skia2/tests/F16StagesTest.cpp
Mike Klein c0bd9f9fe5 rewrite includes to not need so much -Ifoo
Current strategy: everything from the top

Things to look at first are the manual changes:

   - added tools/rewrite_includes.py
   - removed -Idirectives from BUILD.gn
   - various compile.sh simplifications
   - tweak tools/embed_resources.py
   - update gn/find_headers.py to write paths from the top
   - update gn/gn_to_bp.py SkUserConfig.h layout
     so that #include "include/config/SkUserConfig.h" always
     gets the header we want.

No-Presubmit: true
Change-Id: I73a4b181654e0e38d229bc456c0d0854bae3363e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/209706
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Hal Canary <halcanary@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
2019-04-24 16:27:11 +00:00

54 lines
1.9 KiB
C++

/*
* Copyright 2017 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "src/core/SkRasterPipeline.h"
#include "tests/Test.h"
DEF_TEST(F16Stages, r) {
// Make sure SkRasterPipeline::load_f16 and store_f16 can handle a range of
// ordinary (0<=x<=1) and interesting (x<0, x>1) values.
float floats[16] = {
0.0f, 0.25f, 0.5f, 1.0f,
-1.25f, -0.5f, 1.25f, 2.0f,
0,0,0,0, 0,0,0,0, // pad a bit to make sure we qualify for platform-specific code
};
uint64_t halfs[4] = {0,0,0,0};
SkRasterPipeline_MemoryCtx f32 = { floats, 0 },
f16 = { halfs, 0 };
{
SkRasterPipeline_<256> p;
p.append(SkRasterPipeline:: load_f32, &f32);
p.append(SkRasterPipeline::store_f16, &f16);
p.run(0,0,16/4,1);
}
REPORTER_ASSERT(r, ((halfs[0] >> 0) & 0xffff) == 0x0000);
REPORTER_ASSERT(r, ((halfs[0] >> 16) & 0xffff) == 0x3400);
REPORTER_ASSERT(r, ((halfs[0] >> 32) & 0xffff) == 0x3800);
REPORTER_ASSERT(r, ((halfs[0] >> 48) & 0xffff) == 0x3c00);
REPORTER_ASSERT(r, ((halfs[1] >> 0) & 0xffff) == 0xbd00);
REPORTER_ASSERT(r, ((halfs[1] >> 16) & 0xffff) == 0xb800);
REPORTER_ASSERT(r, ((halfs[1] >> 32) & 0xffff) == 0x3d00);
REPORTER_ASSERT(r, ((halfs[1] >> 48) & 0xffff) == 0x4000);
{
SkRasterPipeline_<256> p;
p.append(SkRasterPipeline:: load_f16, &f16);
p.append(SkRasterPipeline::store_f32, &f32);
p.run(0,0,16/4,1);
}
REPORTER_ASSERT(r, floats[0] == 0.00f);
REPORTER_ASSERT(r, floats[1] == 0.25f);
REPORTER_ASSERT(r, floats[2] == 0.50f);
REPORTER_ASSERT(r, floats[3] == 1.00f);
REPORTER_ASSERT(r, floats[4] == -1.25f);
REPORTER_ASSERT(r, floats[5] == -0.50f);
REPORTER_ASSERT(r, floats[6] == 1.25f);
REPORTER_ASSERT(r, floats[7] == 2.00f);
}