Misc fixes and improvements for 2.5.6

1. Fix RCCL unit test
2. Add ROME detection and tuning
3. Change default P2P level
4. Fix search algorithm for XGMI
5. Remove explicit channel duplication with implicit by using half of link speed
6. Add collective trace support
7. Correct Intel Skylake CPU detection and bandwidth
8. Fix topo connect function
9. Disable GDR read and remove unreachable code
10. Disable LL128 kernels
11. Add tuning parameters
12. Use original clock64() implementation which returns RTC counter value
13. Print out timestamp of collective trace
14. Do not use struct ncclColl in kernel launch parameter
15. Fix abort handling and add tracing
17. Add __launch_bounds__ to kernel functions
18. Remove unused abortCount
19. Unset default MIN_NRINGS and MIN_NCHANNELS
20. Do not allocate shared memory when not using LL128 kernels
21. Correct time print out in tuning log


[ROCm/rccl commit: 1e55645d97]

projects/rccl/tools/rccl-prim-test/rccl_prim_test.cpp

@@ -34,7 +34,7 @@ THE SOFTWARE.
 #include "copy_kernel.h"
 
 #define MAX_GPU 8
-#define MAX_WORKGROUPS 16
+#define MAX_WORKGROUPS 32
 #define THREADS 256
 
 #define COPY_UNROLL       4
@@ -56,6 +56,16 @@ THE SOFTWARE.
 #define RTC_CLOCK_FREQ_MI100 2.5E07
 #define RTC_CLOCK_FREQ_DEFAULT 2.7E07
 
+__device__
+inline  __attribute((always_inline))
+long long int __rtc64() {
+#if __HIP__
+  return (long long int) __builtin_amdgcn_s_memrealtime();
+#else
+  return (long long int) __clock_u64();
+#endif
+}
+
 struct transfer_data_t {
   float *dest0[MAX_WORKGROUPS]; //remote fine grain
   float *src0[MAX_WORKGROUPS];  //local fine grain
@@ -116,7 +126,7 @@ __global__ void flag_sync_kernel(struct transfer_data_t* transfer_data, struct p
   __syncthreads();
 
   if (idx == 0) {
-    curr_time = clock64();
+    curr_time = __rtc64();
   }
 
   if (op == OP_COPY) Copy<COPY_UNROLL, THREADS, float>(transfer_data->dest0[bid], transfer_data->src0[bid], n);
@@ -129,7 +139,7 @@ __global__ void flag_sync_kernel(struct transfer_data_t* transfer_data, struct p
   if (op == OP_READ) Copy<COPY_UNROLL, THREADS, float>(transfer_data->src0[bid],transfer_data->dest0[bid], n);
   __syncthreads();
   if (idx == 0) {
-    next_time = clock64();
+    next_time = __rtc64();
     __atomic_fetch_add(&(profiling_data->write_cycles[bid]), next_time - curr_time, __ATOMIC_SEQ_CST);
     __atomic_fetch_add(&(profiling_data->bytes_transferred[bid]), n * sizeof(float), __ATOMIC_SEQ_CST);
   }