Add CARGO_PROFILE support for bench.sh

Document the CARGO_PROFILE override in bench.sh.
Include a tip for using release-nonlto and an
example dfbench command to achieve faster builds
with near-release performance.

Author: kosiew

benchmarks/README.md

@@ -49,6 +49,8 @@ cd ./benchmarks/
 ./bench.sh
 ```
 
+To override the Cargo profile used by the script, set `CARGO_PROFILE` (for example, `CARGO_PROFILE=release-nonlto`).
+
 ## Generating data
 
 You can create / download the data for these benchmarks using the [bench.sh](bench.sh) script:
@@ -265,6 +267,12 @@ Assuming data is in the `data` directory, the `tpch` benchmark can be run with a
 cargo run --release --bin dfbench -- tpch --iterations 3 --path ./data --format tbl --query 1 --batch-size 4096
 ```
 
+Tip: `--profile release-nonlto` builds significantly faster while staying close to release performance. For example:
+
+```bash
+cargo run --profile release-nonlto --bin dfbench -- tpcds --iterations 3 --path ./data --format parquet --query 1
+```
+
 See the help for more details.
 
 ### Different features

benchmarks/bench.sh

@@ -40,7 +40,10 @@ COMMAND=
 BENCHMARK=all
 DATAFUSION_DIR=${DATAFUSION_DIR:-$SCRIPT_DIR/..}
 DATA_DIR=${DATA_DIR:-$SCRIPT_DIR/data}
-CARGO_COMMAND=${CARGO_COMMAND:-"cargo run --release"}
+CARGO_PROFILE=${CARGO_PROFILE:-release}
+if [[ -z "${CARGO_COMMAND}" ]]; then
+    CARGO_COMMAND="cargo run --profile ${CARGO_PROFILE}"
+fi
 PREFER_HASH_JOIN=${PREFER_HASH_JOIN:-true}
 VIRTUAL_ENV=${VIRTUAL_ENV:-$SCRIPT_DIR/venv}
 
@@ -141,6 +144,7 @@ Supported Configuration (Environment Variables)
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 DATA_DIR            directory to store datasets
 CARGO_COMMAND       command that runs the benchmark binary
+CARGO_PROFILE       cargo profile to use when CARGO_COMMAND is unset (default release)
 DATAFUSION_DIR      directory to use (default $DATAFUSION_DIR)
 RESULTS_NAME        folder where the benchmark files are stored
 PREFER_HASH_JOIN    Prefer hash join algorithm (default true)