFSIZE, infile);
-  chars_avail = chars_read;
-  if (chars_avail > BUFFER_SIZE) chars_avail = BUFFER_SIZE;
-
-  matcher.setSrc(srcbuf);
-
-  if(chars_read<4){
+    @decl
+
+    LineColTracker tracker;
+    TagMatcher<SEGMENT_SIZE,OVERLAP_BUFSIZE> matcher;
+
+    uint8_t * src_buf;
+    int block_base=0;
+    int buffer_base=0;
+    int buffer_pos = 0;
+    int block_pos = 0;
+    int chars_avail = 0;
+    int check_pos = 0;
+    int chars_read = 0;
+
+    //////////////////////////////////////////////////////////////////////////////////////////
+    // Buffer Management
+    //////////////////////////////////////////////////////////////////////////////////////////
+    BitBlock buf[(PADDING_SIZE + SEGMENT_SIZE + PADDING_SIZE)/sizeof(BitBlock)];
+    src_buf = (uint8_t *)buf + PADDING_SIZE;
+
+    //ALLOC_STATIC_ALIGNED_BYTE_BUFFER(src_buf, (PADDING_SIZE + SEGMENT_SIZE + PADDING_SIZE));
+
+    buffer_base = buffer_pos;
+    chars_read = fread((void *)src_buf, 1, BUFFER_SIZE, infile);
+    chars_avail = chars_read;
+    if (chars_avail > SEGMENT_SIZE) chars_avail = SEGMENT_SIZE;
+
+    //////////////////////////////////////////////////////////////////////////////////////////
+    // XML Validation / Content Model
+    //////////////////////////////////////////////////////////////////////////////////////////
+    if(chars_read<4){
     fprintf(stderr,"File is too short. Not well formed.\n");
     exit(-1);
-  }
-
-  Entity_Info * e = new Entity_Info;
-  e->AnalyzeSignature((unsigned char *)srcbuf);
-
-  if (e->code_unit_base == ASCII) {
-
-    XML_Decl_Parser<ASCII> decl_parser((unsigned char *)srcbuf);
+    }
+
+    Entity_Info * e = new Entity_Info;
+    e->AnalyzeSignature((unsigned char *)src_buf);
+
+    if (e->code_unit_base == ASCII) {
+
+    XML_Decl_Parser<ASCII> decl_parser((unsigned char *)src_buf);
 
     decl_parser.ReadXMLInfo(*e);
 
     if (e->code_unit_size != SingleByte || (e->has_encoding_decl && (!at_UTF_8(e->encoding)))){
-        fprintf(stderr,"Sorry, this xmlwf demo only works for UTF-8.\n");
-        exit(-1);
-    }
-  }
-  else {
-    fprintf(stderr,"Sorry, this xmlwf demo does not process EBCDIC.\n");
-        exit(-1);
-  }
-
-  if (e->content_start != 0) {
-        memmove(&srcbuf[0], &srcbuf[e->content_start], chars_read - e->content_start);
-        buf_pos = e->content_start;
-        if (chars_avail == BUFFER_SIZE) {
-                chars_read = chars_read - e->content_start +
-                             fread(&srcbuf[chars_read-e->content_start], 1, e->content_start, infile);
-                chars_avail = chars_read;
-                if (chars_avail > BUFFER_SIZE) chars_avail = BUFFER_SIZE;
-        }
-        else {
-          chars_read -=e->content_start;
-          chars_avail -=e->content_start;
-        }
-  }
-
-@stream_stmts
-
-/* Full Buffers */
-
-    while (chars_avail == BUFFER_SIZE) {
+        fprintf(stderr,"Sorry, this xmlwf demo only works for UTF-8.\n");
+        exit(-1);
+    }
+    }
+    else {
+        fprintf(stderr,"Sorry, this xmlwf demo does not process EBCDIC.\n");
+        exit(-1);
+    }
+
+    if (e->content_start != 0) {
+    memmove(&src_buf[0], &src_buf[e->content_start], chars_read - e->content_start);
+    buffer_pos = e->content_start;
+    if (chars_avail == SEGMENT_SIZE) {
+        chars_read = chars_read - e->content_start + fread(&src_buf[chars_read-e->content_start], 1, e->content_start, infile);
+        chars_avail = chars_read;
+        if (chars_avail > SEGMENT_SIZE) chars_avail = SEGMENT_SIZE;
+    }
+    else {
+      chars_read -=e->content_start;
+      chars_avail -=e->content_start;
+    }
+    }
+
+    @stream_stmts
+
+
+    //////////////////////////////////////////////////////////////////////////////////////////
+    // Full Segments
+    //////////////////////////////////////////////////////////////////////////////////////////
+    matcher.setSrc((char *)src_buf);
+    while (chars_avail == SEGMENT_SIZE) {
       PERF_SEC_START(parser_timer);
       for (int blk = 0; blk < SEGMENT_BLOCKS; blk++) {
           block_base = blk*BLOCK_SIZE;
-          s2p_do_block((BytePack *) &srcbuf[block_base], basis_bits);
+          s2p_do_block((BytePack *) &src_buf[block_base], basis_bits);
+
           @block_stmts
-          postprocess_do_block(lex, ctCDPI_Callouts, ref_Callouts, check_streams, chars_avail);
+
+          tracker.StoreNewlines(lex.LF);
+          postprocess_do_block(lex, ctCDPI_Callouts, ref_Callouts, check_streams, (char *)src_buf, buffer_base, block_base, chars_avail, tracker);
+          matcher.store_streams(check_streams.tag_marks, check_streams.name_follows, check_streams.misc_mask, chars_avail);
+          tracker.AdvanceBlock();
       }
       matcher.StreamScan(chars_avail);
@@ -406 +200 @@
 
       int bytes_left = chars_read - chars_avail;
-      memmove(srcbuf, &srcbuf[BUFFER_SIZE], bytes_left);
-      chars_read = fread(&srcbuf[bytes_left],1, BUFFER_SIZE + OVERLAP_BUFSIZE - bytes_left, infile) + bytes_left;
+      memmove(src_buf, &src_buf[SEGMENT_SIZE], bytes_left);
+      chars_read = fread(&src_buf[bytes_left], 1, BUFFER_SIZE - bytes_left, infile) + bytes_left;
       chars_avail = chars_read;
-      if (chars_avail > BUFFER_SIZE) chars_avail = BUFFER_SIZE;
-      buf_pos += chars_avail;
-      buffer_base = buf_pos;
-    }
-/* Final Partial Buffer */
+      if (chars_avail > SEGMENT_SIZE) chars_avail = SEGMENT_SIZE;
+      buffer_pos += chars_avail;
+      buffer_base = buffer_pos;
+    }
+
+    //////////////////////////////////////////////////////////////////////////////////////////
+    // Final Partial Segment
+    //////////////////////////////////////////////////////////////////////////////////////////
     PERF_SEC_START(parser_timer);
 
     block_pos = 0;
     int remaining = chars_avail;
-/* Full Blocks */
+
+    /* Full Blocks */
     while (remaining >= BLOCK_SIZE) {
-          block_base = block_pos;
-          s2p_do_block((BytePack *) &srcbuf[block_pos], basis_bits);
+          block_base = block_pos;
+          s2p_do_block((BytePack *) &src_buf[block_pos], basis_bits);
           @block_stmts
-          postprocess_do_block(lex, ctCDPI_Callouts, ref_Callouts, check_streams, chars_avail);
-          block_pos += BLOCK_SIZE;
+          tracker.StoreNewlines(lex.LF);
+          postprocess_do_block(lex, ctCDPI_Callouts, ref_Callouts, check_streams, (char *)src_buf, buffer_base, block_base, chars_avail, tracker);
+          matcher.store_streams(check_streams.tag_marks, check_streams.name_follows, check_streams.misc_mask, chars_avail);
+          tracker.AdvanceBlock();
+          block_pos += BLOCK_SIZE;
           remaining -= BLOCK_SIZE;
     }
     block_base = block_pos;
+
+    /* Partial Block or Any Carry */
     if (remaining > 0 || @any_carry) {
           EOF_mask = bitblock::srl(simd<1>::constant<1>(), convert(BLOCK_SIZE-remaining));
-          s2p_do_final_block((BytePack *) &srcbuf[block_pos], basis_bits, EOF_mask);
+          s2p_do_final_block((BytePack *) &src_buf[block_pos], basis_bits, EOF_mask);
           @final_block_stmts
-          postprocess_do_block(lex, ctCDPI_Callouts, ref_Callouts, check_streams, chars_avail);
-    }
-    buf_pos += chars_avail;
-    buffer_base = buf_pos;
+          tracker.StoreNewlines(lex.LF);
+          postprocess_do_block(lex, ctCDPI_Callouts, ref_Callouts, check_streams, (char *)src_buf, buffer_base, block_base, chars_avail, tracker);
+          matcher.store_streams(check_streams.tag_marks, check_streams.name_follows, check_streams.misc_mask, chars_avail);
+          tracker.AdvanceBlock();
+
+    }
+
+    buffer_pos += chars_avail;
+    buffer_base = buffer_pos;
 
     matcher.StreamScan(chars_avail);
     matcher.Advance_buffer();
-
 
     PERF_SEC_END(parser_timer, chars_avail);

proto/parabix2/src/TagMatcher.hpp

@@ -4 +4 @@
 #include "../lib/bitblock.hpp"
 
-#define MAX_DEPTH 100
 #include <algorithm>
 #include <iostream>
@@ -10 +9 @@
 using namespace std;
 
-
 #define MAX_DEPTH 100
 #define MAX_ATTS 100
@@ -19 +17 @@
 };
 
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
 class TagMatcher {
-  public:
-  BitBlock tagMarks[BUFFER_SIZE/BLOCK_SIZE];
-  BitBlock miscMarks[BUFFER_SIZE/BLOCK_SIZE];
-  char tags_buf[BUFFER_SIZE];
-  int tags_buf_cur;
-  int stream_index;
-  char * srcbuf;
-  int depth;
-  int inTagPos;
-  int finalStartPos;
-  char* tag_stack[MAX_DEPTH];
-  int tag_lgth_stack[MAX_DEPTH];
-  BitBlock NameFollows[BUFFER_SIZE/BLOCK_SIZE+1]; // 1 extra block for sentinel
-  int buf_base;
-  enum TagMatchState {InStartTag, InEndTag, InAttName, Clear} state;
-  enum TagMatchMode {StartOfFile, InFile} mode;
-  struct attribute Attr[MAX_ATTS];
-  struct attribute InAtt;
-  int att_index;
-  int InFinalEndTag;
-
+
+public:
   TagMatcher();
   ~TagMatcher();
@@ -50 +30 @@
   int does_match(char * s1, char * s2, int lgth);
   int lookup_or_insert(char*s, int lgth);
+
+    int depth;
+
+private:
+  BitBlock tagMarks[BUF_SIZE/BLOCK_SIZE];
+  BitBlock miscMarks[BUF_SIZE/BLOCK_SIZE];
+  char tags_buf[BUF_SIZE];
+  int tags_buf_cur;
+  int stream_index;
+  char * srcbuf;
+
+  int inTagPos;
+  int finalStartPos;
+  char* tag_stack[MAX_DEPTH];
+  int tag_lgth_stack[MAX_DEPTH];
+  BitBlock NameFollows[BUF_SIZE/BLOCK_SIZE+1]; // 1 extra block for sentinel
+  int buf_base;
+  enum TagMatchState {InStartTag, InEndTag, InAttName, Clear} state;
+  enum TagMatchMode {StartOfFile, InFile} mode;
+  struct attribute Attr[MAX_ATTS];
+  struct attribute InAtt;
+  int att_index;
+  int InFinalEndTag;
+
 };
 
-int TagMatcher::lookup_or_insert(char* s, int lgth){
-  for(int i=0; i< att_index; i++)
-    if(lgth == Attr[i].lgth &&  does_match(s,Attr[i].start,lgth))
-      return 1;
-
-  Attr[att_index].start = s;
-  Attr[att_index].lgth = lgth;
-  att_index++;
-  return 0;
-}
-
-int TagMatcher::does_match(char * s1, char * s2, int lgth){
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+int TagMatcher<BUF_SIZE, OVER_SIZE>::does_match(char * s1, char * s2, int lgth){
     int matchlen = 0;
     int i=0;
@@ -69 +63 @@
       /* full 16 byte match */
 
-      if (bitblock::any(simd_xor(bitblock::load_unaligned((BitBlock*)&s1[i]), 
-                                 bitblock::load_unaligned((BitBlock*)&s2[i])))) {
-                return 0;
+      if (bitblock::any(simd_xor(bitblock::load_unaligned((BitBlock*)&s1[i]),
+                 bitblock::load_unaligned((BitBlock*)&s2[i])))) {
+        return 0;
       }
       else {
-                lgth -= sizeof(BitBlock);
-                i +=sizeof(BitBlock);
-      }
-    }
-
-        scanword_t temp = ~hsimd<8>::signmask(simd<8>::eq(bitblock::load_unaligned((BitBlock*)&s1[i]),
+        lgth -= sizeof(BitBlock);
+        i +=sizeof(BitBlock);
+      }
+    }
+
+    scanword_t temp = ~hsimd<8>::signmask(simd<8>::eq(bitblock::load_unaligned((BitBlock*)&s1[i]),
                                                   bitblock::load_unaligned((BitBlock*)&s2[i])));
 
-        return lgth <= scan_forward_zeroes(temp);
-}
-
-
-int TagMatcher:: tag_match(int pos, int chars_avail) {
+    return lgth <= scan_forward_zeroes(temp);
+}
+
+
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+int TagMatcher<BUF_SIZE, OVER_SIZE>::lookup_or_insert(char* s, int lgth){
+  for(int i=0; i< this->att_index; i++)
+    if(lgth == this->Attr[i].lgth &&  this->does_match(s,this->Attr[i].start,lgth))
+      return 1;
+
+  this->Attr[att_index].start = s;
+  this->Attr[att_index].lgth = lgth;
+  this->att_index++;
+  return 0;
+}
+
+
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+int TagMatcher<BUF_SIZE, OVER_SIZE>:: tag_match(int pos, int chars_avail) {
         int rt_val=0;
 //      end tag
-        if(srcbuf[pos]=='/' ){
+    if(this->srcbuf[pos]=='/' ){
           pos++;
-          depth--;
-          if (depth<0)
+      this->depth--;
+      if (this->depth<0)
             return pos;
-          int lgth = tag_lgth_stack[depth];
-
-          if (does_match(tag_stack[depth],&srcbuf[pos],lgth) && ((srcbuf[pos+lgth] == '>') ||(srcbuf[pos+lgth] <= ' '))) rt_val=0;
-          else if (pos + lgth >= BUFFER_SIZE + OVERLAP_BUFSIZE) {
-            state = InEndTag;
-            inTagPos = BUFFER_SIZE - pos;
+      int lgth = this->tag_lgth_stack[depth];
+
+      if (does_match(this->tag_stack[depth],&this->srcbuf[pos],lgth) && ((this->srcbuf[pos+lgth] == '>') ||(this->srcbuf[pos+lgth] <= ' '))) rt_val=0;
+      else if (pos + lgth >= BUF_SIZE + OVER_SIZE) {
+        this->state = InEndTag;
+        this-> inTagPos = BUF_SIZE - pos;
             rt_val=0;
           }
@@ -109 +117 @@
           }
 
-          if (depth == 0){
-            while(srcbuf[pos]!='>'){
+      if (this->depth == 0){
+        while(this->srcbuf[pos]!='>'){
               pos++;
               if(pos>=chars_avail){
-                InFinalEndTag = 1;
+        this->InFinalEndTag = 1;
                 return 0;
               }
             }
-            pos = bitstream_scan(miscMarks,pos+1);
+        pos = bitstream_scan(this->miscMarks,pos+1);
+            if(pos!=chars_avail){
+          fprintf(stderr,"illegal content after root element at position = %i\n",this->buf_base+pos);
+              exit(-1);
+            }
+          }
+          return rt_val;
+        }
+//      empty tag
+    else if(this->srcbuf[pos]=='>'){
+      this->depth--;
+      if (this->depth == 0){
+        while(this->srcbuf[pos]!='>')
+              pos++;
+        pos = bitstream_scan(this->miscMarks,pos+1);
+
             if(pos!=chars_avail){
               fprintf(stderr,"illegal content after root element at position = %i\n",buf_base+pos);
@@ -123 +146 @@
             }
           }
-          return rt_val;
-        }
-//      empty tag
-        else if(srcbuf[pos]=='>'){
-          depth--;
-          if (depth == 0){
-            while(srcbuf[pos]!='>')
-              pos++;
-            pos = bitstream_scan(miscMarks,pos+1);
-
-            if(pos!=chars_avail){
-              fprintf(stderr,"illegal content after root element at position = %i\n",buf_base+pos);
-              exit(-1);
-            }
-          }
         }
 //      start tag
-        else if(srcbuf[pos-1]=='<'){
-          att_index = 0;
-          if(depth<MAX_DEPTH){
-            int end_pos = bitstream_scan(NameFollows,pos);
-            tag_lgth_stack[depth] = end_pos-pos;
-            tag_stack[depth] = &srcbuf[pos];
-            if(end_pos<BUFFER_SIZE){
-              depth++;
+    else if(this->srcbuf[pos-1]=='<'){
+      this->att_index = 0;
+      if(this->depth<MAX_DEPTH){
+        int end_pos = bitstream_scan(this->NameFollows,pos);
+        this->tag_lgth_stack[this->depth] = end_pos-pos;
+        this->tag_stack[depth] = &this->srcbuf[pos];
+        if(end_pos<BUF_SIZE){
+         this->depth++;
             }
             else{
-              state = InStartTag;
-              finalStartPos = pos;
+          this->state = InStartTag;
+          this->finalStartPos = pos;
             }
           }
           else{
-            fprintf(stderr,"Max nesting depth exceeded at position =%i. depth = %i\n",buf_base+pos, depth);
+        fprintf(stderr,"Max nesting depth exceeded at position =%i. depth = %i\n",this->buf_base+pos, this->depth);
             exit(-1);
           }
@@ -161 +169 @@
 //      attribute
         else{
-          int end_pos = bitstream_scan(NameFollows,pos);
-          if(end_pos<BUFFER_SIZE){
-            if(lookup_or_insert(&srcbuf[pos], end_pos-pos)){
-              fprintf(stderr,"Attribute name is not unique at position =%i.\n",buf_base+pos);
+      int end_pos = bitstream_scan(this->NameFollows,pos);
+      if(end_pos<BUF_SIZE){
+        if(lookup_or_insert(&this->srcbuf[pos], end_pos-pos)){
+          fprintf(stderr,"Attribute name is not unique at position =%i.\n",this->buf_base+pos);
               exit(-1);
             }
           }
           else{
-            state = InAttName;
-            InAtt.start = &srcbuf[pos];
-            InAtt.lgth = BUFFER_SIZE-pos;
+        this->state = InAttName;
+        this->InAtt.start = &this->srcbuf[pos];
+        this->InAtt.lgth = BUF_SIZE-pos;
           }
         }
@@ -177 +185 @@
 }
 
-
-int TagMatcher::StreamScan(int chars_avail) {
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+int TagMatcher<BUF_SIZE, OVER_SIZE>::StreamScan(int chars_avail) {
 
         int blk;
@@ -184 +192 @@
         int block_pos = 0;
 
-        if(mode == StartOfFile){
+    if(this->mode == StartOfFile){
           int pos = bitstream_scan(miscMarks,0);
           if (pos==chars_avail){
-            fprintf(stderr,"no element at position =%i.\n",buf_base+pos);
+        fprintf(stderr,"no element at position =%i.\n",this->buf_base+pos);
             exit(-1);
           }
-          if(srcbuf[pos-1]!='<'|| srcbuf[pos]=='!'||srcbuf[pos]=='/'){
+      if(this->srcbuf[pos-1]!='<'|| this->srcbuf[pos]=='!'|| this->srcbuf[pos]=='/'){
 #ifdef DUMP
 print_register<BitBlock>("srcbuf", bitblock::load_unaligned((BitBlock *) srcbuf));
@@ -197 +205 @@
             exit(-1);
           }
-          mode = InFile;
+      this->mode = InFile;
         }
         for (blk = 0; blk < blk_counts; blk++) {
-                scanword_t s = ((scanword_t*)tagMarks)[blk];
+        scanword_t s = ((scanword_t*)this->tagMarks)[blk];
                 while(s) {
                         int code = tag_match(scan_forward_zeroes(s) + block_pos, chars_avail);
@@ -212 +220 @@
 }
 
-void TagMatcher::store_streams(BitBlock tagMark, BitBlock NameFollow, BitBlock miscMark, int chars_avail){
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+void TagMatcher<BUF_SIZE, OVER_SIZE>::store_streams(BitBlock tagMark, BitBlock NameFollow, BitBlock miscMark, int chars_avail){
 #ifdef DUMP
 print_register<BitBlock>("tagMark", tagMark);
@@ -220 +229 @@
 printf("stream_index = %i\n", stream_index);
 #endif
-  tagMarks[stream_index] = tagMark;
-  miscMarks[stream_index] = simd_not(miscMark);
-  NameFollows[stream_index] = NameFollow;
-  stream_index++;
-  if(stream_index==1){
-
-    if (InFinalEndTag == 1){
+  this->tagMarks[this->stream_index] = tagMark;
+  this->miscMarks[this->stream_index] = simd_not(miscMark);
+  this->NameFollows[this->stream_index] = NameFollow;
+  this->stream_index++;
+  if(this->stream_index==1){
+
+    if (this->InFinalEndTag == 1){
       int pos = -1;
-      while(srcbuf[pos]!='>'){
+      while(this->srcbuf[pos]!='>'){
         pos++;
         if(pos>=chars_avail){
-          InFinalEndTag = 1;
+      this->InFinalEndTag = 1;
           return;
         }
       }
-      pos = bitstream_scan(miscMarks,pos+1);
+      pos = bitstream_scan(this->miscMarks,pos+1);
 #ifdef DUMP
 print_register<BitBlock>("miscMarks[0]", miscMarks[0]);
@@ -246 +255 @@
     }
 
-    if(state == InStartTag) {
-      state = Clear;
-      int remain_lgth = bitstream_scan(NameFollows,0);
-      memcpy(&tags_buf[tags_buf_cur],srcbuf,remain_lgth);
-      tag_lgth_stack[depth] += remain_lgth;
-      depth++;
-    }
-    else if (state == InEndTag) {
-      state = Clear;
-      int lgth = tag_lgth_stack[depth];
-      if (does_match(tag_stack[depth]+inTagPos,srcbuf,lgth-inTagPos) && ((srcbuf[lgth-inTagPos] == '>') ||(srcbuf[lgth-inTagPos] <= ' '))) return ;
+    if(this->state == InStartTag) {
+      this->state = this->Clear;
+      int remain_lgth = bitstream_scan(this->NameFollows,0);
+      memcpy(&this->tags_buf[this->tags_buf_cur],this->srcbuf,remain_lgth);
+      this->tag_lgth_stack[this->depth] += remain_lgth;
+      this->depth++;
+    }
+    else if (this->state == InEndTag) {
+      this->state = Clear;
+      int lgth = this->tag_lgth_stack[this->depth];
+      if (does_match(this->tag_stack[this->depth]+this->inTagPos,this->srcbuf,lgth-this->inTagPos) && ((this->srcbuf[lgth-this->inTagPos] == '>') ||(this->srcbuf[lgth-this->inTagPos] <= ' '))) return ;
       else {
           fprintf(stderr,"tag name mismatch at position = %i\n",buf_base);
@@ -262 +271 @@
       }
     }
-    else if (state == InAttName) {
-      state = Clear;
-      int remain_lgth = bitstream_scan(NameFollows,0);
-      memcpy(&tags_buf[tags_buf_cur],srcbuf,remain_lgth);
-      if(lookup_or_insert(InAtt.start, InAtt.lgth+remain_lgth)){
+    else if (this->state == InAttName) {
+      this->state = Clear;
+      int remain_lgth = bitstream_scan(this->NameFollows,0);
+      memcpy(&this->tags_buf[this->tags_buf_cur],this->srcbuf,remain_lgth);
+      if(lookup_or_insert(this->InAtt.start, this->InAtt.lgth+remain_lgth)){
               fprintf(stderr,"Attribute name is not unique at position =%i.\n",buf_base);
               exit(-1);
@@ -274 +283 @@
 }
 
-TagMatcher::TagMatcher(){
-  stream_index = 0;
-  depth = 0;
-  buf_base = 0;
-  state = Clear;
-  mode = StartOfFile;
-  InFinalEndTag = 0;
-  NameFollows[BUFFER_SIZE/BLOCK_SIZE]=simd<1>::constant<1>();// TODO - verify simd_const_1(1);  //sentinel
-}
-
-
-
-TagMatcher::~TagMatcher(){
-
-}
-
-void TagMatcher::setSrc(char * src){
-  srcbuf = src;
-}
-
-void TagMatcher::Advance_buffer(){
-  buf_base += BUFFER_SIZE;
-  stream_index=0;
-  tags_buf_cur = 0;
-  att_index = 0;
-  for(int i=0; i< depth; i++){
-    if(&tags_buf[tags_buf_cur]!=tag_stack[i])
-      memcpy(&tags_buf[tags_buf_cur],tag_stack[i],tag_lgth_stack[i]);
-    tag_stack[i] = &tags_buf[tags_buf_cur];
-    tags_buf_cur += tag_lgth_stack[i];
-  }
-  if(state == InStartTag) {
-      memcpy(&tags_buf[tags_buf_cur],&srcbuf[finalStartPos],tag_lgth_stack[depth]);
-      tag_stack[depth] = &tags_buf[tags_buf_cur];
-      tags_buf_cur += tag_lgth_stack[depth];
-  }
-  else if(state == InEndTag) {
-     memcpy(&tags_buf[tags_buf_cur],tag_stack[depth],tag_lgth_stack[depth]);
-    tag_stack[depth] = &tags_buf[tags_buf_cur];
-    tags_buf_cur += tag_lgth_stack[depth];
-  }
-  else if(state == InAttName) {
-      memcpy(&tags_buf[tags_buf_cur],InAtt.start,InAtt.lgth);
-      InAtt.start = &tags_buf[tags_buf_cur];
-      tags_buf_cur += InAtt.lgth;
-  }
-  srcbuf[-1] = srcbuf[BUFFER_SIZE-1];
-}
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+TagMatcher<BUF_SIZE, OVER_SIZE>::TagMatcher(){
+  this->stream_index = 0;
+  this->depth = 0;
+  this->buf_base = 0;
+  this->state = Clear;
+  this->mode = StartOfFile;
+  this->InFinalEndTag = 0;
+  this->NameFollows[BUF_SIZE/BLOCK_SIZE]=simd<1>::constant<1>();// TODO - verify simd_const_1(1);  //sentinel
+}
+
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+TagMatcher<BUF_SIZE, OVER_SIZE>::~TagMatcher(){
+
+}
+
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+void TagMatcher<BUF_SIZE, OVER_SIZE>::setSrc(char * src){
+  this->srcbuf = src;
+}
+
+template <uint64_t BUF_SIZE, uint64_t OVER_SIZE>
+void TagMatcher<BUF_SIZE, OVER_SIZE>::Advance_buffer(){
+  this->buf_base += BUF_SIZE;
+  this->stream_index=0;
+  this->tags_buf_cur = 0;
+  this->att_index = 0;
+  for(int i=0; i< this->depth; i++){
+    if(&this->tags_buf[this->tags_buf_cur]!=this->tag_stack[i])
+      memcpy(&this->tags_buf[this->tags_buf_cur],this->tag_stack[i],this->tag_lgth_stack[i]);
+    this->tag_stack[i] = &this->tags_buf[tags_buf_cur];
+    this->tags_buf_cur += this->tag_lgth_stack[i];
+  }
+  if(this->state == InStartTag) {
+      memcpy(&this->tags_buf[this->tags_buf_cur],&this->srcbuf[this->finalStartPos],this->tag_lgth_stack[this->depth]);
+      this->tag_stack[depth] = &this->tags_buf[this->tags_buf_cur];
+      this->tags_buf_cur += this->tag_lgth_stack[this->depth];
+  }
+  else if(this->state == InEndTag) {
+     memcpy(&this->tags_buf[this->tags_buf_cur],this->tag_stack[this->depth],this->tag_lgth_stack[this->depth]);
+    this->tag_stack[depth] = &this->tags_buf[this->tags_buf_cur];
+    this->tags_buf_cur += this->tag_lgth_stack[this->depth];
+  }
+  else if(this->state == InAttName) {
+      memcpy(&this->tags_buf[this->tags_buf_cur],this->InAtt.start,this->InAtt.lgth);
+      this->InAtt.start = &this->tags_buf[tags_buf_cur];
+      this->tags_buf_cur += this->InAtt.lgth;
+  }
+  this->srcbuf[-1] = this->srcbuf[BUF_SIZE-1];
+}
+
 
 #endif /* TAGMATCHER_HPP_ */