int main() {
test_setup();
perf_start();
for (int i = 0; i < NUM_ITER; ++i) {
test_clear();
reset_timer();
start_timer();
test_run(i);
stop_timer();
samples[i] = get_time();
}
perf_stop();
int check = test_check();
printf("Correct: %d\n", check);
for (int i = 0; i < NUM_ITER; ++i)
printf("TS[%d]: %d\n", i, samples[i]);
perf_print_all();
return 0;
}
int main(void)
{
int i;
unsigned char key1[16], key2[16], tinit[16];
unsigned char *pt, *ct;
printf("aes_xts_128_enc_perf:\n");
pt = malloc(TEST_LEN);
ct = malloc(TEST_LEN);
if (NULL == pt || NULL == ct) {
printf("malloc of testsize failed\n");
return -1;
}
mk_rand_data(key1, key2, tinit, pt, TEST_LEN);
XTS_AES_128_enc(key2, key1, tinit, TEST_LEN, pt, ct);
struct perf start, stop;
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++) {
XTS_AES_128_enc(key2, key1, tinit, TEST_LEN, pt, ct);
}
perf_stop(&stop);
printf("aes_xts_128_enc" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * i);
return 0;
}
int main(int argc, char *argv[])
{
int i;
void *buf;
uint16_t crc;
struct perf start, stop;
printf("crc16_t10dif_perf:\n");
if (posix_memalign(&buf, 1024, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
printf("Start timed tests\n");
fflush(0);
memset(buf, 0, TEST_LEN);
crc = crc16_t10dif(TEST_SEED, buf, TEST_LEN);
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++) {
crc = crc16_t10dif(TEST_SEED, buf, TEST_LEN);
}
perf_stop(&stop);
printf("crc16_t10dif" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * i);
printf("finish 0x%x\n", crc);
return 0;
}
int mac_load(char *name, MFDB *out, GEM_WINDOW *wprog)
{
FILE *stream ;
size_t taille ;
void *in ;
perf_start( PERF_DISK, &PerfInfo ) ;
stream = fopen(name, "rb") ;
if (stream == NULL) return(-1) ;
fseek(stream, 0x280, SEEK_SET) ;
out->fd_nplanes = 1 ;
out->fd_w = 576 ;
out->fd_h = 720 ;
out->fd_stand = 0 ;
out->fd_wdwidth = out->fd_w/16 ;
out->fd_addr = img_alloc(576, 720, 1) ;
if (out->fd_addr == NULL)
{
fclose(stream) ;
return(-3) ;
}
taille = file_size(name)-sizeof(MACPAINT_HEADER) ;
in = malloc(taille) ;
if (in == NULL)
{
fclose(stream) ;
free(out->fd_addr) ;
return(-3) ;
}
fread(in, taille, 1, stream) ;
fclose(stream) ;
perf_stop( PERF_DISK, &PerfInfo ) ;
perf_start( PERF_COMPRESS, &PerfInfo ) ;
packbits_dcmps(in, out->fd_addr, out->fd_w, out->fd_h, out->fd_nplanes) ;
perf_stop( PERF_COMPRESS, &PerfInfo ) ;
free(in) ;
return(0) ;
}
void
perf_pop (void)
{
struct perf *prev;
struct perf *cur;
prev = get_perf (-2);
cur = get_perf (-1);
ASSERT (cur);
perf_stop (cur);
if (prev)
perf_resume (prev);
pop_perf_index ();
}
static void
eval_repeat(heim_dict_t o)
{
heim_object_t or = heim_dict_get_value(o, HSTR("value"));
heim_number_t n = heim_dict_get_value(o, HSTR("num"));
int i, num;
struct perf perf;
perf_start(&perf);
heim_assert(or != NULL, "value missing");
heim_assert(n != NULL, "num missing");
num = heim_number_get_int(n);
heim_assert(num >= 0, "num >= 0");
for (i = 0; i < num; i++)
eval_object(or);
perf_stop(&perf);
}
int main(int argc, char *argv[])
{
int i;
u8 *buff1, *buff2, gf_const_tbl[64], a = 2;
struct perf start, stop;
printf("gf_vect_mul_perf:\n");
mk_gf_field();
gf_vect_mul_init(a, gf_const_tbl);
// Allocate large mem region
buff1 = (u8*) malloc(TEST_LEN);
buff2 = (u8*) malloc(TEST_LEN);
if (NULL == buff1 || NULL == buff2){
printf("Failed to allocate %dB\n", TEST_LEN);
return 1;
}
memset(buff1, 0, TEST_LEN);
memset(buff2, 0, TEST_LEN);
gf_vect_mul(TEST_LEN, gf_const_tbl, buff1, buff2);
printf("Start timed tests\n");
fflush(0);
gf_vect_mul(TEST_LEN, gf_const_tbl, buff1, buff2);
perf_start(&start);
for(i=0; i<TEST_LOOPS; i++){
gf_vect_mul_init(a, gf_const_tbl);
gf_vect_mul(TEST_LEN, gf_const_tbl, buff1, buff2);
}
perf_stop(&stop);
printf("gf_vect_mul" TEST_TYPE_STR ": ");
perf_print(stop,start,(long long)TEST_LEN*i);
return 0;
}
int main(int argc, char *argv[])
{
int i, j;
void *buf;
u8 g[TEST_SOURCES], g_tbls[TEST_SOURCES * 32], *dest, *dest_ref;
u8 *temp_buff, *buffs[TEST_SOURCES];
struct perf start, stop;
printf(xstr(FUNCTION_UNDER_TEST) ": %dx%d\n", TEST_SOURCES, TEST_LEN);
// Allocate the arrays
for (i = 0; i < TEST_SOURCES; i++) {
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
buffs[i] = buf;
}
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
temp_buff = buf;
// Performance test
for (i = 0; i < TEST_SOURCES; i++)
for (j = 0; j < TEST_LEN; j++)
buffs[i][j] = rand();
memset(dest, 0, TEST_LEN);
memset(temp_buff, 0, TEST_LEN);
memset(dest_ref, 0, TEST_LEN);
memset(g, 0, TEST_SOURCES);
for (i = 0; i < TEST_SOURCES; i++)
g[i] = rand();
for (j = 0; j < TEST_SOURCES; j++)
gf_vect_mul_init(g[j], &g_tbls[j * 32]);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref);
#ifdef DO_REF_PERF
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++) {
for (j = 0; j < TEST_SOURCES; j++)
gf_vect_mul_init(g[j], &g_tbls[j * 32]);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref);
}
perf_stop(&stop);
printf("gf_vect_dot_prod_base" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 1) * i);
#endif
FUNCTION_UNDER_TEST(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest);
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++) {
for (j = 0; j < TEST_SOURCES; j++)
gf_vect_mul_init(g[j], &g_tbls[j * 32]);
FUNCTION_UNDER_TEST(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest);
}
perf_stop(&stop);
printf(xstr(FUNCTION_UNDER_TEST) TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 1) * i);
if (0 != memcmp(dest_ref, dest, TEST_LEN)) {
printf("Fail zero " xstr(FUNCTION_UNDER_TEST) " test\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref, 25);
printf("dprod:");
dump(dest, 25);
return -1;
}
printf("pass perf check\n");
return 0;
}
int main(int argc, char *argv[])
{
int ret;
// For a list of valid performance events, see
// libpfm/examples/showevtinfo -L
// or
// Intel 64 and IA-32 Architectures Software Developer’s Manual,
// Volume 3B: System Programming Guide, Part 2
// Initialize perfmon -------------------------------------------------
char *events[] = {
// on Haswell, can use at most 4, otherwise '0' results
"PERF_COUNT_HW_CPU_CYCLES",
"CACHE-REFERENCES",
"CACHE-MISSES",
"DTLB-LOAD-MISSES",
NULL
};
struct perf_data *data;
ret = perf_init(events, &data);
if (ret < 0) {
fprintf(stderr, "Could not initialize perfmon.\n");
perf_cleanup(data);
return 0;
}
// Example 1: Measure performance for MMM in this file. ---------------
perf_start(data);
for (int i=0; i<N; ++i) {
for (int j=0; j<N; ++j) {
for (int k=0; k<N; ++k) {
COST_INC_ADD(1); COST_INC_MUL(1);
C[i*N+j] = A[i*N+k] * B[k*N+j] + C[i*N+j];
}
}
}
ret = perf_stop(data);
if (ret < 0) {
fprintf(stderr, "Error measuring performance.\n");
perf_cleanup(data);
return 0;
}
// short cycles check
perf_update_values(data);
printf("Cycles after example 1: %llu\n", data[0].value);
// Example 2: Measure external computation ----------------------------
perf_start(data); // if desired, do perf_reset() first
other_computation();
perf_stop(data);
// Report results -----------------------------------------------------
perf_update_values(data);
printf("Overall results:\n");
printf("Performance Counters:\n");
for (struct perf_data *iter = data; iter->name; ++iter) {
printf(" %-50s : %"PRId64"\n", iter->name, iter->value);
}
printf("Cost Model:\n");
printf(" Add: %"PRI_COST"\n", COST_ADD);
printf(" Mul: %"PRI_COST"\n", COST_MUL);
printf(" Div: %"PRI_COST"\n", COST_DIV);
printf(" Pow: %"PRI_COST"\n", COST_POW);
printf(" Abs: %"PRI_COST"\n", COST_ABS);
printf("Derived Results\n");
double flops = COST_ADD + COST_MUL + COST_DIV;
double cycles = data[0].value; // index as in 'events'
double cache_load = data[1].value;
double cache_miss = data[2].value;
printf(" Flops : %.0lf\n", flops);
printf(" Performance : %.3lf\n", flops/cycles);
printf(" Cache Miss Rate : %.3lf\n", cache_miss/cache_load);
// Cleanup ------------------------------------------------------------
perf_cleanup(data);
return 0;
}
int main(int argc, char *argv[])
{
int i,j;
void *buf;
u8 g1[TEST_SOURCES], g2[TEST_SOURCES], g3[TEST_SOURCES];
u8 g_tbls[3*TEST_SOURCES*32], *dest_ptrs[3], *buffs[TEST_SOURCES];
u8 *dest1, *dest2, *dest3, *dest_ref1, *dest_ref2, *dest_ref3;
struct perf start, stop;
printf("gf_3vect_dot_prod_sse: %dx%d\n", TEST_SOURCES, TEST_LEN);
mk_gf_field();
// Allocate the arrays
for(i=0; i<TEST_SOURCES; i++){
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
buffs[i] = buf;
}
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest1 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest2 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest3 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref1 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref2 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref3 = buf;
dest_ptrs[0] = dest1;
dest_ptrs[1] = dest2;
dest_ptrs[2] = dest3;
// Performance test
for(i=0; i<TEST_SOURCES; i++)
for(j=0; j<TEST_LEN; j++)
buffs[i][j] = rand();
memset(dest1, 0, TEST_LEN);
memset(dest2, 0, TEST_LEN);
memset(dest_ref1, 0, TEST_LEN);
memset(dest_ref2, 0, TEST_LEN);
for (i=0; i<TEST_SOURCES; i++){
g1[i] = rand();
g2[i] = rand();
g3[i] = rand();
}
for(j=0; j<TEST_SOURCES; j++){
gf_vect_mul_init(g1[j], &g_tbls[j*32]);
gf_vect_mul_init(g2[j], &g_tbls[(32*TEST_SOURCES) + (j*32)]);
gf_vect_mul_init(g3[j], &g_tbls[(64*TEST_SOURCES) + (j*32)]);
}
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref1);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[32*TEST_SOURCES], buffs, dest_ref2);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[64*TEST_SOURCES], buffs, dest_ref3);
#ifdef DO_REF_PERF
perf_start(&start);
for (i=0; i<TEST_LOOPS/100; i++){
for (j=0; j<TEST_SOURCES; j++){
gf_vect_mul_init(g1[j], &g_tbls[j*32]);
gf_vect_mul_init(g2[j], &g_tbls[(32*TEST_SOURCES) + (j*32)]);
gf_vect_mul_init(g3[j], &g_tbls[(64*TEST_SOURCES) + (j*32)]);
}
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref1);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[32*TEST_SOURCES], buffs, dest_ref2);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[64*TEST_SOURCES], buffs, dest_ref3);
}
perf_stop(&stop);
printf("gf_3vect_dot_prod_base" TEST_TYPE_STR ": ");
perf_print(stop,start,(long long)TEST_LEN*(TEST_SOURCES+3)*i);
#endif
gf_3vect_dot_prod_sse(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest_ptrs);
perf_start(&start);
for (i=0; i<TEST_LOOPS; i++) {
for (j=0; j<TEST_SOURCES; j++){
gf_vect_mul_init(g1[j], &g_tbls[j*32]);
gf_vect_mul_init(g2[j], &g_tbls[(32*TEST_SOURCES) + (j*32)]);
gf_vect_mul_init(g3[j], &g_tbls[(64*TEST_SOURCES) + (j*32)]);
}
gf_3vect_dot_prod_sse(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest_ptrs);
}
perf_stop(&stop);
printf("gf_3vect_dot_prod_sse" TEST_TYPE_STR ": ");
perf_print(stop,start, (long long)TEST_LEN*(TEST_SOURCES+3)*i);
if (0 != memcmp(dest_ref1, dest1, TEST_LEN)){
printf("Fail perf vect_dot_prod_sse test1\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref1, 25);
printf("dprod_sse:");
dump(dest1, 25);
return -1;
}
if (0 != memcmp(dest_ref2, dest2, TEST_LEN)){
printf("Fail perf vect_dot_prod_sse test2\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref2, 25);
printf("dprod_sse:");
dump(dest2, 25);
return -1;
}
if (0 != memcmp(dest_ref3, dest3, TEST_LEN)){
printf("Fail perf vect_dot_prod_sse test3\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref3, 25);
printf("dprod_sse:");
dump(dest3, 25);
return -1;
}
printf("pass perf check\n");
return 0;
}
int main(int argc, char *argv[])
{
int i, j, rtest, m, k, nerrs, r, err;
void *buf;
u8 *temp_buffs[TEST_SOURCES], *buffs[TEST_SOURCES];
u8 a[MMAX*KMAX], b[MMAX*KMAX], c[MMAX*KMAX], d[MMAX*KMAX];
u8 g_tbls[KMAX*TEST_SOURCES*32], src_in_err[TEST_SOURCES];
u8 src_err_list[TEST_SOURCES], *recov[TEST_SOURCES];
struct perf start, stop;
m = 32;
k = 28;
printf("erasure_code_sse_perf: %dx%d ",
m, (TEST_LEN(m)));
// Allocate the arrays
for(i=0; i<TEST_SOURCES; i++) {
if (posix_memalign(&buf, 64, TEST_LEN(m))) {
printf("alloc error: Fail");
return -1;
}
buffs[i] = buf;
}
for (i=0; i<TEST_SOURCES; i++) {
if (posix_memalign(&buf, 64, TEST_LEN(m))) {
printf("alloc error: Fail");
return -1;
}
temp_buffs[i] = buf;
}
// Test erasure code by encode and recovery
// Pick a first test
if (m > MMAX || k > KMAX)
return -1;
// Make random data
for(i=0; i<k; i++)
for(j=0; j<(TEST_LEN(m)); j++)
buffs[i][j] = rand();
memset(src_in_err, 0, TEST_SOURCES);
srand(1);
for (i=0, nerrs=0; i<k && nerrs<m-k; i++) {
err = 1 & rand();
src_in_err[i] = err;
if (err)
src_err_list[nerrs++] = i;
}
if (nerrs == 0) { // should have at least one error
while ((err = (rand() % KMAX)) >= k) ;
src_err_list[nerrs++] = err;
src_in_err[err] = 1;
}
printf("Test erase list = ");
for (i=0; i<nerrs; i++)
printf(" %d", src_err_list[i]);
printf("\n");
perf_start(&start);
for (rtest = 0; rtest < TEST_LOOPS(m); rtest++) {
gf_gen_rs_matrix(a, m, k);
// Make parity vects
ec_init_tables(k, m-k, &a[k*k], g_tbls);
ec_encode_data_sse((TEST_LEN(m)),
k, m-k, g_tbls, buffs, &buffs[k]);
}
perf_stop(&stop);
printf("erasure_code_sse_encode" TEST_TYPE_STR ": ");
perf_print(stop,start,
(long long)(TEST_LEN(m))*(m)*rtest);
perf_start(&start);
for (rtest = 0; rtest < TEST_LOOPS(m); rtest++) {
// Construct b by removing error rows
for(i=0, r=0; i<k; i++, r++) {
while (src_in_err[r])
r++;
for(j=0; j<k; j++)
b[k*i+j] = a[k*r+j];
}
if (gf_invert_matrix(b, d, k) < 0) {
printf("BAD MATRIX\n");
return -1;
}
for(i=0, r=0; i<k; i++, r++) {
while (src_in_err[r])
r++;
recov[i] = buffs[r];
}
for(i=0; i<nerrs; i++) {
for(j=0; j<k; j++) {
c[k*i+j]=d[k*src_err_list[i]+j];
}
}
// Recover data
ec_init_tables(k, nerrs, c, g_tbls);
ec_encode_data_sse((TEST_LEN(m)),
k, nerrs, g_tbls, recov, &temp_buffs[k]);
}
perf_stop(&stop);
for(i=0; i<nerrs; i++) {
if (0 != memcmp(temp_buffs[k+i], buffs[src_err_list[i]],
(TEST_LEN(m)))) {
printf("Fail error recovery (%d, %d, %d) - ",
m, k, nerrs);
printf(" - erase list = ");
for (j=0; j<nerrs; j++)
printf(" %d", src_err_list[j]);
printf("\na:\n");
dump_u8xu8((u8*)a, m, k);
printf("inv b:\n");
dump_u8xu8((u8*)d, k, k);
printf("orig data:\n");
dump_matrix(buffs, m, 25);
printf("orig :");
dump(buffs[src_err_list[i]],25);
printf("recov %d:",src_err_list[i]);
dump(temp_buffs[k+i], 25);
return -1;
}
}
printf("erasure_code_sse_decode" TEST_TYPE_STR ": ");
perf_print(stop,start,
(long long)(TEST_LEN(m))*(k+nerrs)*rtest);
printf("done all: Pass\n");
return 0;
}
int main(void)
{
int i, j, k;
u8 s, vec[TEST_SOURCES], dest1[TEST_LEN], dest2[TEST_LEN];
u8 *matrix[TEST_SOURCES];
struct perf start, stop;
mk_gf_field();
mk_gf_mul_table(gf_mul_table);
//generate random vector and matrix/data
for (i = 0; i < TEST_SOURCES; i++) {
vec[i] = rand();
if (!(matrix[i] = malloc(TEST_LEN))) {
fprintf(stderr, "Error failure\n\n");
return -1;
}
for (j = 0; j < TEST_LEN; j++)
matrix[i][j] = rand();
}
gf_vect_dot_prod_ref(TEST_LEN, TEST_SOURCES, vec, matrix, dest1);
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++)
gf_vect_dot_prod_ref(TEST_LEN, TEST_SOURCES, vec, matrix, dest1);
perf_stop(&stop);
printf("gf_vect_dot_prod_2tbl" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 1) * i);
// Warm up mult tables
for (i = 0; i < TEST_LEN; i++) {
s = 0;
for (j = 0; j < TEST_SOURCES; j++) {
s ^= gf_mul_table[vec[j] * 256 + matrix[j][i]];
}
dest2[i] = s;
}
perf_start(&start);
for (k = 0; k < TEST_LOOPS; k++) {
for (i = 0; i < TEST_LEN; i++) {
s = 0;
for (j = 0; j < TEST_SOURCES; j++) {
s ^= gf_mul_table[vec[j] * 256 + matrix[j][i]];
}
dest2[i] = s;
}
}
perf_stop(&stop);
printf("gf_vect_dot_prod_1tbl" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 1) * k);
// Compare with reference function
if (0 != memcmp(dest1, dest2, TEST_LEN)) {
printf("Error, different results!\n\n");
return -1;
}
printf("Pass functional test\n");
return 0;
}
int main(int argc, char *argv[])
{
int i, j;
void *buf;
u8 g1[TEST_SOURCES], g2[TEST_SOURCES], g3[TEST_SOURCES];
u8 g4[TEST_SOURCES], g5[TEST_SOURCES], g6[TEST_SOURCES], *g_tbls;
u8 *dest1, *dest2, *dest3, *dest4, *dest5, *dest6, *dest_ref1;
u8 *dest_ref2, *dest_ref3, *dest_ref4, *dest_ref5, *dest_ref6;
u8 *dest_ptrs[6], *buffs[TEST_SOURCES];
struct perf start, stop;
printf(xstr(FUNCTION_UNDER_TEST) ": %dx%d\n", TEST_SOURCES, TEST_LEN);
// Allocate the arrays
for (i = 0; i < TEST_SOURCES; i++) {
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
buffs[i] = buf;
}
if (posix_memalign(&buf, 16, 6 * TEST_SOURCES * 32)) {
printf("alloc error: Fail");
return -1;
}
g_tbls = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest1 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest2 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest3 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest4 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest5 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest6 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref1 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref2 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref3 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref4 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref5 = buf;
if (posix_memalign(&buf, 64, TEST_LEN)) {
printf("alloc error: Fail");
return -1;
}
dest_ref6 = buf;
dest_ptrs[0] = dest1;
dest_ptrs[1] = dest2;
dest_ptrs[2] = dest3;
dest_ptrs[3] = dest4;
dest_ptrs[4] = dest5;
dest_ptrs[5] = dest6;
// Performance test
for (i = 0; i < TEST_SOURCES; i++)
for (j = 0; j < TEST_LEN; j++)
buffs[i][j] = rand();
memset(dest1, 0, TEST_LEN);
memset(dest2, 0, TEST_LEN);
memset(dest3, 0, TEST_LEN);
memset(dest4, 0, TEST_LEN);
memset(dest5, 0, TEST_LEN);
memset(dest6, 0, TEST_LEN);
memset(dest_ref1, 0, TEST_LEN);
memset(dest_ref2, 0, TEST_LEN);
memset(dest_ref3, 0, TEST_LEN);
memset(dest_ref4, 0, TEST_LEN);
memset(dest_ref5, 0, TEST_LEN);
memset(dest_ref6, 0, TEST_LEN);
for (i = 0; i < TEST_SOURCES; i++) {
g1[i] = rand();
g2[i] = rand();
g3[i] = rand();
g4[i] = rand();
g5[i] = rand();
g6[i] = rand();
}
for (j = 0; j < TEST_SOURCES; j++) {
gf_vect_mul_init(g1[j], &g_tbls[j * 32]);
gf_vect_mul_init(g2[j], &g_tbls[(32 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g3[j], &g_tbls[(64 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g4[j], &g_tbls[(96 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g5[j], &g_tbls[(128 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g6[j], &g_tbls[(160 * TEST_SOURCES) + (j * 32)]);
}
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref1);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[32 * TEST_SOURCES], buffs,
dest_ref2);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[64 * TEST_SOURCES], buffs,
dest_ref3);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[96 * TEST_SOURCES], buffs,
dest_ref4);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[128 * TEST_SOURCES], buffs,
dest_ref5);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[160 * TEST_SOURCES], buffs,
dest_ref6);
#ifdef DO_REF_PERF
perf_start(&start);
for (i = 0; i < TEST_LOOPS / 20; i++) {
for (j = 0; j < TEST_SOURCES; j++) {
gf_vect_mul_init(g1[j], &g_tbls[j * 32]);
gf_vect_mul_init(g2[j], &g_tbls[(32 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g3[j], &g_tbls[(64 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g4[j], &g_tbls[(96 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g5[j], &g_tbls[(128 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g6[j], &g_tbls[(160 * TEST_SOURCES) + (j * 32)]);
}
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[0], buffs, dest_ref1);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[32 * TEST_SOURCES],
buffs, dest_ref2);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[64 * TEST_SOURCES],
buffs, dest_ref3);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[96 * TEST_SOURCES],
buffs, dest_ref4);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[128 * TEST_SOURCES],
buffs, dest_ref5);
gf_vect_dot_prod_base(TEST_LEN, TEST_SOURCES, &g_tbls[160 * TEST_SOURCES],
buffs, dest_ref6);
}
perf_stop(&stop);
printf("gf_6vect_dot_prod_base" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 6) * i);
#endif
FUNCTION_UNDER_TEST(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest_ptrs);
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++) {
for (j = 0; j < TEST_SOURCES; j++) {
gf_vect_mul_init(g1[j], &g_tbls[j * 32]);
gf_vect_mul_init(g2[j], &g_tbls[(32 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g3[j], &g_tbls[(64 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g4[j], &g_tbls[(96 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g5[j], &g_tbls[(128 * TEST_SOURCES) + (j * 32)]);
gf_vect_mul_init(g6[j], &g_tbls[(160 * TEST_SOURCES) + (j * 32)]);
}
FUNCTION_UNDER_TEST(TEST_LEN, TEST_SOURCES, g_tbls, buffs, dest_ptrs);
}
perf_stop(&stop);
printf(xstr(FUNCTION_UNDER_TEST) TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * (TEST_SOURCES + 6) * i);
if (0 != memcmp(dest_ref1, dest1, TEST_LEN)) {
printf("Fail perf " xstr(FUNCTION_UNDER_TEST) " test1\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref1, 25);
printf("dprod_dut:");
dump(dest1, 25);
return -1;
}
if (0 != memcmp(dest_ref2, dest2, TEST_LEN)) {
printf("Fail perf " xstr(FUNCTION_UNDER_TEST) " test2\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref2, 25);
printf("dprod_dut:");
dump(dest2, 25);
return -1;
}
if (0 != memcmp(dest_ref3, dest3, TEST_LEN)) {
printf("Fail perf " xstr(FUNCTION_UNDER_TEST) " test3\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref3, 25);
printf("dprod_dut:");
dump(dest3, 25);
return -1;
}
if (0 != memcmp(dest_ref4, dest4, TEST_LEN)) {
printf("Fail perf " xstr(FUNCTION_UNDER_TEST) " test4\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref4, 25);
printf("dprod_dut:");
dump(dest4, 25);
return -1;
}
if (0 != memcmp(dest_ref5, dest5, TEST_LEN)) {
printf("Fail perf " xstr(FUNCTION_UNDER_TEST) " test5\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref5, 25);
printf("dprod_dut:");
dump(dest5, 25);
return -1;
}
if (0 != memcmp(dest_ref6, dest6, TEST_LEN)) {
printf("Fail perf " xstr(FUNCTION_UNDER_TEST) " test6\n");
dump_matrix(buffs, 5, TEST_SOURCES);
printf("dprod_base:");
dump(dest_ref6, 25);
printf("dprod_dut:");
dump(dest6, 25);
return -1;
}
printf("pass perf check\n");
return 0;
}
int main(int argc, char *argv[])
{
int i, j, rtest, m, k, nerrs, r;
void *buf;
u8 *temp_buffs[TEST_SOURCES], *buffs[TEST_SOURCES];
u8 a[MMAX * KMAX], b[MMAX * KMAX], c[MMAX * KMAX], d[MMAX * KMAX];
u8 g_tbls[KMAX * TEST_SOURCES * 32], src_in_err[TEST_SOURCES];
u8 src_err_list[TEST_SOURCES], *recov[TEST_SOURCES];
struct perf start, stop;
// Pick test parameters
m = 14;
k = 10;
nerrs = 4;
const u8 err_list[] = {2, 4, 5, 7};
printf("erasure_code_base_perf: %dx%d %d\n", m, TEST_LEN(m), nerrs);
if (m > MMAX || k > KMAX || nerrs > (m - k)){
printf(" Input test parameter error\n");
return -1;
}
memcpy(src_err_list, err_list, nerrs);
memset(src_in_err, 0, TEST_SOURCES);
for (i = 0; i < nerrs; i++)
src_in_err[src_err_list[i]] = 1;
// Allocate the arrays
for (i = 0; i < m; i++) {
if (posix_memalign(&buf, 64, TEST_LEN(m))) {
printf("alloc error: Fail\n");
return -1;
}
buffs[i] = buf;
}
for (i = 0; i < (m - k); i++) {
if (posix_memalign(&buf, 64, TEST_LEN(m))) {
printf("alloc error: Fail\n");
return -1;
}
temp_buffs[i] = buf;
}
// Make random data
for (i = 0; i < k; i++)
for (j = 0; j < TEST_LEN(m); j++)
buffs[i][j] = rand();
gf_gen_rs_matrix(a, m, k);
ec_init_tables(k, m - k, &a[k * k], g_tbls);
ec_encode_data_base(TEST_LEN(m), k, m - k, g_tbls, buffs, &buffs[k]);
// Start encode test
perf_start(&start);
for (rtest = 0; rtest < TEST_LOOPS(m); rtest++) {
// Make parity vects
ec_init_tables(k, m - k, &a[k * k], g_tbls);
ec_encode_data_base(TEST_LEN(m), k, m - k, g_tbls, buffs, &buffs[k]);
}
perf_stop(&stop);
printf("erasure_code_base_encode" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)(TEST_LEN(m)) * (m) * rtest);
// Start decode test
perf_start(&start);
for (rtest = 0; rtest < TEST_LOOPS(m); rtest++) {
// Construct b by removing error rows
for (i = 0, r = 0; i < k; i++, r++) {
while (src_in_err[r])
r++;
recov[i] = buffs[r];
for (j = 0; j < k; j++)
b[k * i + j] = a[k * r + j];
}
if (gf_invert_matrix(b, d, k) < 0) {
printf("BAD MATRIX\n");
return -1;
}
for (i = 0; i < nerrs; i++)
for (j = 0; j < k; j++)
c[k * i + j] = d[k * src_err_list[i] + j];
// Recover data
ec_init_tables(k, nerrs, c, g_tbls);
ec_encode_data_base(TEST_LEN(m), k, nerrs, g_tbls, recov, temp_buffs);
}
perf_stop(&stop);
for (i = 0; i < nerrs; i++) {
if (0 != memcmp(temp_buffs[i], buffs[src_err_list[i]], TEST_LEN(m))) {
printf("Fail error recovery (%d, %d, %d) - ", m, k, nerrs);
return -1;
}
}
printf("erasure_code_base_decode" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)(TEST_LEN(m)) * (k + nerrs) * rtest);
printf("done all: Pass\n");
return 0;
}
int main(int argc, char *argv[])
{
FILE *in, *out = NULL;
unsigned char *inbuf, *outbuf;
int i, infile_size, iterations, outbuf_size;
if (argc > 3 || argc < 2) {
fprintf(stderr, "Usage: igzip_sync_flush_file_perf infile [outfile]\n"
"\t - Runs multiple iterations of igzip on a file to get more accurate time results.\n");
exit(0);
}
in = fopen(argv[1], "rb");
if (!in) {
fprintf(stderr, "Can't open %s for reading\n", argv[1]);
exit(0);
}
if (argc > 2) {
out = fopen(argv[2], "wb");
if (!out) {
fprintf(stderr, "Can't open %s for writing\n", argv[2]);
exit(0);
}
printf("outfile=%s\n", argv[2]);
}
printf("Window Size: %d K\n", HIST_SIZE);
printf("igzip_sync_flush_file_perf: \n");
fflush(0);
/* Allocate space for entire input file and
* output (assuming 1:1 max output size)
*/
infile_size = get_filesize(in);
if (infile_size != 0) {
outbuf_size = infile_size;
iterations = RUN_MEM_SIZE / infile_size;
} else {
outbuf_size = BUF_SIZE;
iterations = MIN_TEST_LOOPS;
}
if (iterations < MIN_TEST_LOOPS)
iterations = MIN_TEST_LOOPS;
inbuf = malloc(infile_size);
if (inbuf == NULL) {
fprintf(stderr, "Can't allocate input buffer memory\n");
exit(0);
}
outbuf = malloc(outbuf_size);
if (outbuf == NULL) {
fprintf(stderr, "Can't allocate output buffer memory\n");
exit(0);
}
printf("igzip_sync_flush_file_perf: %s %d iterations\n", argv[1], iterations);
/* Read complete input file into buffer */
stream.avail_in = (uint32_t) fread(inbuf, 1, infile_size, in);
if (stream.avail_in != infile_size) {
fprintf(stderr, "Couldn't fit all of input file into buffer\n");
exit(0);
}
struct perf start, stop;
perf_start(&start);
for (i = 0; i < iterations; i++) {
isal_deflate_init(&stream);
stream.end_of_stream = 0;
stream.flush = SYNC_FLUSH;
stream.next_in = inbuf;
stream.avail_in = infile_size / 2;
stream.next_out = outbuf;
stream.avail_out = outbuf_size / 2;
isal_deflate(&stream);
if (infile_size == 0)
continue;
stream.avail_in = infile_size - infile_size / 2;
stream.end_of_stream = 1;
stream.next_in = inbuf + stream.total_in;
stream.flush = SYNC_FLUSH;
stream.avail_out = infile_size - outbuf_size / 2;
stream.next_out = outbuf + stream.total_out;
isal_deflate(&stream);
if (stream.avail_in != 0)
break;
}
perf_stop(&stop);
if (stream.avail_in != 0) {
fprintf(stderr, "Could not compress all of inbuf\n");
exit(0);
}
printf(" file %s - in_size=%d out_size=%d iter=%d ratio=%3.1f%%\n", argv[1],
infile_size, stream.total_out, i, 100.0 * stream.total_out / infile_size);
printf("igzip_file: ");
perf_print(stop, start, (long long)infile_size * i);
if (argc > 2 && out) {
printf("writing %s\n", argv[2]);
fwrite(outbuf, 1, stream.total_out, out);
fclose(out);
}
fclose(in);
printf("End of igzip_sync_flush_file_perf\n\n");
fflush(0);
return 0;
}
int main(void)
{
int i;
unsigned char key1[16], key2[16], tinit[16];
unsigned char *pt, *ct, *refct;
struct perf start, stop;
unsigned char keyssl[32]; /* SSL takes both keys together */
/* Initialise our cipher context, which can use same input vectors */
EVP_CIPHER_CTX *ctx;
ctx = EVP_CIPHER_CTX_new();
printf("aes_xts_128_enc_perf:\n");
pt = malloc(TEST_LEN);
ct = malloc(TEST_LEN);
refct = malloc(TEST_LEN);
if (NULL == pt || NULL == ct || NULL == refct) {
printf("malloc of testsize failed\n");
return -1;
}
xts128_mk_rand_data(key1, key2, tinit, pt, TEST_LEN);
/* Set up key for the SSL engine */
for (i = 0; i < 16; i++) {
keyssl[i] = key1[i];
keyssl[i + 16] = key2[i];
}
/* Encrypt and compare output */
XTS_AES_128_enc(key2, key1, tinit, TEST_LEN, pt, ct);
openssl_aes_128_xts_enc(ctx, keyssl, tinit, TEST_LEN, pt, refct);
if (memcmp(ct, refct, TEST_LEN)) {
printf("ISA-L and OpenSSL results don't match\n");
return -1;
}
/* Time ISA-L encryption */
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++)
XTS_AES_128_enc(key2, key1, tinit, TEST_LEN, pt, ct);
perf_stop(&stop);
printf("aes_xts_128_enc" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * i);
/* Time OpenSSL encryption */
perf_start(&start);
for (i = 0; i < TEST_LOOPS; i++)
openssl_aes_128_xts_enc(ctx, keyssl, tinit, TEST_LEN, pt, refct);
perf_stop(&stop);
printf("aes_xts_128_openssl_enc" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * i);
EVP_CIPHER_CTX_free(ctx);
return 0;
}
int main(void)
{
MD5_HASH_CTX_MGR *mgr = NULL;
MD5_HASH_CTX ctxpool[TEST_BUFS];
unsigned char *bufs[TEST_BUFS];
uint32_t i, j, t, fail = 0;
struct perf start, stop;
for (i = 0; i < TEST_BUFS; i++) {
bufs[i] = (unsigned char *)calloc((size_t) TEST_LEN, 1);
if (bufs[i] == NULL) {
printf("calloc failed test aborted\n");
return 1;
}
// Init ctx contents
hash_ctx_init(&ctxpool[i]);
ctxpool[i].user_data = (void *)((uint64_t) i);
}
posix_memalign((void *)&mgr, 16, sizeof(MD5_HASH_CTX_MGR));
md5_ctx_mgr_init(mgr);
// Start OpenSSL tests
perf_start(&start);
for (t = 0; t < TEST_LOOPS; t++) {
for (i = 0; i < TEST_BUFS; i++)
MD5(bufs[i], TEST_LEN, digest_ssl[i]);
}
perf_stop(&stop);
printf("md5_openssl" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * i * t);
// Start mb tests
perf_start(&start);
for (t = 0; t < TEST_LOOPS; t++) {
for (i = 0; i < TEST_BUFS; i++)
md5_ctx_mgr_submit(mgr, &ctxpool[i], bufs[i], TEST_LEN, HASH_ENTIRE);
while (md5_ctx_mgr_flush(mgr)) ;
}
perf_stop(&stop);
printf("multibinary_md5" TEST_TYPE_STR ": ");
perf_print(stop, start, (long long)TEST_LEN * i * t);
for (i = 0; i < TEST_BUFS; i++) {
for (j = 0; j < MD5_DIGEST_NWORDS; j++) {
if (ctxpool[i].job.result_digest[j] != ((uint32_t *) digest_ssl[i])[j]) {
fail++;
printf("Test%d, digest%d fail %08X <=> %08X\n",
i, j, ctxpool[i].job.result_digest[j],
((uint32_t *) digest_ssl[i])[j]);
}
}
}
printf("Multi-buffer md5 test complete %d buffers of %d B with "
"%d iterations\n", TEST_BUFS, TEST_LEN, TEST_LOOPS);
if (fail)
printf("Test failed function check %d\n", fail);
else
printf(" multibinary_md5_ossl_perf: Pass\n");
return fail;
}
