From 13cde8429d761c0042dba29f68f809b59bbbdebc Mon Sep 17 00:00:00 2001
From: Elena Sakhnovitch <elena.sakhnovitch@amd.com>
Date: Tue, 26 Oct 2021 18:39:23 -0400
Subject: [PATCH] [ROCm-SMI] add --showNodesBw

Display min and max bandwidth between gpu nodes

Signed-off-by: Elena Sakhnovitch
Change-Id: I7289fb83f80e2f899996b7d7560ece670cc5f31f
---
 python_smi_tools/rocm_smi.py | 40 ++++++++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)

diff --git a/python_smi_tools/rocm_smi.py b/python_smi_tools/rocm_smi.py
index 8ecdef5af4..c56338ec24 100755
--- a/python_smi_tools/rocm_smi.py
+++ b/python_smi_tools/rocm_smi.py
@@ -2485,6 +2485,43 @@ def showHwTopology(deviceList):
     showNumaTopology(deviceList)
 
 
+def showNodesBw(deviceList):
+    """ Display max and min bandwidth between nodes.
+    Currently supports XGMI only.
+    This reads the HW Topology file and displays the matrix for the nodes
+    @param deviceList: List of DRM devices (can be a single-item list)
+    """
+    devices_ind = range(len(deviceList))
+    minBW = c_uint32()
+    maxBW = c_uint32()
+    gpu_links_type = [[0 for x in devices_ind] for y in devices_ind]
+    printLogSpacer(' Bandwidth ')
+    for srcdevice in deviceList:
+        for destdevice in deviceList:
+            if srcdevice != destdevice:
+                ret = rocmsmi.rsmi_minmax_bandwidth_get(srcdevice, destdevice, byref(minBW), byref(maxBW))
+                if rsmi_ret_ok(ret, " {}  to {}".format(srcdevice, destdevice),None ):
+                    gpu_links_type[srcdevice][destdevice] = "{}-{}".format(minBW.value, maxBW.value)
+            else:
+                gpu_links_type[srcdevice][destdevice] = "N/A"
+    if PRINT_JSON:
+        formatMatrixToJSON(deviceList, "{}-{}".format(minBW.value, maxBW.value),  " min-max bandwidth between DRM devices {} and {}".format(srcdevice, destdevice))
+        return
+    printTableRow(None, '      ')
+    for row in deviceList:
+        tmp = 'GPU%d' % row
+        printTableRow('%-12s', tmp)
+    printEmptyLine()
+    for gpu1 in deviceList:
+        tmp = 'GPU%d' % gpu1
+        printTableRow('%-6s', tmp)
+        for gpu2 in deviceList:
+            printTableRow('%-12s', gpu_links_type[gpu1][gpu2])
+        printEmptyLine()
+    printLog(None,"Format: min-max; Units: mps", None)
+    printLog(None,'"0-0" min-max bandwidth indicates devices are not connected dirrectly', None)
+
+
 def checkAmdGpus(deviceList):
     """ Check if there are any AMD GPUs being queried,
     return False if there are none
@@ -2828,6 +2865,7 @@ if __name__ == '__main__':
     groupDisplay.add_argument('--showtoponuma', help='Shows the numa nodes ', action='store_true')
     groupDisplay.add_argument('--showenergycounter', help='Energy accumulator that stores amount of energy consumed',
                               action='store_true')
+    groupDisplay.add_argument('--shownodesbw', help='Shows the numa nodes ', action='store_true')
 
     groupActionReset.add_argument('-r', '--resetclocks', help='Reset clocks and OverDrive to default',
                                   action='store_true')
@@ -3065,6 +3103,8 @@ if __name__ == '__main__':
         showProductName(deviceList)
     if args.showxgmierr:
         showXgmiErr(deviceList)
+    if args.shownodesbw:
+        showNodesBw(deviceList)
     if args.showtopo:
         showHwTopology(deviceList)
     if args.showtopoaccess: