flashattention_2: Take measurements on second Tc tile

Author: colluca

target/snitch_cluster/sw/apps/dnn/flashattention_2/tcasai/cfg/fp16-opt-vit-base.json

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: Apache-2.0
 
 {
-	L: 64,
-	S: 64,
+	L: 32,
+	S: 128,
     d: 64,
     B_r: 16,
     B_c: 64,

target/snitch_cluster/sw/apps/dnn/flashattention_2/tcasai/cfg/fp16-opt-vit-huge.json

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: Apache-2.0
 
 {
-	L: 64,
-	S: 64,
+	L: 32,
+	S: 128,
     d: 80,
     B_r: 16,
     B_c: 64,

target/snitch_cluster/sw/apps/dnn/flashattention_2/tcasai/cfg/fp32-opt-vit-base.json

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: Apache-2.0
 
 {
-	L: 64,
-	S: 64,
+	L: 32,
+	S: 128,
     d: 64,
     B_r: 16,
     B_c: 64,

target/snitch_cluster/sw/apps/dnn/flashattention_2/tcasai/cfg/fp32-opt-vit-huge.json

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: Apache-2.0
 
 {
-	L: 64,
-	S: 64,
+	L: 32,
+	S: 128,
     d: 80,
     B_r: 16,
     B_c: 64,

target/snitch_cluster/sw/apps/dnn/flashattention_2/tcasai/cfg/fp8-opt-vit-base.json

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: Apache-2.0
 
 {
-	L: 64,
-	S: 64,
+	L: 32,
+	S: 128,
     d: 64,
     B_r: 16,
     B_c: 64,

target/snitch_cluster/sw/apps/dnn/flashattention_2/tcasai/cfg/fp8-opt-vit-huge.json

@@ -3,8 +3,8 @@
 // SPDX-License-Identifier: Apache-2.0
 
 {
-	L: 64,
-	S: 64,
+	L: 32,
+	S: 128,
     d: 80,
     B_r: 16,
     B_c: 64,

target/snitch_cluster/sw/apps/dnn/flashattention_2/tcasai/results.py

@@ -132,7 +132,7 @@ def get_total_runtime(experiment):
     # Define the simulated tc and tr iterations to extract the region runtimes from.
     # The second tc iteration is the worst case, since upon the first iteration
     # several calculations are not performed.
-    tc_iteration = 0
+    tc_iteration = 1
     tr_iteration = 0
 
     # Calculate total runtime