chore: a few small benchmarking and profiling improvements (#3069)

Signed-off-by: Hannah Zhang <hannahz@nvidia.com>

.gitignore

@@ -98,3 +98,6 @@ TensorRT-LLM
 /.cursor/instructions.md.bak
 /CLAUDE.md
 /CLAUDE.md.bak
+
+# Benchmarks
+benchmarks/results
\ No newline at end of file

benchmarks/benchmark.sh

@@ -271,21 +271,9 @@ print_config() {
     echo
 }
 
-clear_output_directory() {
-    if [[ -d "$OUTPUT_DIR" ]]; then
-        echo "🧹 Clearing existing output directory: $OUTPUT_DIR"
-        rm -rf "$OUTPUT_DIR"
-    fi
-    mkdir -p "$OUTPUT_DIR"
-    echo "✅ Output directory prepared: $OUTPUT_DIR"
-}
-
 run_benchmark() {
     echo "🚀 Starting benchmark workflow..."
 
-    # Clear and recreate output directory
-    clear_output_directory
-
     # Change to dynamo root directory
     cd "$DYNAMO_ROOT"
 

docs/benchmarks/pre_deployment_profiling.md

@@ -39,7 +39,7 @@ This sequential approach ensures:
 - **Reliable cleanup** between different TP configuration tests
 - **Accurate SLA compliance verification** for each configuration
 
-After the profiling finishes, two plots will be generated in the `output-dir`. For example, here are the profiling results for `examples/llm/configs/disagg.yaml`:
+After the profiling finishes, two plots will be generated in the `output-dir`. For example, here are the profiling results for `components/backends/vllm/deploy/disagg.yaml`:
 
 ![Prefill Performance](../../docs/images/h100_prefill_performance.png)
 ![Decode Performance](../../docs/images/h100_decode_performance.png)