Fixed #14 the SEGFAULT bug on 64 cores. On SMP server, the number of CPUs or cores...

diff --git a/Changelog.txt b/Changelog.txt
index f4f9c6e..7df8c0c 100644 (file)
--- a/Changelog.txt
+++ b/Changelog.txt
@@ -5,6 +5,12 @@ Version 0.1 alpha2(in development)
 common:
        * Fixed blasint undefined bug in <cblas.h> file. Other software 
          could include this header successfully(Refs issue #13 on github)
+       * Fixed the SEGFAULT bug on 64 cores. On SMP server, the number 
+         of CPUs or cores should be less than or equal to 64.(Refs issue #14 
+         on github)
+       * Support "void goto_set_num_threads(int num_threads)" and "void
+         openblas_set_num_threads(int num_threads)" when USE_OPENMP=1
+
 x86/x86_64:
        * 
 MIPS64:

diff --git a/README b/README
index 9b04f6f..46792c0 100644 (file)
--- a/README
+++ b/README
@@ -39,13 +39,17 @@ export GOTO_NUM_THREADS=4
  or 
 export OMP_NUM_THREADS=4
 
-The priorities are OPENBLAS_NUM_THREAD > GOTO_NUM_THREADS > OMP_NUM_THREADS.
+The priorities are OPENBLAS_NUM_THREADS > GOTO_NUM_THREADS > OMP_NUM_THREADS.
+
+If you compile this lib with USE_OPENMP=1, you should only set OMP_NUM_THREADS environment variable.
 
 4.2 Set the number of threads with calling functions. for example,
 void goto_set_num_threads(int num_threads);
 or
 void openblas_set_num_threads(int num_threads);
 
+If you compile this lib with USE_OPENMP=1, you should use the above functions, too.
+
 5.Report Bugs
 Please add a issue in https://github.com/xianyi/OpenBLAS/issues
 
@@ -56,4 +60,8 @@ Optimization on ICT Loongson 3A CPU
 OpenBLAS users mailing list: http://list.rdcps.ac.cn/mailman/listinfo/openblas
 
 8.ChangeLog
-Please see Changelog.txt to obtain the differences between GotoBLAS2 1.13 BSD version.
\ No newline at end of file
+Please see Changelog.txt to obtain the differences between GotoBLAS2 1.13 BSD version.
+
+9.Known Issues
+* The number of CPUs/Cores should less than or equal to 8*sizeof(unsigned long). On 64 bits, the limit 
+  is 64. On 32 bits, it is 32.

diff --git a/driver/others/blas_server_omp.c b/driver/others/blas_server_omp.c
index 3e70d85..17d886e 100644 (file)
--- a/driver/others/blas_server_omp.c
+++ b/driver/others/blas_server_omp.c
@@ -49,6 +49,26 @@
 
 int blas_server_avail = 0;
 
+void goto_set_num_threads(int num_threads) {
+
+  if (num_threads < 1) num_threads = blas_num_threads;
+
+  if (num_threads > MAX_CPU_NUMBER) num_threads = MAX_CPU_NUMBER;
+
+  if (num_threads > blas_num_threads) {
+    blas_num_threads = num_threads;
+  }
+
+  blas_cpu_number  = num_threads;
+
+  omp_set_num_threads(blas_cpu_number);
+  
+}
+void openblas_set_num_threads(int num_threads) {
+
+       goto_set_num_threads(num_threads);
+}
+
 int blas_thread_init(void){
 
   blas_get_cpu_number();

diff --git a/driver/others/init.c b/driver/others/init.c
index 7ee7dc4..4adba66 100644 (file)
--- a/driver/others/init.c
+++ b/driver/others/init.c
@@ -172,13 +172,20 @@ static inline int rcount(unsigned long number) {
   return count;
 }
 
+/***
+  Known issue: The number of CPUs/cores should less 
+  than sizeof(unsigned long). On 64 bits, the limit 
+  is 64. On 32 bits, it is 32.
+***/
 static inline unsigned long get_cpumap(int node) {
 
   int infile;
   unsigned long affinity;
   char name[160];
+  char cpumap[160];
   char *p, *dummy;
-  
+  int i=0;
+
   sprintf(name, CPUMAP_NAME, node);
   
   infile = open(name, O_RDONLY);
@@ -187,13 +194,19 @@ static inline unsigned long get_cpumap(int node) {
     
   if (infile != -1) {
     
-    read(infile, name, sizeof(name));
-   
+    read(infile, cpumap, sizeof(cpumap));
+    p = cpumap;
+    while (*p != '\n' && i<160){
+      if(*p != ',') {
+       name[i++]=*p;
+      }
+      p++;
+    }
     p = name;
 
-    while ((*p == '0') || (*p == ',')) p++;
+    //    while ((*p == '0') || (*p == ',')) p++;
 
-    affinity = strtol(p, &dummy, 16);
+    affinity = strtoul(p, &dummy, 16);
    
     close(infile);
   }
@@ -347,7 +360,13 @@ static void disable_hyperthread(void) {
   unsigned long share;
   int cpu;
 
-  common -> avail = (1UL << common -> num_procs) - 1;
+  if(common->num_procs > 64){
+    fprintf(stderr, "\nOpenBLAS Warining : The number of CPU/Cores(%d) is beyond the limit(64). Terminated.\n", common->num_procs);
+    exit(1);
+  }else if(common->num_procs == 64){
+    common -> avail = 0xFFFFFFFFFFFFFFFFUL;
+  }else
+    common -> avail = (1UL << common -> num_procs) - 1;
 
 #ifdef DEBUG
   fprintf(stderr, "\nAvail CPUs    : %04lx.\n", common -> avail);
@@ -376,7 +395,13 @@ static void disable_affinity(void) {
     fprintf(stderr, "CPU mask                  : %04lx.\n\n", *(unsigned long *)&cpu_orig_mask[0]);
 #endif
 
-  lprocmask = (1UL << common -> final_num_procs) - 1;
+  if(common->final_num_procs > 64){
+    fprintf(stderr, "\nOpenBLAS Warining : The number of CPU/Cores(%d) is beyond the limit(64). Terminated.\n", common->final_num_procs);
+    exit(1);
+  }else if(common->final_num_procs == 64){
+    lprocmask = 0xFFFFFFFFFFFFFFFFUL;
+  }else
+    lprocmask = (1UL << common -> final_num_procs) - 1;
 
 #ifndef USE_OPENMP
   lprocmask &= *(unsigned long *)&cpu_orig_mask[0];