[Piglit] [PATCH 09/25] arb_shader_image_load_store: Import memory coherency test.
Francisco Jerez
currojerez at riseup.net
Sun Oct 5 14:00:39 PDT 2014
Test the cache coherency of images declared as "coherent" or
"volatile" when accessed from different combinations of dependent
shader stages.
---
tests/all.py | 1 +
tests/quick.py | 2 +
.../arb_shader_image_load_store/CMakeLists.gl.txt | 1 +
tests/spec/arb_shader_image_load_store/coherency.c | 228 +++++++++++++++++++++
4 files changed, 232 insertions(+)
create mode 100644 tests/spec/arb_shader_image_load_store/coherency.c
diff --git a/tests/all.py b/tests/all.py
index 9545034..6ef641a 100644
--- a/tests/all.py
+++ b/tests/all.py
@@ -4248,6 +4248,7 @@ import_glsl_parser_tests(spec['ARB_shader_image_load_store'],
[''])
arb_shader_image_load_store['atomicity'] = concurrent_test('arb_shader_image_load_store-atomicity')
arb_shader_image_load_store['bitcast'] = concurrent_test('arb_shader_image_load_store-bitcast')
+arb_shader_image_load_store['coherency'] = concurrent_test('arb_shader_image_load_store-coherency')
profile.tests['hiz'] = hiz
profile.tests['fast_color_clear'] = fast_color_clear
diff --git a/tests/quick.py b/tests/quick.py
index 9dc0f98..d0b35c1 100644
--- a/tests/quick.py
+++ b/tests/quick.py
@@ -12,3 +12,5 @@ del profile.tests['shaders']['glsl-fs-inline-explosion']
del profile.tests['shaders']['glsl-fs-unroll-explosion']
del profile.tests['shaders']['glsl-vs-inline-explosion']
del profile.tests['shaders']['glsl-vs-unroll-explosion']
+
+profile.tests['spec']['ARB_shader_image_load_store']['coherency'].command += ["--quick"]
diff --git a/tests/spec/arb_shader_image_load_store/CMakeLists.gl.txt b/tests/spec/arb_shader_image_load_store/CMakeLists.gl.txt
index 6b53370..6a6aa25 100644
--- a/tests/spec/arb_shader_image_load_store/CMakeLists.gl.txt
+++ b/tests/spec/arb_shader_image_load_store/CMakeLists.gl.txt
@@ -14,5 +14,6 @@ set(depends image.c grid.c common.c)
piglit_add_executable(arb_shader_image_load_store-atomicity atomicity.c ${depends})
piglit_add_executable(arb_shader_image_load_store-bitcast bitcast.c ${depends})
+piglit_add_executable(arb_shader_image_load_store-coherency coherency.c ${depends})
# vim: ft=cmake:
diff --git a/tests/spec/arb_shader_image_load_store/coherency.c b/tests/spec/arb_shader_image_load_store/coherency.c
new file mode 100644
index 0000000..e8484a7
--- /dev/null
+++ b/tests/spec/arb_shader_image_load_store/coherency.c
@@ -0,0 +1,228 @@
+/*
+ * Copyright (C) 2014 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+/** @file coherency.c
+ *
+ * Images declared as "coherent" or "volatile" give certain guarantees
+ * regarding the visibility of memory writes to other shader invocations
+ * in the pipeline. This test checks the following assertion of the
+ * spec:
+ * "When reading a variable declared as 'coherent', the values returned
+ * will reflect the results of previously completed writes performed by
+ * other shader invocations. When writing a variable declared as
+ * 'coherent', the values written will be reflected in subsequent
+ * coherent reads performed by other shader invocations."
+ *
+ * Together with:
+ * "When executing a shader whose inputs are generated from a previous
+ * programmable stage, the shader invocations from the previous stage
+ * are guaranteed to have executed far enough to generate final values
+ * for all next-stage inputs."
+ *
+ * In order to test this we bind two shader programs at different
+ * execution stages of the pipeline and check that the values written to
+ * an image from the first stage are visible when the same primitive is
+ * dispatched at the second stage. This is repeated for all combinations
+ * of dependent shader stages (what necessarily excludes the compute
+ * shader), for the "coherent" and "volatile" qualifiers (the latter is
+ * supposed to give stricter ordering guarantees), and for different
+ * execution sizes between 4x4 and 1024x1024 to account for
+ * implementations with varying levels of parallelism and with caches of
+ * different sizes.
+ *
+ * Unless running in "quick" mode a series of control tests is
+ * executed repeating the same process without any memory qualifiers.
+ * This is useful to see if the cache is being hit since otherwise the
+ * main test is not meaningful. The control test always passes as it
+ * is expected to misrender.
+ */
+
+#include "common.h"
+
+/** Maximum image width. */
+#define L 1024
+
+/** Maximum number of pixels. */
+#define N (L * L)
+
+PIGLIT_GL_TEST_CONFIG_BEGIN
+
+config.supports_gl_core_version = 32;
+
+config.window_width = L;
+config.window_height = L;
+config.window_visual = PIGLIT_GL_VISUAL_DOUBLE | PIGLIT_GL_VISUAL_RGBA;
+
+PIGLIT_GL_TEST_CONFIG_END
+
+struct image_qualifier_info {
+ /** Test name. */
+ const char *name;
+
+ /** Image qualifier keyword. */
+ const char *qualifier;
+
+ /** Informative "control" test with no memory qualifiers whose
+ * result is ignored. */
+ bool control_test;
+};
+
+const struct image_qualifier_info image_qualifiers[] = {
+ { "control", "", true },
+ { "'coherent' qualifier", "coherent", false },
+ { "'volatile' qualifier", "volatile", false },
+ { 0 }
+};
+
+char *
+qualifier_hunk(const struct image_qualifier_info *qual)
+{
+ char *s = NULL;
+
+ asprintf(&s, "#define IMAGE_Q %s\n", qual->qualifier);
+ return s;
+}
+
+static bool
+init_image(const struct image_info img)
+{
+ static uint32_t pixels[N][4];
+
+ return init_pixels(img, pixels[0], 99, 99, 99, 99) &&
+ upload_image(img, 0, pixels[0]);
+}
+
+static bool
+check(const struct grid_info grid, const struct image_info img)
+{
+ static uint32_t pixels[N][4];
+
+ return download_result(grid, pixels[0]) &&
+ check_pixels(img, pixels[0], 33, 33, 33, 33);
+}
+
+static bool
+run_test(const struct image_qualifier_info *qual,
+ const struct image_stage_info *stage_w,
+ const struct image_stage_info *stage_r,
+ unsigned l)
+{
+ const struct grid_info grid = {
+ stage_w->bit | stage_r->bit,
+ get_image_format(GL_RGBA32UI),
+ { l, l, 1, 1 }
+ };
+ const struct image_info img = image_info_for_grid(grid);
+ GLuint prog = generate_program(
+ grid,
+ /*
+ * Write (11, 22, 33, 44) to some location on the
+ * image from the write stage.
+ */
+ stage_w->stage,
+ concat(qualifier_hunk(qual),
+ image_hunk(img, ""),
+ hunk("IMAGE_Q uniform IMAGE_T img;\n"
+ "\n"
+ "GRID_T op(ivec2 idx, GRID_T x) {\n"
+ " imageStore(img, idx, DATA_T(11, 22, 33, 44));"
+ " return x;"
+ "}\n"), NULL),
+ /*
+ * The same location will read back the expected value
+ * if image access is coherent, as the shader inputs
+ * of the read stage are dependent on the outputs of
+ * the write stage and consequently they are
+ * guaranteed to be executed sequentially.
+ */
+ stage_r->stage,
+ concat(qualifier_hunk(qual),
+ image_hunk(img, ""),
+ hunk("IMAGE_Q uniform IMAGE_T img;\n"
+ "\n"
+ "GRID_T op(ivec2 idx, GRID_T x) {\n"
+ " DATA_T v = imageLoad(img, idx);"
+ " if (v == DATA_T(11, 22, 33, 44))"
+ " return GRID_T(33, 33, 33, 33);"
+ " else"
+ " return GRID_T(77, 77, 77, 77);"
+ "}\n"), NULL));
+ bool ret = prog &&
+ init_fb(grid) &&
+ init_image(img) &&
+ set_uniform_int(prog, "img", 0) &&
+ draw_grid(grid, prog) &&
+ (check(grid, img) || qual->control_test);
+
+ glDeleteProgram(prog);
+ return ret;
+}
+
+void
+piglit_init(int argc, char **argv)
+{
+ const bool quick = (argc >= 2 && !strcmp(argv[1], "--quick"));
+ const struct image_qualifier_info *qual;
+ const struct image_stage_info *stage_w;
+ const struct image_stage_info *stage_r;
+ enum piglit_result status = PIGLIT_PASS;
+ unsigned l;
+
+ piglit_require_extension("GL_ARB_shader_image_load_store");
+
+ for (l = 4; l <= L; l *= (quick ? 8 : 2)) {
+ for (qual = image_qualifiers; qual->name; ++qual) {
+ if (quick && qual->control_test)
+ continue;
+
+ /*
+ * Loop for each pair of shader stages
+ * skipping the compute shader: "coherent"
+ * gives no useful guarantees in that case
+ * since its execution ordering is undefined
+ * with respect to the other shader stages.
+ */
+ for (stage_w = image_stages();
+ stage_w->stage && stage_w->stage != GL_COMPUTE_SHADER;
+ ++stage_w) {
+ for (stage_r = stage_w + 1;
+ stage_r->stage && stage_r->stage != GL_COMPUTE_SHADER;
+ ++stage_r) {
+ subtest(&status, true,
+ run_test(qual, stage_w, stage_r, l),
+ "%s-%s shader/%s coherency test/%dx%d",
+ stage_w->name, stage_r->name,
+ qual->name, l, l);
+ }
+ }
+ }
+ }
+
+ piglit_report_result(status);
+}
+
+enum piglit_result
+piglit_display(void)
+{
+ return PIGLIT_FAIL;
+}
--
2.1.1
More information about the Piglit
mailing list