Add ability to use arbitrary genetic code string (ncbieaa, sncbieaa), make prodigal run 2x faster

.gitignore

@@ -1,2 +1,16 @@
+# Build artifacts
 *.o
 prodigal
+prodigal.exe
+prodigal-table
+prodigal-table.exe
+
+# Release binary names
+prodigal.linux
+prodigal.osx*
+prodigal.windows.exe
+
+# A playground directory
+test/
+
+

Makefile

@@ -25,33 +25,49 @@ CFLAGS  += -pedantic -Wall -O3 -DSUPPORT_GZIP_COMPRESSED
 LFLAGS = -lm $(LDFLAGS) -lz
 
 TARGET  = prodigal
+TABLEUTIL = prodigal-table
 ZTARGET  = zprodigal
 SOURCES = $(shell echo *.c)
-HEADERS = $(shell echo *.h)
+HEADERS = $(shell echo *.h) table-data.hh
 OBJECTS = $(SOURCES:.c=.o)
 ZOBJECTS = $(SOURCES:.c=.oz)
 
 INSTALLDIR  = /usr/local/bin
 
-all: $(TARGET)
+all: $(TARGET) $(TABLEUTIL)
 
 $(TARGET): $(OBJECTS)
 	$(CC) $(CFLAGS) -o $@ $^ $(LFLAGS)
 
+table-data.hh table-data.cc: gc.prt
+	./maketable.sh < gc.prt
+
+table.o: table.c $(HEADERS) table-data.cc
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+# Too big to recompile on every header change
+training.o: training.c training.h table.h
+	$(CC) $(CFLAGS) -c -o $@ $<
+
 %.o: %.c $(HEADERS)
 	$(CC) $(CFLAGS) -c -o $@ $<
 
-install: $(TARGET)
+$(TABLEUTIL): table.c table.h
+	$(CC) $(CFLAGS) -DTABLE_UTIL -Wno-parentheses -o $@ $< $(LFLAGS)
+
+install: $(TARGET) $(TABLEUTIL)
 	install -d -m 0755 $(INSTALLDIR)
 	install -m 0755 $(TARGET) $(INSTALLDIR)
+	install -m 0755 $(TABLEUTIL) $(INSTALLDIR)
 
 uninstall:
 	-rm $(INSTALLDIR)/$(TARGET)
+	-rm $(INSTALLDIR)/$(TABLEUTIL)
 
 clean:
 	-rm -f $(OBJECTS) $(ZOBJECTS)
 
 distclean: clean
-	-rm -f $(TARGET)
+	-rm -f $(TARGET) $(TABLEUTIL)
 
 .PHONY: all install uninstall clean distclean

bitmap.h

@@ -20,10 +20,52 @@
 
 #ifndef BITMAP_H_
 #define BITMAP_H_
+#include <stddef.h>
 
-unsigned char test(unsigned char *, int);
-void clear(unsigned char *, int);
-void set(unsigned char *, int);
-void toggle(unsigned char *, int);
+/* Test a bit, 0 = not set, 1 = set */
+static inline unsigned char test(unsigned char *bm, size_t ndx) {
+  return ( bm[ndx>>3] & (1 << (ndx&0x07))?1:0 ); 
+}
 
+/* Clear a bit (set it to 0) */
+static inline void clear(unsigned char *bm, size_t ndx) {
+  bm[ndx>>3] &= ~(1 << (ndx&0x07));
+}
+
+/* Set a bit to 1 */
+static inline void set(unsigned char *bm, size_t ndx) {
+  bm[ndx>>3] |= (1 << (ndx&0x07));
+}
+
+/* Flip a bit's value 0->1 or 1->0 */
+static inline void toggle(unsigned char *bm, size_t ndx) {
+  bm[ndx>>3] ^= (1 << (ndx&0x07)); 
+}
+
+/* Get two bits */
+static inline unsigned char nuc(unsigned char *bm, size_t ndx) {
+  ndx *= 2;
+  return ((bm[ndx >> 3] >> (ndx & 0x07)) & 0x03);
+}
+
+/* Get six bits, which can be in two bytes
+   out-of-bound read may happen but will not affect result */
+static inline unsigned char trinuc(unsigned char *bm, size_t ndx) {
+  ndx *= 2;
+  unsigned byte = ndx >> 3;
+  unsigned stretch = (bm[byte + 1] << 8) | bm[byte];
+  return ((stretch >> (ndx & 0x07)) & 0x3F);
+}
+
+/* Get twelve bits, which can be in four bytes
+   mer_ndx(6) is relatively common and should be replaced with this
+   (shorter assembly).  unfortunately the order is different, which
+   breaks tinf->gene_dc compatibility! */
+static inline unsigned char hexnuc(unsigned char *bm, int ndx) {
+  ndx *= 2;
+  unsigned byte = ndx >> 3;
+  unsigned stretch = (bm[byte + 3] << 24) | (bm[byte + 2] << 16) |
+                     (bm[byte + 1] << 8) | bm[byte];
+  return ((stretch >> (ndx & 0x07)) & 0x3F);
+}
 #endif

compare.sh

@@ -0,0 +1,68 @@
+#!/bin/bash
+# compare.sh: compare two versions of prodigal
+# ./compare.sh <old_binary> <input_file>
+
+if [[ $# -lt 2 ]]; then
+    echo "Usage: $0 <old_binary> <input_file> [additional opts]"
+    echo "Example: $0 ./prodigal.windows.exe anthus_aco.fas -t test/anthus_aco.train"
+    exit 1
+fi
+
+OLD_BINARY="$1"
+INPUT_FILE="$2"
+OUT_PREFIX="test/$(basename "$INPUT_FILE")"
+shift 2
+
+if ! command -v "$OLD_BINARY" >/dev/null; then
+    echo "Error: Old binary '$OLD_BINARY' is not executable. Did you forget a './' prefix?"
+    exit 1
+fi
+if [[ ! -f $INPUT_FILE ]]; then
+    echo "Error: Input file '$INPUT_FILE' does not exist."
+    exit 1
+fi
+
+mkdir -p "test"
+rm -f "$OUT_PREFIX".old.* "$OUT_PREFIX".new.* "$OUT_PREFIX".*.diff
+
+time "$OLD_BINARY" -i "$INPUT_FILE" -f gbk -o "$OUT_PREFIX".old.gbk -d "$OUT_PREFIX".old.ffn -a "$OUT_PREFIX".old.faa "$@"
+ret=$?
+if ((ret)); then
+    echo "Error: Old binary failed to run with $ret."
+fi
+# rerun if empty (first training creation)
+if [[ ! -s "$OUT_PREFIX".old.gbk || ! -s "$OUT_PREFIX".old.ffn || ! -s "$OUT_PREFIX".old.faa ]]; then
+    echo "Warning: Old binary produced empty output files, retrying..."
+    time "$OLD_BINARY" -i "$INPUT_FILE" -f gbk -o "$OUT_PREFIX".old.gbk -d "$OUT_PREFIX".old.ffn -a "$OUT_PREFIX".old.faa "$@"
+    ret=$?
+    if ((ret)); then
+        echo "Error: Old binary failed to run with $ret on retry."
+    fi
+fi
+
+time ./prodigal -i "$INPUT_FILE" -f gbk -o "$OUT_PREFIX".new.gbk -d "$OUT_PREFIX".new.ffn -a "$OUT_PREFIX".new.faa "$@"
+ret=$?
+if ((ret)); then
+    echo "Error: New binary failed to run with $ret."
+fi
+
+# normalize line endings
+sed -i -e 's/\r//g' "$OUT_PREFIX".{old,new}.*
+
+if ! diff -u "$OUT_PREFIX".old.gbk "$OUT_PREFIX".new.gbk >"$OUT_PREFIX".gbk.diff; then
+    echo "GBK files differ."
+else
+    echo "GBK files are identical."
+fi
+
+if ! diff -u "$OUT_PREFIX".old.ffn "$OUT_PREFIX".new.ffn >"$OUT_PREFIX".ffn.diff; then
+    echo "FFN files differ."
+else
+    echo "FFN files are identical."
+fi
+
+if ! diff -u "$OUT_PREFIX".old.faa "$OUT_PREFIX".new.faa >"$OUT_PREFIX".faa.diff; then
+    echo "FAA files differ."
+else
+    echo "FAA files are identical."
+fi