Reland: Finish up mip opts: 2x3 and 3x2

Original CL:
https://skia-review.googlesource.com/c/10028

mipmap_build_2048x2047 (2x3):
sRGB Float Impl (old): 82.9ms (reference)
8888                 : 12.3ms (reference)
sRGB Before          : 61.9ms
sRGB After           : 53.1ms

mipmap_build_2047x2048 (3x2):
sRGB Float Impl (old): 65.9ms (reference)
8888 Before          : 10.3ms
8888 After           : 8.81ms
sRGB Before          : 47.8ms
sRGB After           : 43.5ms

BUG=skia:

Change-Id: Icf58a857a66102e843ae2585fda8d70c31a2b34d
Reviewed-on: https://skia-review.googlesource.com/13186
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>

diff --git a/bench/MipMapBench.cpp b/bench/MipMapBench.cpp
index cc5c8ae..9b6fc27 100644 (file)
--- a/bench/MipMapBench.cpp
+++ b/bench/MipMapBench.cpp
@@ -71,3 +71,9 @@ DEF_BENCH( return new MipMapBench(2048, 2048,
 DEF_BENCH( return new MipMapBench(2047, 2047, SkDestinationSurfaceColorMode::kLegacy); )
 DEF_BENCH( return new MipMapBench(2047, 2047,
                                   SkDestinationSurfaceColorMode::kGammaAndColorSpaceAware); )
+DEF_BENCH( return new MipMapBench(2048, 2047, SkDestinationSurfaceColorMode::kLegacy); )
+DEF_BENCH( return new MipMapBench(2048, 2047,
+                                  SkDestinationSurfaceColorMode::kGammaAndColorSpaceAware); )
+DEF_BENCH( return new MipMapBench(2047, 2048, SkDestinationSurfaceColorMode::kLegacy); )
+DEF_BENCH( return new MipMapBench(2047, 2048,
+                                  SkDestinationSurfaceColorMode::kGammaAndColorSpaceAware); )

diff --git a/src/core/SkMipMap.cpp b/src/core/SkMipMap.cpp
index 03341c6..d15d2c9 100644 (file)
--- a/src/core/SkMipMap.cpp
+++ b/src/core/SkMipMap.cpp
@@ -242,18 +242,30 @@ template <typename F> void downsample_3_2(void* dst, const void* src, size_t src
     auto p1 = (const typename F::Type*)((const char*)p0 + srcRB);
     auto d = static_cast<typename F::Type*>(dst);
 
-    auto c02 = F::Expand(p0[0]);
-    auto c12 = F::Expand(p1[0]);
+    // Given pixels:
+    // a0 b0 c0 d0 e0 ...
+    // a1 b1 c1 d1 e1 ...
+    // We want:
+    // (a0 + 2*b0 + c0 + a1 + 2*b1 + c1) / 8
+    // (c0 + 2*d0 + e0 + c1 + 2*d1 + e1) / 8
+    // ...
+
+    auto c0 = F::Expand(p0[0]);
+    auto c1 = F::Expand(p1[0]);
+    auto c = c0 + c1;
     for (int i = 0; i < count; ++i) {
-        auto c00 = c02;
-        auto c01 = F::Expand(p0[1]);
-             c02 = F::Expand(p0[2]);
-        auto c10 = c12;
-        auto c11 = F::Expand(p1[1]);
-             c12 = F::Expand(p1[2]);
+        auto a = c;
 
-        auto c = add_121(c00, c01, c02) + add_121(c10, c11, c12);
-        d[i] = F::Compact(shift_right(c, 3));
+        auto b0 = F::Expand(p0[1]);
+        auto b1 = F::Expand(p1[1]);
+        auto b = b0 + b0 + b1 + b1;
+
+        c0 = F::Expand(p0[2]);
+        c1 = F::Expand(p1[2]);
+        c = c0 + c1;
+
+        auto sum = a + b + c;
+        d[i] = F::Compact(shift_right(sum, 3));
         p0 += 2;
         p1 += 2;
     }
@@ -324,7 +336,7 @@ void downsample_2_2_srgb(void* dst, const void* src, size_t srcRB, int count) {
                          sk_linear12_from_srgb[p0[ 8]],
                          sk_linear12_from_srgb[p0[ 9]],
                          sk_linear12_from_srgb[p0[10]],
-                         p0[11] << 4               );
+                         p0[11] << 4                 );
         Sk8h b0d0 = Sk8h(sk_linear12_from_srgb[p0[ 4]],
                          sk_linear12_from_srgb[p0[ 5]],
                          sk_linear12_from_srgb[p0[ 6]],
@@ -371,6 +383,92 @@ void downsample_2_2_srgb(void* dst, const void* src, size_t srcRB, int count) {
     }
 }
 
+void downsample_2_3_srgb(void* dst, const void* src, size_t srcRB, int count) {
+    const uint8_t* p0 = ((const uint8_t*) src);
+    const uint8_t* p1 = p0 + srcRB;
+    const uint8_t* p2 = p1 + srcRB;
+    uint8_t* d = (uint8_t*) dst;
+
+    // Given pixels:
+    // a0 b0 c0 d0 ...
+    // a1 b1 c1 d1 ...
+    // a2 b2 c2 d2 ...
+    // We want:
+    // (a0 + b0 + 2*a1 + 2*b1 + a2 + b2) / 8
+    // (c0 + d0 + 2*c1 + 2*d1 + c2 + d2) / 8
+    // ...
+    while (count >= 2) {
+        Sk8h a0c0 = Sk8h(sk_linear12_from_srgb[p0[ 0]],
+                         sk_linear12_from_srgb[p0[ 1]],
+                         sk_linear12_from_srgb[p0[ 2]],
+                         p0[ 3] << 4                  ,
+                         sk_linear12_from_srgb[p0[ 8]],
+                         sk_linear12_from_srgb[p0[ 9]],
+                         sk_linear12_from_srgb[p0[10]],
+                         p0[11] << 4                 );
+        Sk8h b0d0 = Sk8h(sk_linear12_from_srgb[p0[ 4]],
+                         sk_linear12_from_srgb[p0[ 5]],
+                         sk_linear12_from_srgb[p0[ 6]],
+                         p0[ 7] << 4                  ,
+                         sk_linear12_from_srgb[p0[12]],
+                         sk_linear12_from_srgb[p0[13]],
+                         sk_linear12_from_srgb[p0[14]],
+                         p0[15] << 4                 );
+        Sk8h a1c1 = Sk8h(sk_linear12_from_srgb[p1[ 0]],
+                         sk_linear12_from_srgb[p1[ 1]],
+                         sk_linear12_from_srgb[p1[ 2]],
+                         p1[ 3] << 4                  ,
+                         sk_linear12_from_srgb[p1[ 8]],
+                         sk_linear12_from_srgb[p1[ 9]],
+                         sk_linear12_from_srgb[p1[10]],
+                         p1[11] << 4                 );
+        Sk8h b1d1 = Sk8h(sk_linear12_from_srgb[p1[ 4]],
+                         sk_linear12_from_srgb[p1[ 5]],
+                         sk_linear12_from_srgb[p1[ 6]],
+                         p1[ 7] << 4                  ,
+                         sk_linear12_from_srgb[p1[12]],
+                         sk_linear12_from_srgb[p1[13]],
+                         sk_linear12_from_srgb[p1[14]],
+                         p1[15] << 4                 );
+        Sk8h a2c2 = Sk8h(sk_linear12_from_srgb[p2[ 0]],
+                         sk_linear12_from_srgb[p2[ 1]],
+                         sk_linear12_from_srgb[p2[ 2]],
+                         p2[ 3] << 4                  ,
+                         sk_linear12_from_srgb[p2[ 8]],
+                         sk_linear12_from_srgb[p2[ 9]],
+                         sk_linear12_from_srgb[p2[10]],
+                         p2[11] << 4                 );
+        Sk8h b2d2 = Sk8h(sk_linear12_from_srgb[p2[ 4]],
+                         sk_linear12_from_srgb[p2[ 5]],
+                         sk_linear12_from_srgb[p2[ 6]],
+                         p2[ 7] << 4                  ,
+                         sk_linear12_from_srgb[p2[12]],
+                         sk_linear12_from_srgb[p2[13]],
+                         sk_linear12_from_srgb[p2[14]],
+                         p2[15] << 4                 );
+
+        Sk8h avg = (a0c0 + b0d0 + a1c1 + a1c1 + b1d1 + b1d1 + a2c2 + b2d2) >> 3;
+        d[0] = sk_linear12_to_srgb[avg[0]];
+        d[1] = sk_linear12_to_srgb[avg[1]];
+        d[2] = sk_linear12_to_srgb[avg[2]];
+        d[3] = avg[3] >> 4;
+        d[4] = sk_linear12_to_srgb[avg[4]];
+        d[5] = sk_linear12_to_srgb[avg[5]];
+        d[6] = sk_linear12_to_srgb[avg[6]];
+        d[7] = avg[7] >> 4;
+
+        p0 += 16;
+        p1 += 16;
+        p2 += 16;
+        d += 8;
+        count -= 2;
+    }
+
+    if (count) {
+        downsample_2_3<ColorTypeFilter_S32>(d, p0, srcRB, count);
+    }
+}
+
 ///////////////////////////////////////////////////////////////////////////////////////////////////
 
 size_t SkMipMap::AllocLevelsSize(int levelCount, size_t pixelSize) {
@@ -410,7 +508,7 @@ SkMipMap* SkMipMap::Build(const SkPixmap& src, SkDestinationSurfaceColorMode col
                 proc_1_3 = downsample_1_3<ColorTypeFilter_S32>;
                 proc_2_1 = downsample_2_1<ColorTypeFilter_S32>;
                 proc_2_2 = downsample_2_2_srgb;
-                proc_2_3 = downsample_2_3<ColorTypeFilter_S32>;
+                proc_2_3 = downsample_2_3_srgb;
                 proc_3_1 = downsample_3_1<ColorTypeFilter_S32>;
                 proc_3_2 = downsample_3_2<ColorTypeFilter_S32>;
                 proc_3_3 = downsample_3_3<ColorTypeFilter_S32>;