Feat: cmp with faster mmap compare

Author: Gunter Schmidt

.vscode/cspell.dictionaries/jargon.wordlist.txt

@@ -1,5 +1,6 @@
 # * diffutils project
-
+mmap
+memmap
 
 # *** the following part is a copy of coreutils ***
 AFAICT

Cargo.lock

@@ -91,7 +91,6 @@ license = "MIT"
 readme = "README.package.md"
 version = "0.5.1"
 
-
 [workspace.dependencies]
 assert_cmd = "2.2.0"
 bytecount = "0.6.9"
@@ -104,6 +103,7 @@ divan = { version = "4.3.0", package = "codspeed-divan-compat" }
 itertools = "0.14.0"
 itoa = "1.0.18"
 libc = "0.2.183"
+memmap2 = "0.9.10"
 nix = { version = "0.31.2", default-features = false }
 num-traits = "0.2.19"
 phf = "0.13.1"

Cargo.toml

@@ -25,10 +25,13 @@ path = "src/main.rs"
 path = "src/cmp.rs"
 
 [features]
-default = ["feat_run_binary_bench" ]
+default = ["feat_fast_mmap2_compare","feat_run_binary_bench"]
+# default = ["feat_run_binary_bench"]
+# default = ["feat_run_binary_bench"]
 # The cmd benchmarks start the binaries and take a lot of runtime on the github checks.
 # Only run them locally.
 feat_run_binary_bench = []
+feat_fast_mmap2_compare = []
 
 
 [dependencies]
@@ -37,6 +40,7 @@ clap.workspace = true
 diff_crate.workspace = true
 fluent.workspace = true
 itoa.workspace = true
+memmap2.workspace = true
 regex.workspace = true
 same-file.workspace = true
 uucore = { workspace = true, features = [

src/uu/cmp/Cargo.toml

@@ -11,10 +11,11 @@
 //! File generation up to 1 GB is really fast, Benchmarking above 100 MB takes very long.
 
 /// Generate test files with these sizes in KB.
-const FILE_SIZES_IN_KILO_BYTES: [u64; 4] = [100, 1 * MB, 10 * MB, 25 * MB];
+const FILE_SIZES_IN_KILO_BYTES: [u64; 5] = [100, 1 * MB, 10 * MB, 25 * MB, 100 * MB];
 const NUM_DIFF: u64 = 4;
 // Empty String to use TempDir (files will be removed after test) or specify dir to keep generated files
 const TEMP_DIR: &str = "";
+// const TEMP_DIR: &str = "/home/gunnar/Downloads/tmp";
 // just for FILE_SIZE_KILO_BYTES
 const MB: u64 = 1_000;
 
@@ -60,11 +61,30 @@ fn cmp_compare_files_equal(bencher: Bencher, kb: u64) {
     let fp = get_context().get_files_equal_kb(kb).unwrap();
     let cmd = format!("cmp {} {}", fp.from, fp.to);
     let args = str_to_args(&cmd).into_iter();
+    let matches =
+        uudiff::clap_localization::handle_clap_result_with_exit_code(uu_app(), args, 2).unwrap();
+    let params: Params = matches.try_into().unwrap();
 
     bencher
         // .with_inputs(|| prepare::cmp_params_identical_testfiles(lines))
-        .with_inputs(|| args.clone())
-        .bench_refs(|params| uu_cmp::uumain(params.peekable()));
+        .with_inputs(|| params.clone())
+        .bench_refs(|params| uu_cmp::cmp_compare(params));
+}
+
+// bench equal, full file read
+#[divan::bench(args = FILE_SIZES_IN_KILO_BYTES)]
+fn mmap_cmp_compare_files_equal(bencher: Bencher, kb: u64) {
+    let fp = get_context().get_files_equal_kb(kb).unwrap();
+    let cmd = format!("cmp {} {}", fp.from, fp.to);
+    let args = str_to_args(&cmd).into_iter();
+    let matches =
+        uudiff::clap_localization::handle_clap_result_with_exit_code(uu_app(), args, 2).unwrap();
+    let params: Params = matches.try_into().unwrap();
+
+    bencher
+        // .with_inputs(|| prepare::cmp_params_identical_testfiles(lines))
+        .with_inputs(|| params.clone())
+        .bench_refs(|params| uu_cmp::cmp_compare_mmap(params));
 }
 
 // bench different; cmp exits on first difference