diff --git a/src/gpu/GrDrawTarget.cpp b/src/gpu/GrDrawTarget.cpp
index 549ab37..7565cd1 100644
--- a/src/gpu/GrDrawTarget.cpp
+++ b/src/gpu/GrDrawTarget.cpp
@@ -100,13 +100,10 @@
         return false;
     }
     SkIPoint dstPoint = {0, 0};
-    if (this->copySurface(copy, rt, copyRect, dstPoint)) {
-        dstCopy->setTexture(copy);
-        dstCopy->setOffset(copyRect.fLeft, copyRect.fTop);
-        return true;
-    } else {
-        return false;
-    }
+    this->copySurface(copy, rt, copyRect, dstPoint);
+    dstCopy->setTexture(copy);
+    dstCopy->setOffset(copyRect.fLeft, copyRect.fTop);
+    return true;
 }
 
 void GrDrawTarget::flush() {
@@ -421,7 +418,7 @@
 }
 }
 
-bool GrDrawTarget::copySurface(GrSurface* dst,
+void GrDrawTarget::copySurface(GrSurface* dst,
                                GrSurface* src,
                                const SkIRect& srcRect,
                                const SkIPoint& dstPoint) {
@@ -437,56 +434,10 @@
                                    dstPoint,
                                    &clippedSrcRect,
                                    &clippedDstPoint)) {
-        return true;
+        return;
     }
 
-    if (this->getGpu()->canCopySurface(dst, src, clippedSrcRect, clippedDstPoint)) {
-        this->onCopySurface(dst, src, clippedSrcRect, clippedDstPoint);
-        return true;
-    }
-
-    GrRenderTarget* rt = dst->asRenderTarget();
-    GrTexture* tex = src->asTexture();
-
-    if ((dst == src) || !rt || !tex) {
-        return false;
-    }
-
-    GrPipelineBuilder pipelineBuilder;
-    pipelineBuilder.setRenderTarget(rt);
-    SkMatrix matrix;
-    matrix.setTranslate(SkIntToScalar(clippedSrcRect.fLeft - clippedDstPoint.fX),
-                        SkIntToScalar(clippedSrcRect.fTop - clippedDstPoint.fY));
-    matrix.postIDiv(tex->width(), tex->height());
-    pipelineBuilder.addColorTextureProcessor(tex, matrix);
-    SkIRect dstRect = SkIRect::MakeXYWH(clippedDstPoint.fX,
-                                        clippedDstPoint.fY,
-                                        clippedSrcRect.width(),
-                                        clippedSrcRect.height());
-    this->drawSimpleRect(&pipelineBuilder, GrColor_WHITE, SkMatrix::I(), dstRect);
-    return true;
-}
-
-bool GrDrawTarget::canCopySurface(const GrSurface* dst,
-                                  const GrSurface* src,
-                                  const SkIRect& srcRect,
-                                  const SkIPoint& dstPoint) {
-    SkASSERT(dst);
-    SkASSERT(src);
-
-    SkIRect clippedSrcRect;
-    SkIPoint clippedDstPoint;
-    // If the rect is outside the src or dst then we're guaranteed success
-    if (!clip_srcrect_and_dstpoint(dst,
-                                   src,
-                                   srcRect,
-                                   dstPoint,
-                                   &clippedSrcRect,
-                                   &clippedDstPoint)) {
-        return true;
-    }
-    return ((dst != src) && dst->asRenderTarget() && src->asTexture()) ||
-           this->getGpu()->canCopySurface(dst, src, clippedSrcRect, clippedDstPoint);
+    this->onCopySurface(dst, src, clippedSrcRect, clippedDstPoint);
 }
 
 void GrDrawTarget::setupPipeline(const PipelineInfo& pipelineInfo,
diff --git a/src/gpu/GrDrawTarget.h b/src/gpu/GrDrawTarget.h
index 84f7eb2..0157354 100644
--- a/src/gpu/GrDrawTarget.h
+++ b/src/gpu/GrDrawTarget.h
@@ -178,25 +178,15 @@
      * copied are specified by srcRect. They are copied to a rect of the same
      * size in dst with top left at dstPoint. If the src rect is clipped by the
      * src bounds then  pixel values in the dst rect corresponding to area clipped
-     * by the src rect are not overwritten. This method can fail and return false
+     * by the src rect are not overwritten. This method is not guaranteed to succeed
      * depending on the type of surface, configs, etc, and the backend-specific
-     * limitations. If rect is clipped out entirely by the src or dst bounds then
-     * true is returned since there is no actual copy necessary to succeed.
+     * limitations.
      */
-    bool copySurface(GrSurface* dst,
+    void copySurface(GrSurface* dst,
                      GrSurface* src,
                      const SkIRect& srcRect,
                      const SkIPoint& dstPoint);
     /**
-     * Function that determines whether a copySurface call would succeed without actually
-     * performing the copy.
-     */
-    bool canCopySurface(const GrSurface* dst,
-                        const GrSurface* src,
-                        const SkIRect& srcRect,
-                        const SkIPoint& dstPoint);
-
-    /**
      * Release any resources that are cached but not currently in use. This
      * is intended to give an application some recourse when resources are low.
      */
diff --git a/src/gpu/GrGpu.h b/src/gpu/GrGpu.h
index f2ad10f..e8d4939 100644
--- a/src/gpu/GrGpu.h
+++ b/src/gpu/GrGpu.h
@@ -257,16 +257,6 @@
                                   const GrPipeline&,
                                   const GrBatchTracker&) const = 0;
 
-    // Called to determine whether a copySurface call would succeed or not. Derived
-    // classes must keep this consistent with their implementation of onCopySurface(). Fallbacks
-    // to issuing a draw from the src to dst take place at the GrDrawTarget level and this function
-    // should only return true if a faster copy path exists. The rect and point are pre-clipped. The
-    // src rect and implied dst rect are guaranteed to be within the src/dst bounds and non-empty.
-    virtual bool canCopySurface(const GrSurface* dst,
-                                const GrSurface* src,
-                                const SkIRect& srcRect,
-                                const SkIPoint& dstPoint) = 0;
-
     // Called to perform a surface to surface copy. Fallbacks to issuing a draw from the src to dst
     // take place at the GrDrawTarget level and this function implement faster copy paths. The rect
     // and point are pre-clipped. The src rect and implied dst rect are guaranteed to be within the
diff --git a/src/gpu/GrInOrderDrawBuffer.cpp b/src/gpu/GrInOrderDrawBuffer.cpp
index 6b6490c..8cc3658 100644
--- a/src/gpu/GrInOrderDrawBuffer.cpp
+++ b/src/gpu/GrInOrderDrawBuffer.cpp
@@ -121,7 +121,6 @@
                                         GrSurface* src,
                                         const SkIRect& srcRect,
                                         const SkIPoint& dstPoint) {
-    SkASSERT(this->getGpu()->canCopySurface(dst, src, srcRect, dstPoint));
     GrTargetCommands::Cmd* cmd = fCommands->recordCopySurface(dst, src, srcRect, dstPoint);
     this->recordTraceMarkersIfNecessary(cmd);
 }
diff --git a/src/gpu/GrTest.cpp b/src/gpu/GrTest.cpp
index 0b3d030..6504acf 100644
--- a/src/gpu/GrTest.cpp
+++ b/src/gpu/GrTest.cpp
@@ -155,11 +155,6 @@
 
     void discard(GrRenderTarget*) override {}
 
-    bool canCopySurface(const GrSurface* dst,
-                        const GrSurface* src,
-                        const SkIRect& srcRect,
-                        const SkIPoint& dstPoint) override { return false; };
-
     bool copySurface(GrSurface* dst,
                      GrSurface* src,
                      const SkIRect& srcRect,
diff --git a/src/gpu/gl/GrGLCaps.cpp b/src/gpu/gl/GrGLCaps.cpp
index 06ebe3d..5ee7546 100644
--- a/src/gpu/gl/GrGLCaps.cpp
+++ b/src/gpu/gl/GrGLCaps.cpp
@@ -942,6 +942,7 @@
 
 GrGLSLCaps& GrGLSLCaps::operator= (const GrGLSLCaps& caps) {
     INHERITED::operator=(caps);
+    fGLSLGeneration = caps.fGLSLGeneration;
     fDropsTileOnZeroDivide = caps.fDropsTileOnZeroDivide;
     fFBFetchSupport = caps.fFBFetchSupport;
     fFBFetchNeedsCustomOutput = caps.fFBFetchNeedsCustomOutput;
@@ -960,6 +961,8 @@
         return false;
     }
 
+    fGLSLGeneration = ctxInfo.glslGeneration();
+    
     GrGLStandard standard = ctxInfo.standard();
     GrGLVersion version = ctxInfo.version();
 
diff --git a/src/gpu/gl/GrGLCaps.h b/src/gpu/gl/GrGLCaps.h
index 25c7889..dfc6233 100644
--- a/src/gpu/gl/GrGLCaps.h
+++ b/src/gpu/gl/GrGLCaps.h
@@ -374,6 +374,7 @@
     typedef GrDrawTargetCaps INHERITED;
 };
 
+#include "GrGLSL.h"
 
 class GrGLSLCaps : public GrShaderCaps {
 public:
@@ -438,6 +439,12 @@
     bool mustEnableSpecificAdvBlendEqs() const {
         return fAdvBlendEqInteraction == kSpecificEnables_AdvBlendEqInteraction;
     }
+    
+    bool mustDeclareFragmentShaderOutput() const {
+        return fGLSLGeneration > k110_GrGLSLGeneration;
+    }
+
+    GrGLSLGeneration generation() const { return fGLSLGeneration; }
 
     /**
     * Returns a string containing the caps info.
@@ -448,6 +455,8 @@
     // Must be called after fGeometryShaderSupport is initialized.
     void initShaderPrecisionTable(const GrGLContextInfo&, const GrGLInterface*);
 
+    GrGLSLGeneration fGLSLGeneration;
+    
     bool fDropsTileOnZeroDivide : 1;
     bool fFBFetchSupport : 1;
     bool fFBFetchNeedsCustomOutput : 1;
diff --git a/src/gpu/gl/GrGLGpu.cpp b/src/gpu/gl/GrGLGpu.cpp
index 6d97804..56b3d42 100644
--- a/src/gpu/gl/GrGLGpu.cpp
+++ b/src/gpu/gl/GrGLGpu.cpp
@@ -17,6 +17,7 @@
 #include "GrTexturePriv.h"
 #include "GrTypes.h"
 #include "GrVertices.h"
+#include "builders/GrGLShaderStringBuilder.h"
 #include "SkStrokeRec.h"
 #include "SkTemplates.h"
 
@@ -200,13 +201,14 @@
     if (this->glCaps().shaderCaps()->pathRenderingSupport()) {
         fPathRendering.reset(new GrGLPathRendering(this));
     }
+
+    this->createCopyProgram();
 }
 
 GrGLGpu::~GrGLGpu() {
     if (0 != fHWProgramID) {
         // detach the current program so there is no confusion on OpenGL's part
         // that we want it to be deleted
-        SkASSERT(fHWProgramID == fCurrentProgram->programID());
         GL_CALL(UseProgram(0));
     }
 
@@ -220,6 +222,14 @@
         GL_CALL(DeleteFramebuffers(1, &fStencilClearFBOID));
     }
 
+    if (0 != fCopyProgram.fArrayBuffer) {
+        GL_CALL(DeleteBuffers(1, &fCopyProgram.fArrayBuffer));
+    }
+
+    if (0 != fCopyProgram.fProgram) {
+        GL_CALL(DeleteProgram(fCopyProgram.fProgram));
+    }
+
     delete fProgramCache;
 }
 
@@ -230,6 +240,8 @@
     fTempSrcFBOID = 0;
     fTempDstFBOID = 0;
     fStencilClearFBOID = 0;
+    fCopyProgram.fArrayBuffer = 0;
+    fCopyProgram.fProgram = 0;
     if (this->glCaps().shaderCaps()->pathRenderingSupport()) {
         this->glPathRendering()->abandonGpuResources();
     }
@@ -1417,15 +1429,13 @@
     this->flushColorWrite(blendInfo.fWriteColor);
     this->flushDrawFace(pipeline.getDrawFace());
 
-    fCurrentProgram.reset(fProgramCache->getProgram(args));
-    if (NULL == fCurrentProgram.get()) {
+    SkAutoTUnref<GrGLProgram> program(fProgramCache->refProgram(args));
+    if (!program) {
         GrContextDebugf(this->getContext(), "Failed to create program!\n");
         return false;
     }
 
-    fCurrentProgram.get()->ref();
-
-    GrGLuint programID = fCurrentProgram->programID();
+    GrGLuint programID = program->programID();
     if (fHWProgramID != programID) {
         GL_CALL(UseProgram(programID));
         fHWProgramID = programID;
@@ -1435,7 +1445,7 @@
         this->flushBlend(blendInfo);
     }
 
-    fCurrentProgram->setData(*args.fPrimitiveProcessor, pipeline, *args.fBatchTracker);
+    program->setData(*args.fPrimitiveProcessor, pipeline, *args.fBatchTracker);
 
     GrGLRenderTarget* glRT = static_cast<GrGLRenderTarget*>(pipeline.getRenderTarget());
     this->flushStencil(pipeline.getStencil());
@@ -1490,7 +1500,7 @@
             GrVertexAttribType attribType = attrib.fType;
             attribState->set(this,
                              attribIndex,
-                             vbuf,
+                             vbuf->bufferID(),
                              GrGLAttribTypeToLayout(attribType).fCount,
                              GrGLAttribTypeToLayout(attribType).fType,
                              GrGLAttribTypeToLayout(attribType).fNormalized,
@@ -2618,7 +2628,16 @@
 }
 
 bool GrGLGpu::initCopySurfaceDstDesc(const GrSurface* src, GrSurfaceDesc* desc) {
-    // In here we look for opportunities to use CopyTexSubImage, or fbo blit. If neither are
+    // If the src is a texture, we can implement the blit as a draw assuming the config is
+    // renderable.
+    if (src->asTexture() && this->caps()->isConfigRenderable(src->config(), false)) {
+        desc->fOrigin = kDefault_GrSurfaceOrigin;
+        desc->fFlags = kRenderTarget_GrSurfaceFlag;
+        desc->fConfig = src->config();
+        return true;
+    }
+
+    // We look for opportunities to use CopyTexSubImage, or fbo blit. If neither are
     // possible and we return false to fallback to creating a render target dst for render-to-
     // texture. This code prefers CopyTexSubImage to fbo blit and avoids triggering temporary fbo
     // creation. It isn't clear that avoiding temporary fbo creation is actually optimal.
@@ -2664,127 +2683,295 @@
                           GrSurface* src,
                           const SkIRect& srcRect,
                           const SkIPoint& dstPoint) {
-    bool copied = false;
-    if (can_copy_texsubimage(dst, src, this)) {
-        GrGLuint srcFBO;
-        GrGLIRect srcVP;
-        srcFBO = this->bindSurfaceAsFBO(src, GR_GL_FRAMEBUFFER, &srcVP, kSrc_TempFBOTarget);
-        GrGLTexture* dstTex = static_cast<GrGLTexture*>(dst->asTexture());
-        SkASSERT(dstTex);
-        // We modified the bound FBO
-        fHWBoundRenderTargetUniqueID = SK_InvalidUniqueID;
-        GrGLIRect srcGLRect;
-        srcGLRect.setRelativeTo(srcVP,
-                                srcRect.fLeft,
-                                srcRect.fTop,
-                                srcRect.width(),
-                                srcRect.height(),
-                                src->origin());
-
-        this->setScratchTextureUnit();
-        GL_CALL(BindTexture(GR_GL_TEXTURE_2D, dstTex->textureID()));
-        GrGLint dstY;
-        if (kBottomLeft_GrSurfaceOrigin == dst->origin()) {
-            dstY = dst->height() - (dstPoint.fY + srcGLRect.fHeight);
-        } else {
-            dstY = dstPoint.fY;
-        }
-        GL_CALL(CopyTexSubImage2D(GR_GL_TEXTURE_2D, 0,
-                                  dstPoint.fX, dstY,
-                                  srcGLRect.fLeft, srcGLRect.fBottom,
-                                  srcGLRect.fWidth, srcGLRect.fHeight));
-        copied = true;
-        if (srcFBO) {
-            this->unbindTextureFromFBO(GR_GL_FRAMEBUFFER);
-        }
-    } else if (can_blit_framebuffer(dst, src, this)) {
-        SkIRect dstRect = SkIRect::MakeXYWH(dstPoint.fX, dstPoint.fY,
-                                            srcRect.width(), srcRect.height());
-        bool selfOverlap = false;
-        if (dst == src) {
-            selfOverlap = SkIRect::IntersectsNoEmptyCheck(dstRect, srcRect);
-        }
-
-        if (!selfOverlap) {
-            GrGLuint dstFBO;
-            GrGLuint srcFBO;
-            GrGLIRect dstVP;
-            GrGLIRect srcVP;
-            dstFBO = this->bindSurfaceAsFBO(dst, GR_GL_DRAW_FRAMEBUFFER, &dstVP,
-                                            kDst_TempFBOTarget);
-            srcFBO = this->bindSurfaceAsFBO(src, GR_GL_READ_FRAMEBUFFER, &srcVP,
-                                            kSrc_TempFBOTarget);
-            // We modified the bound FBO
-            fHWBoundRenderTargetUniqueID = SK_InvalidUniqueID;
-            GrGLIRect srcGLRect;
-            GrGLIRect dstGLRect;
-            srcGLRect.setRelativeTo(srcVP,
-                                    srcRect.fLeft,
-                                    srcRect.fTop,
-                                    srcRect.width(),
-                                    srcRect.height(),
-                                    src->origin());
-            dstGLRect.setRelativeTo(dstVP,
-                                    dstRect.fLeft,
-                                    dstRect.fTop,
-                                    dstRect.width(),
-                                    dstRect.height(),
-                                    dst->origin());
-
-            // BlitFrameBuffer respects the scissor, so disable it.
-            this->disableScissor();
-
-            GrGLint srcY0;
-            GrGLint srcY1;
-            // Does the blit need to y-mirror or not?
-            if (src->origin() == dst->origin()) {
-                srcY0 = srcGLRect.fBottom;
-                srcY1 = srcGLRect.fBottom + srcGLRect.fHeight;
-            } else {
-                srcY0 = srcGLRect.fBottom + srcGLRect.fHeight;
-                srcY1 = srcGLRect.fBottom;
-            }
-            GL_CALL(BlitFramebuffer(srcGLRect.fLeft,
-                                    srcY0,
-                                    srcGLRect.fLeft + srcGLRect.fWidth,
-                                    srcY1,
-                                    dstGLRect.fLeft,
-                                    dstGLRect.fBottom,
-                                    dstGLRect.fLeft + dstGLRect.fWidth,
-                                    dstGLRect.fBottom + dstGLRect.fHeight,
-                                    GR_GL_COLOR_BUFFER_BIT, GR_GL_NEAREST));
-            if (dstFBO) {
-                this->unbindTextureFromFBO(GR_GL_DRAW_FRAMEBUFFER);
-            }
-            if (srcFBO) {
-                this->unbindTextureFromFBO(GR_GL_READ_FRAMEBUFFER);
-            }
-            copied = true;
-        }
-    }
-    return copied;
-}
-
-bool GrGLGpu::canCopySurface(const GrSurface* dst,
-                             const GrSurface* src,
-                             const SkIRect& srcRect,
-                             const SkIPoint& dstPoint) {
-    // This mirrors the logic in onCopySurface.
-    if (can_copy_texsubimage(dst, src, this)) {
+    if (src->asTexture() && dst->asRenderTarget()) {
+        this->copySurfaceAsDraw(dst, src, srcRect, dstPoint);
         return true;
     }
+    
+    if (can_copy_texsubimage(dst, src, this)) {
+        this->copySurfaceAsCopyTexSubImage(dst, src, srcRect, dstPoint);
+        return true;
+    }
+
     if (can_blit_framebuffer(dst, src, this)) {
-        if (dst == src) {
-            SkIRect dstRect = SkIRect::MakeXYWH(dstPoint.fX, dstPoint.fY,
-                                                srcRect.width(), srcRect.height());
-            if(!SkIRect::IntersectsNoEmptyCheck(dstRect, srcRect)) {
-                return true;
-            }
-        } else {
-            return true;
+        return this->copySurfaceAsBlitFramebuffer(dst, src, srcRect, dstPoint);
+    }
+
+    return false;
+}
+
+
+void GrGLGpu::createCopyProgram() {
+    const char* version = GrGetGLSLVersionDecl(this->ctxInfo());
+
+    GrGLShaderVar aVertex("a_vertex", kVec2f_GrSLType, GrShaderVar::kAttribute_TypeModifier);
+    GrGLShaderVar uTexCoordXform("u_texCoordXform", kVec4f_GrSLType,
+                                 GrShaderVar::kUniform_TypeModifier);
+    GrGLShaderVar uPosXform("u_posXform", kVec4f_GrSLType, GrShaderVar::kUniform_TypeModifier);
+    GrGLShaderVar uTexture("u_texture", kSampler2D_GrSLType, GrShaderVar::kUniform_TypeModifier);
+    GrGLShaderVar vTexCoord("v_texCoord", kVec2f_GrSLType, GrShaderVar::kVaryingOut_TypeModifier);
+    GrGLShaderVar oFragColor("o_FragColor", kVec4f_GrSLType, GrShaderVar::kOut_TypeModifier);
+    
+    SkString vshaderTxt(version);
+    aVertex.appendDecl(this->ctxInfo(), &vshaderTxt);
+    vshaderTxt.append(";");
+    uTexCoordXform.appendDecl(this->ctxInfo(), &vshaderTxt);
+    vshaderTxt.append(";");
+    uPosXform.appendDecl(this->ctxInfo(), &vshaderTxt);
+    vshaderTxt.append(";");
+    vTexCoord.appendDecl(this->ctxInfo(), &vshaderTxt);
+    vshaderTxt.append(";");
+    
+    vshaderTxt.append(
+        "// Copy Program VS\n"
+        "void main() {"
+        "  v_texCoord = a_vertex.xy * u_texCoordXform.xy + u_texCoordXform.zw;"
+        "  gl_Position.xy = a_vertex * u_posXform.xy + u_posXform.zw;"
+        "  gl_Position.zw = vec2(0, 1);"
+        "}"
+    );
+
+    SkString fshaderTxt(version);
+    GrGLSLAppendDefaultFloatPrecisionDeclaration(kDefault_GrSLPrecision, this->glStandard(),
+                                                 &fshaderTxt);
+    vTexCoord.setTypeModifier(GrShaderVar::kVaryingIn_TypeModifier);
+    vTexCoord.appendDecl(this->ctxInfo(), &fshaderTxt);
+    fshaderTxt.append(";");
+    uTexture.appendDecl(this->ctxInfo(), &fshaderTxt);
+    fshaderTxt.append(";");
+    const char* fsOutName;
+    if (this->glCaps().glslCaps()->mustDeclareFragmentShaderOutput()) {
+        oFragColor.appendDecl(this->ctxInfo(), &fshaderTxt);
+        fshaderTxt.append(";");
+        fsOutName = oFragColor.c_str();
+    } else {
+        fsOutName = "gl_FragColor";
+    }
+    fshaderTxt.appendf(
+        "// Copy Program FS\n"
+        "void main() {"
+        "  %s = %s(u_texture, v_texCoord);"
+        "}",
+        fsOutName,
+        GrGLSLTexture2DFunctionName(kVec2f_GrSLType, this->glslGeneration())
+    );
+    
+    GL_CALL_RET(fCopyProgram.fProgram, CreateProgram());
+    const char* str;
+    GrGLint length;
+
+    str = vshaderTxt.c_str();
+    length = SkToInt(vshaderTxt.size());
+    GrGLuint vshader = GrGLCompileAndAttachShader(fGLContext, fCopyProgram.fProgram,
+                                                  GR_GL_VERTEX_SHADER, &str, &length, 1, &fStats);
+
+    str = fshaderTxt.c_str();
+    length = SkToInt(fshaderTxt.size());
+    GrGLuint fshader = GrGLCompileAndAttachShader(fGLContext, fCopyProgram.fProgram,
+                                                  GR_GL_FRAGMENT_SHADER, &str, &length, 1, &fStats);
+
+    GL_CALL(LinkProgram(fCopyProgram.fProgram));
+
+    GL_CALL_RET(fCopyProgram.fTextureUniform, GetUniformLocation(fCopyProgram.fProgram,
+                                                                 "u_texture"));
+    GL_CALL_RET(fCopyProgram.fPosXformUniform, GetUniformLocation(fCopyProgram.fProgram,
+                                                                  "u_posXform"));
+    GL_CALL_RET(fCopyProgram.fTexCoordXformUniform, GetUniformLocation(fCopyProgram.fProgram,
+                                                                       "u_texCoordXform"));
+
+    GL_CALL(BindAttribLocation(fCopyProgram.fProgram, 0, "a_vertex"));
+
+    GL_CALL(DeleteShader(vshader));
+    GL_CALL(DeleteShader(fshader));
+
+    GL_CALL(GenBuffers(1, &fCopyProgram.fArrayBuffer));
+    fHWGeometryState.setVertexBufferID(this, fCopyProgram.fArrayBuffer);
+    static const GrGLfloat vdata[] = {
+        0, 0,
+        0, 1,
+        1, 0,
+        1, 1
+    };
+    GL_ALLOC_CALL(this->glInterface(),
+                  BufferData(GR_GL_ARRAY_BUFFER,
+                             (GrGLsizeiptr) sizeof(vdata),
+                             vdata,  // data ptr
+                             GR_GL_STATIC_DRAW));
+}
+
+void GrGLGpu::copySurfaceAsDraw(GrSurface* dst,
+                                GrSurface* src,
+                                const SkIRect& srcRect,
+                                const SkIPoint& dstPoint) {
+    int w = srcRect.width();
+    int h = srcRect.height();
+
+    GrGLTexture* srcTex = static_cast<GrGLTexture*>(src->asTexture());
+    GrTextureParams params(SkShader::kClamp_TileMode, GrTextureParams::kNone_FilterMode);
+    this->bindTexture(0, params, srcTex);
+
+    GrGLRenderTarget* dstRT = static_cast<GrGLRenderTarget*>(dst->asRenderTarget());
+    SkIRect dstRect = SkIRect::MakeXYWH(dstPoint.fX, dstPoint.fY, w, h);
+    this->flushRenderTarget(dstRT, &dstRect);
+
+    GL_CALL(UseProgram(fCopyProgram.fProgram));
+    fHWProgramID = fCopyProgram.fProgram;
+
+    fHWGeometryState.setVertexArrayID(this, 0);
+
+    GrGLAttribArrayState* attribs =
+        fHWGeometryState.bindArrayAndBufferToDraw(this, fCopyProgram.fArrayBuffer);
+    attribs->set(this, 0, fCopyProgram.fArrayBuffer, 2, GR_GL_FLOAT, false,
+                    2 * sizeof(GrGLfloat), 0);
+
+
+    // dst rect edges in NDC (-1 to 1)
+    int dw = dst->width();
+    int dh = dst->height();
+    GrGLfloat dx0 = 2.f * dstPoint.fX / dw - 1.f;
+    GrGLfloat dx1 = 2.f * (dstPoint.fX + w) / dw - 1.f;
+    GrGLfloat dy0 = 2.f * dstPoint.fY / dh - 1.f;
+    GrGLfloat dy1 = 2.f * (dstPoint.fY + h) / dh - 1.f;
+    if (kBottomLeft_GrSurfaceOrigin == dst->origin()) {
+        dy0 = -dy0;
+        dy1 = -dy1;
+    }
+
+    // src rect edges in normalized texture space (0 to 1)
+    int sw = src->width();
+    int sh = src->height();
+    GrGLfloat sx0 = (GrGLfloat)srcRect.fLeft / sw;
+    GrGLfloat sx1 = (GrGLfloat)(srcRect.fLeft + w) / sw;
+    GrGLfloat sy0 = (GrGLfloat)srcRect.fTop / sh;
+    GrGLfloat sy1 = (GrGLfloat)(srcRect.fTop + h) / sh;
+    if (kBottomLeft_GrSurfaceOrigin == src->origin()) {
+        sy0 = 1.f - sy0;
+        sy1 = 1.f - sy1;
+    }
+
+    GL_CALL(Uniform4f(fCopyProgram.fPosXformUniform, dx1 - dx0, dy1 - dy0, dx0, dy0));
+    GL_CALL(Uniform4f(fCopyProgram.fTexCoordXformUniform, sx1 - sx0, sy1 - sy0, sx0, sy0));
+    GL_CALL(Uniform1i(fCopyProgram.fTextureUniform, 0));
+
+    GrXferProcessor::BlendInfo blendInfo;
+    blendInfo.reset();
+    this->flushBlend(blendInfo);
+    this->flushColorWrite(true);
+    this->flushDither(false);
+    this->flushDrawFace(GrPipelineBuilder::kBoth_DrawFace);
+    this->flushHWAAState(dstRT, false);
+    this->disableScissor();
+    GrStencilSettings stencil;
+    stencil.setDisabled();
+    this->flushStencil(stencil);
+
+    GL_CALL(DrawArrays(GR_GL_TRIANGLE_STRIP, 0, 4));
+}
+
+void GrGLGpu::copySurfaceAsCopyTexSubImage(GrSurface* dst,
+                                           GrSurface* src,
+                                           const SkIRect& srcRect,
+                                           const SkIPoint& dstPoint) {
+    SkASSERT(can_copy_texsubimage(dst, src, this));
+    GrGLuint srcFBO;
+    GrGLIRect srcVP;
+    srcFBO = this->bindSurfaceAsFBO(src, GR_GL_FRAMEBUFFER, &srcVP, kSrc_TempFBOTarget);
+    GrGLTexture* dstTex = static_cast<GrGLTexture*>(dst->asTexture());
+    SkASSERT(dstTex);
+    // We modified the bound FBO
+    fHWBoundRenderTargetUniqueID = SK_InvalidUniqueID;
+    GrGLIRect srcGLRect;
+    srcGLRect.setRelativeTo(srcVP,
+                            srcRect.fLeft,
+                            srcRect.fTop,
+                            srcRect.width(),
+                            srcRect.height(),
+                            src->origin());
+
+    this->setScratchTextureUnit();
+    GL_CALL(BindTexture(GR_GL_TEXTURE_2D, dstTex->textureID()));
+    GrGLint dstY;
+    if (kBottomLeft_GrSurfaceOrigin == dst->origin()) {
+        dstY = dst->height() - (dstPoint.fY + srcGLRect.fHeight);
+    } else {
+        dstY = dstPoint.fY;
+    }
+    GL_CALL(CopyTexSubImage2D(GR_GL_TEXTURE_2D, 0,
+                                dstPoint.fX, dstY,
+                                srcGLRect.fLeft, srcGLRect.fBottom,
+                                srcGLRect.fWidth, srcGLRect.fHeight));
+    if (srcFBO) {
+        this->unbindTextureFromFBO(GR_GL_FRAMEBUFFER);
+    }
+}
+
+bool GrGLGpu::copySurfaceAsBlitFramebuffer(GrSurface* dst,
+                                           GrSurface* src,
+                                           const SkIRect& srcRect,
+                                           const SkIPoint& dstPoint) {
+    SkASSERT(can_blit_framebuffer(dst, src, this));
+    SkIRect dstRect = SkIRect::MakeXYWH(dstPoint.fX, dstPoint.fY,
+                                        srcRect.width(), srcRect.height());
+    if (dst == src) {
+        if (SkIRect::IntersectsNoEmptyCheck(dstRect, srcRect)) {
+            return false;
         }
     }
-    return false;
+
+    GrGLuint dstFBO;
+    GrGLuint srcFBO;
+    GrGLIRect dstVP;
+    GrGLIRect srcVP;
+    dstFBO = this->bindSurfaceAsFBO(dst, GR_GL_DRAW_FRAMEBUFFER, &dstVP,
+                                    kDst_TempFBOTarget);
+    srcFBO = this->bindSurfaceAsFBO(src, GR_GL_READ_FRAMEBUFFER, &srcVP,
+                                    kSrc_TempFBOTarget);
+    // We modified the bound FBO
+    fHWBoundRenderTargetUniqueID = SK_InvalidUniqueID;
+    GrGLIRect srcGLRect;
+    GrGLIRect dstGLRect;
+    srcGLRect.setRelativeTo(srcVP,
+                            srcRect.fLeft,
+                            srcRect.fTop,
+                            srcRect.width(),
+                            srcRect.height(),
+                            src->origin());
+    dstGLRect.setRelativeTo(dstVP,
+                            dstRect.fLeft,
+                            dstRect.fTop,
+                            dstRect.width(),
+                            dstRect.height(),
+                            dst->origin());
+
+    // BlitFrameBuffer respects the scissor, so disable it.
+    this->disableScissor();
+
+    GrGLint srcY0;
+    GrGLint srcY1;
+    // Does the blit need to y-mirror or not?
+    if (src->origin() == dst->origin()) {
+        srcY0 = srcGLRect.fBottom;
+        srcY1 = srcGLRect.fBottom + srcGLRect.fHeight;
+    } else {
+        srcY0 = srcGLRect.fBottom + srcGLRect.fHeight;
+        srcY1 = srcGLRect.fBottom;
+    }
+    GL_CALL(BlitFramebuffer(srcGLRect.fLeft,
+                            srcY0,
+                            srcGLRect.fLeft + srcGLRect.fWidth,
+                            srcY1,
+                            dstGLRect.fLeft,
+                            dstGLRect.fBottom,
+                            dstGLRect.fLeft + dstGLRect.fWidth,
+                            dstGLRect.fBottom + dstGLRect.fHeight,
+                            GR_GL_COLOR_BUFFER_BIT, GR_GL_NEAREST));
+    if (dstFBO) {
+        this->unbindTextureFromFBO(GR_GL_DRAW_FRAMEBUFFER);
+    }
+    if (srcFBO) {
+        this->unbindTextureFromFBO(GR_GL_READ_FRAMEBUFFER);
+    }
+    return true;
 }
 
 void GrGLGpu::xferBarrier(GrRenderTarget* rt, GrXferBarrierType type) {
@@ -2832,26 +3019,52 @@
 }
 
 ///////////////////////////////////////////////////////////////////////////////
-
 GrGLAttribArrayState* GrGLGpu::HWGeometryState::bindArrayAndBuffersToDraw(
                                                 GrGLGpu* gpu,
                                                 const GrGLVertexBuffer* vbuffer,
                                                 const GrGLIndexBuffer* ibuffer) {
     SkASSERT(vbuffer);
+    GrGLuint vbufferID = vbuffer->bufferID();
+    GrGLuint* ibufferIDPtr = NULL;
+    GrGLuint ibufferID;
+    if (ibuffer) {
+        ibufferID = ibuffer->bufferID();
+        ibufferIDPtr = &ibufferID;
+    }
+    return this->internalBind(gpu, vbufferID, ibufferIDPtr);
+}
+
+GrGLAttribArrayState* GrGLGpu::HWGeometryState::bindArrayAndBufferToDraw(GrGLGpu* gpu,
+                                                                         GrGLuint vbufferID) {
+    return this->internalBind(gpu, vbufferID, NULL);
+}
+
+GrGLAttribArrayState* GrGLGpu::HWGeometryState::bindArrayAndBuffersToDraw(GrGLGpu* gpu,
+                                                                          GrGLuint vbufferID,
+                                                                          GrGLuint ibufferID) {
+    return this->internalBind(gpu, vbufferID, &ibufferID);
+}
+
+GrGLAttribArrayState* GrGLGpu::HWGeometryState::internalBind(GrGLGpu* gpu,
+                                                             GrGLuint vbufferID,
+                                                             GrGLuint* ibufferID) {
     GrGLAttribArrayState* attribState;
 
-    // We use a vertex array if we're on a core profile and the verts are in a VBO.
-    if (gpu->glCaps().isCoreProfile() && !vbuffer->isCPUBacked()) {
+    if (gpu->glCaps().isCoreProfile() && 0 != vbufferID) {
         if (!fVBOVertexArray) {
             GrGLuint arrayID;
             GR_GL_CALL(gpu->glInterface(), GenVertexArrays(1, &arrayID));
             int attrCount = gpu->glCaps().maxVertexAttributes();
             fVBOVertexArray = SkNEW_ARGS(GrGLVertexArray, (arrayID, attrCount));
         }
-        attribState = fVBOVertexArray->bindWithIndexBuffer(gpu, ibuffer);
+        if (ibufferID) {
+            attribState = fVBOVertexArray->bindWithIndexBuffer(gpu, *ibufferID);
+        } else {
+            attribState = fVBOVertexArray->bind(gpu);
+        }
     } else {
-        if (ibuffer) {
-            this->setIndexBufferIDOnDefaultVertexArray(gpu, ibuffer->bufferID());
+        if (ibufferID) {
+            this->setIndexBufferIDOnDefaultVertexArray(gpu, *ibufferID);
         } else {
             this->setVertexArrayID(gpu, 0);
         }
diff --git a/src/gpu/gl/GrGLGpu.h b/src/gpu/gl/GrGLGpu.h
index 3323ece..986045a 100644
--- a/src/gpu/gl/GrGLGpu.h
+++ b/src/gpu/gl/GrGLGpu.h
@@ -101,11 +101,6 @@
                      const SkIRect& srcRect,
                      const SkIPoint& dstPoint) override;
 
-    bool canCopySurface(const GrSurface* dst,
-                        const GrSurface* src,
-                        const SkIRect& srcRect,
-                        const SkIPoint& dstPoint) override;
-
     void xferBarrier(GrRenderTarget*, GrXferBarrierType) override;
 
     void buildProgramDesc(GrProgramDesc*,
@@ -183,6 +178,19 @@
     // Subclasses should call this to flush the blend state.
     void flushBlend(const GrXferProcessor::BlendInfo& blendInfo);
 
+    void copySurfaceAsDraw(GrSurface* dst,
+                           GrSurface* src,
+                           const SkIRect& srcRect,
+                           const SkIPoint& dstPoint);
+    void copySurfaceAsCopyTexSubImage(GrSurface* dst,
+                                      GrSurface* src,
+                                      const SkIRect& srcRect,
+                                      const SkIPoint& dstPoint);
+    bool copySurfaceAsBlitFramebuffer(GrSurface* dst,
+                                      GrSurface* src,
+                                      const SkIRect& srcRect,
+                                      const SkIPoint& dstPoint);
+
     bool hasExtension(const char* ext) const { return fGLContext.hasExtension(ext); }
 
     static bool BlendCoeffReferencesConstant(GrBlendCoeff coeff);
@@ -193,7 +201,7 @@
         ~ProgramCache();
 
         void abandon();
-        GrGLProgram* getProgram(const DrawArgs&);
+        GrGLProgram* refProgram(const DrawArgs&);
 
     private:
         enum {
@@ -294,11 +302,12 @@
 
     void unbindTextureFromFBO(GrGLenum fboTarget);
 
+    void createCopyProgram();
+
     GrGLContext fGLContext;
 
     // GL program-related state
     ProgramCache*               fProgramCache;
-    SkAutoTUnref<GrGLProgram>   fCurrentProgram;
 
     ///////////////////////////////////////////////////////////////////////////
     ///@name Caching of GL State
@@ -420,7 +429,17 @@
                                                         const GrGLVertexBuffer* vbuffer,
                                                         const GrGLIndexBuffer* ibuffer);
 
+        /** Variants of the above that takes GL buffer IDs. Note that 0 does not imply that a 
+            buffer won't be bound. The "default buffer" will be bound, which is used for client-side
+            array rendering. */
+        GrGLAttribArrayState* bindArrayAndBufferToDraw(GrGLGpu* gpu, GrGLuint vbufferID);
+        GrGLAttribArrayState* bindArrayAndBuffersToDraw(GrGLGpu* gpu,
+                                                        GrGLuint vbufferID,
+                                                        GrGLuint ibufferID);
+
     private:
+        GrGLAttribArrayState* internalBind(GrGLGpu* gpu, GrGLuint vbufferID, GrGLuint* ibufferID);
+
         GrGLuint                fBoundVertexArrayID;
         GrGLuint                fBoundVertexBufferID;
         bool                    fBoundVertexArrayIDIsValid;
@@ -454,6 +473,15 @@
         }
     } fHWBlendState;
 
+    /** IDs for copy surface program. */
+    struct {
+        GrGLuint    fProgram;
+        GrGLint     fTextureUniform;
+        GrGLint     fTexCoordXformUniform;
+        GrGLint     fPosXformUniform;
+        GrGLuint    fArrayBuffer;
+    } fCopyProgram;
+
     TriState fMSAAEnabled;
 
     GrStencilSettings           fHWStencilSettings;
diff --git a/src/gpu/gl/GrGLGpuProgramCache.cpp b/src/gpu/gl/GrGLGpuProgramCache.cpp
index 911fd93..06efe29 100644
--- a/src/gpu/gl/GrGLGpuProgramCache.cpp
+++ b/src/gpu/gl/GrGLGpuProgramCache.cpp
@@ -90,7 +90,7 @@
     return SkTSearch(fEntries, fCount, desc, sizeof(Entry*), less);
 }
 
-GrGLProgram* GrGLGpu::ProgramCache::getProgram(const DrawArgs& args) {
+GrGLProgram* GrGLGpu::ProgramCache::refProgram(const DrawArgs& args) {
 #ifdef PROGRAM_CACHE_STATS
     ++fTotalRequests;
 #endif
@@ -193,5 +193,5 @@
         }
     }
     ++fCurrLRUStamp;
-    return entry->fProgram;
+    return SkRef(entry->fProgram.get());
 }
diff --git a/src/gpu/gl/GrGLSL.cpp b/src/gpu/gl/GrGLSL.cpp
index d388d91..f4caf52 100644
--- a/src/gpu/gl/GrGLSL.cpp
+++ b/src/gpu/gl/GrGLSL.cpp
@@ -91,6 +91,25 @@
     }
 }
 
+void GrGLSLAppendDefaultFloatPrecisionDeclaration(GrSLPrecision p, GrGLStandard s, SkString* out) {
+    // Desktop GLSL has added precision qualifiers but they don't do anything.
+    if (kGLES_GrGLStandard == s) {
+        switch (p) {
+            case kHigh_GrSLPrecision:
+                out->append("precision highp float;\n");
+                break;
+            case kMedium_GrSLPrecision:
+                out->append("precision mediump float;\n");
+                break;
+            case kLow_GrSLPrecision:
+                out->append("precision lowp float;\n");
+                break;
+            default:
+                SkFAIL("Unknown precision value.");
+        }
+    }
+}
+
 void GrGLSLMulVarBy4f(SkString* outAppend, const char* vec4VarName, const GrGLSLExpr4& mulFactor) {
     if (mulFactor.isOnes()) {
         *outAppend = SkString();
diff --git a/src/gpu/gl/GrGLSL.h b/src/gpu/gl/GrGLSL.h
index 99671c1..db8e1bd 100644
--- a/src/gpu/gl/GrGLSL.h
+++ b/src/gpu/gl/GrGLSL.h
@@ -57,6 +57,24 @@
 const char* GrGetGLSLVersionDecl(const GrGLContextInfo&);
 
 /**
+ * Adds a line of GLSL code to declare the default precision for float types.
+ */
+void GrGLSLAppendDefaultFloatPrecisionDeclaration(GrSLPrecision, GrGLStandard, SkString* out);
+
+/**
+ * Gets the name of the function that should be used to sample a 2D texture. Coord type is used
+ * to indicate whether the texture is sampled using projective textured (kVec3f) or not (kVec2f).
+ */
+inline const char* GrGLSLTexture2DFunctionName(GrSLType coordType, GrGLSLGeneration glslGen) {
+    if (kVec2f_GrSLType == coordType) {
+        return glslGen >= k130_GrGLSLGeneration ? "texture" : "texture2D";
+    } else {
+        SkASSERT(kVec3f_GrSLType == coordType);
+        return glslGen >= k130_GrGLSLGeneration ? "textureProj" : "texture2DProj";
+    }
+}
+
+/**
  * Converts a GrSLType to a string containing the name of the equivalent GLSL type.
  */
 static inline const char* GrGLSLTypeString(GrSLType t) {
diff --git a/src/gpu/gl/GrGLVertexArray.cpp b/src/gpu/gl/GrGLVertexArray.cpp
index e20dbb5..265b5b3 100644
--- a/src/gpu/gl/GrGLVertexArray.cpp
+++ b/src/gpu/gl/GrGLVertexArray.cpp
@@ -9,9 +9,10 @@
 #include "GrGLGpu.h"
 
 
-void GrGLAttribArrayState::set(const GrGLGpu* gpu,
+
+void GrGLAttribArrayState::set(GrGLGpu* gpu,
                                int index,
-                               GrGLVertexBuffer* buffer,
+                               GrGLuint vertexBufferID,
                                GrGLint size,
                                GrGLenum type,
                                GrGLboolean normalized,
@@ -25,13 +26,13 @@
         array->fEnabled = true;
     }
     if (!array->fAttribPointerIsValid ||
-        array->fVertexBufferID != buffer->bufferID() ||
+        array->fVertexBufferID != vertexBufferID ||
         array->fSize != size ||
         array->fNormalized != normalized ||
         array->fStride != stride ||
         array->fOffset != offset) {
 
-        buffer->bind();
+        gpu->bindVertexBuffer(vertexBufferID);
         GR_GL_CALL(gpu->glInterface(), VertexAttribPointer(index,
                                                            size,
                                                            type,
@@ -39,7 +40,7 @@
                                                            stride,
                                                            offset));
         array->fAttribPointerIsValid = true;
-        array->fVertexBufferID = buffer->bufferID();
+        array->fVertexBufferID = vertexBufferID;
         array->fSize = size;
         array->fNormalized = normalized;
         array->fStride = stride;
@@ -80,15 +81,13 @@
     return &fAttribArrays;
 }
 
-GrGLAttribArrayState* GrGLVertexArray::bindWithIndexBuffer(GrGLGpu* gpu,
-                                                           const GrGLIndexBuffer* buffer) {
+GrGLAttribArrayState* GrGLVertexArray::bindWithIndexBuffer(GrGLGpu* gpu, GrGLuint ibufferID) {
     GrGLAttribArrayState* state = this->bind(gpu);
-    if (state && buffer) {
-        GrGLuint bufferID = buffer->bufferID();
-        if (!fIndexBufferIDIsValid || bufferID != fIndexBufferID) {            
-            GR_GL_CALL(gpu->glInterface(), BindBuffer(GR_GL_ELEMENT_ARRAY_BUFFER, bufferID));
+    if (state) {
+        if (!fIndexBufferIDIsValid || ibufferID != fIndexBufferID) {            
+            GR_GL_CALL(gpu->glInterface(), BindBuffer(GR_GL_ELEMENT_ARRAY_BUFFER, ibufferID));
             fIndexBufferIDIsValid = true;
-            fIndexBufferID = bufferID;
+            fIndexBufferID = ibufferID;
         }
     }
     return state;
diff --git a/src/gpu/gl/GrGLVertexArray.h b/src/gpu/gl/GrGLVertexArray.h
index a8feb73..afb23c5 100644
--- a/src/gpu/gl/GrGLVertexArray.h
+++ b/src/gpu/gl/GrGLVertexArray.h
@@ -67,9 +67,9 @@
      * assumed that the GrGLAttribArrayState is tracking the state of the currently bound vertex
      * array object.
      */
-    void set(const GrGLGpu*,
-             int index,
-             GrGLVertexBuffer*,
+    void set(GrGLGpu*,
+             int attribIndex,
+             GrGLuint vertexBufferID,
              GrGLint size,
              GrGLenum type,
              GrGLboolean normalized,
@@ -147,7 +147,7 @@
      * This is a version of the above function that also binds an index buffer to the vertex
      * array object.
      */
-    GrGLAttribArrayState* bindWithIndexBuffer(GrGLGpu* gpu, const GrGLIndexBuffer*);
+    GrGLAttribArrayState* bindWithIndexBuffer(GrGLGpu* gpu, GrGLuint indexBufferID);
 
     void notifyIndexBufferDelete(GrGLuint bufferID);
 
diff --git a/src/gpu/gl/builders/GrGLFragmentShaderBuilder.cpp b/src/gpu/gl/builders/GrGLFragmentShaderBuilder.cpp
index 044752f..afb0718 100644
--- a/src/gpu/gl/builders/GrGLFragmentShaderBuilder.cpp
+++ b/src/gpu/gl/builders/GrGLFragmentShaderBuilder.cpp
@@ -15,26 +15,6 @@
 const char* GrGLFragmentShaderBuilder::kDstCopyColorName = "_dstColor";
 static const char* declared_color_output_name() { return "fsColorOut"; }
 static const char* dual_source_output_name() { return "dualSourceOut"; }
-static void append_default_precision_qualifier(GrSLPrecision p,
-                                               GrGLStandard standard,
-                                               SkString* str) {
-    // Desktop GLSL has added precision qualifiers but they don't do anything.
-    if (kGLES_GrGLStandard == standard) {
-        switch (p) {
-            case kHigh_GrSLPrecision:
-                str->append("precision highp float;\n");
-                break;
-            case kMedium_GrSLPrecision:
-                str->append("precision mediump float;\n");
-                break;
-            case kLow_GrSLPrecision:
-                str->append("precision lowp float;\n");
-                break;
-            default:
-                SkFAIL("Unknown precision value.");
-        }
-    }
-}
 
 static const char* specific_layout_qualifier_name(GrBlendEquation equation) {
     SkASSERT(GrBlendEquationIsAdvanced(equation));
@@ -270,9 +250,9 @@
                                                         SkTDArray<GrGLuint>* shaderIds) {
     GrGLGpu* gpu = fProgramBuilder->gpu();
     this->versionDecl() = GrGetGLSLVersionDecl(gpu->ctxInfo());
-    append_default_precision_qualifier(kDefault_GrSLPrecision,
-                                       gpu->glStandard(),
-                                       &this->precisionQualifier());
+    GrGLSLAppendDefaultFloatPrecisionDeclaration(kDefault_GrSLPrecision,
+                                                 gpu->glStandard(),
+                                                 &this->precisionQualifier());
     this->compileAndAppendLayoutQualifiers();
     fProgramBuilder->appendUniformDecls(GrGLProgramBuilder::kFragment_Visibility,
                                         &this->uniforms());
diff --git a/src/gpu/gl/builders/GrGLProgramBuilder.cpp b/src/gpu/gl/builders/GrGLProgramBuilder.cpp
index 28e1fba..95f75ff 100644
--- a/src/gpu/gl/builders/GrGLProgramBuilder.cpp
+++ b/src/gpu/gl/builders/GrGLProgramBuilder.cpp
@@ -343,8 +343,7 @@
         fFS.enableSecondaryOutput();
     }
 
-    // On any post 1.10 GLSL supporting GPU, we declare custom output
-    if (k110_GrGLSLGeneration != fFS.fProgramBuilder->gpu()->glslGeneration()) {
+    if (this->ctxInfo().caps()->glslCaps()->mustDeclareFragmentShaderOutput()) {
         fFS.enableCustomOutput();
     }
 
diff --git a/src/gpu/gl/builders/GrGLShaderBuilder.cpp b/src/gpu/gl/builders/GrGLShaderBuilder.cpp
index 67fbe24..984233d 100644
--- a/src/gpu/gl/builders/GrGLShaderBuilder.cpp
+++ b/src/gpu/gl/builders/GrGLShaderBuilder.cpp
@@ -12,14 +12,6 @@
 #include "../GrGLShaderVar.h"
 
 namespace {
-inline const char* sample_function_name(GrSLType type, GrGLSLGeneration glslGen) {
-    if (kVec2f_GrSLType == type) {
-        return glslGen >= k130_GrGLSLGeneration ? "texture" : "texture2D";
-    } else {
-        SkASSERT(kVec3f_GrSLType == type);
-        return glslGen >= k130_GrGLSLGeneration ? "textureProj" : "texture2DProj";
-    }
-}
 void append_texture_lookup(SkString* out,
                            GrGLGpu* gpu,
                            const char* samplerName,
@@ -30,7 +22,7 @@
     SkASSERT(coordName);
 
     out->appendf("%s(%s, %s)",
-                 sample_function_name(varyingType, gpu->glslGeneration()),
+                 GrGLSLTexture2DFunctionName(varyingType, gpu->glslGeneration()),
                  samplerName,
                  coordName);