jumper, bilinear and bicubic sampling stages

This splits SkImageShaderContext into three parts:
   - SkJumper_GatherCtx: always, already done
   - SkJumper_SamplerCtx: when bilinear or bicubic
   - MiscCtx: other little bits (the matrix, paint color, tiling limits)

Thanks for the snazzy allocator that allows this Herb!

Both SkJumper and SkRasterPipeline_opts.h should be speaking all the
same types now.

I've copied the comments about bilinear/bicubic to SkJumper with little
typo fixes and clarifications.

Change-Id: I4ba7b7c02feba3f65f5292169a22c060e34933c6
Reviewed-on: https://skia-review.googlesource.com/13269
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

8 files changed, 3428 insertions, 92 deletions

diff --git a/src/image/SkImageShader.cpp b/src/image/SkImageShader.cpp
index cbba65efa8..ce0d6f504b 100644
--- a/src/image/SkImageShader.cpp
+++ b/src/image/SkImageShader.cpp
@@ -13,10 +13,10 @@
 #include "SkEmptyShader.h"
 #include "SkImage_Base.h"
 #include "SkImageShader.h"
-#include "SkImageShaderContext.h"
 #include "SkPM4fPriv.h"
 #include "SkReadBuffer.h"
 #include "SkWriteBuffer.h"
+#include "../jumper/SkJumper.h"
 
 SkImageShader::SkImageShader(sk_sp<SkImage> img, TileMode tmx, TileMode tmy, const SkMatrix* matrix)
     : INHERITED(matrix)
@@ -257,41 +257,51 @@ bool SkImageShader::onAppendStages(SkRasterPipeline* p, SkColorSpace* dst, SkAre
         }
     }
 
-    auto ctx = scratch->make<SkImageShaderContext>();
-    ctx->pixels  = pm.addr();
-    ctx->ctable  = pm.ctable() ? pm.ctable()->readColors() : nullptr;
-    ctx->stride  = pm.rowBytesAsPixels();
-    ctx->color4f = SkColor4f_from_SkColor(paint.getColor(), dst);
-    ctx->width   = (float)pm.width();
-    ctx->height  = (float)pm.height();
-    ctx->state   = std::move(state);  // Extend lifetime to match the pipeline's.
-    if (matrix.asAffine(ctx->matrix)) {
-        p->append(SkRasterPipeline::matrix_2x3, ctx->matrix);
+
+    struct MiscCtx {
+        std::unique_ptr<SkBitmapController::State> state;
+        SkColor4f paint_color;
+        float     width;
+        float     height;
+        float     matrix[9];
+    };
+    auto misc = scratch->make<MiscCtx>();
+    misc->state       = std::move(state);  // Extend lifetime to match the pipeline's.
+    misc->paint_color = SkColor4f_from_SkColor(paint.getColor(), dst);
+    misc->width       = (float)pm.width();
+    misc->height      = (float)pm.height();
+    if (matrix.asAffine(misc->matrix)) {
+        p->append(SkRasterPipeline::matrix_2x3, misc->matrix);
     } else {
-        matrix.get9(ctx->matrix);
-        p->append(SkRasterPipeline::matrix_perspective, ctx->matrix);
+        matrix.get9(misc->matrix);
+        p->append(SkRasterPipeline::matrix_perspective, misc->matrix);
     }
 
+    auto gather = scratch->make<SkJumper_GatherCtx>();
+    gather->pixels  = pm.addr();
+    gather->ctable  = pm.ctable() ? pm.ctable()->readColors() : nullptr;
+    gather->stride  = pm.rowBytesAsPixels();
+
     auto append_tiling_and_gather = [&] {
         switch (fTileModeX) {
-            case kClamp_TileMode:  p->append(SkRasterPipeline::clamp_x,  &ctx->width); break;
-            case kMirror_TileMode: p->append(SkRasterPipeline::mirror_x, &ctx->width); break;
-            case kRepeat_TileMode: p->append(SkRasterPipeline::repeat_x, &ctx->width); break;
+            case kClamp_TileMode:  p->append(SkRasterPipeline::clamp_x,  &misc->width); break;
+            case kMirror_TileMode: p->append(SkRasterPipeline::mirror_x, &misc->width); break;
+            case kRepeat_TileMode: p->append(SkRasterPipeline::repeat_x, &misc->width); break;
         }
         switch (fTileModeY) {
-            case kClamp_TileMode:  p->append(SkRasterPipeline::clamp_y,  &ctx->height); break;
-            case kMirror_TileMode: p->append(SkRasterPipeline::mirror_y, &ctx->height); break;
-            case kRepeat_TileMode: p->append(SkRasterPipeline::repeat_y, &ctx->height); break;
+            case kClamp_TileMode:  p->append(SkRasterPipeline::clamp_y,  &misc->height); break;
+            case kMirror_TileMode: p->append(SkRasterPipeline::mirror_y, &misc->height); break;
+            case kRepeat_TileMode: p->append(SkRasterPipeline::repeat_y, &misc->height); break;
         }
         switch (info.colorType()) {
-            case kAlpha_8_SkColorType:   p->append(SkRasterPipeline::gather_a8,   ctx); break;
-            case kIndex_8_SkColorType:   p->append(SkRasterPipeline::gather_i8,   ctx); break;
-            case kGray_8_SkColorType:    p->append(SkRasterPipeline::gather_g8,   ctx); break;
-            case kRGB_565_SkColorType:   p->append(SkRasterPipeline::gather_565,  ctx); break;
-            case kARGB_4444_SkColorType: p->append(SkRasterPipeline::gather_4444, ctx); break;
+            case kAlpha_8_SkColorType:   p->append(SkRasterPipeline::gather_a8,   gather); break;
+            case kIndex_8_SkColorType:   p->append(SkRasterPipeline::gather_i8,   gather); break;
+            case kGray_8_SkColorType:    p->append(SkRasterPipeline::gather_g8,   gather); break;
+            case kRGB_565_SkColorType:   p->append(SkRasterPipeline::gather_565,  gather); break;
+            case kARGB_4444_SkColorType: p->append(SkRasterPipeline::gather_4444, gather); break;
             case kRGBA_8888_SkColorType:
-            case kBGRA_8888_SkColorType: p->append(SkRasterPipeline::gather_8888, ctx); break;
-            case kRGBA_F16_SkColorType:  p->append(SkRasterPipeline::gather_f16,  ctx); break;
+            case kBGRA_8888_SkColorType: p->append(SkRasterPipeline::gather_8888, gather); break;
+            case kRGBA_F16_SkColorType:  p->append(SkRasterPipeline::gather_f16,  gather); break;
             default: SkASSERT(false);
         }
         if (info.gammaCloseToSRGB() && dst != nullptr) {
@@ -299,18 +309,23 @@ bool SkImageShader::onAppendStages(SkRasterPipeline* p, SkColorSpace* dst, SkAre
         }
     };
 
+    SkJumper_SamplerCtx* sampler = nullptr;
+    if (quality != kNone_SkFilterQuality) {
+        sampler = scratch->make<SkJumper_SamplerCtx>();
+    }
+
     auto sample = [&](SkRasterPipeline::StockStage setup_x,
                       SkRasterPipeline::StockStage setup_y) {
-        p->append(setup_x, ctx);
-        p->append(setup_y, ctx);
+        p->append(setup_x, sampler);
+        p->append(setup_y, sampler);
         append_tiling_and_gather();
-        p->append(SkRasterPipeline::accumulate, ctx);
+        p->append(SkRasterPipeline::accumulate, sampler);
     };
 
     if (quality == kNone_SkFilterQuality) {
         append_tiling_and_gather();
     } else if (quality == kLow_SkFilterQuality) {
-        p->append(SkRasterPipeline::save_xy, ctx);
+        p->append(SkRasterPipeline::save_xy, sampler);
 
         sample(SkRasterPipeline::bilinear_nx, SkRasterPipeline::bilinear_ny);
         sample(SkRasterPipeline::bilinear_px, SkRasterPipeline::bilinear_ny);
@@ -319,7 +334,7 @@ bool SkImageShader::onAppendStages(SkRasterPipeline* p, SkColorSpace* dst, SkAre
 
         p->append(SkRasterPipeline::move_dst_src);
     } else {
-        p->append(SkRasterPipeline::save_xy, ctx);
+        p->append(SkRasterPipeline::save_xy, sampler);
 
         sample(SkRasterPipeline::bicubic_n3x, SkRasterPipeline::bicubic_n3y);
         sample(SkRasterPipeline::bicubic_n1x, SkRasterPipeline::bicubic_n3y);
@@ -352,7 +367,7 @@ bool SkImageShader::onAppendStages(SkRasterPipeline* p, SkColorSpace* dst, SkAre
         p->append(SkRasterPipeline::swap_rb);
     }
     if (info.colorType() == kAlpha_8_SkColorType) {
-        p->append(SkRasterPipeline::set_rgb, &ctx->color4f);
+        p->append(SkRasterPipeline::set_rgb, &misc->paint_color);
     }
     if (info.colorType() == kAlpha_8_SkColorType || info.alphaType() == kUnpremul_SkAlphaType) {
         p->append(SkRasterPipeline::premul);
diff --git a/src/image/SkImageShaderContext.h b/src/image/SkImageShaderContext.h
deleted file mode 100644
index 7a8ba6369f..0000000000
--- a/src/image/SkImageShaderContext.h
+++ /dev/null
@@ -1,37 +0,0 @@
-/*
- * Copyright 2016 Google Inc.
- *
- * Use of this source code is governed by a BSD-style license that can be
- * found in the LICENSE file.
- */
-
-#ifndef SkImageShaderContext_DEFINED
-#define SkImageShaderContext_DEFINED
-
-#include "SkBitmapController.h"
-#include "SkColor.h"
-#include "SkColorTable.h"
-#include <memory>
-
-// Definition used by SkImageShader.cpp and SkRasterPipeline_opts.h.
-// Otherwise, completely uninteresting.
-
-struct SkImageShaderContext {
-    const void*     pixels;
-    const uint32_t* ctable;
-    int             stride;
-    SkColor4f       color4f;
-    float           width;
-    float           height;
-    float           matrix[9];
-    float           x[8];
-    float           y[8];
-    float           fx[8];
-    float           fy[8];
-    float           scalex[8];
-    float           scaley[8];
-
-    std::unique_ptr<SkBitmapController::State> state;
-};
-
-#endif//SkImageShaderContext_DEFINED
diff --git a/src/jumper/SkJumper.cpp b/src/jumper/SkJumper.cpp
index 736dad6f06..0d4446ce47 100644
--- a/src/jumper/SkJumper.cpp
+++ b/src/jumper/SkJumper.cpp
@@ -123,6 +123,11 @@ static K kConstants = {
     M(repeat_y)           \
     M(mirror_x)           \
     M(mirror_y)           \
+    M(save_xy)            \
+    M(accumulate)         \
+    M(bilinear_nx) M(bilinear_px) M(bilinear_ny) M(bilinear_py)  \
+    M(bicubic_n3x) M(bicubic_n1x) M(bicubic_p1x) M(bicubic_p3x)  \
+    M(bicubic_n3y) M(bicubic_n1y) M(bicubic_p1y) M(bicubic_p3y)  \
     M(linear_gradient)    \
     M(linear_gradient_2stops)
 
diff --git a/src/jumper/SkJumper.h b/src/jumper/SkJumper.h
index b440391f38..7a3f4e85f5 100644
--- a/src/jumper/SkJumper.h
+++ b/src/jumper/SkJumper.h
@@ -52,8 +52,10 @@
 //   - the _i and _f user-defined literal operators call C() for you in a prettier way; or
 //   - you can load values from this struct.
 
+static const int SkJumper_kMaxStride = 8;
+
 struct SkJumper_constants {
-    float iota[8];      //  0,1,2,3,4,5,6,7
+    float iota[SkJumper_kMaxStride];   //  0,1,2,3,4,...
 };
 
 struct SkJumper_GatherCtx {
@@ -62,4 +64,14 @@ struct SkJumper_GatherCtx {
     int             stride;
 };
 
+// State shared by save_xy, accumulate, and bilinear_* / bicubic_*.
+struct SkJumper_SamplerCtx {
+    float      x[SkJumper_kMaxStride];
+    float      y[SkJumper_kMaxStride];
+    float     fx[SkJumper_kMaxStride];
+    float     fy[SkJumper_kMaxStride];
+    float scalex[SkJumper_kMaxStride];
+    float scaley[SkJumper_kMaxStride];
+};
+
 #endif//SkJumper_DEFINED
diff --git a/src/jumper/SkJumper_generated.S b/src/jumper/SkJumper_generated.S
index db0d5933e3..80a08b990f 100644
--- a/src/jumper/SkJumper_generated.S
+++ b/src/jumper/SkJumper_generated.S
@@ -2648,6 +2648,284 @@ _sk_linear_gradient_2stops_aarch64:
   .long  0x4f911001                          // fmla          v1.4s, v0.4s, v17.s[0]
   .long  0x4eb01e00                          // mov           v0.16b, v16.16b
   .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_save_xy_aarch64
+.globl _sk_save_xy_aarch64
+_sk_save_xy_aarch64:
+  .long  0x4f0167f0                          // movi          v16.4s, #0x3f, lsl #24
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x4e30d411                          // fadd          v17.4s, v0.4s, v16.4s
+  .long  0x4e30d430                          // fadd          v16.4s, v1.4s, v16.4s
+  .long  0x4e219a32                          // frintm        v18.4s, v17.4s
+  .long  0x4eb2d631                          // fsub          v17.4s, v17.4s, v18.4s
+  .long  0x4e219a12                          // frintm        v18.4s, v16.4s
+  .long  0x4eb2d610                          // fsub          v16.4s, v16.4s, v18.4s
+  .long  0x3d800100                          // str           q0, [x8]
+  .long  0x3d800901                          // str           q1, [x8, #32]
+  .long  0x3d801111                          // str           q17, [x8, #64]
+  .long  0x3d801910                          // str           q16, [x8, #96]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_accumulate_aarch64
+.globl _sk_accumulate_aarch64
+_sk_accumulate_aarch64:
+  .long  0xa8c10c28                          // ldp           x8, x3, [x1], #16
+  .long  0x3dc02110                          // ldr           q16, [x8, #128]
+  .long  0x3dc02911                          // ldr           q17, [x8, #160]
+  .long  0x6e31de10                          // fmul          v16.4s, v16.4s, v17.4s
+  .long  0x4e30cc04                          // fmla          v4.4s, v0.4s, v16.4s
+  .long  0x4e30cc25                          // fmla          v5.4s, v1.4s, v16.4s
+  .long  0x4e30cc46                          // fmla          v6.4s, v2.4s, v16.4s
+  .long  0x4e30cc67                          // fmla          v7.4s, v3.4s, v16.4s
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bilinear_nx_aarch64
+.globl _sk_bilinear_nx_aarch64
+_sk_bilinear_nx_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x4f03f611                          // fmov          v17.4s, #1.000000000000000000e+00
+  .long  0x3dc01100                          // ldr           q0, [x8, #64]
+  .long  0x3dc00110                          // ldr           q16, [x8]
+  .long  0x4ea0d620                          // fsub          v0.4s, v17.4s, v0.4s
+  .long  0x3d802100                          // str           q0, [x8, #128]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f0567e0                          // movi          v0.4s, #0xbf, lsl #24
+  .long  0x4e20d600                          // fadd          v0.4s, v16.4s, v0.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bilinear_px_aarch64
+.globl _sk_bilinear_px_aarch64
+_sk_bilinear_px_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x3dc01100                          // ldr           q0, [x8, #64]
+  .long  0x3dc00110                          // ldr           q16, [x8]
+  .long  0x3d802100                          // str           q0, [x8, #128]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f0167e0                          // movi          v0.4s, #0x3f, lsl #24
+  .long  0x4e20d600                          // fadd          v0.4s, v16.4s, v0.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bilinear_ny_aarch64
+.globl _sk_bilinear_ny_aarch64
+_sk_bilinear_ny_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x4f03f611                          // fmov          v17.4s, #1.000000000000000000e+00
+  .long  0x3dc01901                          // ldr           q1, [x8, #96]
+  .long  0x3dc00910                          // ldr           q16, [x8, #32]
+  .long  0x4ea1d621                          // fsub          v1.4s, v17.4s, v1.4s
+  .long  0x3d802901                          // str           q1, [x8, #160]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f0567e1                          // movi          v1.4s, #0xbf, lsl #24
+  .long  0x4e21d601                          // fadd          v1.4s, v16.4s, v1.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bilinear_py_aarch64
+.globl _sk_bilinear_py_aarch64
+_sk_bilinear_py_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x3dc01901                          // ldr           q1, [x8, #96]
+  .long  0x3dc00910                          // ldr           q16, [x8, #32]
+  .long  0x3d802901                          // str           q1, [x8, #160]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f0167e1                          // movi          v1.4s, #0x3f, lsl #24
+  .long  0x4e21d601                          // fadd          v1.4s, v16.4s, v1.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_n3x_aarch64
+.globl _sk_bicubic_n3x_aarch64
+_sk_bicubic_n3x_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52a7d8e9                          // mov           w9, #0x3ec70000
+  .long  0x72838e49                          // movk          w9, #0x1c72
+  .long  0x4e040d30                          // dup           v16.4s, w9
+  .long  0x3dc01111                          // ldr           q17, [x8, #64]
+  .long  0x52b7d549                          // mov           w9, #0xbeaa0000
+  .long  0x4f03f600                          // fmov          v0.4s, #1.000000000000000000e+00
+  .long  0x72955569                          // movk          w9, #0xaaab
+  .long  0x4e040d32                          // dup           v18.4s, w9
+  .long  0x4eb1d400                          // fsub          v0.4s, v0.4s, v17.4s
+  .long  0x6e20dc11                          // fmul          v17.4s, v0.4s, v0.4s
+  .long  0x4e20ce12                          // fmla          v18.4s, v16.4s, v0.4s
+  .long  0x6e32de20                          // fmul          v0.4s, v17.4s, v18.4s
+  .long  0x3dc00113                          // ldr           q19, [x8]
+  .long  0x3d802100                          // str           q0, [x8, #128]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f07f700                          // fmov          v0.4s, #-1.500000000000000000e+00
+  .long  0x4e20d660                          // fadd          v0.4s, v19.4s, v0.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_n1x_aarch64
+.globl _sk_bicubic_n1x_aarch64
+_sk_bicubic_n1x_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52b7f2a9                          // mov           w9, #0xbf950000
+  .long  0x4f03f600                          // fmov          v0.4s, #1.000000000000000000e+00
+  .long  0x728aaaa9                          // movk          w9, #0x5555
+  .long  0x3dc01110                          // ldr           q16, [x8, #64]
+  .long  0x4f03f711                          // fmov          v17.4s, #1.500000000000000000e+00
+  .long  0x4f0167f2                          // movi          v18.4s, #0x3f, lsl #24
+  .long  0x4eb0d400                          // fsub          v0.4s, v0.4s, v16.4s
+  .long  0x4e040d30                          // dup           v16.4s, w9
+  .long  0x52a7ac69                          // mov           w9, #0x3d630000
+  .long  0x7291c729                          // movk          w9, #0x8e39
+  .long  0x4e20ce11                          // fmla          v17.4s, v16.4s, v0.4s
+  .long  0x4e20ce32                          // fmla          v18.4s, v17.4s, v0.4s
+  .long  0x4e040d31                          // dup           v17.4s, w9
+  .long  0x4e20ce51                          // fmla          v17.4s, v18.4s, v0.4s
+  .long  0x3dc00110                          // ldr           q16, [x8]
+  .long  0x3d802111                          // str           q17, [x8, #128]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f0567e0                          // movi          v0.4s, #0xbf, lsl #24
+  .long  0x4e20d600                          // fadd          v0.4s, v16.4s, v0.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_p1x_aarch64
+.globl _sk_bicubic_p1x_aarch64
+_sk_bicubic_p1x_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52b7f2a9                          // mov           w9, #0xbf950000
+  .long  0x728aaaa9                          // movk          w9, #0x5555
+  .long  0x4f03f711                          // fmov          v17.4s, #1.500000000000000000e+00
+  .long  0x3dc01112                          // ldr           q18, [x8, #64]
+  .long  0x3dc00100                          // ldr           q0, [x8]
+  .long  0x4e040d33                          // dup           v19.4s, w9
+  .long  0x52a7ac69                          // mov           w9, #0x3d630000
+  .long  0x4f0167f0                          // movi          v16.4s, #0x3f, lsl #24
+  .long  0x7291c729                          // movk          w9, #0x8e39
+  .long  0x4e32ce71                          // fmla          v17.4s, v19.4s, v18.4s
+  .long  0x4e30d400                          // fadd          v0.4s, v0.4s, v16.4s
+  .long  0x4e32ce30                          // fmla          v16.4s, v17.4s, v18.4s
+  .long  0x4e040d31                          // dup           v17.4s, w9
+  .long  0x4e32ce11                          // fmla          v17.4s, v16.4s, v18.4s
+  .long  0x3d802111                          // str           q17, [x8, #128]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_p3x_aarch64
+.globl _sk_bicubic_p3x_aarch64
+_sk_bicubic_p3x_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52a7d8e9                          // mov           w9, #0x3ec70000
+  .long  0x72838e49                          // movk          w9, #0x1c72
+  .long  0x4e040d20                          // dup           v0.4s, w9
+  .long  0x3dc01110                          // ldr           q16, [x8, #64]
+  .long  0x52b7d549                          // mov           w9, #0xbeaa0000
+  .long  0x72955569                          // movk          w9, #0xaaab
+  .long  0x4e040d31                          // dup           v17.4s, w9
+  .long  0x6e30de13                          // fmul          v19.4s, v16.4s, v16.4s
+  .long  0x4e30cc11                          // fmla          v17.4s, v0.4s, v16.4s
+  .long  0x6e31de60                          // fmul          v0.4s, v19.4s, v17.4s
+  .long  0x3dc00112                          // ldr           q18, [x8]
+  .long  0x3d802100                          // str           q0, [x8, #128]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f03f700                          // fmov          v0.4s, #1.500000000000000000e+00
+  .long  0x4e20d640                          // fadd          v0.4s, v18.4s, v0.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_n3y_aarch64
+.globl _sk_bicubic_n3y_aarch64
+_sk_bicubic_n3y_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52a7d8e9                          // mov           w9, #0x3ec70000
+  .long  0x72838e49                          // movk          w9, #0x1c72
+  .long  0x4e040d30                          // dup           v16.4s, w9
+  .long  0x3dc01911                          // ldr           q17, [x8, #96]
+  .long  0x52b7d549                          // mov           w9, #0xbeaa0000
+  .long  0x4f03f601                          // fmov          v1.4s, #1.000000000000000000e+00
+  .long  0x72955569                          // movk          w9, #0xaaab
+  .long  0x4e040d32                          // dup           v18.4s, w9
+  .long  0x4eb1d421                          // fsub          v1.4s, v1.4s, v17.4s
+  .long  0x6e21dc31                          // fmul          v17.4s, v1.4s, v1.4s
+  .long  0x4e21ce12                          // fmla          v18.4s, v16.4s, v1.4s
+  .long  0x6e32de21                          // fmul          v1.4s, v17.4s, v18.4s
+  .long  0x3dc00913                          // ldr           q19, [x8, #32]
+  .long  0x3d802901                          // str           q1, [x8, #160]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f07f701                          // fmov          v1.4s, #-1.500000000000000000e+00
+  .long  0x4e21d661                          // fadd          v1.4s, v19.4s, v1.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_n1y_aarch64
+.globl _sk_bicubic_n1y_aarch64
+_sk_bicubic_n1y_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52b7f2a9                          // mov           w9, #0xbf950000
+  .long  0x4f03f601                          // fmov          v1.4s, #1.000000000000000000e+00
+  .long  0x728aaaa9                          // movk          w9, #0x5555
+  .long  0x3dc01910                          // ldr           q16, [x8, #96]
+  .long  0x4f03f711                          // fmov          v17.4s, #1.500000000000000000e+00
+  .long  0x4f0167f2                          // movi          v18.4s, #0x3f, lsl #24
+  .long  0x4eb0d421                          // fsub          v1.4s, v1.4s, v16.4s
+  .long  0x4e040d30                          // dup           v16.4s, w9
+  .long  0x52a7ac69                          // mov           w9, #0x3d630000
+  .long  0x7291c729                          // movk          w9, #0x8e39
+  .long  0x4e21ce11                          // fmla          v17.4s, v16.4s, v1.4s
+  .long  0x4e21ce32                          // fmla          v18.4s, v17.4s, v1.4s
+  .long  0x4e040d31                          // dup           v17.4s, w9
+  .long  0x4e21ce51                          // fmla          v17.4s, v18.4s, v1.4s
+  .long  0x3dc00910                          // ldr           q16, [x8, #32]
+  .long  0x3d802911                          // str           q17, [x8, #160]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f0567e1                          // movi          v1.4s, #0xbf, lsl #24
+  .long  0x4e21d601                          // fadd          v1.4s, v16.4s, v1.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_p1y_aarch64
+.globl _sk_bicubic_p1y_aarch64
+_sk_bicubic_p1y_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52b7f2a9                          // mov           w9, #0xbf950000
+  .long  0x728aaaa9                          // movk          w9, #0x5555
+  .long  0x4f03f711                          // fmov          v17.4s, #1.500000000000000000e+00
+  .long  0x3dc01912                          // ldr           q18, [x8, #96]
+  .long  0x3dc00901                          // ldr           q1, [x8, #32]
+  .long  0x4e040d33                          // dup           v19.4s, w9
+  .long  0x52a7ac69                          // mov           w9, #0x3d630000
+  .long  0x4f0167f0                          // movi          v16.4s, #0x3f, lsl #24
+  .long  0x7291c729                          // movk          w9, #0x8e39
+  .long  0x4e32ce71                          // fmla          v17.4s, v19.4s, v18.4s
+  .long  0x4e30d421                          // fadd          v1.4s, v1.4s, v16.4s
+  .long  0x4e32ce30                          // fmla          v16.4s, v17.4s, v18.4s
+  .long  0x4e040d31                          // dup           v17.4s, w9
+  .long  0x4e32ce11                          // fmla          v17.4s, v16.4s, v18.4s
+  .long  0x3d802911                          // str           q17, [x8, #160]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
+
+HIDDEN _sk_bicubic_p3y_aarch64
+.globl _sk_bicubic_p3y_aarch64
+_sk_bicubic_p3y_aarch64:
+  .long  0xf9400028                          // ldr           x8, [x1]
+  .long  0x52a7d8e9                          // mov           w9, #0x3ec70000
+  .long  0x72838e49                          // movk          w9, #0x1c72
+  .long  0x4e040d21                          // dup           v1.4s, w9
+  .long  0x3dc01910                          // ldr           q16, [x8, #96]
+  .long  0x52b7d549                          // mov           w9, #0xbeaa0000
+  .long  0x72955569                          // movk          w9, #0xaaab
+  .long  0x4e040d31                          // dup           v17.4s, w9
+  .long  0x6e30de13                          // fmul          v19.4s, v16.4s, v16.4s
+  .long  0x4e30cc31                          // fmla          v17.4s, v1.4s, v16.4s
+  .long  0x6e31de61                          // fmul          v1.4s, v19.4s, v17.4s
+  .long  0x3dc00912                          // ldr           q18, [x8, #32]
+  .long  0x3d802901                          // str           q1, [x8, #160]
+  .long  0xf9400423                          // ldr           x3, [x1, #8]
+  .long  0x4f03f701                          // fmov          v1.4s, #1.500000000000000000e+00
+  .long  0x4e21d641                          // fadd          v1.4s, v18.4s, v1.4s
+  .long  0x91004021                          // add           x1, x1, #0x10
+  .long  0xd61f0060                          // br            x3
 #elif defined(__arm__)
 .balign 4
 
@@ -5496,6 +5774,316 @@ _sk_linear_gradient_2stops_vfp4:
   .long  0xf22001b0                          // vorr          d0, d16, d16
   .long  0xe8bd4010                          // pop           {r4, lr}
   .long  0xe12fff1c                          // bx            ip
+
+HIDDEN _sk_save_xy_vfp4
+.globl _sk_save_xy_vfp4
+_sk_save_xy_vfp4:
+  .long  0xf2c3061f                          // vmov.i32      d16, #1056964608
+  .long  0xeddf7b17                          // vldr          d23, [pc, #92]
+  .long  0xf2c06010                          // vmov.i32      d22, #0
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2401d20                          // vadd.f32      d17, d0, d16
+  .long  0xf2410d20                          // vadd.f32      d16, d1, d16
+  .long  0xed830b00                          // vstr          d0, [r3]
+  .long  0xed831b08                          // vstr          d1, [r3, #32]
+  .long  0xf3fb2721                          // vcvt.s32.f32  d18, d17
+  .long  0xf3fb3720                          // vcvt.s32.f32  d19, d16
+  .long  0xf3fb2622                          // vcvt.f32.s32  d18, d18
+  .long  0xf3fb3623                          // vcvt.f32.s32  d19, d19
+  .long  0xf3624ea1                          // vcgt.f32      d20, d18, d17
+  .long  0xf3635ea0                          // vcgt.f32      d21, d19, d16
+  .long  0xf35741b6                          // vbsl          d20, d23, d22
+  .long  0xf35751b6                          // vbsl          d21, d23, d22
+  .long  0xf2622da4                          // vsub.f32      d18, d18, d20
+  .long  0xf2633da5                          // vsub.f32      d19, d19, d21
+  .long  0xf2611da2                          // vsub.f32      d17, d17, d18
+  .long  0xf2600da3                          // vsub.f32      d16, d16, d19
+  .long  0xedc31b10                          // vstr          d17, [r3, #64]
+  .long  0xedc30b18                          // vstr          d16, [r3, #96]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0x3f800000                          // .word         0x3f800000
+  .long  0x3f800000                          // .word         0x3f800000
+
+HIDDEN _sk_accumulate_vfp4
+.globl _sk_accumulate_vfp4
+_sk_accumulate_vfp4:
+  .long  0xe8911008                          // ldm           r1, {r3, ip}
+  .long  0xe2811008                          // add           r1, r1, #8
+  .long  0xedd31b28                          // vldr          d17, [r3, #160]
+  .long  0xedd30b20                          // vldr          d16, [r3, #128]
+  .long  0xf3400db1                          // vmul.f32      d16, d16, d17
+  .long  0xf2004c90                          // vfma.f32      d4, d16, d0
+  .long  0xf2005c91                          // vfma.f32      d5, d16, d1
+  .long  0xf2006c92                          // vfma.f32      d6, d16, d2
+  .long  0xf2007c93                          // vfma.f32      d7, d16, d3
+  .long  0xe12fff1c                          // bx            ip
+
+HIDDEN _sk_bilinear_nx_vfp4
+.globl _sk_bilinear_nx_vfp4
+_sk_bilinear_nx_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c70f10                          // vmov.f32      d16, #1
+  .long  0xedd32b10                          // vldr          d18, [r3, #64]
+  .long  0xf2600da2                          // vsub.f32      d16, d16, d18
+  .long  0xedd31b00                          // vldr          d17, [r3]
+  .long  0xf3c3261f                          // vmov.i32      d18, #-1090519040
+  .long  0xf2010da2                          // vadd.f32      d0, d17, d18
+  .long  0xedc30b20                          // vstr          d16, [r3, #128]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+
+HIDDEN _sk_bilinear_px_vfp4
+.globl _sk_bilinear_px_vfp4
+_sk_bilinear_px_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c3061f                          // vmov.i32      d16, #1056964608
+  .long  0xedd31b00                          // vldr          d17, [r3]
+  .long  0xedd32b10                          // vldr          d18, [r3, #64]
+  .long  0xf2010da0                          // vadd.f32      d0, d17, d16
+  .long  0xedc32b20                          // vstr          d18, [r3, #128]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+
+HIDDEN _sk_bilinear_ny_vfp4
+.globl _sk_bilinear_ny_vfp4
+_sk_bilinear_ny_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c70f10                          // vmov.f32      d16, #1
+  .long  0xedd32b18                          // vldr          d18, [r3, #96]
+  .long  0xf2600da2                          // vsub.f32      d16, d16, d18
+  .long  0xedd31b08                          // vldr          d17, [r3, #32]
+  .long  0xf3c3261f                          // vmov.i32      d18, #-1090519040
+  .long  0xf2011da2                          // vadd.f32      d1, d17, d18
+  .long  0xedc30b28                          // vstr          d16, [r3, #160]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+
+HIDDEN _sk_bilinear_py_vfp4
+.globl _sk_bilinear_py_vfp4
+_sk_bilinear_py_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c3061f                          // vmov.i32      d16, #1056964608
+  .long  0xedd31b08                          // vldr          d17, [r3, #32]
+  .long  0xedd32b18                          // vldr          d18, [r3, #96]
+  .long  0xf2011da0                          // vadd.f32      d1, d17, d16
+  .long  0xedc32b28                          // vstr          d18, [r3, #160]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+
+HIDDEN _sk_bicubic_n3x_vfp4
+.globl _sk_bicubic_n3x_vfp4
+_sk_bicubic_n3x_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c70f10                          // vmov.f32      d16, #1
+  .long  0xeddf3b10                          // vldr          d19, [pc, #64]
+  .long  0xedd32b10                          // vldr          d18, [r3, #64]
+  .long  0xf2600da2                          // vsub.f32      d16, d16, d18
+  .long  0xeddf2b0b                          // vldr          d18, [pc, #44]
+  .long  0xedd31b00                          // vldr          d17, [r3]
+  .long  0xf2403cb2                          // vfma.f32      d19, d16, d18
+  .long  0xf3400db0                          // vmul.f32      d16, d16, d16
+  .long  0xf3c72f18                          // vmov.f32      d18, #-1.5
+  .long  0xf2010da2                          // vadd.f32      d0, d17, d18
+  .long  0xf3400db3                          // vmul.f32      d16, d16, d19
+  .long  0xedc30b20                          // vstr          d16, [r3, #128]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xe320f000                          // nop           {0}
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
+
+HIDDEN _sk_bicubic_n1x_vfp4
+.globl _sk_bicubic_n1x_vfp4
+_sk_bicubic_n1x_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c70f10                          // vmov.f32      d16, #1
+  .long  0xf2c73f18                          // vmov.f32      d19, #1.5
+  .long  0xedd32b10                          // vldr          d18, [r3, #64]
+  .long  0xf2600da2                          // vsub.f32      d16, d16, d18
+  .long  0xeddf2b0d                          // vldr          d18, [pc, #52]
+  .long  0xedd31b00                          // vldr          d17, [r3]
+  .long  0xf2403cb2                          // vfma.f32      d19, d16, d18
+  .long  0xf2c3261f                          // vmov.i32      d18, #1056964608
+  .long  0xf2402cb3                          // vfma.f32      d18, d16, d19
+  .long  0xeddf3b0a                          // vldr          d19, [pc, #40]
+  .long  0xf2403cb2                          // vfma.f32      d19, d16, d18
+  .long  0xf3c3061f                          // vmov.i32      d16, #-1090519040
+  .long  0xf2010da0                          // vadd.f32      d0, d17, d16
+  .long  0xedc33b20                          // vstr          d19, [r3, #128]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xe320f000                          // nop           {0}
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0x3d638e39                          // .word         0x3d638e39
+  .long  0x3d638e39                          // .word         0x3d638e39
+
+HIDDEN _sk_bicubic_p1x_vfp4
+.globl _sk_bicubic_p1x_vfp4
+_sk_bicubic_p1x_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c71f18                          // vmov.f32      d17, #1.5
+  .long  0xeddf0b0c                          // vldr          d16, [pc, #48]
+  .long  0xedd33b10                          // vldr          d19, [r3, #64]
+  .long  0xf2431cb0                          // vfma.f32      d17, d19, d16
+  .long  0xedd32b00                          // vldr          d18, [r3]
+  .long  0xf2c3061f                          // vmov.i32      d16, #1056964608
+  .long  0xf2020da0                          // vadd.f32      d0, d18, d16
+  .long  0xf2430cb1                          // vfma.f32      d16, d19, d17
+  .long  0xeddf1b07                          // vldr          d17, [pc, #28]
+  .long  0xf2431cb0                          // vfma.f32      d17, d19, d16
+  .long  0xedc31b20                          // vstr          d17, [r3, #128]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0x3d638e39                          // .word         0x3d638e39
+  .long  0x3d638e39                          // .word         0x3d638e39
+
+HIDDEN _sk_bicubic_p3x_vfp4
+.globl _sk_bicubic_p3x_vfp4
+_sk_bicubic_p3x_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xeddf0b0d                          // vldr          d16, [pc, #52]
+  .long  0xeddf3b0e                          // vldr          d19, [pc, #56]
+  .long  0xedd32b10                          // vldr          d18, [r3, #64]
+  .long  0xf2423cb0                          // vfma.f32      d19, d18, d16
+  .long  0xedd31b00                          // vldr          d17, [r3]
+  .long  0xf3420db2                          // vmul.f32      d16, d18, d18
+  .long  0xf2c72f18                          // vmov.f32      d18, #1.5
+  .long  0xf2010da2                          // vadd.f32      d0, d17, d18
+  .long  0xf3400db3                          // vmul.f32      d16, d16, d19
+  .long  0xedc30b20                          // vstr          d16, [r3, #128]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xe320f000                          // nop           {0}
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
+
+HIDDEN _sk_bicubic_n3y_vfp4
+.globl _sk_bicubic_n3y_vfp4
+_sk_bicubic_n3y_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c70f10                          // vmov.f32      d16, #1
+  .long  0xeddf3b10                          // vldr          d19, [pc, #64]
+  .long  0xedd32b18                          // vldr          d18, [r3, #96]
+  .long  0xf2600da2                          // vsub.f32      d16, d16, d18
+  .long  0xeddf2b0b                          // vldr          d18, [pc, #44]
+  .long  0xedd31b08                          // vldr          d17, [r3, #32]
+  .long  0xf2403cb2                          // vfma.f32      d19, d16, d18
+  .long  0xf3400db0                          // vmul.f32      d16, d16, d16
+  .long  0xf3c72f18                          // vmov.f32      d18, #-1.5
+  .long  0xf2011da2                          // vadd.f32      d1, d17, d18
+  .long  0xf3400db3                          // vmul.f32      d16, d16, d19
+  .long  0xedc30b28                          // vstr          d16, [r3, #160]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xe320f000                          // nop           {0}
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
+
+HIDDEN _sk_bicubic_n1y_vfp4
+.globl _sk_bicubic_n1y_vfp4
+_sk_bicubic_n1y_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c70f10                          // vmov.f32      d16, #1
+  .long  0xf2c73f18                          // vmov.f32      d19, #1.5
+  .long  0xedd32b18                          // vldr          d18, [r3, #96]
+  .long  0xf2600da2                          // vsub.f32      d16, d16, d18
+  .long  0xeddf2b0d                          // vldr          d18, [pc, #52]
+  .long  0xedd31b08                          // vldr          d17, [r3, #32]
+  .long  0xf2403cb2                          // vfma.f32      d19, d16, d18
+  .long  0xf2c3261f                          // vmov.i32      d18, #1056964608
+  .long  0xf2402cb3                          // vfma.f32      d18, d16, d19
+  .long  0xeddf3b0a                          // vldr          d19, [pc, #40]
+  .long  0xf2403cb2                          // vfma.f32      d19, d16, d18
+  .long  0xf3c3061f                          // vmov.i32      d16, #-1090519040
+  .long  0xf2011da0                          // vadd.f32      d1, d17, d16
+  .long  0xedc33b28                          // vstr          d19, [r3, #160]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xe320f000                          // nop           {0}
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0x3d638e39                          // .word         0x3d638e39
+  .long  0x3d638e39                          // .word         0x3d638e39
+
+HIDDEN _sk_bicubic_p1y_vfp4
+.globl _sk_bicubic_p1y_vfp4
+_sk_bicubic_p1y_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xf2c71f18                          // vmov.f32      d17, #1.5
+  .long  0xeddf0b0c                          // vldr          d16, [pc, #48]
+  .long  0xedd33b18                          // vldr          d19, [r3, #96]
+  .long  0xf2431cb0                          // vfma.f32      d17, d19, d16
+  .long  0xedd32b08                          // vldr          d18, [r3, #32]
+  .long  0xf2c3061f                          // vmov.i32      d16, #1056964608
+  .long  0xf2021da0                          // vadd.f32      d1, d18, d16
+  .long  0xf2430cb1                          // vfma.f32      d16, d19, d17
+  .long  0xeddf1b07                          // vldr          d17, [pc, #28]
+  .long  0xf2431cb0                          // vfma.f32      d17, d19, d16
+  .long  0xedc31b28                          // vstr          d17, [r3, #160]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0xbf955555                          // .word         0xbf955555
+  .long  0x3d638e39                          // .word         0x3d638e39
+  .long  0x3d638e39                          // .word         0x3d638e39
+
+HIDDEN _sk_bicubic_p3y_vfp4
+.globl _sk_bicubic_p3y_vfp4
+_sk_bicubic_p3y_vfp4:
+  .long  0xe5913000                          // ldr           r3, [r1]
+  .long  0xeddf0b0d                          // vldr          d16, [pc, #52]
+  .long  0xeddf3b0e                          // vldr          d19, [pc, #56]
+  .long  0xedd32b18                          // vldr          d18, [r3, #96]
+  .long  0xf2423cb0                          // vfma.f32      d19, d18, d16
+  .long  0xedd31b08                          // vldr          d17, [r3, #32]
+  .long  0xf3420db2                          // vmul.f32      d16, d18, d18
+  .long  0xf2c72f18                          // vmov.f32      d18, #1.5
+  .long  0xf2011da2                          // vadd.f32      d1, d17, d18
+  .long  0xf3400db3                          // vmul.f32      d16, d16, d19
+  .long  0xedc30b28                          // vstr          d16, [r3, #160]
+  .long  0xe2813008                          // add           r3, r1, #8
+  .long  0xe591c004                          // ldr           ip, [r1, #4]
+  .long  0xe1a01003                          // mov           r1, r3
+  .long  0xe12fff1c                          // bx            ip
+  .long  0xe320f000                          // nop           {0}
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0x3ec71c72                          // .word         0x3ec71c72
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
+  .long  0xbeaaaaab                          // .word         0xbeaaaaab
 #elif defined(__x86_64__)
 
 HIDDEN _sk_start_pipeline_hsw
@@ -7849,7 +8437,7 @@ _sk_load_4444_hsw:
   .byte  255                                 // (bad)
   .byte  255                                 // (bad)
   .byte  255                                 // (bad)
-  .byte  233,255,255,255,225                 // jmpq          ffffffffe2002284 <_sk_linear_gradient_2stops_hsw+0xffffffffe1fff152>
+  .byte  233,255,255,255,225                 // jmpq          ffffffffe2002284 <_sk_bicubic_p3y_hsw+0xffffffffe1ffecd5>
   .byte  255                                 // (bad)
   .byte  255                                 // (bad)
   .byte  255                                 // (bad)
@@ -8835,6 +9423,304 @@ _sk_linear_gradient_2stops_hsw:
   .byte  197,124,41,192                      // vmovaps       %ymm8,%ymm0
   .byte  255,224                             // jmpq          *%rax
 
+HIDDEN _sk_save_xy_hsw
+.globl _sk_save_xy_hsw
+_sk_save_xy_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  197,60,88,200                       // vaddps        %ymm0,%ymm8,%ymm9
+  .byte  196,67,125,8,209,1                  // vroundps      $0x1,%ymm9,%ymm10
+  .byte  196,65,52,92,202                    // vsubps        %ymm10,%ymm9,%ymm9
+  .byte  197,60,88,193                       // vaddps        %ymm1,%ymm8,%ymm8
+  .byte  196,67,125,8,208,1                  // vroundps      $0x1,%ymm8,%ymm10
+  .byte  196,65,60,92,194                    // vsubps        %ymm10,%ymm8,%ymm8
+  .byte  197,252,17,0                        // vmovups       %ymm0,(%rax)
+  .byte  197,252,17,72,32                    // vmovups       %ymm1,0x20(%rax)
+  .byte  197,124,17,72,64                    // vmovups       %ymm9,0x40(%rax)
+  .byte  197,124,17,64,96                    // vmovups       %ymm8,0x60(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_accumulate_hsw
+.globl _sk_accumulate_hsw
+_sk_accumulate_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  197,124,16,128,128,0,0,0            // vmovups       0x80(%rax),%ymm8
+  .byte  197,60,89,128,160,0,0,0             // vmulps        0xa0(%rax),%ymm8,%ymm8
+  .byte  196,226,61,184,224                  // vfmadd231ps   %ymm0,%ymm8,%ymm4
+  .byte  196,226,61,184,233                  // vfmadd231ps   %ymm1,%ymm8,%ymm5
+  .byte  196,226,61,184,242                  // vfmadd231ps   %ymm2,%ymm8,%ymm6
+  .byte  196,98,101,168,199                  // vfmadd213ps   %ymm7,%ymm3,%ymm8
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  197,124,41,199                      // vmovaps       %ymm8,%ymm7
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_nx_hsw
+.globl _sk_bilinear_nx_hsw
+_sk_bilinear_nx_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,226,125,88,192                  // vpbroadcastd  %xmm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  197,60,92,64,64                     // vsubps        0x40(%rax),%ymm8,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_px_hsw
+.globl _sk_bilinear_px_hsw
+_sk_bilinear_px_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,226,125,88,192                  // vpbroadcastd  %xmm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  197,124,16,64,64                    // vmovups       0x40(%rax),%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_ny_hsw
+.globl _sk_bilinear_ny_hsw
+_sk_bilinear_ny_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,226,125,88,201                  // vpbroadcastd  %xmm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  197,60,92,64,96                     // vsubps        0x60(%rax),%ymm8,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_py_hsw
+.globl _sk_bilinear_py_hsw
+_sk_bilinear_py_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,226,125,88,201                  // vpbroadcastd  %xmm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  197,124,16,64,96                    // vmovups       0x60(%rax),%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3x_hsw
+.globl _sk_bicubic_n3x_hsw
+_sk_bicubic_n3x_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,191                  // mov           $0xbfc00000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,226,125,88,192                  // vpbroadcastd  %xmm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  197,60,92,64,64                     // vsubps        0x40(%rax),%ymm8,%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,66,125,88,219                   // vpbroadcastd  %xmm11,%ymm11
+  .byte  196,66,61,168,211                   // vfmadd213ps   %ymm11,%ymm8,%ymm10
+  .byte  196,65,44,89,193                    // vmulps        %ymm9,%ymm10,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1x_hsw
+.globl _sk_bicubic_n1x_hsw
+_sk_bicubic_n1x_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,226,125,88,192                  // vpbroadcastd  %xmm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  197,60,92,64,64                     // vsubps        0x40(%rax),%ymm8,%ymm8
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,200                  // vmovd         %r8d,%xmm9
+  .byte  196,66,125,88,201                   // vpbroadcastd  %xmm9,%ymm9
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  196,66,61,168,202                   // vfmadd213ps   %ymm10,%ymm8,%ymm9
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  196,66,61,184,209                   // vfmadd231ps   %ymm9,%ymm8,%ymm10
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,200                  // vmovd         %r8d,%xmm9
+  .byte  196,66,125,88,201                   // vpbroadcastd  %xmm9,%ymm9
+  .byte  196,66,61,184,202                   // vfmadd231ps   %ymm10,%ymm8,%ymm9
+  .byte  197,124,17,136,128,0,0,0            // vmovups       %ymm9,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1x_hsw
+.globl _sk_bicubic_p1x_hsw
+_sk_bicubic_p1x_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,98,125,88,192                   // vpbroadcastd  %xmm0,%ymm8
+  .byte  197,188,88,0                        // vaddps        (%rax),%ymm8,%ymm0
+  .byte  197,124,16,72,64                    // vmovups       0x40(%rax),%ymm9
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,66,125,88,219                   // vpbroadcastd  %xmm11,%ymm11
+  .byte  196,66,53,168,211                   // vfmadd213ps   %ymm11,%ymm9,%ymm10
+  .byte  196,66,53,168,208                   // vfmadd213ps   %ymm8,%ymm9,%ymm10
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  196,66,53,184,194                   // vfmadd231ps   %ymm10,%ymm9,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3x_hsw
+.globl _sk_bicubic_p3x_hsw
+_sk_bicubic_p3x_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,226,125,88,192                  // vpbroadcastd  %xmm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  197,124,16,64,64                    // vmovups       0x40(%rax),%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,66,125,88,219                   // vpbroadcastd  %xmm11,%ymm11
+  .byte  196,66,61,168,211                   // vfmadd213ps   %ymm11,%ymm8,%ymm10
+  .byte  196,65,52,89,194                    // vmulps        %ymm10,%ymm9,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3y_hsw
+.globl _sk_bicubic_n3y_hsw
+_sk_bicubic_n3y_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,191                  // mov           $0xbfc00000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,226,125,88,201                  // vpbroadcastd  %xmm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  197,60,92,64,96                     // vsubps        0x60(%rax),%ymm8,%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,66,125,88,219                   // vpbroadcastd  %xmm11,%ymm11
+  .byte  196,66,61,168,211                   // vfmadd213ps   %ymm11,%ymm8,%ymm10
+  .byte  196,65,44,89,193                    // vmulps        %ymm9,%ymm10,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1y_hsw
+.globl _sk_bicubic_n1y_hsw
+_sk_bicubic_n1y_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,226,125,88,201                  // vpbroadcastd  %xmm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  197,60,92,64,96                     // vsubps        0x60(%rax),%ymm8,%ymm8
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,200                  // vmovd         %r8d,%xmm9
+  .byte  196,66,125,88,201                   // vpbroadcastd  %xmm9,%ymm9
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  196,66,61,168,202                   // vfmadd213ps   %ymm10,%ymm8,%ymm9
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  196,66,61,184,209                   // vfmadd231ps   %ymm9,%ymm8,%ymm10
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,200                  // vmovd         %r8d,%xmm9
+  .byte  196,66,125,88,201                   // vpbroadcastd  %xmm9,%ymm9
+  .byte  196,66,61,184,202                   // vfmadd231ps   %ymm10,%ymm8,%ymm9
+  .byte  197,124,17,136,160,0,0,0            // vmovups       %ymm9,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1y_hsw
+.globl _sk_bicubic_p1y_hsw
+_sk_bicubic_p1y_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,98,125,88,193                   // vpbroadcastd  %xmm1,%ymm8
+  .byte  197,188,88,72,32                    // vaddps        0x20(%rax),%ymm8,%ymm1
+  .byte  197,124,16,72,96                    // vmovups       0x60(%rax),%ymm9
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,66,125,88,219                   // vpbroadcastd  %xmm11,%ymm11
+  .byte  196,66,53,168,211                   // vfmadd213ps   %ymm11,%ymm9,%ymm10
+  .byte  196,66,53,168,208                   // vfmadd213ps   %ymm8,%ymm9,%ymm10
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,66,125,88,192                   // vpbroadcastd  %xmm8,%ymm8
+  .byte  196,66,53,184,194                   // vfmadd231ps   %ymm10,%ymm9,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3y_hsw
+.globl _sk_bicubic_p3y_hsw
+_sk_bicubic_p3y_hsw:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,226,125,88,201                  // vpbroadcastd  %xmm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  197,124,16,64,96                    // vmovups       0x60(%rax),%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,66,125,88,210                   // vpbroadcastd  %xmm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,66,125,88,219                   // vpbroadcastd  %xmm11,%ymm11
+  .byte  196,66,61,168,211                   // vfmadd213ps   %ymm11,%ymm8,%ymm10
+  .byte  196,65,52,89,194                    // vmulps        %ymm10,%ymm9,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
 HIDDEN _sk_start_pipeline_avx
 .globl _sk_start_pipeline_avx
 _sk_start_pipeline_avx:
@@ -12924,6 +13810,364 @@ _sk_linear_gradient_2stops_avx:
   .byte  197,124,41,192                      // vmovaps       %ymm8,%ymm0
   .byte  255,224                             // jmpq          *%rax
 
+HIDDEN _sk_save_xy_avx
+.globl _sk_save_xy_avx
+_sk_save_xy_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,67,121,4,192,0                  // vpermilps     $0x0,%xmm8,%xmm8
+  .byte  196,67,61,24,192,1                  // vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  .byte  197,60,88,200                       // vaddps        %ymm0,%ymm8,%ymm9
+  .byte  196,67,125,8,209,1                  // vroundps      $0x1,%ymm9,%ymm10
+  .byte  196,65,52,92,202                    // vsubps        %ymm10,%ymm9,%ymm9
+  .byte  197,60,88,193                       // vaddps        %ymm1,%ymm8,%ymm8
+  .byte  196,67,125,8,208,1                  // vroundps      $0x1,%ymm8,%ymm10
+  .byte  196,65,60,92,194                    // vsubps        %ymm10,%ymm8,%ymm8
+  .byte  197,252,17,0                        // vmovups       %ymm0,(%rax)
+  .byte  197,252,17,72,32                    // vmovups       %ymm1,0x20(%rax)
+  .byte  197,124,17,72,64                    // vmovups       %ymm9,0x40(%rax)
+  .byte  197,124,17,64,96                    // vmovups       %ymm8,0x60(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_accumulate_avx
+.globl _sk_accumulate_avx
+_sk_accumulate_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  197,124,16,128,128,0,0,0            // vmovups       0x80(%rax),%ymm8
+  .byte  197,60,89,128,160,0,0,0             // vmulps        0xa0(%rax),%ymm8,%ymm8
+  .byte  197,60,89,200                       // vmulps        %ymm0,%ymm8,%ymm9
+  .byte  197,180,88,228                      // vaddps        %ymm4,%ymm9,%ymm4
+  .byte  197,60,89,201                       // vmulps        %ymm1,%ymm8,%ymm9
+  .byte  197,180,88,237                      // vaddps        %ymm5,%ymm9,%ymm5
+  .byte  197,60,89,202                       // vmulps        %ymm2,%ymm8,%ymm9
+  .byte  197,180,88,246                      // vaddps        %ymm6,%ymm9,%ymm6
+  .byte  197,60,89,195                       // vmulps        %ymm3,%ymm8,%ymm8
+  .byte  197,188,88,255                      // vaddps        %ymm7,%ymm8,%ymm7
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_nx_avx
+.globl _sk_bilinear_nx_avx
+_sk_bilinear_nx_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,227,121,4,192,0                 // vpermilps     $0x0,%xmm0,%xmm0
+  .byte  196,227,125,24,192,1                // vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,67,121,4,192,0                  // vpermilps     $0x0,%xmm8,%xmm8
+  .byte  196,67,61,24,192,1                  // vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  .byte  197,60,92,64,64                     // vsubps        0x40(%rax),%ymm8,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_px_avx
+.globl _sk_bilinear_px_avx
+_sk_bilinear_px_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,227,121,4,192,0                 // vpermilps     $0x0,%xmm0,%xmm0
+  .byte  196,227,125,24,192,1                // vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  .byte  197,124,16,64,64                    // vmovups       0x40(%rax),%ymm8
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_ny_avx
+.globl _sk_bilinear_ny_avx
+_sk_bilinear_ny_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,227,121,4,201,0                 // vpermilps     $0x0,%xmm1,%xmm1
+  .byte  196,227,117,24,201,1                // vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,67,121,4,192,0                  // vpermilps     $0x0,%xmm8,%xmm8
+  .byte  196,67,61,24,192,1                  // vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  .byte  197,60,92,64,96                     // vsubps        0x60(%rax),%ymm8,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_py_avx
+.globl _sk_bilinear_py_avx
+_sk_bilinear_py_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,227,121,4,201,0                 // vpermilps     $0x0,%xmm1,%xmm1
+  .byte  196,227,117,24,201,1                // vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  .byte  197,124,16,64,96                    // vmovups       0x60(%rax),%ymm8
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3x_avx
+.globl _sk_bicubic_n3x_avx
+_sk_bicubic_n3x_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,191                  // mov           $0xbfc00000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,227,121,4,192,0                 // vpermilps     $0x0,%xmm0,%xmm0
+  .byte  196,227,125,24,192,1                // vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,67,121,4,192,0                  // vpermilps     $0x0,%xmm8,%xmm8
+  .byte  196,67,61,24,192,1                  // vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  .byte  197,60,92,64,64                     // vsubps        0x40(%rax),%ymm8,%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,67,121,4,219,0                  // vpermilps     $0x0,%xmm11,%xmm11
+  .byte  196,67,37,24,219,1                  // vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  .byte  196,65,44,89,192                    // vmulps        %ymm8,%ymm10,%ymm8
+  .byte  196,65,60,88,195                    // vaddps        %ymm11,%ymm8,%ymm8
+  .byte  196,65,52,89,192                    // vmulps        %ymm8,%ymm9,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1x_avx
+.globl _sk_bicubic_n1x_avx
+_sk_bicubic_n1x_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,227,121,4,192,0                 // vpermilps     $0x0,%xmm0,%xmm0
+  .byte  196,227,125,24,192,1                // vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,67,121,4,192,0                  // vpermilps     $0x0,%xmm8,%xmm8
+  .byte  196,67,61,24,192,1                  // vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  .byte  197,60,92,64,64                     // vsubps        0x40(%rax),%ymm8,%ymm8
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,200                  // vmovd         %r8d,%xmm9
+  .byte  196,67,121,4,201,0                  // vpermilps     $0x0,%xmm9,%xmm9
+  .byte  196,67,53,24,201,1                  // vinsertf128   $0x1,%xmm9,%ymm9,%ymm9
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,52,89,200                    // vmulps        %ymm8,%ymm9,%ymm9
+  .byte  196,65,52,88,202                    // vaddps        %ymm10,%ymm9,%ymm9
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,60,89,201                    // vmulps        %ymm9,%ymm8,%ymm9
+  .byte  196,65,44,88,201                    // vaddps        %ymm9,%ymm10,%ymm9
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,60,89,193                    // vmulps        %ymm9,%ymm8,%ymm8
+  .byte  196,65,44,88,192                    // vaddps        %ymm8,%ymm10,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1x_avx
+.globl _sk_bicubic_p1x_avx
+_sk_bicubic_p1x_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,227,121,4,192,0                 // vpermilps     $0x0,%xmm0,%xmm0
+  .byte  196,99,125,24,192,1                 // vinsertf128   $0x1,%xmm0,%ymm0,%ymm8
+  .byte  197,188,88,0                        // vaddps        (%rax),%ymm8,%ymm0
+  .byte  197,124,16,72,64                    // vmovups       0x40(%rax),%ymm9
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,67,121,4,219,0                  // vpermilps     $0x0,%xmm11,%xmm11
+  .byte  196,67,37,24,219,1                  // vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  .byte  196,65,52,89,210                    // vmulps        %ymm10,%ymm9,%ymm10
+  .byte  196,65,44,88,211                    // vaddps        %ymm11,%ymm10,%ymm10
+  .byte  196,65,52,89,210                    // vmulps        %ymm10,%ymm9,%ymm10
+  .byte  196,65,60,88,194                    // vaddps        %ymm10,%ymm8,%ymm8
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,52,89,192                    // vmulps        %ymm8,%ymm9,%ymm8
+  .byte  196,65,44,88,192                    // vaddps        %ymm8,%ymm10,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3x_avx
+.globl _sk_bicubic_p3x_avx
+_sk_bicubic_p3x_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,193,121,110,192                 // vmovd         %r8d,%xmm0
+  .byte  196,227,121,4,192,0                 // vpermilps     $0x0,%xmm0,%xmm0
+  .byte  196,227,125,24,192,1                // vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  .byte  197,252,88,0                        // vaddps        (%rax),%ymm0,%ymm0
+  .byte  197,124,16,64,64                    // vmovups       0x40(%rax),%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,67,121,4,219,0                  // vpermilps     $0x0,%xmm11,%xmm11
+  .byte  196,67,37,24,219,1                  // vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  .byte  196,65,60,89,194                    // vmulps        %ymm10,%ymm8,%ymm8
+  .byte  196,65,60,88,195                    // vaddps        %ymm11,%ymm8,%ymm8
+  .byte  196,65,52,89,192                    // vmulps        %ymm8,%ymm9,%ymm8
+  .byte  197,124,17,128,128,0,0,0            // vmovups       %ymm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3y_avx
+.globl _sk_bicubic_n3y_avx
+_sk_bicubic_n3y_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,191                  // mov           $0xbfc00000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,227,121,4,201,0                 // vpermilps     $0x0,%xmm1,%xmm1
+  .byte  196,227,117,24,201,1                // vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,67,121,4,192,0                  // vpermilps     $0x0,%xmm8,%xmm8
+  .byte  196,67,61,24,192,1                  // vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  .byte  197,60,92,64,96                     // vsubps        0x60(%rax),%ymm8,%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,67,121,4,219,0                  // vpermilps     $0x0,%xmm11,%xmm11
+  .byte  196,67,37,24,219,1                  // vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  .byte  196,65,44,89,192                    // vmulps        %ymm8,%ymm10,%ymm8
+  .byte  196,65,60,88,195                    // vaddps        %ymm11,%ymm8,%ymm8
+  .byte  196,65,52,89,192                    // vmulps        %ymm8,%ymm9,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1y_avx
+.globl _sk_bicubic_n1y_avx
+_sk_bicubic_n1y_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,191                    // mov           $0xbf000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,227,121,4,201,0                 // vpermilps     $0x0,%xmm1,%xmm1
+  .byte  196,227,117,24,201,1                // vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  65,184,0,0,128,63                   // mov           $0x3f800000,%r8d
+  .byte  196,65,121,110,192                  // vmovd         %r8d,%xmm8
+  .byte  196,67,121,4,192,0                  // vpermilps     $0x0,%xmm8,%xmm8
+  .byte  196,67,61,24,192,1                  // vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  .byte  197,60,92,64,96                     // vsubps        0x60(%rax),%ymm8,%ymm8
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,200                  // vmovd         %r8d,%xmm9
+  .byte  196,67,121,4,201,0                  // vpermilps     $0x0,%xmm9,%xmm9
+  .byte  196,67,53,24,201,1                  // vinsertf128   $0x1,%xmm9,%ymm9,%ymm9
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,52,89,200                    // vmulps        %ymm8,%ymm9,%ymm9
+  .byte  196,65,52,88,202                    // vaddps        %ymm10,%ymm9,%ymm9
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,60,89,201                    // vmulps        %ymm9,%ymm8,%ymm9
+  .byte  196,65,44,88,201                    // vaddps        %ymm9,%ymm10,%ymm9
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,60,89,193                    // vmulps        %ymm9,%ymm8,%ymm8
+  .byte  196,65,44,88,192                    // vaddps        %ymm8,%ymm10,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1y_avx
+.globl _sk_bicubic_p1y_avx
+_sk_bicubic_p1y_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,0,63                     // mov           $0x3f000000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,227,121,4,201,0                 // vpermilps     $0x0,%xmm1,%xmm1
+  .byte  196,99,117,24,193,1                 // vinsertf128   $0x1,%xmm1,%ymm1,%ymm8
+  .byte  197,188,88,72,32                    // vaddps        0x20(%rax),%ymm8,%ymm1
+  .byte  197,124,16,72,96                    // vmovups       0x60(%rax),%ymm9
+  .byte  65,184,85,85,149,191                // mov           $0xbf955555,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,67,121,4,219,0                  // vpermilps     $0x0,%xmm11,%xmm11
+  .byte  196,67,37,24,219,1                  // vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  .byte  196,65,52,89,210                    // vmulps        %ymm10,%ymm9,%ymm10
+  .byte  196,65,44,88,211                    // vaddps        %ymm11,%ymm10,%ymm10
+  .byte  196,65,52,89,210                    // vmulps        %ymm10,%ymm9,%ymm10
+  .byte  196,65,60,88,194                    // vaddps        %ymm10,%ymm8,%ymm8
+  .byte  65,184,57,142,99,61                 // mov           $0x3d638e39,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  196,65,52,89,192                    // vmulps        %ymm8,%ymm9,%ymm8
+  .byte  196,65,44,88,192                    // vaddps        %ymm8,%ymm10,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3y_avx
+.globl _sk_bicubic_p3y_avx
+_sk_bicubic_p3y_avx:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  65,184,0,0,192,63                   // mov           $0x3fc00000,%r8d
+  .byte  196,193,121,110,200                 // vmovd         %r8d,%xmm1
+  .byte  196,227,121,4,201,0                 // vpermilps     $0x0,%xmm1,%xmm1
+  .byte  196,227,117,24,201,1                // vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  .byte  197,244,88,72,32                    // vaddps        0x20(%rax),%ymm1,%ymm1
+  .byte  197,124,16,64,96                    // vmovups       0x60(%rax),%ymm8
+  .byte  196,65,60,89,200                    // vmulps        %ymm8,%ymm8,%ymm9
+  .byte  65,184,114,28,199,62                // mov           $0x3ec71c72,%r8d
+  .byte  196,65,121,110,208                  // vmovd         %r8d,%xmm10
+  .byte  196,67,121,4,210,0                  // vpermilps     $0x0,%xmm10,%xmm10
+  .byte  196,67,45,24,210,1                  // vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  .byte  65,184,171,170,170,190              // mov           $0xbeaaaaab,%r8d
+  .byte  196,65,121,110,216                  // vmovd         %r8d,%xmm11
+  .byte  196,67,121,4,219,0                  // vpermilps     $0x0,%xmm11,%xmm11
+  .byte  196,67,37,24,219,1                  // vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  .byte  196,65,60,89,194                    // vmulps        %ymm10,%ymm8,%ymm8
+  .byte  196,65,60,88,195                    // vaddps        %ymm11,%ymm8,%ymm8
+  .byte  196,65,52,89,192                    // vmulps        %ymm8,%ymm9,%ymm8
+  .byte  197,124,17,128,160,0,0,0            // vmovups       %ymm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
 HIDDEN _sk_start_pipeline_sse41
 .globl _sk_start_pipeline_sse41
 _sk_start_pipeline_sse41:
@@ -16164,6 +17408,346 @@ _sk_linear_gradient_2stops_sse41:
   .byte  72,173                              // lods          %ds:(%rsi),%rax
   .byte  255,224                             // jmpq          *%rax
 
+HIDDEN _sk_save_xy_sse41
+.globl _sk_save_xy_sse41
+_sk_save_xy_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,40,200                        // movaps        %xmm8,%xmm9
+  .byte  68,15,88,200                        // addps         %xmm0,%xmm9
+  .byte  102,69,15,58,8,209,1                // roundps       $0x1,%xmm9,%xmm10
+  .byte  69,15,92,202                        // subps         %xmm10,%xmm9
+  .byte  68,15,88,193                        // addps         %xmm1,%xmm8
+  .byte  102,69,15,58,8,208,1                // roundps       $0x1,%xmm8,%xmm10
+  .byte  69,15,92,194                        // subps         %xmm10,%xmm8
+  .byte  15,17,0                             // movups        %xmm0,(%rax)
+  .byte  15,17,72,32                         // movups        %xmm1,0x20(%rax)
+  .byte  68,15,17,72,64                      // movups        %xmm9,0x40(%rax)
+  .byte  68,15,17,64,96                      // movups        %xmm8,0x60(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_accumulate_sse41
+.globl _sk_accumulate_sse41
+_sk_accumulate_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  68,15,16,128,128,0,0,0              // movups        0x80(%rax),%xmm8
+  .byte  68,15,16,136,160,0,0,0              // movups        0xa0(%rax),%xmm9
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,40,193                        // movaps        %xmm9,%xmm8
+  .byte  68,15,89,192                        // mulps         %xmm0,%xmm8
+  .byte  65,15,88,224                        // addps         %xmm8,%xmm4
+  .byte  69,15,40,193                        // movaps        %xmm9,%xmm8
+  .byte  68,15,89,193                        // mulps         %xmm1,%xmm8
+  .byte  65,15,88,232                        // addps         %xmm8,%xmm5
+  .byte  69,15,40,193                        // movaps        %xmm9,%xmm8
+  .byte  68,15,89,194                        // mulps         %xmm2,%xmm8
+  .byte  65,15,88,240                        // addps         %xmm8,%xmm6
+  .byte  68,15,89,203                        // mulps         %xmm3,%xmm9
+  .byte  65,15,88,249                        // addps         %xmm9,%xmm7
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_nx_sse41
+.globl _sk_bilinear_nx_sse41
+_sk_bilinear_nx_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  68,15,17,128,128,0,0,0              // movups        %xmm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_px_sse41
+.globl _sk_bilinear_px_sse41
+_sk_bilinear_px_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  68,15,17,136,128,0,0,0              // movups        %xmm9,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_ny_sse41
+.globl _sk_bilinear_ny_sse41
+_sk_bilinear_ny_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  68,15,17,128,160,0,0,0              // movups        %xmm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_py_sse41
+.globl _sk_bilinear_py_sse41
+_sk_bilinear_py_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  68,15,17,136,160,0,0,0              // movups        %xmm9,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3x_sse41
+.globl _sk_bicubic_n3x_sse41
+_sk_bicubic_n3x_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,191                     // mov           $0xbfc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,89,192                        // mulps         %xmm8,%xmm8
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  68,15,17,136,128,0,0,0              // movups        %xmm9,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1x_sse41
+.globl _sk_bicubic_n1x_sse41
+_sk_bicubic_n1x_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  68,15,17,136,128,0,0,0              // movups        %xmm9,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1x_sse41
+.globl _sk_bicubic_p1x_sse41
+_sk_bicubic_p1x_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,217                   // movd          %ecx,%xmm11
+  .byte  69,15,198,219,0                     // shufps        $0x0,%xmm11,%xmm11
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,211                        // addps         %xmm11,%xmm10
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  68,15,17,144,128,0,0,0              // movups        %xmm10,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3x_sse41
+.globl _sk_bicubic_p3x_sse41
+_sk_bicubic_p3x_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  69,15,89,201                        // mulps         %xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,88,194                        // addps         %xmm10,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  68,15,17,128,128,0,0,0              // movups        %xmm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3y_sse41
+.globl _sk_bicubic_n3y_sse41
+_sk_bicubic_n3y_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,191                     // mov           $0xbfc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,89,192                        // mulps         %xmm8,%xmm8
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  68,15,17,136,160,0,0,0              // movups        %xmm9,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1y_sse41
+.globl _sk_bicubic_n1y_sse41
+_sk_bicubic_n1y_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  68,15,17,136,160,0,0,0              // movups        %xmm9,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1y_sse41
+.globl _sk_bicubic_p1y_sse41
+_sk_bicubic_p1y_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,217                   // movd          %ecx,%xmm11
+  .byte  69,15,198,219,0                     // shufps        $0x0,%xmm11,%xmm11
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,211                        // addps         %xmm11,%xmm10
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  68,15,17,144,160,0,0,0              // movups        %xmm10,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3y_sse41
+.globl _sk_bicubic_p3y_sse41
+_sk_bicubic_p3y_sse41:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  69,15,89,201                        // mulps         %xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,88,194                        // addps         %xmm10,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  68,15,17,128,160,0,0,0              // movups        %xmm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
 HIDDEN _sk_start_pipeline_sse2
 .globl _sk_start_pipeline_sse2
 _sk_start_pipeline_sse2:
@@ -19638,4 +21222,357 @@ _sk_linear_gradient_2stops_sse2:
   .byte  65,15,88,217                        // addps         %xmm9,%xmm3
   .byte  72,173                              // lods          %ds:(%rsi),%rax
   .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_save_xy_sse2
+.globl _sk_save_xy_sse2
+_sk_save_xy_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,40,200                        // movaps        %xmm8,%xmm9
+  .byte  68,15,88,200                        // addps         %xmm0,%xmm9
+  .byte  243,69,15,91,209                    // cvttps2dq     %xmm9,%xmm10
+  .byte  69,15,91,210                        // cvtdq2ps      %xmm10,%xmm10
+  .byte  69,15,40,217                        // movaps        %xmm9,%xmm11
+  .byte  69,15,194,218,1                     // cmpltps       %xmm10,%xmm11
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,225                   // movd          %ecx,%xmm12
+  .byte  69,15,198,228,0                     // shufps        $0x0,%xmm12,%xmm12
+  .byte  69,15,84,220                        // andps         %xmm12,%xmm11
+  .byte  69,15,92,211                        // subps         %xmm11,%xmm10
+  .byte  69,15,92,202                        // subps         %xmm10,%xmm9
+  .byte  68,15,88,193                        // addps         %xmm1,%xmm8
+  .byte  243,69,15,91,208                    // cvttps2dq     %xmm8,%xmm10
+  .byte  69,15,91,210                        // cvtdq2ps      %xmm10,%xmm10
+  .byte  69,15,40,216                        // movaps        %xmm8,%xmm11
+  .byte  69,15,194,218,1                     // cmpltps       %xmm10,%xmm11
+  .byte  69,15,84,220                        // andps         %xmm12,%xmm11
+  .byte  69,15,92,211                        // subps         %xmm11,%xmm10
+  .byte  69,15,92,194                        // subps         %xmm10,%xmm8
+  .byte  15,17,0                             // movups        %xmm0,(%rax)
+  .byte  15,17,72,32                         // movups        %xmm1,0x20(%rax)
+  .byte  68,15,17,72,64                      // movups        %xmm9,0x40(%rax)
+  .byte  68,15,17,64,96                      // movups        %xmm8,0x60(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_accumulate_sse2
+.globl _sk_accumulate_sse2
+_sk_accumulate_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  68,15,16,128,128,0,0,0              // movups        0x80(%rax),%xmm8
+  .byte  68,15,16,136,160,0,0,0              // movups        0xa0(%rax),%xmm9
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,40,193                        // movaps        %xmm9,%xmm8
+  .byte  68,15,89,192                        // mulps         %xmm0,%xmm8
+  .byte  65,15,88,224                        // addps         %xmm8,%xmm4
+  .byte  69,15,40,193                        // movaps        %xmm9,%xmm8
+  .byte  68,15,89,193                        // mulps         %xmm1,%xmm8
+  .byte  65,15,88,232                        // addps         %xmm8,%xmm5
+  .byte  69,15,40,193                        // movaps        %xmm9,%xmm8
+  .byte  68,15,89,194                        // mulps         %xmm2,%xmm8
+  .byte  65,15,88,240                        // addps         %xmm8,%xmm6
+  .byte  68,15,89,203                        // mulps         %xmm3,%xmm9
+  .byte  65,15,88,249                        // addps         %xmm9,%xmm7
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_nx_sse2
+.globl _sk_bilinear_nx_sse2
+_sk_bilinear_nx_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  68,15,17,128,128,0,0,0              // movups        %xmm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_px_sse2
+.globl _sk_bilinear_px_sse2
+_sk_bilinear_px_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  68,15,17,136,128,0,0,0              // movups        %xmm9,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_ny_sse2
+.globl _sk_bilinear_ny_sse2
+_sk_bilinear_ny_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  68,15,17,128,160,0,0,0              // movups        %xmm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bilinear_py_sse2
+.globl _sk_bilinear_py_sse2
+_sk_bilinear_py_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  68,15,17,136,160,0,0,0              // movups        %xmm9,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3x_sse2
+.globl _sk_bicubic_n3x_sse2
+_sk_bicubic_n3x_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,191                     // mov           $0xbfc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,89,192                        // mulps         %xmm8,%xmm8
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  68,15,17,136,128,0,0,0              // movups        %xmm9,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1x_sse2
+.globl _sk_bicubic_n1x_sse2
+_sk_bicubic_n1x_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  68,15,17,136,128,0,0,0              // movups        %xmm9,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1x_sse2
+.globl _sk_bicubic_p1x_sse2
+_sk_bicubic_p1x_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,217                   // movd          %ecx,%xmm11
+  .byte  69,15,198,219,0                     // shufps        $0x0,%xmm11,%xmm11
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,211                        // addps         %xmm11,%xmm10
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  68,15,17,144,128,0,0,0              // movups        %xmm10,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3x_sse2
+.globl _sk_bicubic_p3x_sse2
+_sk_bicubic_p3x_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,0                             // movups        (%rax),%xmm0
+  .byte  68,15,16,72,64                      // movups        0x40(%rax),%xmm9
+  .byte  65,15,88,192                        // addps         %xmm8,%xmm0
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  69,15,89,201                        // mulps         %xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,88,194                        // addps         %xmm10,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  68,15,17,128,128,0,0,0              // movups        %xmm8,0x80(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n3y_sse2
+.globl _sk_bicubic_n3y_sse2
+_sk_bicubic_n3y_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,191                     // mov           $0xbfc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,89,192                        // mulps         %xmm8,%xmm8
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  68,15,17,136,160,0,0,0              // movups        %xmm9,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_n1y_sse2
+.globl _sk_bicubic_n1y_sse2
+_sk_bicubic_n1y_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,191                       // mov           $0xbf000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,0,0,128,63                      // mov           $0x3f800000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,92,193                        // subps         %xmm9,%xmm8
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,201                   // movd          %ecx,%xmm9
+  .byte  69,15,198,201,0                     // shufps        $0x0,%xmm9,%xmm9
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,89,200                        // mulps         %xmm8,%xmm9
+  .byte  69,15,88,202                        // addps         %xmm10,%xmm9
+  .byte  68,15,17,136,160,0,0,0              // movups        %xmm9,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p1y_sse2
+.globl _sk_bicubic_p1y_sse2
+_sk_bicubic_p1y_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,0,63                        // mov           $0x3f000000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,85,85,149,191                   // mov           $0xbf955555,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,217                   // movd          %ecx,%xmm11
+  .byte  69,15,198,219,0                     // shufps        $0x0,%xmm11,%xmm11
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,211                        // addps         %xmm11,%xmm10
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  185,57,142,99,61                    // mov           $0x3d638e39,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,209                        // mulps         %xmm9,%xmm10
+  .byte  69,15,88,208                        // addps         %xmm8,%xmm10
+  .byte  68,15,17,144,160,0,0,0              // movups        %xmm10,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
+
+HIDDEN _sk_bicubic_p3y_sse2
+.globl _sk_bicubic_p3y_sse2
+_sk_bicubic_p3y_sse2:
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  185,0,0,192,63                      // mov           $0x3fc00000,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  15,16,72,32                         // movups        0x20(%rax),%xmm1
+  .byte  68,15,16,72,96                      // movups        0x60(%rax),%xmm9
+  .byte  65,15,88,200                        // addps         %xmm8,%xmm1
+  .byte  185,114,28,199,62                   // mov           $0x3ec71c72,%ecx
+  .byte  102,68,15,110,193                   // movd          %ecx,%xmm8
+  .byte  69,15,198,192,0                     // shufps        $0x0,%xmm8,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  69,15,89,201                        // mulps         %xmm9,%xmm9
+  .byte  185,171,170,170,190                 // mov           $0xbeaaaaab,%ecx
+  .byte  102,68,15,110,209                   // movd          %ecx,%xmm10
+  .byte  69,15,198,210,0                     // shufps        $0x0,%xmm10,%xmm10
+  .byte  69,15,88,194                        // addps         %xmm10,%xmm8
+  .byte  69,15,89,193                        // mulps         %xmm9,%xmm8
+  .byte  68,15,17,128,160,0,0,0              // movups        %xmm8,0xa0(%rax)
+  .byte  72,173                              // lods          %ds:(%rsi),%rax
+  .byte  255,224                             // jmpq          *%rax
 #endif
diff --git a/src/jumper/SkJumper_generated_win.S b/src/jumper/SkJumper_generated_win.S
index 043da8576b..b305f23943 100644
--- a/src/jumper/SkJumper_generated_win.S
+++ b/src/jumper/SkJumper_generated_win.S
@@ -1357,7 +1357,7 @@ _sk_lerp_565_hsw LABEL PROC
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
-  DB  233,255,255,255,225                 ; jmpq          ffffffffe2001478 <_sk_linear_gradient_2stops_hsw+0xffffffffe1ffe296>
+  DB  233,255,255,255,225                 ; jmpq          ffffffffe2001478 <_sk_bicubic_p3y_hsw+0xffffffffe1ffde19>
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
@@ -2328,7 +2328,7 @@ _sk_load_4444_hsw LABEL PROC
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
-  DB  233,255,255,255,225                 ; jmpq          ffffffffe2002334 <_sk_linear_gradient_2stops_hsw+0xffffffffe1fff152>
+  DB  233,255,255,255,225                 ; jmpq          ffffffffe2002334 <_sk_bicubic_p3y_hsw+0xffffffffe1ffecd5>
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
   DB  255                                 ; (bad)
@@ -3289,6 +3289,290 @@ _sk_linear_gradient_2stops_hsw LABEL PROC
   DB  197,124,41,192                      ; vmovaps       %ymm8,%ymm0
   DB  255,224                             ; jmpq          *%rax
 
+PUBLIC _sk_save_xy_hsw
+_sk_save_xy_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  197,60,88,200                       ; vaddps        %ymm0,%ymm8,%ymm9
+  DB  196,67,125,8,209,1                  ; vroundps      $0x1,%ymm9,%ymm10
+  DB  196,65,52,92,202                    ; vsubps        %ymm10,%ymm9,%ymm9
+  DB  197,60,88,193                       ; vaddps        %ymm1,%ymm8,%ymm8
+  DB  196,67,125,8,208,1                  ; vroundps      $0x1,%ymm8,%ymm10
+  DB  196,65,60,92,194                    ; vsubps        %ymm10,%ymm8,%ymm8
+  DB  197,252,17,0                        ; vmovups       %ymm0,(%rax)
+  DB  197,252,17,72,32                    ; vmovups       %ymm1,0x20(%rax)
+  DB  197,124,17,72,64                    ; vmovups       %ymm9,0x40(%rax)
+  DB  197,124,17,64,96                    ; vmovups       %ymm8,0x60(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_accumulate_hsw
+_sk_accumulate_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  197,124,16,128,128,0,0,0            ; vmovups       0x80(%rax),%ymm8
+  DB  197,60,89,128,160,0,0,0             ; vmulps        0xa0(%rax),%ymm8,%ymm8
+  DB  196,226,61,184,224                  ; vfmadd231ps   %ymm0,%ymm8,%ymm4
+  DB  196,226,61,184,233                  ; vfmadd231ps   %ymm1,%ymm8,%ymm5
+  DB  196,226,61,184,242                  ; vfmadd231ps   %ymm2,%ymm8,%ymm6
+  DB  196,98,101,168,199                  ; vfmadd213ps   %ymm7,%ymm3,%ymm8
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  197,124,41,199                      ; vmovaps       %ymm8,%ymm7
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_nx_hsw
+_sk_bilinear_nx_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,226,125,88,192                  ; vpbroadcastd  %xmm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  197,60,92,64,64                     ; vsubps        0x40(%rax),%ymm8,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_px_hsw
+_sk_bilinear_px_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,226,125,88,192                  ; vpbroadcastd  %xmm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  197,124,16,64,64                    ; vmovups       0x40(%rax),%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_ny_hsw
+_sk_bilinear_ny_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,226,125,88,201                  ; vpbroadcastd  %xmm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  197,60,92,64,96                     ; vsubps        0x60(%rax),%ymm8,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_py_hsw
+_sk_bilinear_py_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,226,125,88,201                  ; vpbroadcastd  %xmm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  197,124,16,64,96                    ; vmovups       0x60(%rax),%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3x_hsw
+_sk_bicubic_n3x_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,191                  ; mov           $0xbfc00000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,226,125,88,192                  ; vpbroadcastd  %xmm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  197,60,92,64,64                     ; vsubps        0x40(%rax),%ymm8,%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,66,125,88,219                   ; vpbroadcastd  %xmm11,%ymm11
+  DB  196,66,61,168,211                   ; vfmadd213ps   %ymm11,%ymm8,%ymm10
+  DB  196,65,44,89,193                    ; vmulps        %ymm9,%ymm10,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1x_hsw
+_sk_bicubic_n1x_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,226,125,88,192                  ; vpbroadcastd  %xmm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  197,60,92,64,64                     ; vsubps        0x40(%rax),%ymm8,%ymm8
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,200                  ; vmovd         %r8d,%xmm9
+  DB  196,66,125,88,201                   ; vpbroadcastd  %xmm9,%ymm9
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  196,66,61,168,202                   ; vfmadd213ps   %ymm10,%ymm8,%ymm9
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  196,66,61,184,209                   ; vfmadd231ps   %ymm9,%ymm8,%ymm10
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,200                  ; vmovd         %r8d,%xmm9
+  DB  196,66,125,88,201                   ; vpbroadcastd  %xmm9,%ymm9
+  DB  196,66,61,184,202                   ; vfmadd231ps   %ymm10,%ymm8,%ymm9
+  DB  197,124,17,136,128,0,0,0            ; vmovups       %ymm9,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1x_hsw
+_sk_bicubic_p1x_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,98,125,88,192                   ; vpbroadcastd  %xmm0,%ymm8
+  DB  197,188,88,0                        ; vaddps        (%rax),%ymm8,%ymm0
+  DB  197,124,16,72,64                    ; vmovups       0x40(%rax),%ymm9
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,66,125,88,219                   ; vpbroadcastd  %xmm11,%ymm11
+  DB  196,66,53,168,211                   ; vfmadd213ps   %ymm11,%ymm9,%ymm10
+  DB  196,66,53,168,208                   ; vfmadd213ps   %ymm8,%ymm9,%ymm10
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  196,66,53,184,194                   ; vfmadd231ps   %ymm10,%ymm9,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3x_hsw
+_sk_bicubic_p3x_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,226,125,88,192                  ; vpbroadcastd  %xmm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  197,124,16,64,64                    ; vmovups       0x40(%rax),%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,66,125,88,219                   ; vpbroadcastd  %xmm11,%ymm11
+  DB  196,66,61,168,211                   ; vfmadd213ps   %ymm11,%ymm8,%ymm10
+  DB  196,65,52,89,194                    ; vmulps        %ymm10,%ymm9,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3y_hsw
+_sk_bicubic_n3y_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,191                  ; mov           $0xbfc00000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,226,125,88,201                  ; vpbroadcastd  %xmm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  197,60,92,64,96                     ; vsubps        0x60(%rax),%ymm8,%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,66,125,88,219                   ; vpbroadcastd  %xmm11,%ymm11
+  DB  196,66,61,168,211                   ; vfmadd213ps   %ymm11,%ymm8,%ymm10
+  DB  196,65,44,89,193                    ; vmulps        %ymm9,%ymm10,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1y_hsw
+_sk_bicubic_n1y_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,226,125,88,201                  ; vpbroadcastd  %xmm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  197,60,92,64,96                     ; vsubps        0x60(%rax),%ymm8,%ymm8
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,200                  ; vmovd         %r8d,%xmm9
+  DB  196,66,125,88,201                   ; vpbroadcastd  %xmm9,%ymm9
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  196,66,61,168,202                   ; vfmadd213ps   %ymm10,%ymm8,%ymm9
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  196,66,61,184,209                   ; vfmadd231ps   %ymm9,%ymm8,%ymm10
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,200                  ; vmovd         %r8d,%xmm9
+  DB  196,66,125,88,201                   ; vpbroadcastd  %xmm9,%ymm9
+  DB  196,66,61,184,202                   ; vfmadd231ps   %ymm10,%ymm8,%ymm9
+  DB  197,124,17,136,160,0,0,0            ; vmovups       %ymm9,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1y_hsw
+_sk_bicubic_p1y_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,98,125,88,193                   ; vpbroadcastd  %xmm1,%ymm8
+  DB  197,188,88,72,32                    ; vaddps        0x20(%rax),%ymm8,%ymm1
+  DB  197,124,16,72,96                    ; vmovups       0x60(%rax),%ymm9
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,66,125,88,219                   ; vpbroadcastd  %xmm11,%ymm11
+  DB  196,66,53,168,211                   ; vfmadd213ps   %ymm11,%ymm9,%ymm10
+  DB  196,66,53,168,208                   ; vfmadd213ps   %ymm8,%ymm9,%ymm10
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,66,125,88,192                   ; vpbroadcastd  %xmm8,%ymm8
+  DB  196,66,53,184,194                   ; vfmadd231ps   %ymm10,%ymm9,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3y_hsw
+_sk_bicubic_p3y_hsw LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,226,125,88,201                  ; vpbroadcastd  %xmm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  197,124,16,64,96                    ; vmovups       0x60(%rax),%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,66,125,88,210                   ; vpbroadcastd  %xmm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,66,125,88,219                   ; vpbroadcastd  %xmm11,%ymm11
+  DB  196,66,61,168,211                   ; vfmadd213ps   %ymm11,%ymm8,%ymm10
+  DB  196,65,52,89,194                    ; vmulps        %ymm10,%ymm9,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
 PUBLIC _sk_start_pipeline_avx
 _sk_start_pipeline_avx LABEL PROC
   DB  65,87                               ; push          %r15
@@ -7321,6 +7605,350 @@ _sk_linear_gradient_2stops_avx LABEL PROC
   DB  197,124,41,192                      ; vmovaps       %ymm8,%ymm0
   DB  255,224                             ; jmpq          *%rax
 
+PUBLIC _sk_save_xy_avx
+_sk_save_xy_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,67,121,4,192,0                  ; vpermilps     $0x0,%xmm8,%xmm8
+  DB  196,67,61,24,192,1                  ; vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  DB  197,60,88,200                       ; vaddps        %ymm0,%ymm8,%ymm9
+  DB  196,67,125,8,209,1                  ; vroundps      $0x1,%ymm9,%ymm10
+  DB  196,65,52,92,202                    ; vsubps        %ymm10,%ymm9,%ymm9
+  DB  197,60,88,193                       ; vaddps        %ymm1,%ymm8,%ymm8
+  DB  196,67,125,8,208,1                  ; vroundps      $0x1,%ymm8,%ymm10
+  DB  196,65,60,92,194                    ; vsubps        %ymm10,%ymm8,%ymm8
+  DB  197,252,17,0                        ; vmovups       %ymm0,(%rax)
+  DB  197,252,17,72,32                    ; vmovups       %ymm1,0x20(%rax)
+  DB  197,124,17,72,64                    ; vmovups       %ymm9,0x40(%rax)
+  DB  197,124,17,64,96                    ; vmovups       %ymm8,0x60(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_accumulate_avx
+_sk_accumulate_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  197,124,16,128,128,0,0,0            ; vmovups       0x80(%rax),%ymm8
+  DB  197,60,89,128,160,0,0,0             ; vmulps        0xa0(%rax),%ymm8,%ymm8
+  DB  197,60,89,200                       ; vmulps        %ymm0,%ymm8,%ymm9
+  DB  197,180,88,228                      ; vaddps        %ymm4,%ymm9,%ymm4
+  DB  197,60,89,201                       ; vmulps        %ymm1,%ymm8,%ymm9
+  DB  197,180,88,237                      ; vaddps        %ymm5,%ymm9,%ymm5
+  DB  197,60,89,202                       ; vmulps        %ymm2,%ymm8,%ymm9
+  DB  197,180,88,246                      ; vaddps        %ymm6,%ymm9,%ymm6
+  DB  197,60,89,195                       ; vmulps        %ymm3,%ymm8,%ymm8
+  DB  197,188,88,255                      ; vaddps        %ymm7,%ymm8,%ymm7
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_nx_avx
+_sk_bilinear_nx_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,227,121,4,192,0                 ; vpermilps     $0x0,%xmm0,%xmm0
+  DB  196,227,125,24,192,1                ; vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,67,121,4,192,0                  ; vpermilps     $0x0,%xmm8,%xmm8
+  DB  196,67,61,24,192,1                  ; vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  DB  197,60,92,64,64                     ; vsubps        0x40(%rax),%ymm8,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_px_avx
+_sk_bilinear_px_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,227,121,4,192,0                 ; vpermilps     $0x0,%xmm0,%xmm0
+  DB  196,227,125,24,192,1                ; vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  DB  197,124,16,64,64                    ; vmovups       0x40(%rax),%ymm8
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_ny_avx
+_sk_bilinear_ny_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,227,121,4,201,0                 ; vpermilps     $0x0,%xmm1,%xmm1
+  DB  196,227,117,24,201,1                ; vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,67,121,4,192,0                  ; vpermilps     $0x0,%xmm8,%xmm8
+  DB  196,67,61,24,192,1                  ; vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  DB  197,60,92,64,96                     ; vsubps        0x60(%rax),%ymm8,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_py_avx
+_sk_bilinear_py_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,227,121,4,201,0                 ; vpermilps     $0x0,%xmm1,%xmm1
+  DB  196,227,117,24,201,1                ; vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  DB  197,124,16,64,96                    ; vmovups       0x60(%rax),%ymm8
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3x_avx
+_sk_bicubic_n3x_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,191                  ; mov           $0xbfc00000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,227,121,4,192,0                 ; vpermilps     $0x0,%xmm0,%xmm0
+  DB  196,227,125,24,192,1                ; vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,67,121,4,192,0                  ; vpermilps     $0x0,%xmm8,%xmm8
+  DB  196,67,61,24,192,1                  ; vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  DB  197,60,92,64,64                     ; vsubps        0x40(%rax),%ymm8,%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,67,121,4,219,0                  ; vpermilps     $0x0,%xmm11,%xmm11
+  DB  196,67,37,24,219,1                  ; vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  DB  196,65,44,89,192                    ; vmulps        %ymm8,%ymm10,%ymm8
+  DB  196,65,60,88,195                    ; vaddps        %ymm11,%ymm8,%ymm8
+  DB  196,65,52,89,192                    ; vmulps        %ymm8,%ymm9,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1x_avx
+_sk_bicubic_n1x_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,227,121,4,192,0                 ; vpermilps     $0x0,%xmm0,%xmm0
+  DB  196,227,125,24,192,1                ; vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,67,121,4,192,0                  ; vpermilps     $0x0,%xmm8,%xmm8
+  DB  196,67,61,24,192,1                  ; vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  DB  197,60,92,64,64                     ; vsubps        0x40(%rax),%ymm8,%ymm8
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,200                  ; vmovd         %r8d,%xmm9
+  DB  196,67,121,4,201,0                  ; vpermilps     $0x0,%xmm9,%xmm9
+  DB  196,67,53,24,201,1                  ; vinsertf128   $0x1,%xmm9,%ymm9,%ymm9
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,52,89,200                    ; vmulps        %ymm8,%ymm9,%ymm9
+  DB  196,65,52,88,202                    ; vaddps        %ymm10,%ymm9,%ymm9
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,60,89,201                    ; vmulps        %ymm9,%ymm8,%ymm9
+  DB  196,65,44,88,201                    ; vaddps        %ymm9,%ymm10,%ymm9
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,60,89,193                    ; vmulps        %ymm9,%ymm8,%ymm8
+  DB  196,65,44,88,192                    ; vaddps        %ymm8,%ymm10,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1x_avx
+_sk_bicubic_p1x_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,227,121,4,192,0                 ; vpermilps     $0x0,%xmm0,%xmm0
+  DB  196,99,125,24,192,1                 ; vinsertf128   $0x1,%xmm0,%ymm0,%ymm8
+  DB  197,188,88,0                        ; vaddps        (%rax),%ymm8,%ymm0
+  DB  197,124,16,72,64                    ; vmovups       0x40(%rax),%ymm9
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,67,121,4,219,0                  ; vpermilps     $0x0,%xmm11,%xmm11
+  DB  196,67,37,24,219,1                  ; vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  DB  196,65,52,89,210                    ; vmulps        %ymm10,%ymm9,%ymm10
+  DB  196,65,44,88,211                    ; vaddps        %ymm11,%ymm10,%ymm10
+  DB  196,65,52,89,210                    ; vmulps        %ymm10,%ymm9,%ymm10
+  DB  196,65,60,88,194                    ; vaddps        %ymm10,%ymm8,%ymm8
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,52,89,192                    ; vmulps        %ymm8,%ymm9,%ymm8
+  DB  196,65,44,88,192                    ; vaddps        %ymm8,%ymm10,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3x_avx
+_sk_bicubic_p3x_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,193,121,110,192                 ; vmovd         %r8d,%xmm0
+  DB  196,227,121,4,192,0                 ; vpermilps     $0x0,%xmm0,%xmm0
+  DB  196,227,125,24,192,1                ; vinsertf128   $0x1,%xmm0,%ymm0,%ymm0
+  DB  197,252,88,0                        ; vaddps        (%rax),%ymm0,%ymm0
+  DB  197,124,16,64,64                    ; vmovups       0x40(%rax),%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,67,121,4,219,0                  ; vpermilps     $0x0,%xmm11,%xmm11
+  DB  196,67,37,24,219,1                  ; vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  DB  196,65,60,89,194                    ; vmulps        %ymm10,%ymm8,%ymm8
+  DB  196,65,60,88,195                    ; vaddps        %ymm11,%ymm8,%ymm8
+  DB  196,65,52,89,192                    ; vmulps        %ymm8,%ymm9,%ymm8
+  DB  197,124,17,128,128,0,0,0            ; vmovups       %ymm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3y_avx
+_sk_bicubic_n3y_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,191                  ; mov           $0xbfc00000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,227,121,4,201,0                 ; vpermilps     $0x0,%xmm1,%xmm1
+  DB  196,227,117,24,201,1                ; vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,67,121,4,192,0                  ; vpermilps     $0x0,%xmm8,%xmm8
+  DB  196,67,61,24,192,1                  ; vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  DB  197,60,92,64,96                     ; vsubps        0x60(%rax),%ymm8,%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,67,121,4,219,0                  ; vpermilps     $0x0,%xmm11,%xmm11
+  DB  196,67,37,24,219,1                  ; vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  DB  196,65,44,89,192                    ; vmulps        %ymm8,%ymm10,%ymm8
+  DB  196,65,60,88,195                    ; vaddps        %ymm11,%ymm8,%ymm8
+  DB  196,65,52,89,192                    ; vmulps        %ymm8,%ymm9,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1y_avx
+_sk_bicubic_n1y_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,191                    ; mov           $0xbf000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,227,121,4,201,0                 ; vpermilps     $0x0,%xmm1,%xmm1
+  DB  196,227,117,24,201,1                ; vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  65,184,0,0,128,63                   ; mov           $0x3f800000,%r8d
+  DB  196,65,121,110,192                  ; vmovd         %r8d,%xmm8
+  DB  196,67,121,4,192,0                  ; vpermilps     $0x0,%xmm8,%xmm8
+  DB  196,67,61,24,192,1                  ; vinsertf128   $0x1,%xmm8,%ymm8,%ymm8
+  DB  197,60,92,64,96                     ; vsubps        0x60(%rax),%ymm8,%ymm8
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,200                  ; vmovd         %r8d,%xmm9
+  DB  196,67,121,4,201,0                  ; vpermilps     $0x0,%xmm9,%xmm9
+  DB  196,67,53,24,201,1                  ; vinsertf128   $0x1,%xmm9,%ymm9,%ymm9
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,52,89,200                    ; vmulps        %ymm8,%ymm9,%ymm9
+  DB  196,65,52,88,202                    ; vaddps        %ymm10,%ymm9,%ymm9
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,60,89,201                    ; vmulps        %ymm9,%ymm8,%ymm9
+  DB  196,65,44,88,201                    ; vaddps        %ymm9,%ymm10,%ymm9
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,60,89,193                    ; vmulps        %ymm9,%ymm8,%ymm8
+  DB  196,65,44,88,192                    ; vaddps        %ymm8,%ymm10,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1y_avx
+_sk_bicubic_p1y_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,0,63                     ; mov           $0x3f000000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,227,121,4,201,0                 ; vpermilps     $0x0,%xmm1,%xmm1
+  DB  196,99,117,24,193,1                 ; vinsertf128   $0x1,%xmm1,%ymm1,%ymm8
+  DB  197,188,88,72,32                    ; vaddps        0x20(%rax),%ymm8,%ymm1
+  DB  197,124,16,72,96                    ; vmovups       0x60(%rax),%ymm9
+  DB  65,184,85,85,149,191                ; mov           $0xbf955555,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,67,121,4,219,0                  ; vpermilps     $0x0,%xmm11,%xmm11
+  DB  196,67,37,24,219,1                  ; vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  DB  196,65,52,89,210                    ; vmulps        %ymm10,%ymm9,%ymm10
+  DB  196,65,44,88,211                    ; vaddps        %ymm11,%ymm10,%ymm10
+  DB  196,65,52,89,210                    ; vmulps        %ymm10,%ymm9,%ymm10
+  DB  196,65,60,88,194                    ; vaddps        %ymm10,%ymm8,%ymm8
+  DB  65,184,57,142,99,61                 ; mov           $0x3d638e39,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  196,65,52,89,192                    ; vmulps        %ymm8,%ymm9,%ymm8
+  DB  196,65,44,88,192                    ; vaddps        %ymm8,%ymm10,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3y_avx
+_sk_bicubic_p3y_avx LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  65,184,0,0,192,63                   ; mov           $0x3fc00000,%r8d
+  DB  196,193,121,110,200                 ; vmovd         %r8d,%xmm1
+  DB  196,227,121,4,201,0                 ; vpermilps     $0x0,%xmm1,%xmm1
+  DB  196,227,117,24,201,1                ; vinsertf128   $0x1,%xmm1,%ymm1,%ymm1
+  DB  197,244,88,72,32                    ; vaddps        0x20(%rax),%ymm1,%ymm1
+  DB  197,124,16,64,96                    ; vmovups       0x60(%rax),%ymm8
+  DB  196,65,60,89,200                    ; vmulps        %ymm8,%ymm8,%ymm9
+  DB  65,184,114,28,199,62                ; mov           $0x3ec71c72,%r8d
+  DB  196,65,121,110,208                  ; vmovd         %r8d,%xmm10
+  DB  196,67,121,4,210,0                  ; vpermilps     $0x0,%xmm10,%xmm10
+  DB  196,67,45,24,210,1                  ; vinsertf128   $0x1,%xmm10,%ymm10,%ymm10
+  DB  65,184,171,170,170,190              ; mov           $0xbeaaaaab,%r8d
+  DB  196,65,121,110,216                  ; vmovd         %r8d,%xmm11
+  DB  196,67,121,4,219,0                  ; vpermilps     $0x0,%xmm11,%xmm11
+  DB  196,67,37,24,219,1                  ; vinsertf128   $0x1,%xmm11,%ymm11,%ymm11
+  DB  196,65,60,89,194                    ; vmulps        %ymm10,%ymm8,%ymm8
+  DB  196,65,60,88,195                    ; vaddps        %ymm11,%ymm8,%ymm8
+  DB  196,65,52,89,192                    ; vmulps        %ymm8,%ymm9,%ymm8
+  DB  197,124,17,128,160,0,0,0            ; vmovups       %ymm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
 PUBLIC _sk_start_pipeline_sse41
 _sk_start_pipeline_sse41 LABEL PROC
   DB  65,87                               ; push          %r15
@@ -10512,6 +11140,332 @@ _sk_linear_gradient_2stops_sse41 LABEL PROC
   DB  72,173                              ; lods          %ds:(%rsi),%rax
   DB  255,224                             ; jmpq          *%rax
 
+PUBLIC _sk_save_xy_sse41
+_sk_save_xy_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,40,200                        ; movaps        %xmm8,%xmm9
+  DB  68,15,88,200                        ; addps         %xmm0,%xmm9
+  DB  102,69,15,58,8,209,1                ; roundps       $0x1,%xmm9,%xmm10
+  DB  69,15,92,202                        ; subps         %xmm10,%xmm9
+  DB  68,15,88,193                        ; addps         %xmm1,%xmm8
+  DB  102,69,15,58,8,208,1                ; roundps       $0x1,%xmm8,%xmm10
+  DB  69,15,92,194                        ; subps         %xmm10,%xmm8
+  DB  15,17,0                             ; movups        %xmm0,(%rax)
+  DB  15,17,72,32                         ; movups        %xmm1,0x20(%rax)
+  DB  68,15,17,72,64                      ; movups        %xmm9,0x40(%rax)
+  DB  68,15,17,64,96                      ; movups        %xmm8,0x60(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_accumulate_sse41
+_sk_accumulate_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  68,15,16,128,128,0,0,0              ; movups        0x80(%rax),%xmm8
+  DB  68,15,16,136,160,0,0,0              ; movups        0xa0(%rax),%xmm9
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,40,193                        ; movaps        %xmm9,%xmm8
+  DB  68,15,89,192                        ; mulps         %xmm0,%xmm8
+  DB  65,15,88,224                        ; addps         %xmm8,%xmm4
+  DB  69,15,40,193                        ; movaps        %xmm9,%xmm8
+  DB  68,15,89,193                        ; mulps         %xmm1,%xmm8
+  DB  65,15,88,232                        ; addps         %xmm8,%xmm5
+  DB  69,15,40,193                        ; movaps        %xmm9,%xmm8
+  DB  68,15,89,194                        ; mulps         %xmm2,%xmm8
+  DB  65,15,88,240                        ; addps         %xmm8,%xmm6
+  DB  68,15,89,203                        ; mulps         %xmm3,%xmm9
+  DB  65,15,88,249                        ; addps         %xmm9,%xmm7
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_nx_sse41
+_sk_bilinear_nx_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  68,15,17,128,128,0,0,0              ; movups        %xmm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_px_sse41
+_sk_bilinear_px_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  68,15,17,136,128,0,0,0              ; movups        %xmm9,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_ny_sse41
+_sk_bilinear_ny_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  68,15,17,128,160,0,0,0              ; movups        %xmm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_py_sse41
+_sk_bilinear_py_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  68,15,17,136,160,0,0,0              ; movups        %xmm9,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3x_sse41
+_sk_bicubic_n3x_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,191                     ; mov           $0xbfc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,89,192                        ; mulps         %xmm8,%xmm8
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  68,15,17,136,128,0,0,0              ; movups        %xmm9,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1x_sse41
+_sk_bicubic_n1x_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  68,15,17,136,128,0,0,0              ; movups        %xmm9,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1x_sse41
+_sk_bicubic_p1x_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,217                   ; movd          %ecx,%xmm11
+  DB  69,15,198,219,0                     ; shufps        $0x0,%xmm11,%xmm11
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,211                        ; addps         %xmm11,%xmm10
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  68,15,17,144,128,0,0,0              ; movups        %xmm10,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3x_sse41
+_sk_bicubic_p3x_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  69,15,89,201                        ; mulps         %xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,88,194                        ; addps         %xmm10,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  68,15,17,128,128,0,0,0              ; movups        %xmm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3y_sse41
+_sk_bicubic_n3y_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,191                     ; mov           $0xbfc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,89,192                        ; mulps         %xmm8,%xmm8
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  68,15,17,136,160,0,0,0              ; movups        %xmm9,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1y_sse41
+_sk_bicubic_n1y_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  68,15,17,136,160,0,0,0              ; movups        %xmm9,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1y_sse41
+_sk_bicubic_p1y_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,217                   ; movd          %ecx,%xmm11
+  DB  69,15,198,219,0                     ; shufps        $0x0,%xmm11,%xmm11
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,211                        ; addps         %xmm11,%xmm10
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  68,15,17,144,160,0,0,0              ; movups        %xmm10,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3y_sse41
+_sk_bicubic_p3y_sse41 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  69,15,89,201                        ; mulps         %xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,88,194                        ; addps         %xmm10,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  68,15,17,128,160,0,0,0              ; movups        %xmm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
 PUBLIC _sk_start_pipeline_sse2
 _sk_start_pipeline_sse2 LABEL PROC
   DB  65,87                               ; push          %r15
@@ -13933,5 +14887,344 @@ _sk_linear_gradient_2stops_sse2 LABEL PROC
   DB  65,15,88,217                        ; addps         %xmm9,%xmm3
   DB  72,173                              ; lods          %ds:(%rsi),%rax
   DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_save_xy_sse2
+_sk_save_xy_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,40,200                        ; movaps        %xmm8,%xmm9
+  DB  68,15,88,200                        ; addps         %xmm0,%xmm9
+  DB  243,69,15,91,209                    ; cvttps2dq     %xmm9,%xmm10
+  DB  69,15,91,210                        ; cvtdq2ps      %xmm10,%xmm10
+  DB  69,15,40,217                        ; movaps        %xmm9,%xmm11
+  DB  69,15,194,218,1                     ; cmpltps       %xmm10,%xmm11
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,225                   ; movd          %ecx,%xmm12
+  DB  69,15,198,228,0                     ; shufps        $0x0,%xmm12,%xmm12
+  DB  69,15,84,220                        ; andps         %xmm12,%xmm11
+  DB  69,15,92,211                        ; subps         %xmm11,%xmm10
+  DB  69,15,92,202                        ; subps         %xmm10,%xmm9
+  DB  68,15,88,193                        ; addps         %xmm1,%xmm8
+  DB  243,69,15,91,208                    ; cvttps2dq     %xmm8,%xmm10
+  DB  69,15,91,210                        ; cvtdq2ps      %xmm10,%xmm10
+  DB  69,15,40,216                        ; movaps        %xmm8,%xmm11
+  DB  69,15,194,218,1                     ; cmpltps       %xmm10,%xmm11
+  DB  69,15,84,220                        ; andps         %xmm12,%xmm11
+  DB  69,15,92,211                        ; subps         %xmm11,%xmm10
+  DB  69,15,92,194                        ; subps         %xmm10,%xmm8
+  DB  15,17,0                             ; movups        %xmm0,(%rax)
+  DB  15,17,72,32                         ; movups        %xmm1,0x20(%rax)
+  DB  68,15,17,72,64                      ; movups        %xmm9,0x40(%rax)
+  DB  68,15,17,64,96                      ; movups        %xmm8,0x60(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_accumulate_sse2
+_sk_accumulate_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  68,15,16,128,128,0,0,0              ; movups        0x80(%rax),%xmm8
+  DB  68,15,16,136,160,0,0,0              ; movups        0xa0(%rax),%xmm9
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,40,193                        ; movaps        %xmm9,%xmm8
+  DB  68,15,89,192                        ; mulps         %xmm0,%xmm8
+  DB  65,15,88,224                        ; addps         %xmm8,%xmm4
+  DB  69,15,40,193                        ; movaps        %xmm9,%xmm8
+  DB  68,15,89,193                        ; mulps         %xmm1,%xmm8
+  DB  65,15,88,232                        ; addps         %xmm8,%xmm5
+  DB  69,15,40,193                        ; movaps        %xmm9,%xmm8
+  DB  68,15,89,194                        ; mulps         %xmm2,%xmm8
+  DB  65,15,88,240                        ; addps         %xmm8,%xmm6
+  DB  68,15,89,203                        ; mulps         %xmm3,%xmm9
+  DB  65,15,88,249                        ; addps         %xmm9,%xmm7
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_nx_sse2
+_sk_bilinear_nx_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  68,15,17,128,128,0,0,0              ; movups        %xmm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_px_sse2
+_sk_bilinear_px_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  68,15,17,136,128,0,0,0              ; movups        %xmm9,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_ny_sse2
+_sk_bilinear_ny_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  68,15,17,128,160,0,0,0              ; movups        %xmm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bilinear_py_sse2
+_sk_bilinear_py_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  68,15,17,136,160,0,0,0              ; movups        %xmm9,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3x_sse2
+_sk_bicubic_n3x_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,191                     ; mov           $0xbfc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,89,192                        ; mulps         %xmm8,%xmm8
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  68,15,17,136,128,0,0,0              ; movups        %xmm9,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1x_sse2
+_sk_bicubic_n1x_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  68,15,17,136,128,0,0,0              ; movups        %xmm9,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1x_sse2
+_sk_bicubic_p1x_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,217                   ; movd          %ecx,%xmm11
+  DB  69,15,198,219,0                     ; shufps        $0x0,%xmm11,%xmm11
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,211                        ; addps         %xmm11,%xmm10
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  68,15,17,144,128,0,0,0              ; movups        %xmm10,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3x_sse2
+_sk_bicubic_p3x_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,0                             ; movups        (%rax),%xmm0
+  DB  68,15,16,72,64                      ; movups        0x40(%rax),%xmm9
+  DB  65,15,88,192                        ; addps         %xmm8,%xmm0
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  69,15,89,201                        ; mulps         %xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,88,194                        ; addps         %xmm10,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  68,15,17,128,128,0,0,0              ; movups        %xmm8,0x80(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n3y_sse2
+_sk_bicubic_n3y_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,191                     ; mov           $0xbfc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,89,192                        ; mulps         %xmm8,%xmm8
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  68,15,17,136,160,0,0,0              ; movups        %xmm9,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_n1y_sse2
+_sk_bicubic_n1y_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,191                       ; mov           $0xbf000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,0,0,128,63                      ; mov           $0x3f800000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,92,193                        ; subps         %xmm9,%xmm8
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,201                   ; movd          %ecx,%xmm9
+  DB  69,15,198,201,0                     ; shufps        $0x0,%xmm9,%xmm9
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,89,200                        ; mulps         %xmm8,%xmm9
+  DB  69,15,88,202                        ; addps         %xmm10,%xmm9
+  DB  68,15,17,136,160,0,0,0              ; movups        %xmm9,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p1y_sse2
+_sk_bicubic_p1y_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,0,63                        ; mov           $0x3f000000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,85,85,149,191                   ; mov           $0xbf955555,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,217                   ; movd          %ecx,%xmm11
+  DB  69,15,198,219,0                     ; shufps        $0x0,%xmm11,%xmm11
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,211                        ; addps         %xmm11,%xmm10
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  185,57,142,99,61                    ; mov           $0x3d638e39,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,209                        ; mulps         %xmm9,%xmm10
+  DB  69,15,88,208                        ; addps         %xmm8,%xmm10
+  DB  68,15,17,144,160,0,0,0              ; movups        %xmm10,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
+
+PUBLIC _sk_bicubic_p3y_sse2
+_sk_bicubic_p3y_sse2 LABEL PROC
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  185,0,0,192,63                      ; mov           $0x3fc00000,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  15,16,72,32                         ; movups        0x20(%rax),%xmm1
+  DB  68,15,16,72,96                      ; movups        0x60(%rax),%xmm9
+  DB  65,15,88,200                        ; addps         %xmm8,%xmm1
+  DB  185,114,28,199,62                   ; mov           $0x3ec71c72,%ecx
+  DB  102,68,15,110,193                   ; movd          %ecx,%xmm8
+  DB  69,15,198,192,0                     ; shufps        $0x0,%xmm8,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  69,15,89,201                        ; mulps         %xmm9,%xmm9
+  DB  185,171,170,170,190                 ; mov           $0xbeaaaaab,%ecx
+  DB  102,68,15,110,209                   ; movd          %ecx,%xmm10
+  DB  69,15,198,210,0                     ; shufps        $0x0,%xmm10,%xmm10
+  DB  69,15,88,194                        ; addps         %xmm10,%xmm8
+  DB  69,15,89,193                        ; mulps         %xmm9,%xmm8
+  DB  68,15,17,128,160,0,0,0              ; movups        %xmm8,0xa0(%rax)
+  DB  72,173                              ; lods          %ds:(%rsi),%rax
+  DB  255,224                             ; jmpq          *%rax
 ENDIF
 END
diff --git a/src/jumper/SkJumper_stages.cpp b/src/jumper/SkJumper_stages.cpp
index 82191c91b8..2e6746c338 100644
--- a/src/jumper/SkJumper_stages.cpp
+++ b/src/jumper/SkJumper_stages.cpp
@@ -910,3 +910,115 @@ STAGE(linear_gradient_2stops) {
     b = mad(t, c->f[2], c->b[2]);
     a = mad(t, c->f[3], c->b[3]);
 }
+
+STAGE(save_xy) {
+    auto c = (SkJumper_SamplerCtx*)ctx;
+
+    // Whether bilinear or bicubic, all sample points are at the same fractional offset (fx,fy).
+    // They're either the 4 corners of a logical 1x1 pixel or the 16 corners of a 3x3 grid
+    // surrounding (x,y) at (0.5,0.5) off-center.
+    auto fract = [](F v) { return v - floor_(v); };
+    F fx = fract(r + 0.5_f),
+      fy = fract(g + 0.5_f);
+
+    // Samplers will need to load x and fx, or y and fy.
+    memcpy(c->x,  &r,  sizeof(F));
+    memcpy(c->y,  &g,  sizeof(F));
+    memcpy(c->fx, &fx, sizeof(F));
+    memcpy(c->fy, &fy, sizeof(F));
+}
+
+STAGE(accumulate) {
+    auto c = (const SkJumper_SamplerCtx*)ctx;
+
+    // Bilinear and bicubic filters are both separable, so we produce independent contributions
+    // from x and y, multiplying them together here to get each pixel's total scale factor.
+    auto scale = unaligned_load<F>(c->scalex)
+               * unaligned_load<F>(c->scaley);
+    dr = mad(scale, r, dr);
+    dg = mad(scale, g, dg);
+    db = mad(scale, b, db);
+    da = mad(scale, a, da);
+}
+
+// In bilinear interpolation, the 4 pixels at +/- 0.5 offsets from the sample pixel center
+// are combined in direct proportion to their area overlapping that logical query pixel.
+// At positive offsets, the x-axis contribution to that rectangle is fx, or (1-fx) at negative x.
+// The y-axis is symmetric.
+
+template <int kScale>
+SI void bilinear_x(SkJumper_SamplerCtx* ctx, F* x) {
+    *x = unaligned_load<F>(ctx->x) + C(kScale * 0.5f);
+    F fx = unaligned_load<F>(ctx->fx);
+
+    F scalex;
+    if (kScale == -1) { scalex = 1.0_f - fx; }
+    if (kScale == +1) { scalex =         fx; }
+    memcpy(ctx->scalex, &scalex, sizeof(F));
+}
+template <int kScale>
+SI void bilinear_y(SkJumper_SamplerCtx* ctx, F* y) {
+    *y = unaligned_load<F>(ctx->y) + C(kScale * 0.5f);
+    F fy = unaligned_load<F>(ctx->fy);
+
+    F scaley;
+    if (kScale == -1) { scaley = 1.0_f - fy; }
+    if (kScale == +1) { scaley =         fy; }
+    memcpy(ctx->scaley, &scaley, sizeof(F));
+}
+
+STAGE(bilinear_nx) { bilinear_x<-1>(ctx, &r); }
+STAGE(bilinear_px) { bilinear_x<+1>(ctx, &r); }
+STAGE(bilinear_ny) { bilinear_y<-1>(ctx, &g); }
+STAGE(bilinear_py) { bilinear_y<+1>(ctx, &g); }
+
+
+// In bicubic interpolation, the 16 pixels and +/- 0.5 and +/- 1.5 offsets from the sample
+// pixel center are combined with a non-uniform cubic filter, with higher values near the center.
+//
+// We break this function into two parts, one for near 0.5 offsets and one for far 1.5 offsets.
+// See GrCubicEffect for details of this particular filter.
+
+SI F bicubic_near(F t) {
+    // 1/18 + 9/18t + 27/18t^2 - 21/18t^3 == t ( t ( -21/18t + 27/18) + 9/18) + 1/18
+    return mad(t, mad(t, mad(C(-21/18.0f), t, C(27/18.0f)), C(9/18.0f)), C(1/18.0f));
+}
+SI F bicubic_far(F t) {
+    // 0/18 + 0/18*t - 6/18t^2 + 7/18t^3 == t^2 (7/18t - 6/18)
+    return (t*t)*mad(C(7/18.0f), t, C(-6/18.0f));
+}
+
+template <int kScale>
+SI void bicubic_x(SkJumper_SamplerCtx* ctx, F* x) {
+    *x = unaligned_load<F>(ctx->x) + C(kScale * 0.5f);
+    F fx = unaligned_load<F>(ctx->fx);
+
+    F scalex;
+    if (kScale == -3) { scalex = bicubic_far (1.0_f - fx); }
+    if (kScale == -1) { scalex = bicubic_near(1.0_f - fx); }
+    if (kScale == +1) { scalex = bicubic_near(        fx); }
+    if (kScale == +3) { scalex = bicubic_far (        fx); }
+    memcpy(ctx->scalex, &scalex, sizeof(F));
+}
+template <int kScale>
+SI void bicubic_y(SkJumper_SamplerCtx* ctx, F* y) {
+    *y = unaligned_load<F>(ctx->y) + C(kScale * 0.5f);
+    F fy = unaligned_load<F>(ctx->fy);
+
+    F scaley;
+    if (kScale == -3) { scaley = bicubic_far (1.0_f - fy); }
+    if (kScale == -1) { scaley = bicubic_near(1.0_f - fy); }
+    if (kScale == +1) { scaley = bicubic_near(        fy); }
+    if (kScale == +3) { scaley = bicubic_far (        fy); }
+    memcpy(ctx->scaley, &scaley, sizeof(F));
+}
+
+STAGE(bicubic_n3x) { bicubic_x<-3>(ctx, &r); }
+STAGE(bicubic_n1x) { bicubic_x<-1>(ctx, &r); }
+STAGE(bicubic_p1x) { bicubic_x<+1>(ctx, &r); }
+STAGE(bicubic_p3x) { bicubic_x<+3>(ctx, &r); }
+
+STAGE(bicubic_n3y) { bicubic_y<-3>(ctx, &g); }
+STAGE(bicubic_n1y) { bicubic_y<-1>(ctx, &g); }
+STAGE(bicubic_p1y) { bicubic_y<+1>(ctx, &g); }
+STAGE(bicubic_p3y) { bicubic_y<+3>(ctx, &g); }
diff --git a/src/opts/SkRasterPipeline_opts.h b/src/opts/SkRasterPipeline_opts.h
index fd7a9e5b1a..1146b3d7fb 100644
--- a/src/opts/SkRasterPipeline_opts.h
+++ b/src/opts/SkRasterPipeline_opts.h
@@ -13,7 +13,6 @@
 #include "SkColorSpaceXform_A2B.h"
 #include "SkColorSpaceXformPriv.h"
 #include "SkHalf.h"
-#include "SkImageShaderContext.h"
 #include "SkMSAN.h"
 #include "SkPM4f.h"
 #include "SkPM4fPriv.h"
@@ -883,7 +882,7 @@ STAGE_CTX( clamp_y, const float*) { g = clamp (g, *ctx); }
 STAGE_CTX(repeat_y, const float*) { g = repeat(g, *ctx); }
 STAGE_CTX(mirror_y, const float*) { g = mirror(g, *ctx); }
 
-STAGE_CTX(save_xy, SkImageShaderContext*) {
+STAGE_CTX(save_xy, SkJumper_SamplerCtx*) {
     r.store(ctx->x);
     g.store(ctx->y);
 
@@ -895,7 +894,7 @@ STAGE_CTX(save_xy, SkImageShaderContext*) {
     fract(g + 0.5f).store(ctx->fy);
 }
 
-STAGE_CTX(accumulate, const SkImageShaderContext*) {
+STAGE_CTX(accumulate, const SkJumper_SamplerCtx*) {
     // Bilinear and bicubic filtering are both separable, so we'll end up with independent
     // scale contributions in x and y that we multiply together to get each pixel's scale factor.
     auto scale = SkNf::Load(ctx->scalex) * SkNf::Load(ctx->scaley);
@@ -910,21 +909,21 @@ STAGE_CTX(accumulate, const SkImageShaderContext*) {
 // At positive offsets, the x-axis contribution to that rectangular area is fx; (1-fx)
 // at negative x offsets.  The y-axis is treated symmetrically.
 template <int Scale>
-SI void bilinear_x(SkImageShaderContext* ctx, SkNf* x) {
+SI void bilinear_x(SkJumper_SamplerCtx* ctx, SkNf* x) {
     *x = SkNf::Load(ctx->x) + Scale*0.5f;
     auto fx = SkNf::Load(ctx->fx);
     (Scale > 0 ? fx : (1.0f - fx)).store(ctx->scalex);
 }
 template <int Scale>
-SI void bilinear_y(SkImageShaderContext* ctx, SkNf* y) {
+SI void bilinear_y(SkJumper_SamplerCtx* ctx, SkNf* y) {
     *y = SkNf::Load(ctx->y) + Scale*0.5f;
     auto fy = SkNf::Load(ctx->fy);
     (Scale > 0 ? fy : (1.0f - fy)).store(ctx->scaley);
 }
-STAGE_CTX(bilinear_nx, SkImageShaderContext*) { bilinear_x<-1>(ctx, &r); }
-STAGE_CTX(bilinear_px, SkImageShaderContext*) { bilinear_x<+1>(ctx, &r); }
-STAGE_CTX(bilinear_ny, SkImageShaderContext*) { bilinear_y<-1>(ctx, &g); }
-STAGE_CTX(bilinear_py, SkImageShaderContext*) { bilinear_y<+1>(ctx, &g); }
+STAGE_CTX(bilinear_nx, SkJumper_SamplerCtx*) { bilinear_x<-1>(ctx, &r); }
+STAGE_CTX(bilinear_px, SkJumper_SamplerCtx*) { bilinear_x<+1>(ctx, &r); }
+STAGE_CTX(bilinear_ny, SkJumper_SamplerCtx*) { bilinear_y<-1>(ctx, &g); }
+STAGE_CTX(bilinear_py, SkJumper_SamplerCtx*) { bilinear_y<+1>(ctx, &g); }
 
 
 // In bilinear interpolation, the 16 pixels at +/- 0.5 and +/- 1.5 offsets from the sample
@@ -945,7 +944,7 @@ SI SkNf bicubic_far(const SkNf& t) {
 }
 
 template <int Scale>
-SI void bicubic_x(SkImageShaderContext* ctx, SkNf* x) {
+SI void bicubic_x(SkJumper_SamplerCtx* ctx, SkNf* x) {
     *x = SkNf::Load(ctx->x) + Scale*0.5f;
     auto fx = SkNf::Load(ctx->fx);
     if (Scale == -3) { return bicubic_far (1.0f - fx).store(ctx->scalex); }
@@ -955,7 +954,7 @@ SI void bicubic_x(SkImageShaderContext* ctx, SkNf* x) {
     SkDEBUGFAIL("unreachable");
 }
 template <int Scale>
-SI void bicubic_y(SkImageShaderContext* ctx, SkNf* y) {
+SI void bicubic_y(SkJumper_SamplerCtx* ctx, SkNf* y) {
     *y = SkNf::Load(ctx->y) + Scale*0.5f;
     auto fy = SkNf::Load(ctx->fy);
     if (Scale == -3) { return bicubic_far (1.0f - fy).store(ctx->scaley); }
@@ -964,15 +963,15 @@ SI void bicubic_y(SkImageShaderContext* ctx, SkNf* y) {
     if (Scale == +3) { return bicubic_far (       fy).store(ctx->scaley); }
     SkDEBUGFAIL("unreachable");
 }
-STAGE_CTX(bicubic_n3x, SkImageShaderContext*) { bicubic_x<-3>(ctx, &r); }
-STAGE_CTX(bicubic_n1x, SkImageShaderContext*) { bicubic_x<-1>(ctx, &r); }
-STAGE_CTX(bicubic_p1x, SkImageShaderContext*) { bicubic_x<+1>(ctx, &r); }
-STAGE_CTX(bicubic_p3x, SkImageShaderContext*) { bicubic_x<+3>(ctx, &r); }
+STAGE_CTX(bicubic_n3x, SkJumper_SamplerCtx*) { bicubic_x<-3>(ctx, &r); }
+STAGE_CTX(bicubic_n1x, SkJumper_SamplerCtx*) { bicubic_x<-1>(ctx, &r); }
+STAGE_CTX(bicubic_p1x, SkJumper_SamplerCtx*) { bicubic_x<+1>(ctx, &r); }
+STAGE_CTX(bicubic_p3x, SkJumper_SamplerCtx*) { bicubic_x<+3>(ctx, &r); }
 
-STAGE_CTX(bicubic_n3y, SkImageShaderContext*) { bicubic_y<-3>(ctx, &g); }
-STAGE_CTX(bicubic_n1y, SkImageShaderContext*) { bicubic_y<-1>(ctx, &g); }
-STAGE_CTX(bicubic_p1y, SkImageShaderContext*) { bicubic_y<+1>(ctx, &g); }
-STAGE_CTX(bicubic_p3y, SkImageShaderContext*) { bicubic_y<+3>(ctx, &g); }
+STAGE_CTX(bicubic_n3y, SkJumper_SamplerCtx*) { bicubic_y<-3>(ctx, &g); }
+STAGE_CTX(bicubic_n1y, SkJumper_SamplerCtx*) { bicubic_y<-1>(ctx, &g); }
+STAGE_CTX(bicubic_p1y, SkJumper_SamplerCtx*) { bicubic_y<+1>(ctx, &g); }
+STAGE_CTX(bicubic_p3y, SkJumper_SamplerCtx*) { bicubic_y<+3>(ctx, &g); }
 
 
 template <typename T>