diff --git a/Projet/CODE/apm/Makefile b/Projet/CODE/apm/Makefile
index 2b3cde15d79f8095b5f8690d205ccb3eff2abcb6..02bc6ae708963cd043ab1275933f66a90f0589e1 100644
--- a/Projet/CODE/apm/Makefile
+++ b/Projet/CODE/apm/Makefile
@@ -24,6 +24,9 @@ apm_gpu: src/apm_gpu.cu
# you need to create the file dna/test.fa
test: all
+ ./apm 0 dna/small_chrY.fa TTGA TTGA TTGA
+ ./apm_omp 0 dna/small_chrY.fa TTGA TTGA TTGA
+ ./apm_gpu 0 dna/small_chrY.fa TTGA TTGA TTGA
./apm 0 dna/test.fa test123
./apm_omp 0 dna/test.fa test123
./apm_gpu 0 dna/test.fa test123
@@ -31,6 +34,7 @@ test: all
./apm_omp 0 dna/xenoRefMrna.fa ttggaa
./apm_gpu 0 dna/xenoRefMrna.fa ttggaa
+
.PHONY: clean
clean:
rm -rf apm apm_omp apm_gpu
diff --git a/Projet/CODE/apm/src/apm.c b/Projet/CODE/apm/src/apm.c
index 802e016712c498e6804d2c5ccda1b566ddd0c1a7..1847f6b6767078fd412c82964f1d02725bd70d15 100644
--- a/Projet/CODE/apm/src/apm.c
+++ b/Projet/CODE/apm/src/apm.c
@@ -14,6 +14,11 @@
#define APM_DEBUG 0
+#define MAX_BUFFER_SIZE 1024 * 1024 * 1024 // 1Go per reads at most
+//#define MAX_BUFFER_SIZE 512 // debug
+
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+#define MIN3(a, b, c) ((a)<(b) ? ((a)<(c) ? (a) : (c)) : ((b)<(c) ? (b) : (c)))
// can't be inline as it is a variadic function
void
@@ -31,55 +36,72 @@ __checkErrors(const int isBad, const char *fmt, ...)
}
}
+struct openfile
+{
+ int fd; // file descriptor
+ size_t size; // size of the file
+ size_t buf_start; // index of the start of the buffer in the file
+ char *buf; // buffer being read
+ unsigned int buf_alloc_size; // allocateed size of the buffer
+ unsigned int buf_size; // size of the datas in the buffer
+};
-char *
-read_input_file(char *filename, size_t *size)
+void
+open_input_file(struct openfile *file, char *filename)
{
- char *buf;
- struct stat fs;
- size_t fsize;
- int fd = 0;
- size_t read_bytes = 0;
- size_t total_bytes = 0;
- /* Open the text file */
- fd = open(filename, O_RDONLY);
- __checkErrors(fd == -1, "Unable to open the text file <%s>\n", filename);
+ // openning the file
+ file->fd = open(filename, O_RDONLY);
+ __checkErrors(file->fd < 0, "Unable to open the text file <%s>\n", filename);
- /* Get the number of characters in the textfile */
- int stat = fstat(fd, &fs);
- __checkErrors(stat < 0, "Stat of file <%s> failed\n", filename);
- fsize = fs.st_size;
+ // size of the file
+ struct stat fs;
+ int res = fstat(file->fd, &fs);
+ __checkErrors(res < 0, "Stat of file <%s> failed\n", filename);
+ file->size = fs.st_size;
+
+ // buffer allocation
+ file->buf_alloc_size = MIN(file->size, MAX_BUFFER_SIZE);
+ file->buf = (char *) malloc(file->buf_alloc_size * sizeof(char));
+ __checkErrors(file->buf == NULL,
+ "Unable to allocate %ld byte(s) for main array\n",
+ file->buf_alloc_size);
+
+ file->buf_start = 0;
+ file->buf_size = 0;
+}
-#if APM_DEBUG
- printf("File length: %ld\n", fsize);
-#endif
+int
+read_input_file(struct openfile *file, size_t recovering)
+{
+ // how mutch do we realy go back
+ size_t go_back = MIN(file->buf_start, recovering);
- /* Allocate data to copy the target text */
- buf = malloc(fsize * sizeof(char));
- __checkErrors(buf == NULL,
- "Unable to allocate %ld byte(s) for main array\n", fsize);
+ // we read the min between the max size of the buffer
+ // and what is left in the file
+ int to_read = MIN(file->buf_alloc_size - go_back,
+ file->size - file->buf_start);
- do
- {
- read_bytes = read(fd, buf, fsize - total_bytes);
- total_bytes += read_bytes;
- }
- while (read_bytes != 0);
-
- __checkErrors(total_bytes != fsize, "Unable to copy %ld byte(s) "
- "from text file (%ld byte(s) copied)\n", fsize, total_bytes);
+ // we copy the recovery from the end of the buffer
+ memcpy(file->buf, file->buf + file->buf_size - go_back, go_back);
-#if APM_DEBUG
- printf("Number of read bytes: %ld\n", total_bytes);
-#endif
+ // we read the file
+ int read_size = read(file->fd, file->buf + go_back, to_read);
+ __checkErrors(read_size != to_read, "Unable to copy %ld byte(s) "
+ "from text file (%ld byte(s) copied)\n", to_read, read_size);
+ file->buf_start += read_size;
+ file->buf_size = go_back + read_size;
- *size = total_bytes;
- close(fd);
- return buf;
+ return read_size;
}
-#define MIN3(a, b, c) ((a)<(b) ? ((a)<(c) ? (a) : (c)) : ((b)<(c) ? (b) : (c)))
+void
+close_input_file(struct openfile *file)
+{
+ int res = close(file->fd);
+ __checkErrors(res < 0, "Faile to close file descriptor (%d)\n.", file->fd);
+ free(file->buf);
+}
int
levenshtein(char *s1, char *s2, unsigned int len, unsigned int *column)
@@ -113,65 +135,77 @@ levenshtein(char *s1, char *s2, unsigned int len, unsigned int *column)
int
main(int argc, char **argv)
{
- char **pattern;
- char *filename;
+ // arguemnts
int approx_factor = 0;
- size_t nb_patterns = 0;
- size_t i, j;
- char *buf;
+ char *filename;
+ int nb_patterns = 0;
+ char **pattern;
+
+ // time interval and duration for time measurement
struct timeval t1, t2;
double duration;
- size_t n_bytes;
- size_t *n_matches;
- /* Check number of arguments */
- if (argc < 4)
- {
- printf("Usage: %s approximation_factor "
- "dna_database pattern1 pattern2 ...\n",
- argv[0]);
- return 1;
- }
+ // result matches
+ int *n_matches;
- approx_factor = atoi(argv[1]);/* Get the distance factor */
- filename = argv[2];/* Grab the filename containing the target text */
- nb_patterns = argc - 3;/* Get the number of patterns to search for */
- pattern = malloc(nb_patterns * sizeof(char *));
- __checkErrors(pattern == NULL, "Unable to allocate array "
- "of pattern of size %ld\n", nb_patterns);
+ // file
+ struct openfile file;
- for (i = 0; i < nb_patterns; i++) /* Grab the patterns */
+ // Arguments parsing
{
- int l;
- l = strlen(argv[i + 3]);
- __checkErrors(l <= 0, "Error while parsing argument %ld\n", i + 3);
+ if (argc < 4)
+ {
+ printf("Usage: %s approximation_factor "
+ "dna_database pattern1 pattern2 ...\n",
+ argv[0]);
+ return 1;
+ }
- pattern[i] = (char *)malloc((l + 1) * sizeof(char));
- __checkErrors(pattern[i] == NULL, "Unable to allocate string "
- "of size %d\n", l);
+ // reading argument
+ approx_factor = atoi(argv[1]);/* Get the distance factor */
+ filename = argv[2];/* Grab the filename containing the target text */
+ nb_patterns = argc - 3;/* Get the number of patterns to search for */
+ pattern = malloc(nb_patterns * sizeof(char *));
+ __checkErrors(pattern == NULL, "Unable to allocate array "
+ "of pattern of size %ld\n", nb_patterns);
+
+ // copying patterns
+ for (int i = 0; i < nb_patterns; i++) /* Grab the patterns */
+ {
+ int l = strlen(argv[i + 3]);
+ __checkErrors(l <= 0, "Error while parsing argument %ld\n", i + 3);
+
+ pattern[i] = (char *) malloc((l + 1) * sizeof(char));
+ __checkErrors(pattern[i] == NULL, "Unable to allocate string "
+ "of size %d\n", l);
+
+ strncpy(pattern[i], argv[i + 3], (l + 1));
+ }
+ }
- strncpy(pattern[i], argv[i + 3], (l + 1));
+ // results allocations
+ {
+ n_matches = malloc(nb_patterns * sizeof(size_t));
+ __checkErrors(n_matches == NULL, "Error: unable to "
+ "allocate memory for %ldB\n", nb_patterns * sizeof(int));
}
printf("Approximate Pattern Mathing: "
- "looking for %ld pattern(s) in file %s w/ distance of %d\n",
+ "looking for %d pattern(s) in file %s w/ distance of %d\n",
nb_patterns, filename, approx_factor);
- buf = read_input_file(filename, &n_bytes);
- __checkErrors(buf == NULL, "Error: NULL pointer from reading input file.");
-
- n_matches = malloc(nb_patterns * sizeof(size_t));/*Alloc the matches*/
- __checkErrors(n_matches == NULL, "Error: unable to "
- "allocate memory for %ldB\n", nb_patterns * sizeof(int));
-
/*****
* BEGIN MAIN LOOP
******/
+
/* Timer start */
gettimeofday(&t1, NULL);
- for (i = 0; i < nb_patterns; i++)
+ for (int i = 0; i < nb_patterns; i++)
{
+ // file initialisation
+ open_input_file(&file, filename);
+
size_t size_pattern = strlen(pattern[i]);
unsigned int *column;
n_matches[i] = 0;
@@ -180,27 +214,31 @@ main(int argc, char **argv)
"allocate memory for column (%ldB)\n",
(size_pattern + 1) * sizeof(unsigned int));
- for (j = 0; j < n_bytes - size_pattern; j++)
+ while (read_input_file(&file, size_pattern))
{
- int distance = 0;
+ for (unsigned int j = 0; j < file.buf_size - size_pattern; j++)
+ {
+ int distance = 0;
#if APM_DEBUG
- if (j % (n_bytes / 100) == 0)
- {
- printf("Procesing byte %ld (out of %ld)(%ld%%)\n",
- j, n_bytes, j / (n_bytes / 100));
- }
+ if (j % (file.size / 100) == 0)
+ {
+ printf("Procesing byte %d (out of %ld)(%ld%%)\n",
+ j, file.size, j / (file.size / 100));
+ }
#endif
+ distance = levenshtein(pattern[i], &file.buf[j],
+ size_pattern, column);
- distance = levenshtein(pattern[i], &buf[j], size_pattern, column);
-
- if (distance <= approx_factor)
- {
- n_matches[i]++;
+ if (distance <= approx_factor)
+ {
+ n_matches[i]++;
+ }
}
}
+ close_input_file(&file);
free(column);
}
@@ -213,9 +251,9 @@ main(int argc, char **argv)
* END MAIN LOOP
******/
- for (i = 0; i < nb_patterns; i++)
+ for (int i = 0; i < nb_patterns; i++)
{
- printf("Number of matches for pattern <%s>: %ld\n",
+ printf("Number of matches for pattern <%s>: %d\n",
pattern[i], n_matches[i]);
}
diff --git a/Projet/CODE/apm/src/apm_omp.c b/Projet/CODE/apm/src/apm_omp.c
index 4cc6499fadbf020d53215653c18b25df513a03f2..0893108d9275618a1cb3211e483a526ab4902b30 100644
--- a/Projet/CODE/apm/src/apm_omp.c
+++ b/Projet/CODE/apm/src/apm_omp.c
@@ -15,6 +15,11 @@
#include "omp.h"
#define APM_DEBUG 0
+#define MAX_BUFFER_SIZE 1024 * 1024 * 1024 // 1Go per reads at most
+//#define MAX_BUFFER_SIZE 512 // debug
+
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+#define MIN3(a, b, c) ((a)<(b) ? ((a)<(c) ? (a) : (c)) : ((b)<(c) ? (b) : (c)))
// can't be inline as it is a variadic function
void
@@ -32,55 +37,72 @@ __checkErrors(const int isBad, const char *fmt, ...)
}
}
+struct openfile
+{
+ int fd; // file descriptor
+ size_t size; // size of the file
+ size_t buf_start; // index of the start of the buffer in the file
+ char *buf; // buffer being read
+ unsigned int buf_alloc_size; // allocateed size of the buffer
+ unsigned int buf_size; // size of the datas in the buffer
+};
-char *
-read_input_file(char *filename, size_t *size)
+void
+open_input_file(struct openfile *file, char *filename)
{
- char *buf;
- struct stat fs;
- size_t fsize;
- int fd = 0;
- size_t read_bytes = 0;
- size_t total_bytes = 0;
- /* Open the text file */
- fd = open(filename, O_RDONLY);
- __checkErrors(fd == -1, "Unable to open the text file <%s>\n", filename);
+ // openning the file
+ file->fd = open(filename, O_RDONLY);
+ __checkErrors(file->fd < 0, "Unable to open the text file <%s>\n", filename);
- /* Get the number of characters in the textfile */
- int stat = fstat(fd, &fs);
- __checkErrors(stat < 0, "Stat of file <%s> failed\n", filename);
- fsize = fs.st_size;
+ // size of the file
+ struct stat fs;
+ int res = fstat(file->fd, &fs);
+ __checkErrors(res < 0, "Stat of file <%s> failed\n", filename);
+ file->size = fs.st_size;
+
+ // buffer allocation
+ file->buf_alloc_size = MIN(file->size, MAX_BUFFER_SIZE);
+ file->buf = (char *) malloc(file->buf_alloc_size * sizeof(char));
+ __checkErrors(file->buf == NULL,
+ "Unable to allocate %ld byte(s) for main array\n",
+ file->buf_alloc_size);
+
+ file->buf_start = 0;
+ file->buf_size = 0;
+}
-#if APM_DEBUG
- printf("File length: %ld\n", fsize);
-#endif
+int
+read_input_file(struct openfile *file, size_t recovering)
+{
+ // how mutch do we realy go back
+ size_t go_back = MIN(file->buf_start, recovering);
- /* Allocate data to copy the target text */
- buf = malloc(fsize * sizeof(char));
- __checkErrors(buf == NULL,
- "Unable to allocate %ld byte(s) for main array\n", fsize);
+ // we read the min between the max size of the buffer
+ // and what is left in the file
+ int to_read = MIN(file->buf_alloc_size - go_back,
+ file->size - file->buf_start);
- do
- {
- read_bytes = read(fd, buf, fsize - total_bytes);
- total_bytes += read_bytes;
- }
- while (read_bytes != 0);
+ // we copy the recovery from the end of the buffer
+ memcpy(file->buf, file->buf + file->buf_size - go_back, go_back);
- __checkErrors(total_bytes != fsize, "Unable to copy %ld byte(s) "
- "from text file (%ld byte(s) copied)\n", fsize, total_bytes);
+ // we read the file
+ int read_size = read(file->fd, file->buf + go_back, to_read);
+ __checkErrors(read_size != to_read, "Unable to copy %ld byte(s) "
+ "from text file (%ld byte(s) copied)\n", to_read, read_size);
+ file->buf_start += read_size;
+ file->buf_size = go_back + read_size;
-#if APM_DEBUG
- printf("Number of read bytes: %ld\n", total_bytes);
-#endif
-
- *size = total_bytes;
- close(fd);
- return buf;
+ return read_size;
}
-#define MIN3(a, b, c) ((a)<(b) ? ((a)<(c) ? (a) : (c)) : ((b)<(c) ? (b) : (c)))
+void
+close_input_file(struct openfile *file)
+{
+ int res = close(file->fd);
+ __checkErrors(res < 0, "Faile to close file descriptor (%d)\n.", file->fd);
+ free(file->buf);
+}
int
levenshtein(char *s1, char *s2, unsigned int len, unsigned int *column)
@@ -114,105 +136,122 @@ levenshtein(char *s1, char *s2, unsigned int len, unsigned int *column)
int
main(int argc, char **argv)
{
- char **pattern;
- char *filename;
+ // arguemnts
int approx_factor = 0;
- size_t nb_patterns = 0;
- size_t i, j;
- char *buf;
+ char *filename;
+ int nb_patterns = 0;
+ char **pattern;
+
+ // time interval and duration for time measurement
struct timeval t1, t2;
double duration;
- size_t n_bytes;
- size_t *n_matches;
- /* Check number of arguments */
- if (argc < 4)
- {
- printf("Usage: %s approximation_factor "
- "dna_database pattern1 pattern2 ...\n",
- argv[0]);
- return 1;
- }
+ // result matches
+ int *n_matches;
- approx_factor = atoi(argv[1]);/* Get the distance factor */
- filename = argv[2];/* Grab the filename containing the target text */
- nb_patterns = argc - 3;/* Get the number of patterns to search for */
- pattern = malloc(nb_patterns * sizeof(char *));
- __checkErrors(pattern == NULL, "Unable to allocate array "
- "of pattern of size %ld\n", nb_patterns);
+ // file
+ struct openfile file;
- for (i = 0; i < nb_patterns; i++) /* Grab the patterns */
+ // Arguments parsing
{
- int l;
- l = strlen(argv[i + 3]);
- __checkErrors(l <= 0, "Error while parsing argument %ld\n", i + 3);
+ if (argc < 4)
+ {
+ printf("Usage: %s approximation_factor "
+ "dna_database pattern1 pattern2 ...\n",
+ argv[0]);
+ return 1;
+ }
+
+ // reading argument
+ approx_factor = atoi(argv[1]);/* Get the distance factor */
+ filename = argv[2];/* Grab the filename containing the target text */
+ nb_patterns = argc - 3;/* Get the number of patterns to search for */
+ pattern = malloc(nb_patterns * sizeof(char *));
+ __checkErrors(pattern == NULL, "Unable to allocate array "
+ "of pattern of size %ld\n", nb_patterns);
+
+ // copying patterns
+ for (int i = 0; i < nb_patterns; i++) /* Grab the patterns */
+ {
+ int l = strlen(argv[i + 3]);
+ __checkErrors(l <= 0, "Error while parsing argument %ld\n", i + 3);
+
+ pattern[i] = (char *) malloc((l + 1) * sizeof(char));
+ __checkErrors(pattern[i] == NULL, "Unable to allocate string "
+ "of size %d\n", l);
- pattern[i] = (char *)malloc((l + 1) * sizeof(char));
- __checkErrors(pattern[i] == NULL, "Unable to allocate string "
- "of size %d\n", l);
+ strncpy(pattern[i], argv[i + 3], (l + 1));
+ }
+ }
- strncpy(pattern[i], argv[i + 3], (l + 1));
+ // results allocations
+ {
+ n_matches = malloc(nb_patterns * sizeof(size_t));
+ __checkErrors(n_matches == NULL, "Error: unable to "
+ "allocate memory for %ldB\n", nb_patterns * sizeof(int));
}
printf("Approximate Pattern Mathing: "
- "looking for %ld pattern(s) in file %s w/ distance of %d\n",
+ "looking for %d pattern(s) in file %s w/ distance of %d\n",
nb_patterns, filename, approx_factor);
- buf = read_input_file(filename, &n_bytes);
- __checkErrors(buf == NULL, "Error: NULL pointer from reading input file.");
-
- n_matches = malloc(nb_patterns * sizeof(size_t));/*Alloc the matches*/
- __checkErrors(n_matches == NULL, "Error: unable to "
- "allocate memory for %ldB\n", nb_patterns * sizeof(int));
-
/*****
* BEGIN MAIN LOOP
******/
+
/* Timer start */
gettimeofday(&t1, NULL);
- for (i = 0; i < nb_patterns; i++)
+ for (int i = 0; i < nb_patterns; i++)
{
+ // file initialisation
+ open_input_file(&file, filename);
+
size_t size_pattern = strlen(pattern[i]);
n_matches[i] = 0;
- int matches = 0;
- #pragma omp parallel reduction(+:matches)
+ while (read_input_file(&file, size_pattern))
{
- unsigned int *column;
- column = malloc((size_pattern + 1) * sizeof(unsigned int));
- __checkErrors(column == NULL,
- "Error: unable to allocate memory for column (%ldB)\n",
- (size_pattern + 1) * sizeof(unsigned int));
+ int matches = 0;
+ #pragma omp parallel reduction(+:matches)
+ {
+ unsigned int *column;
+ column = malloc((size_pattern + 1) * sizeof(unsigned int));
+ __checkErrors(column == NULL,
+ "Error: unable to allocate memory for column (%ldB)\n",
+ (size_pattern + 1) * sizeof(unsigned int));
- #pragma omp for schedule(dynamic, 40960)
+ #pragma omp for schedule(dynamic, 40960)
- for (j = 0; j < n_bytes - size_pattern; j++)
- {
- int distance = 0;
+ for (unsigned int j = 0; j < file.buf_size - size_pattern; j++)
+ {
+ int distance = 0;
#if APM_DEBUG
- if (j % (n_bytes / 100) == 0)
- {
- printf("Procesing byte %ld (out of %ld)(%ld%%)\n",
- j, n_bytes, j / (n_bytes / 100));
- }
+ if (j % (file.size / 100) == 0)
+ {
+ printf("Procesing byte %d (out of %ld)(%ld%%)\n",
+ j, file.size, j / (file.size / 100));
+ }
#endif
- distance = levenshtein(pattern[i], &buf[j],
- size_pattern, column);
+ distance = levenshtein(pattern[i], &file.buf[j],
+ size_pattern, column);
- if (distance <= approx_factor)
- {
- matches++;
+ if (distance <= approx_factor)
+ {
+ matches++;
+ }
}
- }
- free(column);
+ free(column);
+ }
+ n_matches[i] += matches;
}
- n_matches[i] = matches;
+
+ close_input_file(&file);
}
/* Timer stop */
@@ -224,9 +263,9 @@ main(int argc, char **argv)
* END MAIN LOOP
******/
- for (i = 0; i < nb_patterns; i++)
+ for (int i = 0; i < nb_patterns; i++)
{
- printf("Number of matches for pattern <%s>: %ld\n",
+ printf("Number of matches for pattern <%s>: %d\n",
pattern[i], n_matches[i]);
}