test: image_span_test reduce benchmark load for debug and CI renders (AcademySoftwareFoundation#4951)

Almost all the other unit tests do this, just adding it to
image_span_test: automatic reduction in test load for CI and debug
renders by doing fewer time trials. This cuts down on runtime for CI,
but stil exercises the test code. Command line arguments allow you to
override, in cases where you need really good timing fidelity even for a
debug run.

Also conform simd_test to the usual standard `-iters` used elsewhere
instead of being the one place that uses `-iterations`.

---------

Signed-off-by: Larry Gritz <[email protected]>

Author: lgritz

src/libOpenImageIO/image_span_test.cpp

@@ -5,6 +5,7 @@
 
 #include <OpenImageIO/half.h>
 
+#include <OpenImageIO/argparse.h>
 #include <OpenImageIO/benchmark.h>
 #include <OpenImageIO/fmath.h>
 #include <OpenImageIO/imageio.h>
@@ -18,6 +19,9 @@
 using namespace OIIO;
 
 
+static int ntrials = 5;
+
+
 
 template<typename T>
 void
@@ -195,6 +199,7 @@ test_image_span_copy_image()
 
         // Benchmark old (ptr) versus new (span) copy_image functions
         Benchmarker bench;
+        bench.trials(ntrials);
         bench.units(Benchmarker::Unit::us);
 
         bench(Strutil::format("    copy_image image_span {}", label),
@@ -266,6 +271,7 @@ test_image_span_contiguize()
 
         // Benchmark old (ptr) versus new (span) contiguize functions
         Benchmarker bench;
+        bench.trials(ntrials);
         bench.units(Benchmarker::Unit::us);
 
         bench(Strutil::format("    contiguize image_span {}", label), [&]() {
@@ -332,6 +338,7 @@ test_image_span_convert_image()
 
         // Benchmark old (ptr) versus new (span) contiguize functions
         Benchmarker bench;
+        bench.trials(ntrials);
         bench.units(Benchmarker::Unit::ms);
 
         bench(Strutil::format("    convert_image image_span {}", label),
@@ -418,6 +425,7 @@ benchmark_image_span_passing()
     image_span<const float> ispan(sbuf.data(), nchans, xres, yres, 1);
 
     Benchmarker bench;
+    bench.trials(ntrials);
     bench.units(Benchmarker::Unit::us);
     float sum = 0.0f;
 
@@ -442,6 +450,7 @@ benchmark_image_span_passing()
 
     // Do it all again for a SMALL image
     bench.units(Benchmarker::Unit::ns);
+    bench.trials(ntrials);
     int small = 16;
     image_span<const float> smispan(sbuf.data(), nchans, small, small, 1);
     bench("  pass by value     (small)",
@@ -544,9 +553,36 @@ test_image_span_within_span()
 
 
 
+static void
+getargs(int argc, char* argv[])
+{
+    ArgParse ap;
+    ap.intro(
+          "image_span_test -- unit test and benchmarks for OpenImageIO/image_span.h\n" OIIO_INTRO_STRING)
+        .usage("image_span_test [options]");
+
+    // ap.arg("--iterations %d", &iterations)
+    //     .help(Strutil::fmt::format("Number of iterations (default: {})",
+    //                                iterations));
+    ap.arg("--trials %d", &ntrials).help("Number of trials");
+
+    ap.parse_args(argc, (const char**)argv);
+}
+
+
+
 int
-main(int /*argc*/, char* /*argv*/[])
+main(int argc, char* argv[])
 {
+#if !defined(NDEBUG) || defined(OIIO_CI) || defined(OIIO_CODE_COVERAGE)
+    // For the sake of test time, reduce the default iterations for DEBUG,
+    // CI, and code coverage builds. Explicit use of --trials
+    // will override this, since it comes before the getargs() call.
+    ntrials = 1;
+#endif
+
+    getargs(argc, argv);
+
     test_image_span<float>();
     test_image_span<const float>();
     test_image_span<uint16_t>();

src/libutil/simd_test.cpp

@@ -43,8 +43,10 @@ getargs(int argc, char* argv[])
              OIIO_INTRO_STRING)
       .usage("simd_test [options]");
 
-    ap.arg("--iterations %d", &iterations)
+    ap.arg("--iters %d", &iterations)
       .help(Strutil::fmt::format("Number of iterations (default: {})", iterations));
+    ap.arg("--iterations %d", &iterations)
+      .hidden();  // obsolete synonym for --iters
     ap.arg("--trials %d", &ntrials)
       .help("Number of trials");