sknx refactoring

  - trim unused specializations (Sk4i, Sk2d) and apis (SkNx_dup)
  - expand apis a little
    * v[0] == v.kth<0>()
    * SkNx_shuffle can now convert to different-sized vectors, e.g. Sk2f <-> Sk4f
  - remove anonymous namespace

I believe it's safe to remove the anonymous namespace right now.
We're worried about violating the One Definition Rule; the anonymous namespace protected us from that.

In Release builds, this is mostly moot, as everything tends to inline completely.
In Debug builds, violating the ODR is at worst an inconvenience, time spent trying to figure out why the bot is broken.

Now that we're building with SSE2/NEON everywhere, very few bots have even a chance about getting confused by two definitions of the same type or function.  Where we do compile variants depending on, e.g., SSSE3, we do so in static inline functions.  These are not subject to the ODR.

I plan to follow up with a tedious .kth<...>() -> [...] auto-replace.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1683543002
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Review URL: https://codereview.chromium.org/1683543002

1 files changed, 16 insertions, 20 deletions

diff --git a/tests/SkNxTest.cpp b/tests/SkNxTest.cpp
index 49e920f4b3..13e0a9f61f 100644
--- a/tests/SkNxTest.cpp
+++ b/tests/SkNxTest.cpp
@@ -146,12 +146,12 @@ DEF_TEST(SkNi_min_lt, r) {
     for (int i = 0; i < (1<<16); i++) {
         uint16_t a = rand.nextU() >> 16,
                  b = rand.nextU() >> 16;
-        REPORTER_ASSERT(r, Sk8h::Min(Sk8h(a), Sk8h(b)).kth<0>() == SkTMin(a, b));
+        REPORTER_ASSERT(r, Sk16h::Min(Sk16h(a), Sk16h(b)).kth<0>() == SkTMin(a, b));
     }
 #else
     for (int a = 0; a < (1<<16); a++) {
     for (int b = 0; b < (1<<16); b++) {
-        REPORTER_ASSERT(r, Sk8h::Min(Sk8h(a), Sk8h(b)).kth<0>() == SkTMin(a, b));
+        REPORTER_ASSERT(r, Sk16h::Min(Sk16h(a), Sk16h(b)).kth<0>() == SkTMin(a, b));
     }}
 #endif
 }
@@ -207,16 +207,6 @@ DEF_TEST(Sk4px_widening, r) {
     REPORTER_ASSERT(r, 0 == memcmp(&wideLoHi, &wideLoHiAlt, sizeof(wideLoHi)));
 }
 
-DEF_TEST(SkNx_cast, r) {
-    Sk4f fs(-1.7f, -1.4f, 0.5f, 1.9f);
-    Sk4i is = SkNx_cast<int>(fs);
-
-    REPORTER_ASSERT(r, is.kth<0>() == -1);
-    REPORTER_ASSERT(r, is.kth<1>() == -1);
-    REPORTER_ASSERT(r, is.kth<2>() ==  0);
-    REPORTER_ASSERT(r, is.kth<3>() ==  1);
-}
-
 DEF_TEST(SkNx_abs, r) {
     auto fs = Sk4f(0.0f, -0.0f, 2.0f, -4.0f).abs();
     REPORTER_ASSERT(r, fs.kth<0>() == 0.0f);
@@ -225,20 +215,27 @@ DEF_TEST(SkNx_abs, r) {
     REPORTER_ASSERT(r, fs.kth<3>() == 4.0f);
 }
 
-#include "SkRandom.h"
+DEF_TEST(SkNx_shuffle, r) {
+    Sk4f f4(0,10,20,30);
 
-static void dump(const Sk4f& f4, const Sk4h& h4) {
-    SkDebugf("%g %g %g %g --> %d %d %d %d\n",
-             f4.kth<0>(), f4.kth<1>(), f4.kth<2>(), f4.kth<3>(),
-             h4.kth<0>(), h4.kth<1>(), h4.kth<2>(), h4.kth<3>());
+    Sk2f f2 = SkNx_shuffle<2,1>(f4);
+    REPORTER_ASSERT(r, f2[0] == 20);
+    REPORTER_ASSERT(r, f2[1] == 10);
+
+    f4 = SkNx_shuffle<0,1,1,0>(f2);
+    REPORTER_ASSERT(r, f4[0] == 20);
+    REPORTER_ASSERT(r, f4[1] == 10);
+    REPORTER_ASSERT(r, f4[2] == 10);
+    REPORTER_ASSERT(r, f4[3] == 20);
 }
 
+#include "SkRandom.h"
+
 DEF_TEST(SkNx_u16_float, r) {
     {
         // u16 --> float
         auto h4 = Sk4h(15, 17, 257, 65535);
         auto f4 = SkNx_cast<float>(h4);
-        dump(f4, h4);
         REPORTER_ASSERT(r, f4.kth<0>() == 15.0f);
         REPORTER_ASSERT(r, f4.kth<1>() == 17.0f);
         REPORTER_ASSERT(r, f4.kth<2>() == 257.0f);
@@ -248,7 +245,6 @@ DEF_TEST(SkNx_u16_float, r) {
         // float -> u16
         auto f4 = Sk4f(15, 17, 257, 65535);
         auto h4 = SkNx_cast<uint16_t>(f4);
-        dump(f4, h4);
         REPORTER_ASSERT(r, h4.kth<0>() == 15);
         REPORTER_ASSERT(r, h4.kth<1>() == 17);
         REPORTER_ASSERT(r, h4.kth<2>() == 257);
@@ -258,7 +254,7 @@ DEF_TEST(SkNx_u16_float, r) {
     // starting with any u16 value, we should be able to have a perfect round-trip in/out of floats
     //
     SkRandom rand;
-    for (int i = 0; i < 0; ++i) {
+    for (int i = 0; i < 10000; ++i) {
         const uint16_t s16[4] {
             (uint16_t)rand.nextU16(), (uint16_t)rand.nextU16(),
             (uint16_t)rand.nextU16(), (uint16_t)rand.nextU16(),