skeleton for float <-> half optimized procs
Nothing fancy yet, just calls the serial code in a loop.
I will try to folow this up with at least some of:
- SSE2 version of serial code
- NEON version of serial code
- NEON version using vcvt.f32.f16/vcvt.f16.f32
- F16C (between AVX and AVX2) version using vcvtph2ps/vcvtps2ph
The last two are fastest but need runtime detection.
BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1686543003
Review URL: https://codereview.chromium.org/1686543003
diff --git a/tests/Float16Test.cpp b/tests/Float16Test.cpp
index 0a2c3d5..f437268 100644
--- a/tests/Float16Test.cpp
+++ b/tests/Float16Test.cpp
@@ -8,6 +8,7 @@
#include "Test.h"
#include "SkColor.h"
#include "SkHalf.h"
+#include "SkOpts.h"
#include "SkPixmap.h"
static bool eq_within_half_float(float a, float b) {
@@ -50,3 +51,16 @@
}
}
}
+
+DEF_TEST(float_to_half, reporter) {
+ const float fs[] = { 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0 };
+ const uint16_t hs[] = { 0x3c00, 0x4000, 0x4200, 0x4400, 0x4500, 0x4600, 0x4700 };
+
+ uint16_t hscratch[7];
+ SkOpts::float_to_half(hscratch, fs, 7);
+ REPORTER_ASSERT(reporter, 0 == memcmp(hscratch, hs, sizeof(hs)));
+
+ float fscratch[7];
+ SkOpts::half_to_float(fscratch, hs, 7);
+ REPORTER_ASSERT(reporter, 0 == memcmp(fscratch, fs, sizeof(fs)));
+}