simplify by removing _d stages

CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD

Change-Id: I75e232faee6ad48f65bac5b119a461280b27bbc8
Reviewed-on: https://skia-review.googlesource.com/6661
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

6 files changed, 34 insertions, 52 deletions

diff --git a/bench/SkRasterPipelineBench.cpp b/bench/SkRasterPipelineBench.cpp
index 376efdeb84..6d9bb49f50 100644
--- a/bench/SkRasterPipelineBench.cpp
+++ b/bench/SkRasterPipelineBench.cpp
@@ -45,13 +45,14 @@ public:
         p.append(SkRasterPipeline::load_8888, &src_ctx);
         p.append_from_srgb(kUnpremul_SkAlphaType);
         p.append(SkRasterPipeline::scale_u8, &mask_ctx);
+        p.append(SkRasterPipeline::move_src_dst);
         if (kF16) {
-            p.append(SkRasterPipeline::load_f16_d, &dst_ctx);
+            p.append(SkRasterPipeline::load_f16, &dst_ctx);
         } else {
-            p.append(SkRasterPipeline::load_8888_d, &dst_ctx);
-            p.append_from_srgb_d(kPremul_SkAlphaType);
+            p.append(SkRasterPipeline::load_8888, &dst_ctx);
+            p.append_from_srgb(kPremul_SkAlphaType);
         }
-        p.append(SkRasterPipeline::srcover);
+        p.append(SkRasterPipeline::dstover);
         if (kF16) {
             p.append(SkRasterPipeline::store_f16, &dst_ctx);
         } else {
diff --git a/src/core/SkRasterPipeline.cpp b/src/core/SkRasterPipeline.cpp
index 00ba81e6f3..7b3d49d413 100644
--- a/src/core/SkRasterPipeline.cpp
+++ b/src/core/SkRasterPipeline.cpp
@@ -11,14 +11,7 @@
 SkRasterPipeline::SkRasterPipeline() {}
 
 void SkRasterPipeline::append(StockStage stage, void* ctx) {
-#ifdef SK_DEBUG
-    switch (stage) {
-        case from_srgb:
-        case from_srgb_d:
-            SkDEBUGFAIL("Please use append_srgb[_d]() instead.");
-        default: break;
-    }
-#endif
+    SkASSERT(stage != from_srgb);
     fStages.push_back({stage, ctx});
 }
 
@@ -67,12 +60,3 @@ void SkRasterPipeline::append_from_srgb(SkAlphaType at) {
         this->append(SkRasterPipeline::clamp_a);
     }
 }
-
-void SkRasterPipeline::append_from_srgb_d(SkAlphaType at) {
-    //this->append(from_srgb_d);
-    fStages.push_back({from_srgb_d, nullptr});
-
-    if (at == kPremul_SkAlphaType) {
-        this->append(SkRasterPipeline::clamp_a_d);
-    }
-}
diff --git a/src/core/SkRasterPipeline.h b/src/core/SkRasterPipeline.h
index 3b18b73470..61a44eb1db 100644
--- a/src/core/SkRasterPipeline.h
+++ b/src/core/SkRasterPipeline.h
@@ -58,16 +58,16 @@
 
 #define SK_RASTER_PIPELINE_STAGES(M)                             \
     M(trace) M(registers)                                        \
-    M(move_src_dst) M(move_dst_src) M(swap_rb) M(swap_rb_d)      \
-    M(clamp_0) M(clamp_1) M(clamp_a) M(clamp_a_d)                \
+    M(move_src_dst) M(move_dst_src) M(swap)                      \
+    M(clamp_0) M(clamp_1) M(clamp_a)                             \
     M(unpremul) M(premul)                                        \
-    M(set_rgb)                                                   \
-    M(from_srgb) M(from_srgb_d) M(to_srgb)                       \
+    M(set_rgb) M(swap_rb)                                        \
+    M(from_srgb) M(to_srgb)                                      \
     M(from_2dot2) M(to_2dot2)                                    \
     M(constant_color) M(store_f32)                               \
-    M(load_565)  M(load_565_d)  M(store_565)                     \
-    M(load_f16)  M(load_f16_d)  M(store_f16)                     \
-    M(load_8888) M(load_8888_d) M(store_8888)                    \
+    M(load_565)  M(store_565)                                    \
+    M(load_f16)  M(store_f16)                                    \
+    M(load_8888) M(store_8888)                                   \
     M(load_tables) M(store_tables)                               \
     M(scale_u8) M(scale_1_float)                                 \
     M(lerp_u8) M(lerp_565) M(lerp_1_float)                       \
@@ -122,8 +122,7 @@ public:
 
     // Conversion from sRGB can be subtly tricky when premultiplication is involved.
     // Use these helpers to keep things sane.
-    void append_from_srgb  (SkAlphaType);
-    void append_from_srgb_d(SkAlphaType);
+    void append_from_srgb(SkAlphaType);
 
 private:
     std::vector<Stage> fStages;
diff --git a/src/core/SkRasterPipelineBlitter.cpp b/src/core/SkRasterPipelineBlitter.cpp
index cbdb09faa4..9fad0b4752 100644
--- a/src/core/SkRasterPipelineBlitter.cpp
+++ b/src/core/SkRasterPipelineBlitter.cpp
@@ -186,19 +186,21 @@ SkBlitter* SkRasterPipelineBlitter::Create(const SkPixmap& dst,
 void SkRasterPipelineBlitter::append_load_d(SkRasterPipeline* p) const {
     SkASSERT(supported(fDst.info()));
 
+    p->append(SkRasterPipeline::move_src_dst);
     switch (fDst.info().colorType()) {
-        case kRGB_565_SkColorType:   p->append(SkRasterPipeline::load_565_d,  &fDstPtr); break;
+        case kRGB_565_SkColorType:   p->append(SkRasterPipeline::load_565,  &fDstPtr); break;
         case kBGRA_8888_SkColorType:
-        case kRGBA_8888_SkColorType: p->append(SkRasterPipeline::load_8888_d, &fDstPtr); break;
-        case kRGBA_F16_SkColorType:  p->append(SkRasterPipeline::load_f16_d,  &fDstPtr); break;
+        case kRGBA_8888_SkColorType: p->append(SkRasterPipeline::load_8888, &fDstPtr); break;
+        case kRGBA_F16_SkColorType:  p->append(SkRasterPipeline::load_f16,  &fDstPtr); break;
         default: break;
     }
     if (fDst.info().colorType() == kBGRA_8888_SkColorType) {
-        p->append(SkRasterPipeline::swap_rb_d);
+        p->append(SkRasterPipeline::swap_rb);
     }
     if (fDst.info().gammaCloseToSRGB()) {
-        p->append_from_srgb_d(fDst.info().alphaType());
+        p->append_from_srgb(fDst.info().alphaType());
     }
+    p->append(SkRasterPipeline::swap);
 }
 
 void SkRasterPipelineBlitter::append_store(SkRasterPipeline* p) const {
diff --git a/src/opts/SkRasterPipeline_opts.h b/src/opts/SkRasterPipeline_opts.h
index 9721e2abef..143a9ece79 100644
--- a/src/opts/SkRasterPipeline_opts.h
+++ b/src/opts/SkRasterPipeline_opts.h
@@ -44,8 +44,10 @@ namespace {
     //    &load_8888
     //    (src ptr)
     //    &from_srgb
-    //    &load_f16_d
+    //    &move_src_dst
+    //    &load_f16
     //    (dst ptr)
+    //    &swap
     //    &srcover
     //    &store_f16
     //    (dst ptr)
@@ -92,12 +94,6 @@ SI void SK_VECTORCALL just_return(size_t, void**, SkNf, SkNf, SkNf, SkNf,
         name##_kernel(x_tail/N, x_tail%N, r,g,b,a, dr,dg,db,da);                         \
         next(x_tail,p, r,g,b,a, dr,dg,db,da);                                            \
     }                                                                                    \
-    SI void SK_VECTORCALL name##_d(size_t x_tail, void** p,                              \
-                                   SkNf  r, SkNf  g, SkNf  b, SkNf  a,                   \
-                                   SkNf dr, SkNf dg, SkNf db, SkNf da) {                 \
-        name##_kernel(x_tail/N, x_tail%N, dr,dg,db,da, r,g,b,a);                         \
-        next(x_tail,p, r,g,b,a, dr,dg,db,da);                                            \
-    }                                                                                    \
     static SK_ALWAYS_INLINE void name##_kernel(size_t x, size_t tail,                    \
                                                SkNf&  r, SkNf&  g, SkNf&  b, SkNf&  a,   \
                                                SkNf& dr, SkNf& dg, SkNf& db, SkNf& da)
@@ -113,13 +109,6 @@ SI void SK_VECTORCALL just_return(size_t, void**, SkNf, SkNf, SkNf, SkNf,
         name##_kernel(ctx, x_tail/N, x_tail%N, r,g,b,a, dr,dg,db,da);                    \
         next(x_tail,p, r,g,b,a, dr,dg,db,da);                                            \
     }                                                                                    \
-    SI void SK_VECTORCALL name##_d(size_t x_tail, void** p,                              \
-                                   SkNf  r, SkNf  g, SkNf  b, SkNf  a,                   \
-                                   SkNf dr, SkNf dg, SkNf db, SkNf da) {                 \
-        auto ctx = (Ctx)load_and_increment(&p);                                          \
-        name##_kernel(ctx, x_tail/N, x_tail%N, dr,dg,db,da, r,g,b,a);                    \
-        next(x_tail,p, r,g,b,a, dr,dg,db,da);                                            \
-    }                                                                                    \
     static SK_ALWAYS_INLINE void name##_kernel(Ctx ctx, size_t x, size_t tail,           \
                                                SkNf&  r, SkNf&  g, SkNf&  b, SkNf&  a,   \
                                                SkNf& dr, SkNf& dg, SkNf& db, SkNf& da)
@@ -380,6 +369,7 @@ STAGE_CTX(set_rgb, const float*) {
     g = ctx[1];
     b = ctx[2];
 }
+STAGE(swap_rb) { SkTSwap(r,b); }
 
 STAGE(move_src_dst) {
     dr = r;
@@ -393,8 +383,12 @@ STAGE(move_dst_src) {
     b = db;
     a = da;
 }
-
-STAGE(swap_rb) { SkTSwap(r, b); }
+STAGE(swap) {
+    SkTSwap(r,dr);
+    SkTSwap(g,dg);
+    SkTSwap(b,db);
+    SkTSwap(a,da);
+}
 
 STAGE(from_srgb) {
     r = sk_linear_from_srgb_math(r);
diff --git a/tests/SkRasterPipelineTest.cpp b/tests/SkRasterPipelineTest.cpp
index 50c417f5f0..ddd84e7cce 100644
--- a/tests/SkRasterPipelineTest.cpp
+++ b/tests/SkRasterPipelineTest.cpp
@@ -22,7 +22,9 @@ DEF_TEST(SkRasterPipeline, r) {
 
     SkRasterPipeline p;
     p.append(SkRasterPipeline::load_f16, &load_s_ctx);
-    p.append(SkRasterPipeline::load_f16_d, &load_d_ctx);
+    p.append(SkRasterPipeline::move_src_dst);
+    p.append(SkRasterPipeline::load_f16, &load_d_ctx);
+    p.append(SkRasterPipeline::swap);
     p.append(SkRasterPipeline::srcover);
     p.append(SkRasterPipeline::store_f16, &store_ctx);
     p.run(0,0, 1);