vitalif
/
mdtest
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge pull request #166 from hpc/feature-stonewall-perf-report 

Feature stonewall perf report
master
Glenn K. Lockwood 2019-08-02 08:38:37 -05:00 committed by GitHub
parent b686b6c26a df8355a9bc
commit a0c5dcec89
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 153 additions and 120 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
src/ior-output.c
View File

@ -534,10 +534,14 @@ static void PrintLongSummaryOneOperation(IOR_test_t *test, const int access)
reps = params->repetitions;
double * times = malloc(sizeof(double)* reps);
long long stonewall_avg_data_accessed = 0;
double stonewall_time = 0;
for(int i=0; i < reps; i++){
IOR_point_t *point = (access == WRITE) ? &results[i].write :
&results[i].read;
times[i] = point->time;
stonewall_time += point->stonewall_time;
stonewall_avg_data_accessed += point->stonewall_avg_data_accessed;
}
bw = bw_values(reps, results, times, access);
@ -558,6 +562,13 @@ static void PrintLongSummaryOneOperation(IOR_test_t *test, const int access)
fprintf(out_resultfile, "%10.2f ", ops->mean);
fprintf(out_resultfile, "%10.2f ", ops->sd);
fprintf(out_resultfile, "%10.5f ", mean_of_array_of_doubles(times, reps));
if(test->params.stoneWallingWearOut){
fprintf(out_resultfile, "%10.2f ", stonewall_time / reps);
fprintf(out_resultfile, "%13.2f ", stonewall_avg_data_accessed / stonewall_time / MEBIBYTE);
}else{
fprintf(out_resultfile, "%10s ", "NA");
fprintf(out_resultfile, "%13s ", "NA");
}
fprintf(out_resultfile, "%5d ", params->id);
fprintf(out_resultfile, "%6d ", params->numTasks);
fprintf(out_resultfile, "%3d ", params->tasksPerNode);
@ -603,6 +614,10 @@ static void PrintLongSummaryOneOperation(IOR_test_t *test, const int access)
PrintKeyValDouble("OPsMean", ops->mean);
PrintKeyValDouble("OPsSD", ops->sd);
PrintKeyValDouble("MeanTime", mean_of_array_of_doubles(times, reps));
if(test->params.stoneWallingWearOut){
PrintKeyValDouble("StoneWallTime", stonewall_time / reps);
PrintKeyValDouble("StoneWallbwMeanMIB", stonewall_avg_data_accessed / stonewall_time / MEBIBYTE);
}
PrintKeyValDouble("xsizeMiB", (double) point->aggFileSizeForBW / MEBIBYTE);
PrintEndSection();
}else if (outputFormat == OUTPUT_CSV){
@ -635,10 +650,10 @@ void PrintLongSummaryHeader()
}
fprintf(out_resultfile, "\n");
fprintf(out_resultfile, "%-9s %10s %10s %10s %10s %10s %10s %10s %10s %10s",
fprintf(out_resultfile, "%-9s %10s %10s %10s %10s %10s %10s %10s %10s %10s %10s %13s",
"Operation", "Max(MiB)", "Min(MiB)", "Mean(MiB)", "StdDev",
"Max(OPs)", "Min(OPs)", "Mean(OPs)", "StdDev",
"Mean(s)");
"Mean(s)", "Stonewall(s)", "Stonewall(MiB)");
fprintf(out_resultfile, " Test# #Tasks tPN reps fPP reord reordoff reordrand seed"
" segcnt ");
fprintf(out_resultfile, "%8s %8s %9s %5s", " blksiz", "xsize","aggs(MiB)", "API");

4
src/ior.c
View File

@ -1944,10 +1944,6 @@ static IOR_offset_t WriteOrRead(IOR_param_t *test, IOR_results_t *results,
point->stonewall_min_data_accessed /1024.0 / 1024 / 1024, point->stonewall_avg_data_accessed / 1024.0 / 1024 / 1024 / test->numTasks , point->stonewall_time);
point->stonewall_min_data_accessed *= test->numTasks;
}
if(pairs_accessed_min == pairCnt){
point->stonewall_min_data_accessed = 0;
point->stonewall_avg_data_accessed = 0;
}
if(pairCnt != point->pairs_accessed){
// some work needs still to be done !
for(; pairCnt < point->pairs_accessed; pairCnt++ ) {

42
src/mdtest.c
View File

@ -131,6 +131,7 @@ static uint64_t items_per_dir;
static uint64_t num_dirs_in_tree_calc; /* this is a workaround until the overal code is refactored */
static int directory_loops;
static int print_time;
static int print_rate_and_time;
static int random_seed;
static int shared_file;
static int files_only;
@ -967,10 +968,10 @@ int updateStoneWallIterations(int iteration, rank_progress_t * progress, double
MPI_Reduce(& progress->items_done, & min_accessed, 1, MPI_LONG_LONG_INT, MPI_MIN, 0, testComm);
long long sum_accessed = 0;
MPI_Reduce(& progress->items_done, & sum_accessed, 1, MPI_LONG_LONG_INT, MPI_SUM, 0, testComm);
if(items != (sum_accessed / size)){
summary_table[iteration].stonewall_item_sum[MDTEST_FILE_CREATE_NUM] = sum_accessed;
summary_table[iteration].stonewall_item_min[MDTEST_FILE_CREATE_NUM] = min_accessed * size;
if(items != (sum_accessed / size)){
VERBOSE(0,-1, "Continue stonewall hit min: %lld max: %lld avg: %.1f \n", min_accessed, max_iter, ((double) sum_accessed) / size);
hit = 1;
}
@ -1192,12 +1193,15 @@ void file_test(const int iteration, const int ntasks, const char *path, rank_pro
}
VERBOSE(1,-1," File creation : %14.3f sec, %14.3f ops/sec", t[1] - t[0], summary_table[iteration].rate[4]);
if(summary_table[iteration].stonewall_time[MDTEST_FILE_CREATE_NUM]){
VERBOSE(1,-1," File creation (stonewall): %14.3f sec, %14.3f ops/sec", summary_table[iteration].stonewall_time[MDTEST_FILE_CREATE_NUM], summary_table[iteration].stonewall_item_sum[MDTEST_FILE_CREATE_NUM]);
}
VERBOSE(1,-1," File stat : %14.3f sec, %14.3f ops/sec", t[2] - t[1], summary_table[iteration].rate[5]);
VERBOSE(1,-1," File read : %14.3f sec, %14.3f ops/sec", t[3] - t[2], summary_table[iteration].rate[6]);
VERBOSE(1,-1," File removal : %14.3f sec, %14.3f ops/sec", t[4] - t[3], summary_table[iteration].rate[7]);
}
void summarize_results(int iterations) {
void summarize_results(int iterations, int print_time) {
char access[MAX_PATHLEN];
int i, j, k;
int start, stop, tableSize = MDTEST_LAST_NUM;
@ -1210,14 +1214,13 @@ void summarize_results(int iterations) {
MPI_Barrier(testComm);
for(int i=0; i < iterations; i++){
if(print_time){
MPI_Gather(& summary_table[i].time[0], tableSize, MPI_DOUBLE, & all[i*tableSize*size], tableSize, MPI_DOUBLE, 0, testComm);
}else{
MPI_Gather(& summary_table[i].rate[0], tableSize, MPI_DOUBLE, & all[i*tableSize*size], tableSize, MPI_DOUBLE, 0, testComm);
}
}
if (rank == 0) {
if (rank != 0) {
return;
}
VERBOSE(0,-1,"\nSUMMARY %s: (of %d iterations)", print_time ? "time": "rate", iterations);
VERBOSE(0,-1," Operation Max Min Mean Std Dev");
@ -1290,6 +1293,20 @@ void summarize_results(int iterations) {
}
// TODO generalize once more stonewall timers are supported
double stonewall_time = 0;
uint64_t stonewall_items = 0;
for(int i=0; i < iterations; i++){
if(summary_table[i].stonewall_time[MDTEST_FILE_CREATE_NUM]){
stonewall_time += summary_table[i].stonewall_time[MDTEST_FILE_CREATE_NUM];
stonewall_items += summary_table[i].stonewall_item_sum[MDTEST_FILE_CREATE_NUM];
}
}
if(stonewall_items != 0){
fprintf(out_logfile, " File create (stonewall) : ");
fprintf(out_logfile, "%14s %14s %14.3f %14s\n", "NA", "NA", print_time ? stonewall_time : stonewall_items / stonewall_time, "NA");
}
/* calculate tree create/remove rates */
for (i = 8; i < tableSize; i++) {
min = max = all[i];
@ -1334,7 +1351,6 @@ void summarize_results(int iterations) {
sum = var = 0;
}
}
}
/* Checks to see if the test setup is valid. If it isn't, fail. */
void valid_tests() {
@ -1794,6 +1810,7 @@ void mdtest_init_args(){
items_per_dir = 0;
random_seed = 0;
print_time = 0;
print_rate_and_time = 0;
shared_file = 0;
files_only = 0;
dirs_only = 0;
@ -1864,6 +1881,7 @@ mdtest_results_t * mdtest_run(int argc, char **argv, MPI_Comm world_com, FILE *
{'n', NULL, "every process will creat/stat/read/remove # directories and files", OPTION_OPTIONAL_ARGUMENT, 'l', & items},
{'N', NULL, "stride # between tasks for file/dir operation (local=0; set to 1 to avoid client cache)", OPTION_OPTIONAL_ARGUMENT, 'd', & nstride},
{'p', NULL, "pre-iteration delay (in seconds)", OPTION_OPTIONAL_ARGUMENT, 'd', & pre_delay},
{'P', NULL, "print rate AND time", OPTION_FLAG, 'd', & print_rate_and_time},
{'R', NULL, "random access to files (only for stat)", OPTION_FLAG, 'd', & randomize},
{0, "random-seed", "random seed for -R", OPTION_OPTIONAL_ARGUMENT, 'd', & random_seed},
{'s', NULL, "stride between the number of tasks for each test", OPTION_OPTIONAL_ARGUMENT, 'd', & stride},
@ -1878,7 +1896,6 @@ mdtest_results_t * mdtest_run(int argc, char **argv, MPI_Comm world_com, FILE *
{'x', NULL, "StoneWallingStatusFile; contains the number of iterations of the creation phase, can be used to split phases across runs", OPTION_OPTIONAL_ARGUMENT, 's', & stoneWallingStatusFile},
{'y', NULL, "sync file after writing", OPTION_FLAG, 'd', & sync_file},
{'z', NULL, "depth of hierarchical directory structure", OPTION_OPTIONAL_ARGUMENT, 'd', & depth},
{'Z', NULL, "print time instead of rate", OPTION_FLAG, 'd', & print_time},
LAST_OPTION
};
options_all_t * global_options = airoi_create_all_module_options(options);
@ -2154,7 +2171,12 @@ mdtest_results_t * mdtest_run(int argc, char **argv, MPI_Comm world_com, FILE *
// keep track of the current status for stonewalling
mdtest_iteration(i, j, testgroup, & summary_table[j]);
}
summarize_results(iterations);
if (print_rate_and_time){
summarize_results(iterations, 0);
summarize_results(iterations, 1);
}else{
summarize_results(iterations, print_time);
}
if (i == 1 && stride > 1) {
i = 0;
}