diff --git a/pxr/imaging/hdx/shaders/renderPass.glslfx b/pxr/imaging/hdx/shaders/renderPass.glslfx
index fd54daf976..926993acf2 100644
--- a/pxr/imaging/hdx/shaders/renderPass.glslfx
+++ b/pxr/imaging/hdx/shaders/renderPass.glslfx
@@ -256,7 +256,30 @@ void RenderOutput(vec4 Peye, vec3 Neye, vec4 color, vec4 patchCoord)
     if (ATOMIC_LOAD(PickBuffer[0]) == 0)
         return;
 	
-    // prepare some constants
+    // Get some constants from the Pick Buffer
+    // These are initialized in: HdxPickTask::_ClearPickBuffer()
+    //
+    // The first 8 entries in the PickBuffer describe the buffer layout
+    // [0] == The number of sub-buffers (total max hits allowed / sub-buffer size)
+    //        The total max his allowed is configurable on the pickContext (default = 32,000)
+    // [1] == The max number of hits hashed and stored per sub-buffer (value = 32)
+    // [2] == The TableOffset which is the first entry after this header block.  (value = 8)
+    //        The Table is a collection of fields the length of the number of sub-buffers.
+    //        It holds the number of current hits stored per sub-buffer. 
+    //        These fields are initialized to zero and populated by this shader to track hit storage.
+    // [3] == The StorageOffset which is the starting point where hits are stored.
+    //        Hit results are stored in sub-buffers.
+    //        Each sub-buffer holds up to 32 entries each with 3 entries per hit (primId, instanceId, partId).
+    // [4] == Indicates if faces should be picked (value == 1 or 0)
+    // [5] == Indicates if edges should be picked (value == 1 or 0)
+    // [6] == Indicates if points should be picked (value == 1 or 0)
+    // [7] == Padding
+    // 
+    // [TableOffset -> StorageOffset]
+    //        The table of current entry counts in each sub-buffer (values initialized to 0)
+    // [StorageOffset -> End of Buffer]
+    //        The collection of hits as a (primId, instanceId, partId) tuple. (values initialized to -9)
+
     const int entrySize = 3;
     const int numSubBuffers = ATOMIC_LOAD(PickBuffer[0]);
     const int subBufferCapacity = ATOMIC_LOAD(PickBuffer[1]);
@@ -266,7 +289,7 @@ void RenderOutput(vec4 Peye, vec3 Neye, vec4 color, vec4 patchCoord)
                         ATOMIC_LOAD(PickBuffer[5]) * edgeId +
                         ATOMIC_LOAD(PickBuffer[6]) * pointId;
 
-    // more constants
+    // compute the hash for an instance/element and assign it to a sub-buffer
     const int hashValue = hash3(primId, instanceId, partId);
     const int subBufferNumber = hashValue % numSubBuffers;
     int bufferNumber = subBufferNumber;
diff --git a/pxr/imaging/hgiMetal/shaderGenerator.mm b/pxr/imaging/hgiMetal/shaderGenerator.mm
index 69e32684a2..260ef8abe5 100644
--- a/pxr/imaging/hgiMetal/shaderGenerator.mm
+++ b/pxr/imaging/hgiMetal/shaderGenerator.mm
@@ -303,9 +303,9 @@ void _Init(
         " atomic_fetch_add_explicit(&a, v, memory_order_relaxed)\n"
         "#define ATOMIC_EXCHANGE(a, desired)"
         " atomic_exchange_explicit(&a, desired, memory_order_relaxed)\n"
-        "#define ATOMIC_COMP_SWAP(a, expected, desired) \\"
-        "  atomic_compare_exchange_strong_explicit(&a, &expected, desired, "
-        "memory_order_relaxed, memory_order_relaxed)\n"
+        "#define ATOMIC_COMP_SWAP(a, expected, desired)"
+        "  atomic_compare_exchange_weak_explicit(&a, &expected, desired, "
+        "memory_order_relaxed, memory_order_relaxed) ? expected : expected;\n"
         "\n";
 }