Port SkBlitRow::Color32 to SkOpts.

This was a pre-SkOpts attempt that we can bring under its wing now.

This should be a perf no-op, deo volente.

BUG=skia:4117

Review URL: https://codereview.chromium.org/1314863006

10 files changed, 44 insertions, 52 deletions

diff --git a/include/core/SkBlitRow.h b/include/core/SkBlitRow.h
index 8c7665ac54..56121eba78 100644
--- a/include/core/SkBlitRow.h
+++ b/include/core/SkBlitRow.h
@@ -82,9 +82,6 @@ public:
     static ColorProc16 PlatformColorFactory565(unsigned flags);
 
 private:
-    typedef void (*Color32Proc)(SkPMColor[], const SkPMColor[], int, SkPMColor);
-    static Color32Proc PlatformColor32Proc();
-
     enum {
         kFlags16_Mask = 7,
         kFlags32_Mask = 3
diff --git a/src/core/SkBlitRow_D32.cpp b/src/core/SkBlitRow_D32.cpp
index 0f8cbe8104..75330ae337 100644
--- a/src/core/SkBlitRow_D32.cpp
+++ b/src/core/SkBlitRow_D32.cpp
@@ -8,6 +8,7 @@
 #include "SkBlitRow.h"
 #include "SkBlitMask.h"
 #include "SkColorPriv.h"
+#include "SkOpts.h"
 #include "SkUtils.h"
 
 #define UNROLL
@@ -131,32 +132,10 @@ SkBlitRow::Proc32 SkBlitRow::Factory32(unsigned flags) {
     return proc;
 }
 
-#include "Sk4px.h"
-
-// Color32 uses the blend_256_round_alt algorithm from tests/BlendTest.cpp.
-// It's not quite perfect, but it's never wrong in the interesting edge cases,
-// and it's quite a bit faster than blend_perfect.
-//
-// blend_256_round_alt is our currently blessed algorithm.  Please use it or an analogous one.
 void SkBlitRow::Color32(SkPMColor dst[], const SkPMColor src[], int count, SkPMColor color) {
     switch (SkGetPackedA32(color)) {
         case   0: memmove(dst, src, count * sizeof(SkPMColor)); return;
         case 255: sk_memset32(dst, color, count);               return;
     }
-
-    // This Sk4px impl works great on other platforms or when we have NEON.
-#if defined(SK_CPU_ARM32) && !defined(SK_ARM_HAS_NEON)
-    if (auto proc = PlatformColor32Proc()) { return proc(dst, src, count, color); }
-#endif
-
-    unsigned invA = 255 - SkGetPackedA32(color);
-    invA += invA >> 7;
-    SkASSERT(invA < 256);  // We've already handled alpha == 0 above.
-
-    Sk16h colorHighAndRound = Sk4px::DupPMColor(color).widenHi() + Sk16h(128);
-    Sk16b invA_16x(invA);
-
-    Sk4px::MapSrc(count, dst, src, [&](const Sk4px& src4) -> Sk4px {
-        return (src4 * invA_16x).addNarrowHi(colorHighAndRound);
-    });
+    return SkOpts::blit_row_color32(dst, src, count, color);
 }
diff --git a/src/core/SkOpts.cpp b/src/core/SkOpts.cpp
index 69373c0fd1..492fae3a6b 100644
--- a/src/core/SkOpts.cpp
+++ b/src/core/SkOpts.cpp
@@ -10,6 +10,7 @@
 
 #define SK_OPTS_NS sk_default
 #include "SkBlitMask_opts.h"
+#include "SkBlitRow_opts.h"
 #include "SkBlurImageFilter_opts.h"
 #include "SkColorCubeFilter_opts.h"
 #include "SkFloatingPoint_opts.h"
@@ -55,6 +56,8 @@ namespace SkOpts {
 
     decltype(blit_mask_d32_a8) blit_mask_d32_a8 = sk_default::blit_mask_d32_a8;
 
+    decltype(blit_row_color32) blit_row_color32 = sk_default::blit_row_color32;
+
     // Each Init_foo() is defined in src/opts/SkOpts_foo.cpp.
     void Init_ssse3();
     void Init_sse41();
diff --git a/src/core/SkOpts.h b/src/core/SkOpts.h
index 04aa9b1753..d73319a73c 100644
--- a/src/core/SkOpts.h
+++ b/src/core/SkOpts.h
@@ -44,6 +44,7 @@ namespace SkOpts {
     extern bool (*fill_block_dimensions)(SkTextureCompressor::Format, int* x, int* y);
 
     extern void (*blit_mask_d32_a8)(SkPMColor*, size_t, const SkAlpha*, size_t, SkColor, int, int);
+    extern void (*blit_row_color32)(SkPMColor*, const SkPMColor*, int, SkPMColor);
 
     // This function is an optimized version of SkColorCubeFilter::filterSpan
     extern void (*color_cube_filter_span)(const SkPMColor[],
diff --git a/src/opts/SkBlitRow_opts.h b/src/opts/SkBlitRow_opts.h
new file mode 100644
index 0000000000..f000f30fd0
--- /dev/null
+++ b/src/opts/SkBlitRow_opts.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright 2015 Google Inc.
+ *
+ * Use of this source code is governed by a BSD-style license that can be
+ * found in the LICENSE file.
+ */
+
+#ifndef SkBlitRow_opts_DEFINED
+#define SkBlitRow_opts_DEFINED
+
+#include "Sk4px.h"
+
+namespace SK_OPTS_NS {
+
+// Color32 uses the blend_256_round_alt algorithm from tests/BlendTest.cpp.
+// It's not quite perfect, but it's never wrong in the interesting edge cases,
+// and it's quite a bit faster than blend_perfect.
+//
+// blend_256_round_alt is our currently blessed algorithm.  Please use it or an analogous one.
+static void blit_row_color32(SkPMColor* dst, const SkPMColor* src, int count, SkPMColor color) {
+    unsigned invA = 255 - SkGetPackedA32(color);
+    invA += invA >> 7;
+    SkASSERT(invA < 256);  // We've should have already handled alpha == 0 externally.
+
+    Sk16h colorHighAndRound = Sk4px::DupPMColor(color).widenHi() + Sk16h(128);
+    Sk16b invA_16x(invA);
+
+    Sk4px::MapSrc(count, dst, src, [&](const Sk4px& src4) -> Sk4px {
+        return (src4 * invA_16x).addNarrowHi(colorHighAndRound);
+    });
+}
+
+}  // SK_OPTS_NS
+
+#endif//SkBlitRow_opts_DEFINED
diff --git a/src/opts/SkBlitRow_opts_arm.cpp b/src/opts/SkBlitRow_opts_arm.cpp
index 5770c1f1d7..24566ba6c5 100644
--- a/src/opts/SkBlitRow_opts_arm.cpp
+++ b/src/opts/SkBlitRow_opts_arm.cpp
@@ -33,8 +33,3 @@ SkBlitRow::ColorProc16 SkBlitRow::PlatformColorFactory565(unsigned flags) {
 SkBlitRow::Proc32 SkBlitRow::PlatformProcs32(unsigned flags) {
     return SK_ARM_NEON_WRAP(sk_blitrow_platform_32_procs_arm)[flags];
 }
-
-SkBlitRow::Color32Proc SkBlitRow::PlatformColor32Proc() {
-    #define sk_blitrow_color32_arm nullptr
-    return SK_ARM_NEON_WRAP(sk_blitrow_color32_arm);
-}
diff --git a/src/opts/SkBlitRow_opts_arm_neon.cpp b/src/opts/SkBlitRow_opts_arm_neon.cpp
index d5260afd50..1f97971ee4 100644
--- a/src/opts/SkBlitRow_opts_arm_neon.cpp
+++ b/src/opts/SkBlitRow_opts_arm_neon.cpp
@@ -1582,20 +1582,3 @@ const SkBlitRow::Proc32 sk_blitrow_platform_32_procs_arm_neon[] = {
     nullptr
 #endif
 };
-
-#include "Sk4px.h"
-
-void sk_blitrow_color32_arm_neon(SkPMColor* dst, const SkPMColor* src, int count, SkPMColor color) {
-    // Until it becomes a little more reasonable to assume we'll be built with NEON,
-    // we copy our Sk4px implementation of SkBlitRow::Color32 here so it picks up NEON at runtime.
-    unsigned invA = 255 - SkGetPackedA32(color);
-    invA += invA >> 7;
-    SkASSERT(invA < 256);  // Our caller has already handled the alpha == 0 case.
-
-    Sk16h colorHighAndRound = Sk4px::DupPMColor(color).widenHi() + Sk16h(128);
-    Sk16b invA_16x(invA);
-
-    Sk4px::MapSrc(count, dst, src, [&](const Sk4px& src4) -> Sk4px {
-        return (src4 * invA_16x).addNarrowHi(colorHighAndRound);
-    });
-}
diff --git a/src/opts/SkBlitRow_opts_arm_neon.h b/src/opts/SkBlitRow_opts_arm_neon.h
index fb3d2e0a7b..159a466c98 100644
--- a/src/opts/SkBlitRow_opts_arm_neon.h
+++ b/src/opts/SkBlitRow_opts_arm_neon.h
@@ -13,6 +13,4 @@ extern const SkBlitRow::Proc16 sk_blitrow_platform_565_procs_arm_neon[];
 extern const SkBlitRow::ColorProc16 sk_blitrow_platform_565_colorprocs_arm_neon[];
 extern const SkBlitRow::Proc32 sk_blitrow_platform_32_procs_arm_neon[];
 
-extern void sk_blitrow_color32_arm_neon(SkPMColor[], const SkPMColor[], int, SkPMColor);
-
 #endif
diff --git a/src/opts/SkBlitRow_opts_none.cpp b/src/opts/SkBlitRow_opts_none.cpp
index 03298e888e..a9abe06580 100644
--- a/src/opts/SkBlitRow_opts_none.cpp
+++ b/src/opts/SkBlitRow_opts_none.cpp
@@ -20,5 +20,3 @@ SkBlitRow::ColorProc16 SkBlitRow::PlatformColorFactory565(unsigned flags) {
 SkBlitRow::Proc32 SkBlitRow::PlatformProcs32(unsigned flags) {
     return nullptr;
 }
-
-SkBlitRow::Color32Proc SkBlitRow::PlatformColor32Proc() { return nullptr; }
diff --git a/src/opts/SkOpts_neon.cpp b/src/opts/SkOpts_neon.cpp
index a05df66762..802d83f69c 100644
--- a/src/opts/SkOpts_neon.cpp
+++ b/src/opts/SkOpts_neon.cpp
@@ -9,6 +9,7 @@
 
 #define SK_OPTS_NS sk_neon
 #include "SkBlitMask_opts.h"
+#include "SkBlitRow_opts.h"
 #include "SkBlurImageFilter_opts.h"
 #include "SkColorCubeFilter_opts.h"
 #include "SkFloatingPoint_opts.h"
@@ -38,6 +39,8 @@ namespace SkOpts {
 
         blit_mask_d32_a8 = sk_neon::blit_mask_d32_a8;
 
+        blit_row_color32 = sk_neon::blit_row_color32;
+
         color_cube_filter_span = sk_neon::color_cube_filter_span;
     }
 }