Merge JDK-8220154 initial metal implementation patch to the jdk sandbox branch
Reviewed-by: avu, prr, kcr
Contributed-by: avu(Jetbrains), aghaisas, jdv
/*
* Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation. Oracle designates this
* particular file as subject to the "Classpath" exception as provided
* by Oracle in the LICENSE file that accompanied this code.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
#ifndef HEADLESS
#include <jlong.h>
#include "MTLBufImgOps.h"
#include "MTLContext.h"
#include "MTLRenderQueue.h"
#include "MTLSurfaceDataBase.h"
#include "GraphicsPrimitiveMgr.h"
/** Evaluates to true if the given bit is set on the local flags variable. */
#define IS_SET(flagbit) \
(((flags) & (flagbit)) != 0)
/**************************** ConvolveOp support ****************************/
/**
* The ConvolveOp shader is fairly straightforward. For each texel in
* the source texture, the shader samples the MxN texels in the surrounding
* area, multiplies each by its corresponding kernel value, and then sums
* them all together to produce a single color result. Finally, the
* resulting value is multiplied by the current OpenGL color, which contains
* the extra alpha value.
*
* Note that this shader source code includes some "holes" marked by "%s".
* This allows us to build different shader programs (e.g. one for
* 3x3, one for 5x5, and so on) simply by filling in these "holes" with
* a call to sprintf(). See the MTLBufImgOps_CreateConvolveProgram() method
* for more details.
*
* REMIND: Currently this shader (and the supporting code in the
* EnableConvolveOp() method) only supports 3x3 and 5x5 filters.
* Early shader-level hardware did not support non-constant sized
* arrays but modern hardware should support them (although I
* don't know of any simple way to find out, other than to compile
* the shader at runtime and see if the drivers complain).
*/
static const char *convolveShaderSource =
// maximum size supported by this shader
"const int MAX_KERNEL_SIZE = %s;"
// image to be convolved
"uniform sampler%s baseImage;"
// image edge limits:
// imgEdge.xy = imgMin.xy (anything < will be treated as edge case)
// imgEdge.zw = imgMax.xy (anything > will be treated as edge case)
"uniform vec4 imgEdge;"
// value for each location in the convolution kernel:
// kernelVals[i].x = offsetX[i]
// kernelVals[i].y = offsetY[i]
// kernelVals[i].z = kernel[i]
"uniform vec3 kernelVals[MAX_KERNEL_SIZE];"
""
"void main(void)"
"{"
" int i;"
" vec4 sum;"
""
" if (any(lessThan(gl_TexCoord[0].st, imgEdge.xy)) ||"
" any(greaterThan(gl_TexCoord[0].st, imgEdge.zw)))"
" {"
// (placeholder for edge condition code)
" %s"
" } else {"
" sum = vec4(0.0);"
" for (i = 0; i < MAX_KERNEL_SIZE; i++) {"
" sum +="
" kernelVals[i].z *"
" texture%s(baseImage,"
" gl_TexCoord[0].st + kernelVals[i].xy);"
" }"
" }"
""
// modulate with gl_Color in order to apply extra alpha
" gl_FragColor = sum * gl_Color;"
"}";
/**
* Flags that can be bitwise-or'ed together to control how the shader
* source code is generated.
*/
#define CONVOLVE_RECT (1 << 0)
#define CONVOLVE_EDGE_ZERO_FILL (1 << 1)
#define CONVOLVE_5X5 (1 << 2)
/**
* The handles to the ConvolveOp fragment program objects. The index to
* the array should be a bitwise-or'ing of the CONVOLVE_* flags defined
* above. Note that most applications will likely need to initialize one
* or two of these elements, so the array is usually sparsely populated.
*/
static GLhandleARB convolvePrograms[8];
/**
* The maximum kernel size supported by the ConvolveOp shader.
*/
#define MAX_KERNEL_SIZE 25
/**
* Compiles and links the ConvolveOp shader program. If successful, this
* function returns a handle to the newly created shader program; otherwise
* returns 0.
*/
static GLhandleARB
MTLBufImgOps_CreateConvolveProgram(jint flags)
{
//TODO
return NULL;
}
void
MTLBufImgOps_EnableConvolveOp(MTLContext *mtlc, jlong pSrcOps,
jboolean edgeZeroFill,
jint kernelWidth, jint kernelHeight,
unsigned char *kernel)
{
//TODO
}
void
MTLBufImgOps_DisableConvolveOp(MTLContext *mtlc)
{
//TODO
J2dTraceLn(J2D_TRACE_INFO, "MTLBufImgOps_DisableConvolveOp");
}
/**************************** RescaleOp support *****************************/
/**
* The RescaleOp shader is one of the simplest possible. Each fragment
* from the source image is multiplied by the user's scale factor and added
* to the user's offset value (these are component-wise operations).
* Finally, the resulting value is multiplied by the current OpenGL color,
* which contains the extra alpha value.
*
* The RescaleOp spec says that the operation is performed regardless of
* whether the source data is premultiplied or non-premultiplied. This is
* a problem for the OpenGL pipeline in that a non-premultiplied
* BufferedImage will have already been converted into premultiplied
* when uploaded to an OpenGL texture. Therefore, we have a special mode
* called RESCALE_NON_PREMULT (used only for source images that were
* originally non-premultiplied) that un-premultiplies the source color
* prior to the rescale operation, then re-premultiplies the resulting
* color before returning from the fragment shader.
*
* Note that this shader source code includes some "holes" marked by "%s".
* This allows us to build different shader programs (e.g. one for
* GL_TEXTURE_2D targets, one for GL_TEXTURE_RECTANGLE_ARB targets, and so on)
* simply by filling in these "holes" with a call to sprintf(). See the
* MTLBufImgOps_CreateRescaleProgram() method for more details.
*/
static const char *rescaleShaderSource =
// image to be rescaled
"uniform sampler%s baseImage;"
// vector containing scale factors
"uniform vec4 scaleFactors;"
// vector containing offsets
"uniform vec4 offsets;"
""
"void main(void)"
"{"
" vec4 srcColor = texture%s(baseImage, gl_TexCoord[0].st);"
// (placeholder for un-premult code)
" %s"
// rescale source value
" vec4 result = (srcColor * scaleFactors) + offsets;"
// (placeholder for re-premult code)
" %s"
// modulate with gl_Color in order to apply extra alpha
" gl_FragColor = result * gl_Color;"
"}";
/**
* Flags that can be bitwise-or'ed together to control how the shader
* source code is generated.
*/
#define RESCALE_RECT (1 << 0)
#define RESCALE_NON_PREMULT (1 << 1)
/**
* The handles to the RescaleOp fragment program objects. The index to
* the array should be a bitwise-or'ing of the RESCALE_* flags defined
* above. Note that most applications will likely need to initialize one
* or two of these elements, so the array is usually sparsely populated.
*/
static GLhandleARB rescalePrograms[4];
/**
* Compiles and links the RescaleOp shader program. If successful, this
* function returns a handle to the newly created shader program; otherwise
* returns 0.
*/
static GLhandleARB
MTLBufImgOps_CreateRescaleProgram(jint flags)
{
//TODO
return NULL;
}
void
MTLBufImgOps_EnableRescaleOp(MTLContext *mtlc, jlong pSrcOps,
jboolean nonPremult,
unsigned char *scaleFactors,
unsigned char *offsets)
{
//TODO
}
void
MTLBufImgOps_DisableRescaleOp(MTLContext *mtlc)
{
//TODO
J2dTraceLn(J2D_TRACE_INFO, "MTLBufImgOps_DisableRescaleOp");
RETURN_IF_NULL(mtlc);
}
/**************************** LookupOp support ******************************/
/**
* The LookupOp shader takes a fragment color (from the source texture) as
* input, subtracts the optional user offset value, and then uses the
* resulting value to index into the lookup table texture to provide
* a new color result. Finally, the resulting value is multiplied by
* the current OpenGL color, which contains the extra alpha value.
*
* The lookup step requires 3 texture accesses (or 4, when alpha is included),
* which is somewhat unfortunate because it's not ideal from a performance
* standpoint, but that sort of thing is getting faster with newer hardware.
* In the 3-band case, we could consider using a three-dimensional texture
* and performing the lookup with a single texture access step. We already
* use this approach in the LCD text shader, and it works well, but for the
* purposes of this LookupOp shader, it's probably overkill. Also, there's
* a difference in that the LCD text shader only needs to populate the 3D LUT
* once, but here we would need to populate it on every invocation, which
* would likely be a waste of VRAM and CPU/GPU cycles.
*
* The LUT texture is currently hardcoded as 4 rows/bands, each containing
* 256 elements. This means that we currently only support user-provided
* tables with no more than 256 elements in each band (this is checked at
* at the Java level). If the user provides a table with less than 256
* elements per band, our shader will still work fine, but if elements are
* accessed with an index >= the size of the LUT, then the shader will simply
* produce undefined values. Typically the user would provide an offset
* value that would prevent this from happening, but it's worth pointing out
* this fact because the software LookupOp implementation would usually
* throw an ArrayIndexOutOfBoundsException in this scenario (although it is
* not something demanded by the spec).
*
* The LookupOp spec says that the operation is performed regardless of
* whether the source data is premultiplied or non-premultiplied. This is
* a problem for the OpenGL pipeline in that a non-premultiplied
* BufferedImage will have already been converted into premultiplied
* when uploaded to an OpenGL texture. Therefore, we have a special mode
* called LOOKUP_NON_PREMULT (used only for source images that were
* originally non-premultiplied) that un-premultiplies the source color
* prior to the lookup operation, then re-premultiplies the resulting
* color before returning from the fragment shader.
*
* Note that this shader source code includes some "holes" marked by "%s".
* This allows us to build different shader programs (e.g. one for
* GL_TEXTURE_2D targets, one for GL_TEXTURE_RECTANGLE_ARB targets, and so on)
* simply by filling in these "holes" with a call to sprintf(). See the
* MTLBufImgOps_CreateLookupProgram() method for more details.
*/
static const char *lookupShaderSource =
// source image (bound to texture unit 0)
"uniform sampler%s baseImage;"
// lookup table (bound to texture unit 1)
"uniform sampler2D lookupTable;"
// offset subtracted from source index prior to lookup step
"uniform vec4 offset;"
""
"void main(void)"
"{"
" vec4 srcColor = texture%s(baseImage, gl_TexCoord[0].st);"
// (placeholder for un-premult code)
" %s"
// subtract offset from original index
" vec4 srcIndex = srcColor - offset;"
// use source value as input to lookup table (note that
// "v" texcoords are hardcoded to hit texel centers of
// each row/band in texture)
" vec4 result;"
" result.r = texture2D(lookupTable, vec2(srcIndex.r, 0.125)).r;"
" result.g = texture2D(lookupTable, vec2(srcIndex.g, 0.375)).r;"
" result.b = texture2D(lookupTable, vec2(srcIndex.b, 0.625)).r;"
// (placeholder for alpha store code)
" %s"
// (placeholder for re-premult code)
" %s"
// modulate with gl_Color in order to apply extra alpha
" gl_FragColor = result * gl_Color;"
"}";
/**
* Flags that can be bitwise-or'ed together to control how the shader
* source code is generated.
*/
#define LOOKUP_RECT (1 << 0)
#define LOOKUP_USE_SRC_ALPHA (1 << 1)
#define LOOKUP_NON_PREMULT (1 << 2)
/**
* The handles to the LookupOp fragment program objects. The index to
* the array should be a bitwise-or'ing of the LOOKUP_* flags defined
* above. Note that most applications will likely need to initialize one
* or two of these elements, so the array is usually sparsely populated.
*/
static GLhandleARB lookupPrograms[8];
/**
* The handle to the lookup table texture object used by the shader.
*/
static GLuint lutTextureID = 0;
/**
* Compiles and links the LookupOp shader program. If successful, this
* function returns a handle to the newly created shader program; otherwise
* returns 0.
*/
static GLhandleARB
MTLBufImgOps_CreateLookupProgram(jint flags)
{
//TODO
return NULL;
}
void
MTLBufImgOps_EnableLookupOp(MTLContext *mtlc, jlong pSrcOps,
jboolean nonPremult, jboolean shortData,
jint numBands, jint bandLength, jint offset,
void *tableValues)
{
//TODO
}
void
MTLBufImgOps_DisableLookupOp(MTLContext *mtlc)
{
//TODO
J2dTraceLn(J2D_TRACE_INFO, "MTLBufImgOps_DisableLookupOp");
}
#endif /* !HEADLESS */