tools/x264_perf_O3: support to disable loop vectorize optimization for get_ref (#1)

Reviewed-on: #1

pyproject.toml

@@ -4,7 +4,9 @@ version = "0.1.0"
 description = "Scripts that helps debugging SPEC CPU benchmark suite."
 readme = "README.md"
 requires-python = ">=3.12"
-dependencies = []
+dependencies = [
+    "click>=8.1.8",
+]
 
 [tool.setuptools]
 packages = { find = { where = ["src"] } }

tools/x264_perf_O3.py

@@ -1,27 +1,64 @@
 from pathlib import Path
 from subprocess import run
 
-from speccpu import get_benchspec_dir, x264
+import click
+
+from speccpu import create_spec_env, find_build, get_benchspec_dir, x264
 from speccpu.flaglib.perf import record as recordflags
 
 
-def main():
+@click.command()
-    x264_run_id = "c56635a4-9f49-4d2b-be79-38d9e08097c5"
+@click.option(
-    x264_outputroot = Path("/home/lyc/work-ts/local/specOutput") / x264_run_id
+    "--get-ref-vectorize/--no-get-ref-vectorize",
-
+    default=True,
-
+    help="Enable or disable mc.c `get_ref` vectorization",
-    x264_dir = get_benchspec_dir(x264_outputroot) / "625.x264_s"
+)
+@click.option(
+    "--spec",
+    type=click.Path(),
+    help="Path to SPEC CPU directory",
+    required=True,
+)
+@click.option(
+    "--x264-run-id",
+    type=str,
+    required=True,
+)
+@click.option(
+    "--outputroot",
+    type=click.Path(),
+    required=True,
+)
+def main(
+    get_ref_vectorize: bool,
+    spec,
+    x264_run_id: str,
+    outputroot,
+):
+    outputroot = Path(outputroot)
+    spec = Path(spec)
+    x264_dir = get_benchspec_dir(outputroot / x264_run_id) / "625.x264_s"
 
     x264_run = x264_dir / "run" / "run_base_refspeed_mytest-m64.0000"
+    x264_exe = x264_run / "x264_s_base.mytest-m64"
 
+    mc_builder = x264.MCBuilder()
 
+    if not get_ref_vectorize:
+        mc_builder.with_disabled_vectorize_get_ref()
+
+    x264_exe = x264.recompile_mc_src(
+        mc_builder.build(),
+        find_build(x264_dir / "build"),
+        create_spec_env(spec),
+    )
     run(
         [
             "perf",
             "record",
             *recordflags(
                 [
-                    str(x264_run / "x264_s_base.mytest-m64"),
+                    str(x264_exe),
                     *x264.specinvoke_args["ref"],
                 ],
                 output=str(Path.cwd() / f"x264-{x264_run_id}perf"),
@@ -31,5 +68,6 @@ def main():
         cwd=x264_run,
     )
 
+
 if __name__ == "__main__":
     main()

uv.lock

@@ -2,6 +2,18 @@ version = 1
 revision = 1
 requires-python = ">=3.12"
 
+[[package]]
+name = "click"
+version = "8.1.8"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "colorama", marker = "sys_platform == 'win32'" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/b9/2e/0090cbf739cee7d23781ad4b89a9894a41538e4fcf4c31dcdd705b78eb8b/click-8.1.8.tar.gz", hash = "sha256:ed53c9d8990d83c2a27deae68e4ee337473f6330c040a31d4225c9574d16096a", size = 226593 }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7e/d4/7ebdbd03970677812aac39c869717059dbb71a4cfc033ca6e5221787892c/click-8.1.8-py3-none-any.whl", hash = "sha256:63c132bbbed01578a06712a2d1f497bb62d9c1c0d329b7903a866228027263b2", size = 98188 },
+]
+
 [[package]]
 name = "colorama"
 version = "0.4.6"
@@ -57,6 +69,9 @@ wheels = [
 name = "speccpulib"
 version = "0.1.0"
 source = { virtual = "." }
+dependencies = [
+    { name = "click" },
+]
 
 [package.dev-dependencies]
 dev = [
@@ -64,6 +79,7 @@ dev = [
 ]
 
 [package.metadata]
+requires-dist = [{ name = "click", specifier = ">=8.1.8" }]
 
 [package.metadata.requires-dev]
 dev = [{ name = "pytest", specifier = ">=8.3.5" }]