Calculation of chunk range limits was ignoring num_cpu

bug introduced by me in #1262

driver/level2/symv_thread.c

@@ -177,7 +177,7 @@ int CNAME(BLASLONG m, FLOAT *alpha, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG i
 
     range_m[num_cpu + 1] = range_m[num_cpu] + width;
     range_n[num_cpu] = num_cpu * (((m + 15) & ~15) + 16);
-    if (range_n[num_cpu] > m) range_n[num_cpu] = m;
+    if (range_n[num_cpu] > m * num_cpu) range_n[num_cpu] = m * num_cpu;
     
     queue[MAX_CPU_NUMBER - num_cpu - 1].mode    = mode;
     queue[MAX_CPU_NUMBER - num_cpu - 1].routine = symv_kernel;
@@ -226,7 +226,7 @@ int CNAME(BLASLONG m, FLOAT *alpha, FLOAT *a, BLASLONG lda, FLOAT *x, BLASLONG i
 
     range_m[num_cpu + 1] = range_m[num_cpu] + width;
     range_n[num_cpu] = num_cpu * (((m + 15) & ~15) + 16);
-    if (range_n[num_cpu] > m) range_n[num_cpu] = m;
+    if (range_n[num_cpu] > m * num_cpu) range_n[num_cpu] = m * num_cpu;
 
     queue[num_cpu].mode    = mode;
     queue[num_cpu].routine = symv_kernel;