Rename CMN memory bandwidth metric to reflect read+write

Summary:
Rename CMN Memory Read Bandwidth to CMN Memory Bandwidth since the underlying CMN event hns_mc_reqs_local_all counts both read and write
  requests to the memory controller. CMN-Cypress does not distinguish read vs write MC requests in this event, so the previous "Read" label was
  misleading.

Differential Revision: D102244354

Author: marziehlenjaniMeta

perfutils/collect_neoversev3_perf_counters.sh

@@ -138,6 +138,8 @@ for cmn_dev in $(find /sys/bus/event_source/devices/ -maxdepth 1 -name 'arm_cmn_
     CMN_SLC_EVENTS+="${cmn_dev}/hns_cache_fill_all/,"
     CMN_SLC_EVENTS+="${cmn_dev}/hns_mc_reqs_local_all/,"
     CMN_SLC_EVENTS+="${cmn_dev}/hns_pocq_reqs_recvd_all/,"
+    CMN_SLC_EVENTS+="${cmn_dev}/hns_qos_pocq_occupancy_read/,"
+    CMN_SLC_EVENTS+="${cmn_dev}/dtc_cycles/,"
 done
 CMN_SLC_EVENTS="${CMN_SLC_EVENTS%,}"
 

perfutils/generate_arm_neoversev3_perf_report.py

@@ -807,23 +807,47 @@ def dispatch_stall_mcq(grouped_df):
 
 
 @skip_if_missing
-def cmn_mem_read_bw_MBps(grouped_df):
-    """Memory read bandwidth from CMN MC request counters.
+def cmn_mem_bw_MBps(grouped_df):
+    """Memory bandwidth (read+write) from CMN MC request counters.
 
-    Each hns_mc_reqs_local_all is a cache-line (64B) request to the memory
-    controller, analogous to Grace's SCF cmem_rd_data.
+    hns_mc_reqs_local_all counts all cache-line (64B) requests to the memory
+    controller, including both reads and writes. CMN-Cypress does not
+    distinguish read vs write MC requests in this event.
     """
     mc_reqs = _sum_cmn_event(grouped_df, "hns_mc_reqs_local_all")
     dur = get_duration_series(grouped_df.get_group("instructions"))
     mc_reqs.index = dur.index
     bw_series = (mc_reqs * 64).div(dur)
     return {
-        "name": "CMN Memory Read Bandwidth (MBps)",
+        "name": "CMN Memory Bandwidth (MBps)",
         "series": bw_series,
         "prefix": 10**-6,
     }
 
 
+@skip_if_missing
+def cmn_mesh_freq_ghz(grouped_df):
+    """CMN mesh clock frequency derived from dtc_cycles per CMN instance."""
+    dur = get_duration_series(grouped_df.get_group("instructions"))
+
+    freqs = []
+    for name, group in grouped_df:
+        if isinstance(name, str) and name.endswith("/dtc_cycles/"):
+            cyc = group.counter_value.reset_index(drop=True)
+            cyc.index = dur.index
+            freqs.append(cyc / dur)
+
+    if not freqs:
+        raise KeyError("dtc_cycles")
+
+    avg_freq = functools.reduce(lambda a, b: a + b, freqs) / len(freqs)
+    return {
+        "name": "CMN Mesh Frequency (GHz)",
+        "series": avg_freq,
+        "prefix": 10**-9,
+    }
+
+
 @skip_if_missing
 def sve_pred_empty_pct(grouped_df):
     """SVE predicated ops with no active lanes (wasted work)."""
@@ -960,8 +984,9 @@ def main(
         sve_pred_empty_pct(grouped_df),
         sve_pred_full_pct(grouped_df),
         sve_pred_partial_pct(grouped_df),
-        # --- CMN uncore (SLC / memory bandwidth) ---
-        cmn_mem_read_bw_MBps(grouped_df),
+        # --- CMN memory metrics ---
+        cmn_mem_bw_MBps(grouped_df),
+        cmn_mesh_freq_ghz(grouped_df),
     ]
 
     filtered_metrics = list(itertools.filterfalse(lambda x: x is None, metrics))