From 43bed07320954f6ac499e7a65bd01685369658cf Mon Sep 17 00:00:00 2001 From: Bill Wendling Date: Wed, 20 Feb 2002 17:35:20 -0500 Subject: [svn-r4997] Purpose: Bug Fix Description: So, for Raw I/O in parallel, if you open a file with truncation by multiple processes, it looks as if one process could open the file and start writing to it while another process also opens the file with truncation, thus wiping out all of the stuff the first process wrote to the file. This is bad. Also added some garbage collection to the pio_perf routine to reclaim the space taken by some of the tables. Solution: Placed an MPI_Barrier() statement after the Raw open()/create() call so that all processes are synced up before they start writing to the file. Added free() calls to the tables which weren't being free'd. Platforms tested: Linux-pp (eirene) --- perform/pio_engine.c | 10 ++++++++++ perform/pio_perf.c | 3 +++ 2 files changed, 13 insertions(+) diff --git a/perform/pio_engine.c b/perform/pio_engine.c index d1ea6d3..efcfa98 100644 --- a/perform/pio_engine.c +++ b/perform/pio_engine.c @@ -926,6 +926,16 @@ do_fopen(iotype iot, char *fname, file_descr *fd /*out*/, int flags) else fd->rawfd = RAWOPEN(fname, O_RDONLY); + /* The perils of raw I/O in a parallel environment. The problem is: + * + * - Process n opens a file with truncation and then starts + * writing to the file. + * - Process m also opens the file with truncation, but after + * process n has already started to write to the file. Thus, + * all of the stuff process n wrote is now lost. + */ + MPI_Barrier(pio_comm_g); + if (fd->rawfd < 0 ) { fprintf(stderr, "Raw File Open failed(%s)\n", fname); GOTOERROR(FAIL); diff --git a/perform/pio_perf.c b/perform/pio_perf.c index 902817c..557f9bd 100644 --- a/perform/pio_perf.c +++ b/perform/pio_perf.c @@ -583,8 +583,11 @@ run_test(FILE *output, iotype iot, parameters parms) output_report(output, "Average Throughput: %.2f MB/s\n", total_mm.sum / total_mm.num); + /* clean up our mess */ free(write_mm_table); free(read_mm_table); + free(write_gross_mm_table); + free(read_gross_mm_table); pio_time_destroy(res.timers); return ret_value; } -- cgit v0.12