SkRasterPipeline: 8x pipelines

Bench runtime changes:
sRGB: 7194 -> 3735  = 1.93x faster 
F16:  6531 -> 2559  = 2.55x faster

Instead of building 4x and 1-3x pipelines and then maybe 8x and 1-7x, instead build either the short ones or the long ones, but not both.  If we just take care to use a compatible run_pipeline(), there's some cross-module type disagreement but everything works out in the end.

Oddly, a few places that looked like they'd be faster using SkNx_fma() or Sk4f_round()/Sk8f_round() are actually faster the long way, e.g. multiply, add 0.5, truncate.  Curious!  In all the other places you see here that I've used SkNx_fma(), it's been a significant speedup.

This folds in a couple refactors and cleanups that I've been meaning to do.  Hope you don't mind... if find the new code considerably easier to read than the old code.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2990
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Change-Id: I1c82e5755d8e44cc0b9c6673d04b117f85d71a3a
Reviewed-on: https://skia-review.googlesource.com/2990
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

1 files changed, 37 insertions, 33 deletions

diff --git a/src/core/SkOpts.cpp b/src/core/SkOpts.cpp
index 7784e7fcf1..a57808da60 100644
--- a/src/core/SkOpts.cpp
+++ b/src/core/SkOpts.cpp
@@ -88,30 +88,32 @@ namespace SkOpts {
     DEFINE_DEFAULT(srcover_srgb_srgb);
 
     DEFINE_DEFAULT(hash_fn);
+
+    DEFINE_DEFAULT(run_pipeline);
 #undef DEFINE_DEFAULT
 
-    // TODO: might be nice to only create one instance of tail-insensitive stages.
+    SkRasterPipeline::Fn body[] = {
+        SK_OPTS_NS::just_return,
 
-    SkRasterPipeline::Fn stages_4[] = {
-        stage_4<SK_OPTS_NS::store_565 , false>,
-        stage_4<SK_OPTS_NS::store_srgb, false>,
-        stage_4<SK_OPTS_NS::store_f16 , false>,
+        SK_OPTS_NS::store_565,
+        SK_OPTS_NS::store_srgb,
+        SK_OPTS_NS::store_f16,
 
-        stage_4<SK_OPTS_NS::load_s_565 , true>,
-        stage_4<SK_OPTS_NS::load_s_srgb, true>,
-        stage_4<SK_OPTS_NS::load_s_f16 , true>,
+        SK_OPTS_NS::load_s_565,
+        SK_OPTS_NS::load_s_srgb,
+        SK_OPTS_NS::load_s_f16,
 
-        stage_4<SK_OPTS_NS::load_d_565 , true>,
-        stage_4<SK_OPTS_NS::load_d_srgb, true>,
-        stage_4<SK_OPTS_NS::load_d_f16 , true>,
+        SK_OPTS_NS::load_d_565,
+        SK_OPTS_NS::load_d_srgb,
+        SK_OPTS_NS::load_d_f16,
 
-        stage_4<SK_OPTS_NS::scale_u8, true>,
+        SK_OPTS_NS::scale_u8,
 
-        stage_4<SK_OPTS_NS::lerp_u8            , true>,
-        stage_4<SK_OPTS_NS::lerp_565           , true>,
-        stage_4<SK_OPTS_NS::lerp_constant_float, true>,
+        SK_OPTS_NS::lerp_u8,
+        SK_OPTS_NS::lerp_565,
+        SK_OPTS_NS::lerp_constant_float,
 
-        stage_4<SK_OPTS_NS::constant_color, true>,
+        SK_OPTS_NS::constant_color,
 
         SK_OPTS_NS::dst,
         SK_OPTS_NS::dstatop,
@@ -138,28 +140,30 @@ namespace SkOpts {
         SK_OPTS_NS::overlay,
         SK_OPTS_NS::softlight,
     };
-    static_assert(SK_ARRAY_COUNT(stages_4) == SkRasterPipeline::kNumStockStages, "");
+    static_assert(SK_ARRAY_COUNT(body) == SkRasterPipeline::kNumStockStages, "");
+
+    SkRasterPipeline::Fn tail[] = {
+        SK_OPTS_NS::just_return,
 
-    SkRasterPipeline::Fn stages_1_3[] = {
-        stage_1_3<SK_OPTS_NS::store_565 , false>,
-        stage_1_3<SK_OPTS_NS::store_srgb, false>,
-        stage_1_3<SK_OPTS_NS::store_f16 , false>,
+        SK_OPTS_NS::store_565_tail,
+        SK_OPTS_NS::store_srgb_tail,
+        SK_OPTS_NS::store_f16_tail,
 
-        stage_1_3<SK_OPTS_NS::load_s_565 , true>,
-        stage_1_3<SK_OPTS_NS::load_s_srgb, true>,
-        stage_1_3<SK_OPTS_NS::load_s_f16 , true>,
+        SK_OPTS_NS::load_s_565_tail,
+        SK_OPTS_NS::load_s_srgb_tail,
+        SK_OPTS_NS::load_s_f16_tail,
 
-        stage_1_3<SK_OPTS_NS::load_d_565 , true>,
-        stage_1_3<SK_OPTS_NS::load_d_srgb, true>,
-        stage_1_3<SK_OPTS_NS::load_d_f16 , true>,
+        SK_OPTS_NS::load_d_565_tail,
+        SK_OPTS_NS::load_d_srgb_tail,
+        SK_OPTS_NS::load_d_f16_tail,
 
-        stage_1_3<SK_OPTS_NS::scale_u8, true>,
+        SK_OPTS_NS::scale_u8_tail,
 
-        stage_1_3<SK_OPTS_NS::lerp_u8            , true>,
-        stage_1_3<SK_OPTS_NS::lerp_565           , true>,
-        stage_1_3<SK_OPTS_NS::lerp_constant_float, true>,
+        SK_OPTS_NS::lerp_u8_tail,
+        SK_OPTS_NS::lerp_565_tail,
+        SK_OPTS_NS::lerp_constant_float,
 
-        stage_1_3<SK_OPTS_NS::constant_color, true>,
+        SK_OPTS_NS::constant_color,
 
         SK_OPTS_NS::dst,
         SK_OPTS_NS::dstatop,
@@ -186,7 +190,7 @@ namespace SkOpts {
         SK_OPTS_NS::overlay,
         SK_OPTS_NS::softlight,
     };
-    static_assert(SK_ARRAY_COUNT(stages_1_3) == SkRasterPipeline::kNumStockStages, "");
+    static_assert(SK_ARRAY_COUNT(tail) == SkRasterPipeline::kNumStockStages, "");
 
     // Each Init_foo() is defined in src/opts/SkOpts_foo.cpp.
     void Init_ssse3();