diff --git a/tests/fmha_v3_stage_c_example9.py b/tests/fmha_v3_stage_c_example9.py
index 0381c543..23008686 100644
--- a/tests/fmha_v3_stage_c_example9.py
+++ b/tests/fmha_v3_stage_c_example9.py
@@ -180,16 +180,12 @@ class FmhaV3StageCMulti:
         tBsK,tBgK = cpasync.tma_partition(tma_k,0,b_lay,cute.group_modes(sK,0,3),cute.group_modes(tCgK,0,3))
         tVsV,tVgV = cpasync.tma_partition(tma_v,0,b_lay,cute.group_modes(sV,0,3),cute.group_modes(tCgV,0,3))
         # DIAG: print shapes before pre-slice
-        cute.printf("tAgQ shape: {} rank: {}\n", cute.shape(tAgQ), tAgQ.rank)
-        cute.printf("tBgK shape: {} rank: {}\n", cute.shape(tBgK), tBgK.rank)
-        cute.printf("tVgV shape: {} rank: {}\n", cute.shape(tVgV), tVgV.rank)
-        cute.printf("tAsQ shape: {} rank: {}\n", cute.shape(tAsQ), tAsQ.rank)
-        cute.printf("tBsK shape: {} rank: {}\n", cute.shape(tBsK), tBsK.rank)
-        cute.printf("tVsV shape: {} rank: {}\n", cute.shape(tVsV), tVsV.rank)
-        for i in range(tBgK.rank):
-            cute.printf("  tBgK mode {} size: {}\n", i, cute.size(tBgK, mode=[i]))
-        for i in range(tVgV.rank):
-            cute.printf("  tVgV mode {} size: {}\n", i, cute.size(tVgV, mode=[i]))
+        cute.printf("tAgQ shape: {}\n", cute.shape(tAgQ))
+        cute.printf("tBgK shape: {}\n", cute.shape(tBgK))
+        cute.printf("tVgV shape: {}\n", cute.shape(tVgV))
+        cute.printf("tAsQ shape: {}\n", cute.shape(tAsQ))
+        cute.printf("tBsK shape: {}\n", cute.shape(tBsK))
+        cute.printf("tVsV shape: {}\n", cute.shape(tVsV))
         tAgQ = tAgQ[(None,0,None,0)]; tBgK = tBgK[(None,None,0,0)]; tVgV = tVgV[(None,0,None,0)]
 
         tCrQ = qk_mma.make_fragment_A(sQ); tCrK = qk_mma.make_fragment_B(sK)