Add an enum layer of indirection for stock raster pipeline stages.

This is handy now, and becomes necessary with fancier backends:
   - most code can't speak the type of AVX pipeline stages,
     so indirection's definitely needed there;
   - if the pipleine is entirely composed of stock stages,
     these enum values become an abstract recipe that can be JITted.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2782
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Change-Id: Iedd62e99ce39e94cf3e6ffc78c428f0ccc182342
Reviewed-on: https://skia-review.googlesource.com/2782
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

2 files changed, 27 insertions, 29 deletions

diff --git a/src/opts/SkOpts_sse41.cpp b/src/opts/SkOpts_sse41.cpp
index e4e3246780..0509b91fdf 100644
--- a/src/opts/SkOpts_sse41.cpp
+++ b/src/opts/SkOpts_sse41.cpp
@@ -22,33 +22,31 @@ namespace SkOpts {
         blit_row_s32a_opaque = sse41::blit_row_s32a_opaque;
 
     #define STAGE(stage, kCallNext) \
-        stage = body<SK_OPTS_NS::stage, kCallNext>
+        stages_4  [SkRasterPipeline::stage] = stage_4  <SK_OPTS_NS::stage, kCallNext>; \
+        stages_1_3[SkRasterPipeline::stage] = stage_1_3<SK_OPTS_NS::stage, kCallNext>
 
-        STAGE(srcover, true);
-        STAGE(constant_color, true);
-        STAGE(lerp_constant_float, true);
-    #undef STAGE
+        STAGE(store_565 , false);
+        STAGE(store_srgb, false);
+        STAGE(store_f16 , false);
 
-    #define STAGE(stage, kCallNext) \
-        stage##_body = body<SK_OPTS_NS::stage, kCallNext>; \
-        stage##_tail = tail<SK_OPTS_NS::stage, kCallNext>
+        STAGE(load_s_565 , true);
+        STAGE(load_s_srgb, true);
+        STAGE(load_s_f16 , true);
 
+        STAGE(load_d_565 , true);
         STAGE(load_d_srgb, true);
-        STAGE(load_s_srgb, true);
-        STAGE( store_srgb, false);
+        STAGE(load_d_f16 , true);
 
-        STAGE(load_d_f16, true);
-        STAGE(load_s_f16, true);
-        STAGE( store_f16, false);
+        STAGE(scale_u8, true);
 
-        STAGE(load_d_565, true);
-        STAGE(load_s_565, true);
-        STAGE( store_565, false);
+        STAGE(lerp_u8            , true);
+        STAGE(lerp_565           , true);
+        STAGE(lerp_constant_float, true);
 
-        STAGE(scale_u8, true);
+        STAGE(constant_color, true);
 
-        STAGE(lerp_u8,  true);
-        STAGE(lerp_565, true);
+        STAGE(srcover, true);
     #undef STAGE
+
     }
 }
diff --git a/src/opts/SkRasterPipeline_opts.h b/src/opts/SkRasterPipeline_opts.h
index 70c4d0c225..fe0fde8139 100644
--- a/src/opts/SkRasterPipeline_opts.h
+++ b/src/opts/SkRasterPipeline_opts.h
@@ -25,26 +25,26 @@ using Kernel_Sk4f = void(void*, size_t, size_t, Sk4f&, Sk4f&, Sk4f&, Sk4f&,
 
 
 template <Kernel_Sk4f kernel, bool kCallNext>
-static inline void SK_VECTORCALL body(SkRasterPipeline::Stage* st, size_t x, size_t t,
-                                      Sk4f  r, Sk4f  g, Sk4f  b, Sk4f  a,
-                                      Sk4f dr, Sk4f dg, Sk4f db, Sk4f da) {
+static inline void SK_VECTORCALL stage_4(SkRasterPipeline::Stage* st, size_t x, size_t tail,
+                                         Sk4f  r, Sk4f  g, Sk4f  b, Sk4f  a,
+                                         Sk4f dr, Sk4f dg, Sk4f db, Sk4f da) {
     // Passing 0 lets the optimizer completely drop any "if (tail) {...}" code in kernel.
     kernel(st->ctx<void*>(), x,0, r,g,b,a, dr,dg,db,da);
     if (kCallNext) {
-        st->next(x,t, r,g,b,a, dr,dg,db,da);  // It's faster to pass t here than 0.
+        st->next(x,tail, r,g,b,a, dr,dg,db,da);  // It's faster to pass t here than 0.
     }
 }
 
 template <Kernel_Sk4f kernel, bool kCallNext>
-static inline void SK_VECTORCALL tail(SkRasterPipeline::Stage* st, size_t x, size_t t,
-                                      Sk4f  r, Sk4f  g, Sk4f  b, Sk4f  a,
-                                      Sk4f dr, Sk4f dg, Sk4f db, Sk4f da) {
+static inline void SK_VECTORCALL stage_1_3(SkRasterPipeline::Stage* st, size_t x, size_t tail,
+                                           Sk4f  r, Sk4f  g, Sk4f  b, Sk4f  a,
+                                           Sk4f dr, Sk4f dg, Sk4f db, Sk4f da) {
 #if defined(__clang__)
-    __builtin_assume(t > 0);  // This flourish lets Clang compile away any tail==0 code.
+    __builtin_assume(tail > 0);  // This flourish lets Clang compile away any tail==0 code.
 #endif
-    kernel(st->ctx<void*>(), x,t, r,g,b,a, dr,dg,db,da);
+    kernel(st->ctx<void*>(), x,tail, r,g,b,a, dr,dg,db,da);
     if (kCallNext) {
-        st->next(x,t, r,g,b,a, dr,dg,db,da);
+        st->next(x,tail, r,g,b,a, dr,dg,db,da);
     }
 }