Merge pull request #175 from wenkaidu/ll_abort

Fix abort handling in LL primitives

src/collectives/device/prims_ll.h

@@ -130,6 +130,7 @@ class ncclLLPrimitives {
       asm volatile ("flat_load_dwordx4 %0, %1, glc\n"
         "s_waitcnt vmcnt(0)\n"
         "buffer_wbinvl1_vol\n" : "=v"(i4) : "v"(src));
+      if (checkAbort(i, 0)) break;
     } while ((i4[1] != flag) || (i4[3] != flag));
     uint64_t val64 = (uint64_t)(i4[0]) + (((uint64_t)i4[2]) << 32);
 #else