Files
renderdoc/renderdoc/driver/gl/gl_shader_refl.cpp
T

2825 lines
93 KiB
C++

/******************************************************************************
* The MIT License (MIT)
*
* Copyright (c) 2014-2019 Baldur Karlsson
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
******************************************************************************/
#include "gl_shader_refl.h"
#include <algorithm>
#include <functional>
#include "3rdparty/glslang/glslang/Public/ShaderLang.h"
#include "driver/shaders/spirv/glslang_compile.h"
#include "gl_driver.h"
template <>
rdcstr DoStringise(const FFVertexOutput &el)
{
BEGIN_ENUM_STRINGISE(FFVertexOutput);
{
STRINGISE_ENUM_CLASS_NAMED(PointSize, "gl_PointSize");
STRINGISE_ENUM_CLASS_NAMED(ClipDistance, "gl_ClipDistance");
STRINGISE_ENUM_CLASS_NAMED(CullDistance, "gl_CullDistance");
STRINGISE_ENUM_CLASS_NAMED(ClipVertex, "gl_ClipVertex");
STRINGISE_ENUM_CLASS_NAMED(FrontColor, "gl_FrontColor");
STRINGISE_ENUM_CLASS_NAMED(BackColor, "gl_BackColor");
STRINGISE_ENUM_CLASS_NAMED(FrontSecondaryColor, "gl_FrontSecondaryColor");
STRINGISE_ENUM_CLASS_NAMED(BackSecondaryColor, "gl_BackSecondaryColor");
STRINGISE_ENUM_CLASS_NAMED(TexCoord, "gl_TexCoord");
STRINGISE_ENUM_CLASS_NAMED(FogFragCoord, "gl_FogFragCoord");
STRINGISE_ENUM_CLASS_NAMED(Count, "gl_Count");
}
END_ENUM_STRINGISE();
}
void namesort(rdcarray<ShaderConstant> &vars)
{
if(vars.empty())
return;
struct name_sort
{
bool operator()(const ShaderConstant &a, const ShaderConstant &b) { return a.name < b.name; }
};
std::sort(vars.begin(), vars.end(), name_sort());
for(size_t i = 0; i < vars.size(); i++)
namesort(vars[i].type.members);
}
void sort(rdcarray<ShaderConstant> &vars)
{
if(vars.empty())
return;
std::sort(vars.begin(), vars.end(), [](const ShaderConstant &a, const ShaderConstant &b) {
return a.byteOffset < b.byteOffset;
});
for(size_t i = 0; i < vars.size(); i++)
sort(vars[i].type.members);
}
void CheckVertexOutputUses(const std::vector<std::string> &sources,
FixedFunctionVertexOutputs &outputUsage)
{
outputUsage = FixedFunctionVertexOutputs();
for(FFVertexOutput output : values<FFVertexOutput>())
{
// we consider an output used if we encounter a '=' before either a ';' or the end of the string
std::string name = ToStr(output);
for(size_t i = 0; i < sources.size(); i++)
{
const std::string &s = sources[i];
size_t offs = 0;
for(;;)
{
offs = s.find(name, offs);
if(offs == std::string::npos)
break;
while(offs < s.length())
{
if(s[offs] == '=')
{
outputUsage.used[(int)output] = true;
break;
}
if(s[offs] == ';')
break;
offs++;
}
}
}
}
}
// little utility function that if necessary emulates glCreateShaderProgramv functionality but using
// glCompileShaderIncludeARB
static GLuint CreateSepProgram(WrappedOpenGL &driver, GLenum type, GLsizei numSources,
const char **sources, GLsizei numPaths, const char **paths)
{
// by the nature of this function, it might fail - we don't want to spew
// false positive looking messages into the log.
driver.SuppressDebugMessages(true);
GLuint program = 0;
// definition of glCreateShaderProgramv from the spec
GLuint shader = driver.glCreateShader(type);
if(shader)
{
driver.glShaderSource(shader, numSources, sources, NULL);
if(paths == NULL)
driver.glCompileShader(shader);
else
driver.glCompileShaderIncludeARB(shader, numPaths, paths, NULL);
program = driver.glCreateProgram();
if(program)
{
GLint compiled = 0;
driver.glGetShaderiv(shader, eGL_COMPILE_STATUS, &compiled);
driver.glProgramParameteri(program, eGL_PROGRAM_SEPARABLE, GL_TRUE);
if(compiled)
{
driver.glAttachShader(program, shader);
driver.glLinkProgram(program);
// we deliberately leave the shaders attached so this program can be re-linked.
// they will be cleaned up when the program is deleted
// driver.glDetachShader(program, shader);
}
}
driver.glDeleteShader(shader);
}
driver.SuppressDebugMessages(false);
return program;
}
static bool isspacetab(char c)
{
return c == '\t' || c == ' ';
}
static bool isnewline(char c)
{
return c == '\r' || c == '\n';
}
static bool iswhitespace(char c)
{
return isspacetab(c) || isnewline(c);
}
GLuint MakeSeparableShaderProgram(WrappedOpenGL &drv, GLenum type, std::vector<std::string> sources,
std::vector<std::string> *includepaths)
{
// in and out blocks are added separately, in case one is there already
const char *blockIdentifiers[2] = {"in gl_PerVertex", "out gl_PerVertex"};
std::string blocks[2] = {"", ""};
if(type == eGL_VERTEX_SHADER)
{
blocks[1] =
"out gl_PerVertex { vec4 gl_Position; float gl_PointSize; float gl_ClipDistance[]; };\n";
}
else if(type == eGL_TESS_CONTROL_SHADER)
{
blocks[0] =
"in gl_PerVertex { vec4 gl_Position; float gl_PointSize; float gl_ClipDistance[]; } "
"gl_in[];\n";
blocks[1] =
"out gl_PerVertex { vec4 gl_Position; float gl_PointSize; float gl_ClipDistance[]; } "
"gl_out[];\n";
}
else
{
blocks[0] =
"in gl_PerVertex { vec4 gl_Position; float gl_PointSize; float gl_ClipDistance[]; } "
"gl_in[];\n";
blocks[1] =
"out gl_PerVertex { vec4 gl_Position; float gl_PointSize; float gl_ClipDistance[]; };\n";
}
const char **strings = new const char *[sources.size()];
for(size_t i = 0; i < sources.size(); i++)
strings[i] = sources[i].c_str();
const char **paths = NULL;
GLsizei numPaths = 0;
if(includepaths)
{
numPaths = (GLsizei)includepaths->size();
paths = new const char *[includepaths->size()];
for(size_t i = 0; i < includepaths->size(); i++)
paths[i] = (*includepaths)[i].c_str();
}
GLuint sepProg = CreateSepProgram(drv, type, (GLsizei)sources.size(), strings, numPaths, paths);
GLint status;
drv.glGetProgramiv(sepProg, eGL_LINK_STATUS, &status);
// allow any vertex processing shader to redeclare gl_PerVertex
// on GLES it is not required
if(!IsGLES && status == 0 && type != eGL_FRAGMENT_SHADER && type != eGL_COMPUTE_SHADER)
{
drv.glDeleteProgram(sepProg);
sepProg = 0;
// try and patch up shader
// naively insert gl_PerVertex block as soon as it's valid (after #version)
// this will fail if e.g. a member of gl_PerVertex is declared at global scope
// (this is probably most likely for clipdistance if it's redeclared with a size)
// we start by concatenating the source strings to make parsing easier.
std::string combined;
for(size_t i = 0; i < sources.size(); i++)
combined += sources[i];
for(int attempt = 0; attempt < 2; attempt++)
{
std::string src = combined;
if(attempt == 1)
{
drv.glDeleteProgram(sepProg);
sepProg = 0;
RDCLOG("Attempting to pre-process shader with glslang to allow patching");
glslang::TShader sh(EShLanguage(ShaderIdx(type)));
const char *c_src = combined.c_str();
sh.setStrings(&c_src, 1);
sh.setEnvInput(glslang::EShSourceGlsl, EShLanguage(ShaderIdx(type)),
glslang::EShClientOpenGL, 100);
sh.setEnvClient(glslang::EShClientOpenGL, glslang::EShTargetOpenGL_450);
sh.setEnvTarget(glslang::EShTargetNone, glslang::EShTargetSpv_1_0);
glslang::TShader::ForbidIncluder incl;
bool success = sh.preprocess(GetDefaultResources(), 100, ENoProfile, false, false,
EShMsgOnlyPreprocessor, &src, incl);
if(!success)
{
RDCLOG("glslang failed:\n\n%s\n\n%s", sh.getInfoLog(), sh.getInfoDebugLog());
continue;
}
}
for(int blocktype = 0; blocktype < 2; blocktype++)
{
// vertex shaders don't have an in block
if(type == eGL_VERTEX_SHADER && blocktype == 0)
continue;
std::string block = blocks[blocktype];
const char *identifier = blockIdentifiers[blocktype];
// if we find the 'identifier' (ie. the block name),
// assume this block is already present and stop.
// only try and insert this block if the shader doesn't already have it
if(src.find(identifier) != std::string::npos)
{
continue;
}
{
size_t len = src.length();
// find if this source contains a #version, accounting for whitespace
size_t it = 0;
while(it != std::string::npos)
{
it = src.find("#", it);
if(it == std::string::npos)
break;
// advance past the #
++it;
// skip whitespace
while(it < len && isspacetab(src[it]))
++it;
if(it + 7 < len && !strncmp(&src[it], "version", 7))
{
it += sizeof("version") - 1;
break;
}
}
// no #version found
if(it == std::string::npos)
{
// insert at the start
it = 0;
}
else
{
// it now points after the #version
// skip whitespace
while(it < len && isspacetab(src[it]))
++it;
// skip number
while(it < len && src[it] >= '0' && src[it] <= '9')
++it;
// skip whitespace
while(it < len && isspacetab(src[it]))
++it;
if(!strncmp(&src[it], "core", 4))
it += sizeof("core") - 1;
if(!strncmp(&src[it], "compatibility", 13))
it += sizeof("compatibility") - 1;
if(!strncmp(&src[it], "es", 2))
it += sizeof("es") - 1;
it++;
// next line after #version, insert the extension declaration
if(it < src.length())
src.insert(it, "#extension GL_ARB_separate_shader_objects : enable\n");
// how deep are we in an #if. We want to place our definition
// outside of any #ifs.
int if_depth = 0;
// now skip past comments, and any #directives
while(it < len)
{
// skip whitespace
while(it < len && iswhitespace(src[it]))
++it;
// skip C++ style comments
if(it + 1 < len && src[it] == '/' && src[it + 1] == '/')
{
// keep going until the next newline
while(it < len && !isnewline(src[it]))
++it;
// skip more things
continue;
}
// skip preprocessor directives
if(src[it] == '#')
{
// skip the '#'
it++;
// skip whitespace
while(it < len && iswhitespace(src[it]))
++it;
// if it's an if, then increase our depth
// This covers:
// #if
// #ifdef
// #ifndef
if(!strncmp(&src[it], "if", 2))
{
if_depth++;
}
else if(!strncmp(&src[it], "endif", 5))
{
if_depth--;
}
// everything else is #extension or #else or #undef or anything
// keep going until the next newline
while(it < len && !isnewline(src[it]))
{
// if we encounter a C-style comment in the middle of a #define
// we can't consume it because then we'd miss the start of it.
// Instead we break out (although we're not technically at the
// end of the pre-processor line) and let it be consumed next.
// Note that we can discount C++-style comments because they
// want to consume to the end of the line too.
if(it + 1 < len && src[it] == '/' && src[it + 1] == '*')
break;
++it;
}
// skip more things
continue;
}
// skip C style comments
if(it + 1 < len && src[it] == '/' && src[it + 1] == '*')
{
// keep going until the we reach a */
while(it + 1 < len && (src[it] != '*' || src[it + 1] != '/'))
++it;
// skip the closing */ too
it += 2;
// skip more things
continue;
}
// see if we have a precision statement, if so skip that
const char precision[] = "precision";
if(it + sizeof(precision) < len && !strncmp(&src[it], precision, sizeof(precision) - 1))
{
// since we're speculating here (although what else could it be?) we don't modify
// it until we're sure.
size_t pit = it + sizeof(precision);
// skip whitespace
while(pit < len && isspacetab(src[pit]))
++pit;
// if we now match any of the precisions, then continue consuming until the next ;
const char lowp[] = "lowp";
const char mediump[] = "mediump";
const char highp[] = "highp";
bool precisionMatch =
(pit + sizeof(lowp) < len && !strncmp(&src[pit], lowp, sizeof(lowp) - 1) &&
isspacetab(src[pit + sizeof(lowp) - 1]));
precisionMatch |= (pit + sizeof(mediump) < len &&
!strncmp(&src[pit], mediump, sizeof(mediump) - 1) &&
isspacetab(src[pit + sizeof(mediump) - 1]));
precisionMatch |=
(pit + sizeof(highp) < len && !strncmp(&src[pit], highp, sizeof(highp) - 1) &&
isspacetab(src[pit + sizeof(highp) - 1]));
if(precisionMatch)
{
it = pit;
while(it < len && src[it] != ';')
++it;
++it; // skip the ; itself
// skip more things
continue;
}
// otherwise just stop here, it's not a precision statement
}
// nothing more to skip, check if we're outside an if
if(if_depth == 0)
break;
// if not, this might not be a comment, etc etc. Just skip to the next line
// so we can keep going to find the #endif
while(it < len && !isnewline(src[it]))
++it;
}
}
if(it < src.length())
src.insert(it, block);
}
}
const char *c_src = src.c_str();
sepProg = CreateSepProgram(drv, type, 1, &c_src, numPaths, paths);
// when we get it to link, bail!
drv.glGetProgramiv(sepProg, eGL_LINK_STATUS, &status);
if(status == 1)
break;
RDCWARN("Couldn't patch separability into shader, attempt #%d", attempt + 1);
}
}
if(sepProg)
drv.glGetProgramiv(sepProg, eGL_LINK_STATUS, &status);
else
status = 0;
if(status == 0)
{
char buffer[1025] = {0};
drv.glGetProgramInfoLog(sepProg, 1024, NULL, buffer);
RDCERR("Couldn't make separable shader program for shader. Errors:\n%s", buffer);
drv.glDeleteProgram(sepProg);
sepProg = 0;
}
delete[] strings;
if(paths)
delete[] paths;
return sepProg;
}
void ReconstructVarTree(GLenum query, GLuint sepProg, GLuint varIdx, GLint numParentBlocks,
rdcarray<ShaderConstant> *parentBlocks,
rdcarray<ShaderConstant> *defaultBlock)
{
const size_t numProps = 9;
GLenum resProps[numProps] = {eGL_TYPE, eGL_NAME_LENGTH, eGL_LOCATION,
eGL_BLOCK_INDEX, eGL_ARRAY_SIZE, eGL_OFFSET,
eGL_IS_ROW_MAJOR, eGL_ARRAY_STRIDE, eGL_MATRIX_STRIDE};
// GL_LOCATION not valid for buffer variables (it's only used if offset comes back -1, which will
// never happen for buffer variables)
if(query == eGL_BUFFER_VARIABLE)
resProps[2] = eGL_OFFSET;
GLint values[numProps] = {-1, -1, -1, -1, -1, -1, -1, -1};
GL.glGetProgramResourceiv(sepProg, query, varIdx, numProps, resProps, numProps, NULL, values);
ShaderConstant var;
var.type.descriptor.elements = RDCMAX(1, values[4]);
// set type (or bail if it's not a variable - sampler or such)
switch(values[0])
{
case eGL_FLOAT_VEC4:
case eGL_FLOAT_VEC3:
case eGL_FLOAT_VEC2:
case eGL_FLOAT:
case eGL_FLOAT_MAT4:
case eGL_FLOAT_MAT3:
case eGL_FLOAT_MAT2:
case eGL_FLOAT_MAT4x2:
case eGL_FLOAT_MAT4x3:
case eGL_FLOAT_MAT3x4:
case eGL_FLOAT_MAT3x2:
case eGL_FLOAT_MAT2x4:
case eGL_FLOAT_MAT2x3: var.type.descriptor.type = VarType::Float; break;
case eGL_DOUBLE_VEC4:
case eGL_DOUBLE_VEC3:
case eGL_DOUBLE_VEC2:
case eGL_DOUBLE:
case eGL_DOUBLE_MAT4:
case eGL_DOUBLE_MAT3:
case eGL_DOUBLE_MAT2:
case eGL_DOUBLE_MAT4x2:
case eGL_DOUBLE_MAT4x3:
case eGL_DOUBLE_MAT3x4:
case eGL_DOUBLE_MAT3x2:
case eGL_DOUBLE_MAT2x4:
case eGL_DOUBLE_MAT2x3: var.type.descriptor.type = VarType::Double; break;
case eGL_UNSIGNED_INT_VEC4:
case eGL_UNSIGNED_INT_VEC3:
case eGL_UNSIGNED_INT_VEC2:
case eGL_UNSIGNED_INT:
case eGL_BOOL_VEC4:
case eGL_BOOL_VEC3:
case eGL_BOOL_VEC2:
case eGL_BOOL: var.type.descriptor.type = VarType::UInt; break;
case eGL_INT_VEC4:
case eGL_INT_VEC3:
case eGL_INT_VEC2:
case eGL_INT: var.type.descriptor.type = VarType::SInt; break;
default:
// not a variable (sampler etc)
return;
}
// set # rows if it's a matrix
var.type.descriptor.rows = 1;
switch(values[0])
{
case eGL_FLOAT_MAT4:
case eGL_DOUBLE_MAT4:
case eGL_FLOAT_MAT2x4:
case eGL_DOUBLE_MAT2x4:
case eGL_FLOAT_MAT3x4:
case eGL_DOUBLE_MAT3x4: var.type.descriptor.rows = 4; break;
case eGL_FLOAT_MAT3:
case eGL_DOUBLE_MAT3:
case eGL_FLOAT_MAT4x3:
case eGL_DOUBLE_MAT4x3:
case eGL_FLOAT_MAT2x3:
case eGL_DOUBLE_MAT2x3: var.type.descriptor.rows = 3; break;
case eGL_FLOAT_MAT2:
case eGL_DOUBLE_MAT2:
case eGL_FLOAT_MAT4x2:
case eGL_DOUBLE_MAT4x2:
case eGL_FLOAT_MAT3x2:
case eGL_DOUBLE_MAT3x2: var.type.descriptor.rows = 2; break;
default: break;
}
// set # columns
switch(values[0])
{
case eGL_FLOAT_VEC4:
case eGL_FLOAT_MAT4:
case eGL_FLOAT_MAT4x2:
case eGL_FLOAT_MAT4x3:
case eGL_DOUBLE_VEC4:
case eGL_DOUBLE_MAT4:
case eGL_DOUBLE_MAT4x2:
case eGL_DOUBLE_MAT4x3:
case eGL_UNSIGNED_INT_VEC4:
case eGL_BOOL_VEC4:
case eGL_INT_VEC4: var.type.descriptor.columns = 4; break;
case eGL_FLOAT_VEC3:
case eGL_FLOAT_MAT3:
case eGL_FLOAT_MAT3x4:
case eGL_FLOAT_MAT3x2:
case eGL_DOUBLE_VEC3:
case eGL_DOUBLE_MAT3:
case eGL_DOUBLE_MAT3x4:
case eGL_DOUBLE_MAT3x2:
case eGL_UNSIGNED_INT_VEC3:
case eGL_BOOL_VEC3:
case eGL_INT_VEC3: var.type.descriptor.columns = 3; break;
case eGL_FLOAT_VEC2:
case eGL_FLOAT_MAT2:
case eGL_FLOAT_MAT2x4:
case eGL_FLOAT_MAT2x3:
case eGL_DOUBLE_VEC2:
case eGL_DOUBLE_MAT2:
case eGL_DOUBLE_MAT2x4:
case eGL_DOUBLE_MAT2x3:
case eGL_UNSIGNED_INT_VEC2:
case eGL_BOOL_VEC2:
case eGL_INT_VEC2: var.type.descriptor.columns = 2; break;
case eGL_FLOAT:
case eGL_DOUBLE:
case eGL_UNSIGNED_INT:
case eGL_INT:
case eGL_BOOL: var.type.descriptor.columns = 1; break;
default: break;
}
// set name
switch(values[0])
{
case eGL_FLOAT_VEC4: var.type.descriptor.name = "vec4"; break;
case eGL_FLOAT_VEC3: var.type.descriptor.name = "vec3"; break;
case eGL_FLOAT_VEC2: var.type.descriptor.name = "vec2"; break;
case eGL_FLOAT: var.type.descriptor.name = "float"; break;
case eGL_FLOAT_MAT4: var.type.descriptor.name = "mat4"; break;
case eGL_FLOAT_MAT3: var.type.descriptor.name = "mat3"; break;
case eGL_FLOAT_MAT2: var.type.descriptor.name = "mat2"; break;
case eGL_FLOAT_MAT4x2: var.type.descriptor.name = "mat4x2"; break;
case eGL_FLOAT_MAT4x3: var.type.descriptor.name = "mat4x3"; break;
case eGL_FLOAT_MAT3x4: var.type.descriptor.name = "mat3x4"; break;
case eGL_FLOAT_MAT3x2: var.type.descriptor.name = "mat3x2"; break;
case eGL_FLOAT_MAT2x4: var.type.descriptor.name = "mat2x4"; break;
case eGL_FLOAT_MAT2x3: var.type.descriptor.name = "mat2x3"; break;
case eGL_DOUBLE_VEC4: var.type.descriptor.name = "dvec4"; break;
case eGL_DOUBLE_VEC3: var.type.descriptor.name = "dvec3"; break;
case eGL_DOUBLE_VEC2: var.type.descriptor.name = "dvec2"; break;
case eGL_DOUBLE: var.type.descriptor.name = "double"; break;
case eGL_DOUBLE_MAT4: var.type.descriptor.name = "dmat4"; break;
case eGL_DOUBLE_MAT3: var.type.descriptor.name = "dmat3"; break;
case eGL_DOUBLE_MAT2: var.type.descriptor.name = "dmat2"; break;
case eGL_DOUBLE_MAT4x2: var.type.descriptor.name = "dmat4x2"; break;
case eGL_DOUBLE_MAT4x3: var.type.descriptor.name = "dmat4x3"; break;
case eGL_DOUBLE_MAT3x4: var.type.descriptor.name = "dmat3x4"; break;
case eGL_DOUBLE_MAT3x2: var.type.descriptor.name = "dmat3x2"; break;
case eGL_DOUBLE_MAT2x4: var.type.descriptor.name = "dmat2x4"; break;
case eGL_DOUBLE_MAT2x3: var.type.descriptor.name = "dmat2x3"; break;
case eGL_UNSIGNED_INT_VEC4: var.type.descriptor.name = "uvec4"; break;
case eGL_UNSIGNED_INT_VEC3: var.type.descriptor.name = "uvec3"; break;
case eGL_UNSIGNED_INT_VEC2: var.type.descriptor.name = "uvec2"; break;
case eGL_UNSIGNED_INT: var.type.descriptor.name = "uint"; break;
case eGL_BOOL_VEC4: var.type.descriptor.name = "bvec4"; break;
case eGL_BOOL_VEC3: var.type.descriptor.name = "bvec3"; break;
case eGL_BOOL_VEC2: var.type.descriptor.name = "bvec2"; break;
case eGL_BOOL: var.type.descriptor.name = "bool"; break;
case eGL_INT_VEC4: var.type.descriptor.name = "ivec4"; break;
case eGL_INT_VEC3: var.type.descriptor.name = "ivec3"; break;
case eGL_INT_VEC2: var.type.descriptor.name = "ivec2"; break;
case eGL_INT: var.type.descriptor.name = "int"; break;
default: break;
}
if(values[5] == -1 && values[2] >= 0)
{
var.byteOffset = values[2];
}
else if(values[5] >= 0)
{
var.byteOffset = values[5];
}
else
{
var.byteOffset = ~0U;
}
var.type.descriptor.rowMajorStorage = (values[6] > 0);
var.type.descriptor.matrixByteStride = (uint8_t)values[8];
RDCASSERTMSG("Stride is too large for uint16_t", values[7] <= 0xffff);
var.type.descriptor.arrayByteStride = RDCMIN((uint32_t)values[7], 0xffffu) & 0xffff;
bool bareUniform = false;
// for plain uniforms we won't get an array/matrix byte stride. Calculate tightly packed strides
if(values[3] == -1)
{
bareUniform = true;
// plain matrices are always column major, so this is the size of a column
var.type.descriptor.rowMajorStorage = false;
const uint32_t elemByteStride = (var.type.descriptor.type == VarType::Double) ? 8 : 4;
var.type.descriptor.matrixByteStride = uint8_t(var.type.descriptor.rows * elemByteStride);
// arrays are fetched as individual glGetUniform calls
var.type.descriptor.arrayByteStride = 0;
}
// set vectors as row major for convenience, since that's how they're stored in the fv array.
switch(values[0])
{
case eGL_FLOAT_VEC4:
case eGL_FLOAT_VEC3:
case eGL_FLOAT_VEC2:
case eGL_FLOAT:
case eGL_DOUBLE_VEC4:
case eGL_DOUBLE_VEC3:
case eGL_DOUBLE_VEC2:
case eGL_DOUBLE:
case eGL_UNSIGNED_INT_VEC4:
case eGL_UNSIGNED_INT_VEC3:
case eGL_UNSIGNED_INT_VEC2:
case eGL_UNSIGNED_INT:
case eGL_BOOL_VEC4:
case eGL_BOOL_VEC3:
case eGL_BOOL_VEC2:
case eGL_BOOL:
case eGL_INT_VEC4:
case eGL_INT_VEC3:
case eGL_INT_VEC2:
case eGL_INT: var.type.descriptor.rowMajorStorage = true; break;
default: break;
}
var.name.resize(values[1] - 1);
GL.glGetProgramResourceName(sepProg, query, varIdx, values[1], NULL, &var.name[0]);
std::string fullname = var.name;
int32_t c = values[1] - 1;
// trim off trailing [0] if it's an array
if(var.name[c - 3] == '[' && var.name[c - 2] == '0' && var.name[c - 1] == ']')
var.name.resize(c - 3);
else
var.type.descriptor.elements = 0;
GLint topLevelStride = 0;
if(query == eGL_BUFFER_VARIABLE)
{
GLenum propName = eGL_TOP_LEVEL_ARRAY_STRIDE;
GL.glGetProgramResourceiv(sepProg, query, varIdx, 1, &propName, 1, NULL, &topLevelStride);
}
rdcarray<ShaderConstant> *parentmembers = defaultBlock;
if(!bareUniform && values[3] < numParentBlocks)
{
parentmembers = &parentBlocks[values[3]];
}
if(parentmembers == NULL)
{
RDCWARN("Found variable '%s' without parent block index '%d'", var.name.c_str(), values[3]);
return;
}
char *nm = &var.name[0];
bool multiDimArray = false;
int arrayIdx = 0;
bool blockLevel = true;
// reverse figure out structures and structure arrays
while(strchr(nm, '.') || strchr(nm, '['))
{
char *base = nm;
while(*nm != '.' && *nm != '[')
nm++;
// determine if we have an array index, and NULL out
// what's after the base variable name
bool isarray = (*nm == '[');
*nm = 0;
nm++;
arrayIdx = 0;
// if it's an array, get the index used
if(isarray)
{
// get array index, it's always a decimal number
while(*nm >= '0' && *nm <= '9')
{
arrayIdx *= 10;
arrayIdx += int(*nm) - int('0');
nm++;
}
RDCASSERT(*nm == ']');
*nm = 0;
nm++;
// skip forward to the child name
if(*nm == '.')
{
*nm = 0;
nm++;
}
else
{
// if there's no . after the array index, this is a multi-dimensional array.
multiDimArray = true;
}
}
// construct a parent variable
ShaderConstant parentVar;
parentVar.name = base;
parentVar.byteOffset = var.byteOffset;
parentVar.type.descriptor.name = "struct";
parentVar.type.descriptor.rows = 0;
parentVar.type.descriptor.columns = 0;
parentVar.type.descriptor.rowMajorStorage = false;
parentVar.type.descriptor.type = var.type.descriptor.type;
parentVar.type.descriptor.elements =
isarray && !multiDimArray ? RDCMAX(1U, uint32_t(arrayIdx + 1)) : 0;
parentVar.type.descriptor.matrixByteStride = 0;
RDCASSERTMSG("Stride is too large for uint16_t", topLevelStride <= 0xffff);
parentVar.type.descriptor.arrayByteStride = RDCMIN((uint32_t)topLevelStride, 0xffffu) & 0xffff;
// consider all block-level SSBO structs to have infinite elements if they are an array at all
if(blockLevel && topLevelStride && isarray)
parentVar.type.descriptor.elements = ~0U;
if(!blockLevel)
topLevelStride = 0;
bool found = false;
// if we can find the base variable already, we recurse into its members
for(size_t i = 0; i < parentmembers->size(); i++)
{
if((*parentmembers)[i].name == base)
{
// if we find the variable, update the # elements to account for this new array index
// and pick the minimum offset of all of our children as the parent offset. This is mostly
// just for sorting
(*parentmembers)[i].type.descriptor.elements =
RDCMAX((*parentmembers)[i].type.descriptor.elements, parentVar.type.descriptor.elements);
(*parentmembers)[i].byteOffset = RDCMIN((*parentmembers)[i].byteOffset, parentVar.byteOffset);
parentmembers = &((*parentmembers)[i].type.members);
found = true;
blockLevel = false;
break;
}
}
// if we didn't find the base variable, add it and recuse inside
if(!found)
{
parentmembers->push_back(parentVar);
parentmembers = &(parentmembers->back().type.members);
}
if(multiDimArray)
{
// if this is a multi-dimensional array, we've now selected the root array.
// We now iterate all the way down to the last element, then break out of the list so it can
// be added as an array.
//
// Note: this means that for float foo[4][3] we won't iterate here - we've already selected
// foo, and outside of the loop we'll push back a float[3] for each of foo[0], foo[1], foo[2],
// foo[3] that we encounter in this iteration process.
//
// For bar[4][3][2] we've selected bar, we'll then push back a [4] member and then outside of
// the loop we'll push back each of bar[.][0], bar[.][1], etc as a float[2]
while(*nm)
{
parentVar.name = StringFormat::Fmt("[%d]", arrayIdx);
found = false;
for(size_t i = 0; i < parentmembers->size(); i++)
{
if((*parentmembers)[i].name == parentVar.name)
{
parentmembers = &((*parentmembers)[i].type.members);
found = true;
break;
}
}
if(!found)
{
parentmembers->push_back(parentVar);
parentmembers = &(parentmembers->back().type.members);
}
arrayIdx = 0;
RDCASSERT(*nm == '[');
nm++;
while(*nm >= '0' && *nm <= '9')
{
arrayIdx *= 10;
arrayIdx += int(*nm) - int('0');
nm++;
}
RDCASSERT(*nm == ']');
*nm = 0;
nm++;
}
break;
}
// the 0th element of each array fills out the actual members, when we
// encounter an index above that we only use it to increase the type.descriptor.elements
// member (which we've done by this point) and can stop recursing
//
// The exception is when we're looking at bare uniforms - there the struct members all have
// individual locations and are generally aggressively stripped by the driver.
// So then it's possible to get foo[0].bar.a and foo[1].bar.b - so higher indices can reveal
// more of the structure.
if(arrayIdx > 0 && !bareUniform)
{
parentmembers = NULL;
break;
}
}
if(parentmembers)
{
// if this is a bare uniform we need to be careful - just above we continued iterating for
// higher indices, because you can have cases that return e.g. foo[0].bar.a and foo[1].bar.b and
// get 'more information' from later indices, the full structure is not revealed under foo[0].
// However as a result of that it means we could see foo[0].bar.a and foo[1].bar.a - we need to
// be careful not to add the final 'a' twice. Check for duplicates and be sure it's really a
// duplicate.
bool duplicate = false;
// nm points into var.name's storage, so copy out to a temporary
std::string n = nm;
var.name = n;
if(bareUniform && !multiDimArray)
{
for(size_t i = 0; i < parentmembers->size(); i++)
{
if((*parentmembers)[i].name == var.name)
{
ShaderVariableDescriptor &oldtype = (*parentmembers)[i].type.descriptor;
ShaderVariableDescriptor &newtype = var.type.descriptor;
if(oldtype.rows != newtype.rows || oldtype.columns != newtype.columns ||
oldtype.type != newtype.type || oldtype.elements != newtype.elements)
{
RDCERR("When reconstructing %s, found duplicate but different final member %s",
fullname.c_str(), (*parentmembers)[i].name.c_str());
}
duplicate = true;
break;
}
}
}
if(duplicate)
return;
// for multidimensional arrays there will be no proper name, so name the variable by the index
if(multiDimArray)
var.name = StringFormat::Fmt("[%d]", arrayIdx);
parentmembers->push_back(var);
}
}
void MakeChildByteOffsetsRelative(ShaderConstant &member)
{
for(ShaderConstant &child : member.type.members)
{
MakeChildByteOffsetsRelative(child);
child.byteOffset -= member.byteOffset;
}
}
int ParseVersionStatement(const char *version)
{
if(strncmp(version, "#version", 8))
return 0;
version += 8;
while(isspace(*version))
version++;
int ret = 0;
while(*version >= '0' && *version <= '9')
{
ret *= 10;
ret += int(*version) - int('0');
version++;
}
return ret;
}
static void AddSigParameter(std::vector<SigParameter> &sigs, uint32_t &regIndex,
const SigParameter &sig, const char *nm, int rows, int arrayIdx)
{
if(rows == 1)
{
SigParameter s = sig;
if(s.regIndex == ~0U)
s.regIndex = regIndex++;
else if(arrayIdx >= 0)
s.regIndex += arrayIdx;
if(arrayIdx >= 0)
{
s.arrayIndex = arrayIdx;
s.varName = StringFormat::Fmt("%s[%d]", nm, arrayIdx);
}
sigs.push_back(s);
}
else
{
for(int r = 0; r < rows; r++)
{
SigParameter s = sig;
if(s.regIndex == ~0U)
s.regIndex = regIndex++;
else if(arrayIdx >= 0)
s.regIndex += rows * arrayIdx + r;
else
s.regIndex += r;
if(arrayIdx >= 0)
{
s.arrayIndex = arrayIdx;
s.varName = StringFormat::Fmt("%s[%d]:row%d", nm, arrayIdx, r);
}
else
{
s.varName = StringFormat::Fmt("%s:row%d", nm, r);
}
sigs.push_back(s);
}
}
}
void MakeShaderReflection(GLenum shadType, GLuint sepProg, ShaderReflection &refl,
const FixedFunctionVertexOutputs &outputUsage)
{
refl.stage = MakeShaderStage(shadType);
refl.entryPoint = "main";
refl.encoding = ShaderEncoding::GLSL;
refl.debugInfo.encoding = ShaderEncoding::GLSL;
if(shadType == eGL_COMPUTE_SHADER)
{
GL.glGetProgramiv(sepProg, eGL_COMPUTE_WORK_GROUP_SIZE, (GLint *)refl.dispatchThreadsDimension);
}
else
{
RDCEraseEl(refl.dispatchThreadsDimension);
}
rdcarray<ShaderResource> &roresources = refl.readOnlyResources;
rdcarray<ShaderResource> &rwresources = refl.readWriteResources;
GLint numUniforms = 0;
GL.glGetProgramInterfaceiv(sepProg, eGL_UNIFORM, eGL_ACTIVE_RESOURCES, &numUniforms);
const size_t numProps = 7;
GLenum resProps[numProps] = {
eGL_TYPE, eGL_NAME_LENGTH, eGL_LOCATION, eGL_BLOCK_INDEX,
eGL_ARRAY_SIZE, eGL_OFFSET, eGL_IS_ROW_MAJOR,
};
for(GLint u = 0; u < numUniforms; u++)
{
GLint values[numProps];
GL.glGetProgramResourceiv(sepProg, eGL_UNIFORM, u, numProps, resProps, numProps, NULL, values);
ShaderResource res;
res.isReadOnly = true;
res.isTexture = true;
res.variableType.descriptor.rows = 1;
res.variableType.descriptor.columns = 4;
res.variableType.descriptor.elements = 0;
res.variableType.descriptor.rowMajorStorage = false;
res.variableType.descriptor.arrayByteStride = 0;
res.variableType.descriptor.matrixByteStride = 0;
// float samplers
if(values[0] == eGL_SAMPLER_BUFFER)
{
res.resType = TextureType::Buffer;
res.variableType.descriptor.name = "samplerBuffer";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_1D)
{
res.resType = TextureType::Texture1D;
res.variableType.descriptor.name = "sampler1D";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_1D_ARRAY)
{
res.resType = TextureType::Texture1DArray;
res.variableType.descriptor.name = "sampler1DArray";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_1D_SHADOW)
{
res.resType = TextureType::Texture1D;
res.variableType.descriptor.name = "sampler1DShadow";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_1D_ARRAY_SHADOW)
{
res.resType = TextureType::Texture1DArray;
res.variableType.descriptor.name = "sampler1DArrayShadow";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D)
{
res.resType = TextureType::Texture2D;
res.variableType.descriptor.name = "sampler2D";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D_ARRAY)
{
res.resType = TextureType::Texture2DArray;
res.variableType.descriptor.name = "sampler2DArray";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D_SHADOW)
{
res.resType = TextureType::Texture2D;
res.variableType.descriptor.name = "sampler2DShadow";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D_ARRAY_SHADOW)
{
res.resType = TextureType::Texture2DArray;
res.variableType.descriptor.name = "sampler2DArrayShadow";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D_RECT)
{
res.resType = TextureType::TextureRect;
res.variableType.descriptor.name = "sampler2DRect";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D_RECT_SHADOW)
{
res.resType = TextureType::TextureRect;
res.variableType.descriptor.name = "sampler2DRectShadow";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_3D)
{
res.resType = TextureType::Texture3D;
res.variableType.descriptor.name = "sampler3D";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_CUBE)
{
res.resType = TextureType::TextureCube;
res.variableType.descriptor.name = "samplerCube";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_CUBE_SHADOW)
{
res.resType = TextureType::TextureCube;
res.variableType.descriptor.name = "samplerCubeShadow";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_CUBE_MAP_ARRAY)
{
res.resType = TextureType::TextureCubeArray;
res.variableType.descriptor.name = "samplerCubeArray";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D_MULTISAMPLE)
{
res.resType = TextureType::Texture2DMS;
res.variableType.descriptor.name = "sampler2DMS";
res.variableType.descriptor.type = VarType::Float;
}
else if(values[0] == eGL_SAMPLER_2D_MULTISAMPLE_ARRAY)
{
res.resType = TextureType::Texture2DMSArray;
res.variableType.descriptor.name = "sampler2DMSArray";
res.variableType.descriptor.type = VarType::Float;
}
// int samplers
else if(values[0] == eGL_INT_SAMPLER_BUFFER)
{
res.resType = TextureType::Buffer;
res.variableType.descriptor.name = "isamplerBuffer";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_1D)
{
res.resType = TextureType::Texture1D;
res.variableType.descriptor.name = "isampler1D";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_1D_ARRAY)
{
res.resType = TextureType::Texture1DArray;
res.variableType.descriptor.name = "isampler1DArray";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_2D)
{
res.resType = TextureType::Texture2D;
res.variableType.descriptor.name = "isampler2D";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_2D_ARRAY)
{
res.resType = TextureType::Texture2DArray;
res.variableType.descriptor.name = "isampler2DArray";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_2D_RECT)
{
res.resType = TextureType::TextureRect;
res.variableType.descriptor.name = "isampler2DRect";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_3D)
{
res.resType = TextureType::Texture3D;
res.variableType.descriptor.name = "isampler3D";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_CUBE)
{
res.resType = TextureType::TextureCube;
res.variableType.descriptor.name = "isamplerCube";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_CUBE_MAP_ARRAY)
{
res.resType = TextureType::TextureCubeArray;
res.variableType.descriptor.name = "isamplerCubeArray";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_2D_MULTISAMPLE)
{
res.resType = TextureType::Texture2DMS;
res.variableType.descriptor.name = "isampler2DMS";
res.variableType.descriptor.type = VarType::SInt;
}
else if(values[0] == eGL_INT_SAMPLER_2D_MULTISAMPLE_ARRAY)
{
res.resType = TextureType::Texture2DMSArray;
res.variableType.descriptor.name = "isampler2DMSArray";
res.variableType.descriptor.type = VarType::SInt;
}
// unsigned int samplers
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_BUFFER)
{
res.resType = TextureType::Buffer;
res.variableType.descriptor.name = "usamplerBuffer";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_1D)
{
res.resType = TextureType::Texture1D;
res.variableType.descriptor.name = "usampler1D";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_1D_ARRAY)
{
res.resType = TextureType::Texture1DArray;
res.variableType.descriptor.name = "usampler1DArray";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_2D)
{
res.resType = TextureType::Texture2D;
res.variableType.descriptor.name = "usampler2D";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_2D_ARRAY)
{
res.resType = TextureType::Texture2DArray;
res.variableType.descriptor.name = "usampler2DArray";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_2D_RECT)
{
res.resType = TextureType::TextureRect;
res.variableType.descriptor.name = "usampler2DRect";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_3D)
{
res.resType = TextureType::Texture3D;
res.variableType.descriptor.name = "usampler3D";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_CUBE)
{
res.resType = TextureType::TextureCube;
res.variableType.descriptor.name = "usamplerCube";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_CUBE_MAP_ARRAY)
{
res.resType = TextureType::TextureCubeArray;
res.variableType.descriptor.name = "usamplerCubeArray";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE)
{
res.resType = TextureType::Texture2DMS;
res.variableType.descriptor.name = "usampler2DMS";
res.variableType.descriptor.type = VarType::UInt;
}
else if(values[0] == eGL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE_ARRAY)
{
res.resType = TextureType::Texture2DMSArray;
res.variableType.descriptor.name = "usampler2DMSArray";
res.variableType.descriptor.type = VarType::UInt;
}
// float images
else if(values[0] == eGL_IMAGE_BUFFER)
{
res.resType = TextureType::Buffer;
res.variableType.descriptor.name = "imageBuffer";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_1D)
{
res.resType = TextureType::Texture1D;
res.variableType.descriptor.name = "image1D";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_1D_ARRAY)
{
res.resType = TextureType::Texture1DArray;
res.variableType.descriptor.name = "image1DArray";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_2D)
{
res.resType = TextureType::Texture2D;
res.variableType.descriptor.name = "image2D";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_2D_ARRAY)
{
res.resType = TextureType::Texture2DArray;
res.variableType.descriptor.name = "image2DArray";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_2D_RECT)
{
res.resType = TextureType::TextureRect;
res.variableType.descriptor.name = "image2DRect";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_3D)
{
res.resType = TextureType::Texture3D;
res.variableType.descriptor.name = "image3D";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_CUBE)
{
res.resType = TextureType::TextureCube;
res.variableType.descriptor.name = "imageCube";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_CUBE_MAP_ARRAY)
{
res.resType = TextureType::TextureCubeArray;
res.variableType.descriptor.name = "imageCubeArray";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_2D_MULTISAMPLE)
{
res.resType = TextureType::Texture2DMS;
res.variableType.descriptor.name = "image2DMS";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
else if(values[0] == eGL_IMAGE_2D_MULTISAMPLE_ARRAY)
{
res.resType = TextureType::Texture2DMSArray;
res.variableType.descriptor.name = "image2DMSArray";
res.variableType.descriptor.type = VarType::Float;
res.isReadOnly = false;
}
// int images
else if(values[0] == eGL_INT_IMAGE_BUFFER)
{
res.resType = TextureType::Buffer;
res.variableType.descriptor.name = "iimageBuffer";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_1D)
{
res.resType = TextureType::Texture1D;
res.variableType.descriptor.name = "iimage1D";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_1D_ARRAY)
{
res.resType = TextureType::Texture1DArray;
res.variableType.descriptor.name = "iimage1DArray";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_2D)
{
res.resType = TextureType::Texture2D;
res.variableType.descriptor.name = "iimage2D";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_2D_ARRAY)
{
res.resType = TextureType::Texture2DArray;
res.variableType.descriptor.name = "iimage2DArray";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_2D_RECT)
{
res.resType = TextureType::TextureRect;
res.variableType.descriptor.name = "iimage2DRect";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_3D)
{
res.resType = TextureType::Texture3D;
res.variableType.descriptor.name = "iimage3D";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_CUBE)
{
res.resType = TextureType::TextureCube;
res.variableType.descriptor.name = "iimageCube";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_CUBE_MAP_ARRAY)
{
res.resType = TextureType::TextureCubeArray;
res.variableType.descriptor.name = "iimageCubeArray";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_2D_MULTISAMPLE)
{
res.resType = TextureType::Texture2DMS;
res.variableType.descriptor.name = "iimage2DMS";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_INT_IMAGE_2D_MULTISAMPLE_ARRAY)
{
res.resType = TextureType::Texture2DMSArray;
res.variableType.descriptor.name = "iimage2DMSArray";
res.variableType.descriptor.type = VarType::SInt;
res.isReadOnly = false;
}
// unsigned int images
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_BUFFER)
{
res.resType = TextureType::Buffer;
res.variableType.descriptor.name = "uimageBuffer";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_1D)
{
res.resType = TextureType::Texture1D;
res.variableType.descriptor.name = "uimage1D";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_1D_ARRAY)
{
res.resType = TextureType::Texture1DArray;
res.variableType.descriptor.name = "uimage1DArray";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_2D)
{
res.resType = TextureType::Texture2D;
res.variableType.descriptor.name = "uimage2D";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_2D_ARRAY)
{
res.resType = TextureType::Texture2DArray;
res.variableType.descriptor.name = "uimage2DArray";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_2D_RECT)
{
res.resType = TextureType::TextureRect;
res.variableType.descriptor.name = "uimage2DRect";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_3D)
{
res.resType = TextureType::Texture3D;
res.variableType.descriptor.name = "uimage3D";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_CUBE)
{
res.resType = TextureType::TextureCube;
res.variableType.descriptor.name = "uimageCube";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_CUBE_MAP_ARRAY)
{
res.resType = TextureType::TextureCubeArray;
res.variableType.descriptor.name = "uimageCubeArray";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_2D_MULTISAMPLE)
{
res.resType = TextureType::Texture2DMS;
res.variableType.descriptor.name = "uimage2DMS";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
else if(values[0] == eGL_UNSIGNED_INT_IMAGE_2D_MULTISAMPLE_ARRAY)
{
res.resType = TextureType::Texture2DMSArray;
res.variableType.descriptor.name = "uimage2DMSArray";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
}
// atomic counter
else if(values[0] == eGL_UNSIGNED_INT_ATOMIC_COUNTER)
{
res.resType = TextureType::Buffer;
res.variableType.descriptor.name = "atomic_uint";
res.variableType.descriptor.type = VarType::UInt;
res.isReadOnly = false;
res.isTexture = false;
res.variableType.descriptor.columns = 1;
}
else
{
// not a sampler
continue;
}
char *namebuf = new char[values[1] + 1];
GL.glGetProgramResourceName(sepProg, eGL_UNIFORM, u, values[1], NULL, namebuf);
namebuf[values[1]] = 0;
std::string name = namebuf;
delete[] namebuf;
res.name = name;
rdcarray<ShaderResource> &reslist = (res.isReadOnly ? roresources : rwresources);
res.bindPoint = (int32_t)reslist.size();
reslist.push_back(res);
// array of samplers
if(values[4] > 1)
{
name = name.substr(0, name.length() - 3); // trim off [0] on the end
for(int i = 1; i < values[4]; i++)
{
std::string arrname = StringFormat::Fmt("%s[%d]", name.c_str(), i);
res.bindPoint = (int32_t)reslist.size();
res.name = arrname;
reslist.push_back(res);
}
}
}
std::vector<int32_t> ssbos;
uint32_t ssboMembers = 0;
GLint numSSBOs = 0;
if(HasExt[ARB_shader_storage_buffer_object])
{
GL.glGetProgramInterfaceiv(sepProg, eGL_SHADER_STORAGE_BLOCK, eGL_ACTIVE_RESOURCES, &numSSBOs);
for(GLint u = 0; u < numSSBOs; u++)
{
GLenum propName = eGL_NAME_LENGTH;
GLint len;
GL.glGetProgramResourceiv(sepProg, eGL_SHADER_STORAGE_BLOCK, u, 1, &propName, 1, NULL, &len);
char *nm = new char[len + 1];
GL.glGetProgramResourceName(sepProg, eGL_SHADER_STORAGE_BLOCK, u, len + 1, NULL, nm);
ShaderResource res;
res.isReadOnly = false;
res.isTexture = false;
res.resType = TextureType::Buffer;
res.variableType.descriptor.rows = 0;
res.variableType.descriptor.columns = 0;
res.variableType.descriptor.elements = 0;
res.variableType.descriptor.rowMajorStorage = false;
res.variableType.descriptor.arrayByteStride = 0;
res.variableType.descriptor.matrixByteStride = 0;
res.variableType.descriptor.name = "buffer";
res.variableType.descriptor.type = VarType::UInt;
res.bindPoint = (int32_t)rwresources.size();
res.name = nm;
GLint numMembers = 0;
propName = eGL_NUM_ACTIVE_VARIABLES;
GL.glGetProgramResourceiv(sepProg, eGL_SHADER_STORAGE_BLOCK, u, 1, &propName, 1, NULL,
(GLint *)&numMembers);
char *arr = strchr(nm, '[');
const bool isArray = (arr != NULL);
uint32_t arrayIdx = 0;
if(isArray)
{
arr++;
while(*arr >= '0' && *arr <= '9')
{
arrayIdx *= 10;
arrayIdx += int(*arr) - int('0');
arr++;
}
}
rwresources.push_back(res);
ssbos.push_back(res.bindPoint);
// only count members from the first array index
if(!isArray || arrayIdx == 0)
ssboMembers += numMembers;
delete[] nm;
}
}
{
rdcarray<ShaderConstant> *members = new rdcarray<ShaderConstant>[ssbos.size()];
for(uint32_t i = 0; i < ssboMembers; i++)
{
ReconstructVarTree(eGL_BUFFER_VARIABLE, sepProg, i, (GLint)ssbos.size(), members, NULL);
}
// if we have members in an array ssbo, broadcast this to any ssbo with the same basename
// without members, since the variables will only be reported as belonging to one block
for(size_t ssbo = 0; ssbo < ssbos.size(); ssbo++)
{
rdcstr basename = rwresources[ssbos[ssbo]].name;
int arrOffs = basename.indexOf('[');
if(arrOffs > 0)
{
basename.erase(arrOffs, basename.size());
if(!members[ssbo].empty())
{
for(size_t ssbo2 = 0; ssbo2 < ssbos.size(); ssbo2++)
{
if(!members[ssbo2].empty())
continue;
rdcstr basename2 = rwresources[ssbos[ssbo2]].name;
arrOffs = basename2.indexOf('[');
if(arrOffs > 0)
{
basename2.erase(arrOffs, basename2.size());
if(basename == basename2)
{
members[ssbo2] = members[ssbo];
}
}
}
}
}
}
for(size_t ssbo = 0; ssbo < ssbos.size(); ssbo++)
{
sort(members[ssbo]);
rdcstr basename = rwresources[ssbos[ssbo]].name;
int arrOffs = basename.indexOf('[');
if(arrOffs > 0)
basename.erase(arrOffs, basename.size());
if(!members[ssbo].empty() && basename == members[ssbo][0].name)
std::swap(rwresources[ssbos[ssbo]].variableType.members, members[ssbo][0].type.members);
else
std::swap(rwresources[ssbos[ssbo]].variableType.members, members[ssbo]);
}
// patch-up reflection data. For top-level arrays use the stride & rough size to calculate the
// number of elements, and make all child byteOffset values relative to their parent
for(size_t ssbo = 0; ssbo < ssbos.size(); ssbo++)
{
rdcarray<ShaderConstant> &ssboVars = rwresources[ssbos[ssbo]].variableType.members;
for(size_t rootMember = 0; rootMember + 1 < ssboVars.size(); rootMember++)
{
ShaderConstant &member = ssboVars[rootMember];
const uint32_t memberSizeBound = ssboVars[rootMember + 1].byteOffset - member.byteOffset;
const uint32_t stride = member.type.descriptor.arrayByteStride;
if(stride != 0 && member.type.descriptor.elements <= 1 && memberSizeBound > 2 * stride)
{
member.type.descriptor.elements = memberSizeBound / stride;
}
}
for(ShaderConstant &member : ssboVars)
MakeChildByteOffsetsRelative(member);
}
delete[] members;
}
rdcarray<ShaderConstant> globalUniforms;
GLint numUBOs = 0;
std::vector<std::string> uboNames;
rdcarray<ShaderConstant> *ubos = NULL;
{
GL.glGetProgramInterfaceiv(sepProg, eGL_UNIFORM_BLOCK, eGL_ACTIVE_RESOURCES, &numUBOs);
ubos = new rdcarray<ShaderConstant>[numUBOs];
uboNames.resize(numUBOs);
for(GLint u = 0; u < numUBOs; u++)
{
GLenum nameLen = eGL_NAME_LENGTH;
GLint len;
GL.glGetProgramResourceiv(sepProg, eGL_UNIFORM_BLOCK, u, 1, &nameLen, 1, NULL, &len);
char *nm = new char[len + 1];
GL.glGetProgramResourceName(sepProg, eGL_UNIFORM_BLOCK, u, len + 1, NULL, nm);
uboNames[u] = nm;
delete[] nm;
}
}
for(GLint u = 0; u < numUniforms; u++)
{
ReconstructVarTree(eGL_UNIFORM, sepProg, u, numUBOs, ubos, &globalUniforms);
}
refl.constantBlocks.reserve(numUBOs + (globalUniforms.empty() ? 0 : 1));
if(ubos)
{
for(int i = 0; i < numUBOs; i++)
{
if(!ubos[i].empty())
{
ConstantBlock cblock;
cblock.name = uboNames[i];
cblock.bufferBacked = true;
cblock.bindPoint = (int32_t)refl.constantBlocks.size();
GLenum bufSize = eGL_BUFFER_DATA_SIZE;
GL.glGetProgramResourceiv(sepProg, eGL_UNIFORM_BLOCK, i, 1, &bufSize, 1, NULL,
(GLint *)&cblock.byteSize);
sort(ubos[i]);
for(ShaderConstant &member : ubos[i])
MakeChildByteOffsetsRelative(member);
std::swap(cblock.variables, ubos[i]);
refl.constantBlocks.push_back(cblock);
}
}
}
if(!globalUniforms.empty())
{
ConstantBlock globals;
globals.name = "$Globals";
globals.bufferBacked = false;
globals.bindPoint = (int32_t)refl.constantBlocks.size();
// global uniforms have no defined order, location will be per implementation, so sort instead
// alphabetically
namesort(globalUniforms);
std::swap(globals.variables, globalUniforms);
refl.constantBlocks.push_back(globals);
}
delete[] ubos;
for(int sigType = 0; sigType < 2; sigType++)
{
GLenum sigEnum = (sigType == 0 ? eGL_PROGRAM_INPUT : eGL_PROGRAM_OUTPUT);
rdcarray<SigParameter> *sigArray = (sigType == 0 ? &refl.inputSignature : &refl.outputSignature);
GLint numInputs;
GL.glGetProgramInterfaceiv(sepProg, sigEnum, eGL_ACTIVE_RESOURCES, &numInputs);
if(numInputs > 0)
{
std::vector<SigParameter> sigs;
sigs.reserve(numInputs);
uint32_t regIndex = 0;
for(GLint i = 0; i < numInputs; i++)
{
GLenum props[] = {eGL_NAME_LENGTH, eGL_TYPE, eGL_LOCATION, eGL_ARRAY_SIZE,
eGL_LOCATION_COMPONENT};
GLint values[] = {0, 0, 0, 0, 0};
GLsizei numSigProps = (GLsizei)ARRAY_COUNT(props);
// GL_LOCATION_COMPONENT not supported on core <4.4 (or without GL_ARB_enhanced_layouts)
// on GLES, or when we don't have native program interface query
if(!HasExt[ARB_enhanced_layouts] || !HasExt[ARB_program_interface_query])
numSigProps--;
GL.glGetProgramResourceiv(sepProg, sigEnum, i, numSigProps, props, numSigProps, NULL, values);
char *nm = new char[values[0] + 1];
GL.glGetProgramResourceName(sepProg, sigEnum, i, values[0] + 1, NULL, nm);
SigParameter sig;
sig.varName = nm;
sig.semanticIndex = 0;
sig.needSemanticIndex = false;
sig.stream = 0;
int rows = 1;
switch(values[1])
{
case eGL_DOUBLE:
case eGL_DOUBLE_VEC2:
case eGL_DOUBLE_VEC3:
case eGL_DOUBLE_VEC4:
case eGL_DOUBLE_MAT4:
case eGL_DOUBLE_MAT4x3:
case eGL_DOUBLE_MAT4x2:
case eGL_DOUBLE_MAT3:
case eGL_DOUBLE_MAT3x4:
case eGL_DOUBLE_MAT3x2:
case eGL_DOUBLE_MAT2:
case eGL_DOUBLE_MAT2x3:
case eGL_DOUBLE_MAT2x4: sig.compType = CompType::Double; break;
case eGL_FLOAT:
case eGL_FLOAT_VEC2:
case eGL_FLOAT_VEC3:
case eGL_FLOAT_VEC4:
case eGL_FLOAT_MAT4:
case eGL_FLOAT_MAT4x3:
case eGL_FLOAT_MAT4x2:
case eGL_FLOAT_MAT3:
case eGL_FLOAT_MAT3x4:
case eGL_FLOAT_MAT3x2:
case eGL_FLOAT_MAT2:
case eGL_FLOAT_MAT2x3:
case eGL_FLOAT_MAT2x4: sig.compType = CompType::Float; break;
case eGL_INT:
case eGL_INT_VEC2:
case eGL_INT_VEC3:
case eGL_INT_VEC4: sig.compType = CompType::SInt; break;
case eGL_UNSIGNED_INT:
case eGL_BOOL:
case eGL_UNSIGNED_INT_VEC2:
case eGL_BOOL_VEC2:
case eGL_UNSIGNED_INT_VEC3:
case eGL_BOOL_VEC3:
case eGL_UNSIGNED_INT_VEC4:
case eGL_BOOL_VEC4: sig.compType = CompType::UInt; break;
default:
sig.compType = CompType::Float;
RDCWARN("Unhandled signature element type %s", ToStr((GLenum)values[1]).c_str());
}
switch(values[1])
{
case eGL_FLOAT:
case eGL_DOUBLE:
case eGL_INT:
case eGL_UNSIGNED_INT:
case eGL_BOOL:
sig.compCount = 1;
sig.regChannelMask = 0x1;
break;
case eGL_FLOAT_VEC2:
case eGL_DOUBLE_VEC2:
case eGL_INT_VEC2:
case eGL_UNSIGNED_INT_VEC2:
case eGL_BOOL_VEC2:
sig.compCount = 2;
sig.regChannelMask = 0x3;
break;
case eGL_FLOAT_VEC3:
case eGL_DOUBLE_VEC3:
case eGL_INT_VEC3:
case eGL_UNSIGNED_INT_VEC3:
case eGL_BOOL_VEC3:
sig.compCount = 3;
sig.regChannelMask = 0x7;
break;
case eGL_FLOAT_VEC4:
case eGL_DOUBLE_VEC4:
case eGL_INT_VEC4:
case eGL_UNSIGNED_INT_VEC4:
case eGL_BOOL_VEC4:
sig.compCount = 4;
sig.regChannelMask = 0xf;
break;
case eGL_FLOAT_MAT4:
case eGL_DOUBLE_MAT4:
sig.compCount = 4;
rows = 4;
sig.regChannelMask = 0xf;
break;
case eGL_FLOAT_MAT4x3:
case eGL_DOUBLE_MAT4x3:
sig.compCount = 4;
rows = 3;
sig.regChannelMask = 0xf;
break;
case eGL_FLOAT_MAT4x2:
case eGL_DOUBLE_MAT4x2:
sig.compCount = 4;
rows = 2;
sig.regChannelMask = 0xf;
break;
case eGL_FLOAT_MAT3:
case eGL_DOUBLE_MAT3:
sig.compCount = 3;
rows = 3;
sig.regChannelMask = 0x7;
break;
case eGL_FLOAT_MAT3x4:
case eGL_DOUBLE_MAT3x4:
sig.compCount = 3;
rows = 2;
sig.regChannelMask = 0x7;
break;
case eGL_FLOAT_MAT3x2:
case eGL_DOUBLE_MAT3x2:
sig.compCount = 3;
rows = 2;
sig.regChannelMask = 0x7;
break;
case eGL_FLOAT_MAT2:
case eGL_DOUBLE_MAT2:
sig.compCount = 2;
rows = 2;
sig.regChannelMask = 0x3;
break;
case eGL_FLOAT_MAT2x3:
case eGL_DOUBLE_MAT2x3:
sig.compCount = 2;
rows = 3;
sig.regChannelMask = 0x3;
break;
case eGL_FLOAT_MAT2x4:
case eGL_DOUBLE_MAT2x4:
sig.compCount = 2;
rows = 4;
sig.regChannelMask = 0x3;
break;
default:
RDCWARN("Unhandled signature element type %s", ToStr((GLenum)values[1]).c_str());
sig.compCount = 4;
sig.regChannelMask = 0xf;
break;
}
sig.systemValue = ShaderBuiltin::Undefined;
const char *varname = nm;
if(!strncmp(varname, "gl_PerVertex.", 13))
varname += 13;
#define IS_BUILTIN(builtin) !strncmp(varname, builtin, sizeof(builtin) - 1)
// some vertex outputs can be reflected (especially by glslang) if they're just declared and
// not used, which is quite common with redeclaring outputs for separable programs - either
// by the program or by us. So instead use our manual quick-and-dirty usage check to skip
// potential false-positives.
bool unused = false;
for(FFVertexOutput ffoutput : ::values<FFVertexOutput>())
{
// we consider an output used if we encounter a '=' before either a ';' or the end of the
// string
std::string outName = ToStr(ffoutput);
// we do a substring search so that gl_ClipDistance matches gl_ClipDistance[0]
if(strstr(varname, outName.c_str()))
{
unused = !outputUsage.used[(int)ffoutput];
break;
}
}
if(unused)
{
delete[] nm;
continue;
}
// VS built-in inputs
if(IS_BUILTIN("gl_VertexID"))
sig.systemValue = ShaderBuiltin::VertexIndex;
if(IS_BUILTIN("gl_InstanceID"))
sig.systemValue = ShaderBuiltin::InstanceIndex;
// VS built-in outputs
if(IS_BUILTIN("gl_Position"))
sig.systemValue = ShaderBuiltin::Position;
if(IS_BUILTIN("gl_PointSize"))
sig.systemValue = ShaderBuiltin::PointSize;
if(IS_BUILTIN("gl_ClipDistance"))
sig.systemValue = ShaderBuiltin::ClipDistance;
// TCS built-in inputs
if(IS_BUILTIN("gl_PatchVerticesIn"))
sig.systemValue = ShaderBuiltin::PatchNumVertices;
if(IS_BUILTIN("gl_PrimitiveID"))
sig.systemValue = ShaderBuiltin::PrimitiveIndex;
if(IS_BUILTIN("gl_InvocationID"))
sig.systemValue = ShaderBuiltin::OutputControlPointIndex;
// TCS built-in outputs
if(IS_BUILTIN("gl_TessLevelOuter"))
sig.systemValue = ShaderBuiltin::OuterTessFactor;
if(IS_BUILTIN("gl_TessLevelInner"))
sig.systemValue = ShaderBuiltin::InsideTessFactor;
// TES built-in inputs
if(IS_BUILTIN("gl_TessCoord"))
sig.systemValue = ShaderBuiltin::DomainLocation;
if(IS_BUILTIN("gl_PatchVerticesIn"))
sig.systemValue = ShaderBuiltin::PatchNumVertices;
if(IS_BUILTIN("gl_PrimitiveID"))
sig.systemValue = ShaderBuiltin::PrimitiveIndex;
// GS built-in inputs
if(IS_BUILTIN("gl_PrimitiveIDIn"))
sig.systemValue = ShaderBuiltin::PrimitiveIndex;
if(IS_BUILTIN("gl_InvocationID") && shadType == eGL_GEOMETRY_SHADER)
sig.systemValue = ShaderBuiltin::GSInstanceIndex;
if(IS_BUILTIN("gl_Layer"))
sig.systemValue = ShaderBuiltin::RTIndex;
if(IS_BUILTIN("gl_ViewportIndex"))
sig.systemValue = ShaderBuiltin::ViewportIndex;
// GS built-in outputs
if(IS_BUILTIN("gl_Layer"))
sig.systemValue = ShaderBuiltin::RTIndex;
if(IS_BUILTIN("gl_ViewportIndex"))
sig.systemValue = ShaderBuiltin::ViewportIndex;
// PS built-in inputs
if(IS_BUILTIN("gl_FragCoord"))
sig.systemValue = ShaderBuiltin::Position;
if(IS_BUILTIN("gl_FrontFacing"))
sig.systemValue = ShaderBuiltin::IsFrontFace;
if(IS_BUILTIN("gl_PointCoord"))
sig.systemValue = ShaderBuiltin::RTIndex;
if(IS_BUILTIN("gl_SampleID"))
sig.systemValue = ShaderBuiltin::MSAASampleIndex;
if(IS_BUILTIN("gl_SamplePosition"))
sig.systemValue = ShaderBuiltin::MSAASamplePosition;
if(IS_BUILTIN("gl_SampleMaskIn"))
sig.systemValue = ShaderBuiltin::MSAACoverage;
// PS built-in outputs
if(IS_BUILTIN("gl_FragDepth"))
sig.systemValue = ShaderBuiltin::DepthOutput;
if(IS_BUILTIN("gl_SampleMask"))
sig.systemValue = ShaderBuiltin::MSAACoverage;
if(IS_BUILTIN("gl_FragStencilRefARB"))
sig.systemValue = ShaderBuiltin::StencilReference;
// CS built-in inputs
if(IS_BUILTIN("gl_NumWorkGroups"))
sig.systemValue = ShaderBuiltin::DispatchSize;
if(IS_BUILTIN("gl_WorkGroupID"))
sig.systemValue = ShaderBuiltin::GroupIndex;
if(IS_BUILTIN("gl_LocalInvocationID"))
sig.systemValue = ShaderBuiltin::GroupThreadIndex;
if(IS_BUILTIN("gl_GlobalInvocationID"))
sig.systemValue = ShaderBuiltin::DispatchThreadIndex;
if(IS_BUILTIN("gl_LocalInvocationIndex"))
sig.systemValue = ShaderBuiltin::GroupFlatIndex;
#undef IS_BUILTIN
if(sig.systemValue == ShaderBuiltin::Undefined)
sig.regIndex = values[2] >= 0 ? values[2] : ~0U;
else
sig.regIndex = 0;
if(shadType == eGL_FRAGMENT_SHADER && sigEnum == eGL_PROGRAM_OUTPUT &&
sig.systemValue == ShaderBuiltin::Undefined)
sig.systemValue = ShaderBuiltin::ColorOutput;
// don't apply location component for built-ins
if(sig.systemValue == ShaderBuiltin::Undefined)
sig.regChannelMask <<= values[4];
sig.channelUsedMask = sig.regChannelMask;
if(values[3] <= 1)
{
AddSigParameter(sigs, regIndex, sig, nm, rows, -1);
}
else
{
std::string basename = nm;
if(basename[basename.size() - 3] == '[' && basename[basename.size() - 2] == '0' &&
basename[basename.size() - 1] == ']')
{
basename.erase(basename.size() - 3);
for(int a = 0; a < values[3]; a++)
AddSigParameter(sigs, regIndex, sig, basename.c_str(), rows, a);
}
else
{
RDCWARN("Got signature parameter %s with array size %d but no [0] suffix", nm, values[3]);
AddSigParameter(sigs, regIndex, sig, nm, rows, -1);
}
}
delete[] nm;
}
struct sig_param_sort
{
bool operator()(const SigParameter &a, const SigParameter &b)
{
if(a.systemValue == b.systemValue)
{
if(a.regIndex != b.regIndex)
return a.regIndex < b.regIndex;
return a.varName < b.varName;
}
if(a.systemValue == ShaderBuiltin::Undefined)
return false;
if(b.systemValue == ShaderBuiltin::Undefined)
return true;
return a.systemValue < b.systemValue;
}
};
std::sort(sigs.begin(), sigs.end(), sig_param_sort());
*sigArray = sigs;
}
}
// TODO: fill in Interfaces with shader subroutines?
}
void GetBindpointMapping(GLuint curProg, int shadIdx, const ShaderReflection *refl,
ShaderBindpointMapping &mapping)
{
if(!refl)
{
mapping = ShaderBindpointMapping();
return;
}
// in case of bugs, we readback into this array instead of a single int
GLint dummyReadback[32];
#if ENABLED(RDOC_DEVEL)
for(size_t i = 1; i < ARRAY_COUNT(dummyReadback); i++)
dummyReadback[i] = 0x6c7b8a9d;
#endif
const GLenum refEnum[] = {
eGL_REFERENCED_BY_VERTEX_SHADER, eGL_REFERENCED_BY_TESS_CONTROL_SHADER,
eGL_REFERENCED_BY_TESS_EVALUATION_SHADER, eGL_REFERENCED_BY_GEOMETRY_SHADER,
eGL_REFERENCED_BY_FRAGMENT_SHADER, eGL_REFERENCED_BY_COMPUTE_SHADER,
};
mapping.readOnlyResources.resize(refl->readOnlyResources.size());
for(size_t i = 0; i < refl->readOnlyResources.size(); i++)
{
if(refl->readOnlyResources[i].isTexture)
{
// normal sampler or image load/store
GLint loc = GL.glGetUniformLocation(curProg, refl->readOnlyResources[i].name.c_str());
if(loc >= 0)
{
GL.glGetUniformiv(curProg, loc, dummyReadback);
mapping.readOnlyResources[i].bindset = 0;
mapping.readOnlyResources[i].bind = dummyReadback[0];
mapping.readOnlyResources[i].arraySize = 1;
}
// handle sampler arrays, use the base name
std::string name = refl->readOnlyResources[i].name.c_str();
if(name.back() == ']')
{
do
{
name.pop_back();
} while(name.back() != '[');
name.pop_back();
}
GLuint idx = 0;
idx = GL.glGetProgramResourceIndex(curProg, eGL_UNIFORM, name.c_str());
if(idx == GL_INVALID_INDEX)
{
mapping.readOnlyResources[i].used = false;
}
else
{
GLint used = 0;
GL.glGetProgramResourceiv(curProg, eGL_UNIFORM, idx, 1, &refEnum[shadIdx], 1, NULL, &used);
mapping.readOnlyResources[i].used = (used != 0);
}
}
else
{
mapping.readOnlyResources[i].bindset = -1;
mapping.readOnlyResources[i].bind = -1;
mapping.readOnlyResources[i].used = false;
mapping.readOnlyResources[i].arraySize = 1;
}
}
mapping.readWriteResources.resize(refl->readWriteResources.size());
for(size_t i = 0; i < refl->readWriteResources.size(); i++)
{
if(refl->readWriteResources[i].isTexture)
{
// image load/store
GLint loc = GL.glGetUniformLocation(curProg, refl->readWriteResources[i].name.c_str());
if(loc >= 0)
{
GL.glGetUniformiv(curProg, loc, dummyReadback);
mapping.readWriteResources[i].bindset = 0;
mapping.readWriteResources[i].bind = dummyReadback[0];
mapping.readWriteResources[i].arraySize = 1;
}
// handle sampler arrays, use the base name
std::string name = refl->readWriteResources[i].name.c_str();
if(name.back() == ']')
{
do
{
name.pop_back();
} while(name.back() != '[');
name.pop_back();
}
GLuint idx = 0;
idx = GL.glGetProgramResourceIndex(curProg, eGL_UNIFORM, name.c_str());
if(idx == GL_INVALID_INDEX)
{
mapping.readWriteResources[i].used = false;
}
else
{
GLint used = 0;
GL.glGetProgramResourceiv(curProg, eGL_UNIFORM, idx, 1, &refEnum[shadIdx], 1, NULL, &used);
mapping.readWriteResources[i].used = (used != 0);
}
}
else if(!refl->readWriteResources[i].isTexture)
{
if(refl->readWriteResources[i].variableType.descriptor.columns == 1 &&
refl->readWriteResources[i].variableType.descriptor.rows == 1 &&
refl->readWriteResources[i].variableType.descriptor.type == VarType::UInt)
{
// atomic uint
GLuint idx = GL.glGetProgramResourceIndex(curProg, eGL_UNIFORM,
refl->readWriteResources[i].name.c_str());
if(idx == GL_INVALID_INDEX)
{
mapping.readWriteResources[i].bindset = -1;
mapping.readWriteResources[i].bind = -1;
mapping.readWriteResources[i].used = false;
mapping.readWriteResources[i].arraySize = 1;
}
else
{
GLenum prop = eGL_ATOMIC_COUNTER_BUFFER_INDEX;
GLuint atomicIndex;
GL.glGetProgramResourceiv(curProg, eGL_UNIFORM, idx, 1, &prop, 1, NULL,
(GLint *)&atomicIndex);
if(atomicIndex == GL_INVALID_INDEX)
{
mapping.readWriteResources[i].bindset = -1;
mapping.readWriteResources[i].bind = -1;
mapping.readWriteResources[i].used = false;
mapping.readWriteResources[i].arraySize = 1;
}
else
{
if(IsGLES)
{
prop = eGL_BUFFER_BINDING;
mapping.readWriteResources[i].bindset = 0;
GL.glGetProgramResourceiv(curProg, eGL_ATOMIC_COUNTER_BUFFER, atomicIndex, 1, &prop,
1, NULL, &mapping.readWriteResources[i].bind);
GLint used = 0;
GL.glGetProgramResourceiv(curProg, eGL_ATOMIC_COUNTER_BUFFER, atomicIndex, 1,
&refEnum[shadIdx], 1, NULL, &used);
mapping.readWriteResources[i].used = (used != 0);
mapping.readWriteResources[i].arraySize = 1;
}
else
{
const GLenum atomicRefEnum[] = {
eGL_ATOMIC_COUNTER_BUFFER_REFERENCED_BY_VERTEX_SHADER,
eGL_ATOMIC_COUNTER_BUFFER_REFERENCED_BY_TESS_CONTROL_SHADER,
eGL_ATOMIC_COUNTER_BUFFER_REFERENCED_BY_TESS_EVALUATION_SHADER,
eGL_ATOMIC_COUNTER_BUFFER_REFERENCED_BY_GEOMETRY_SHADER,
eGL_ATOMIC_COUNTER_BUFFER_REFERENCED_BY_FRAGMENT_SHADER,
eGL_ATOMIC_COUNTER_BUFFER_REFERENCED_BY_COMPUTE_SHADER,
};
mapping.readWriteResources[i].bindset = 0;
GL.glGetActiveAtomicCounterBufferiv(curProg, atomicIndex,
eGL_ATOMIC_COUNTER_BUFFER_BINDING,
&mapping.readWriteResources[i].bind);
GLint used = 0;
GL.glGetActiveAtomicCounterBufferiv(curProg, atomicIndex, atomicRefEnum[shadIdx],
&used);
mapping.readWriteResources[i].used = (used != 0);
mapping.readWriteResources[i].arraySize = 1;
}
}
}
}
else
{
// shader storage buffer object
GLuint idx = GL.glGetProgramResourceIndex(curProg, eGL_SHADER_STORAGE_BLOCK,
refl->readWriteResources[i].name.c_str());
if(idx == GL_INVALID_INDEX)
{
mapping.readWriteResources[i].bindset = -1;
mapping.readWriteResources[i].bind = -1;
mapping.readWriteResources[i].used = false;
mapping.readWriteResources[i].arraySize = 1;
}
else
{
GLenum prop = eGL_BUFFER_BINDING;
mapping.readWriteResources[i].bindset = 0;
GL.glGetProgramResourceiv(curProg, eGL_SHADER_STORAGE_BLOCK, idx, 1, &prop, 1, NULL,
&mapping.readWriteResources[i].bind);
GLint used = 0;
GL.glGetProgramResourceiv(curProg, eGL_SHADER_STORAGE_BLOCK, idx, 1, &refEnum[shadIdx], 1,
NULL, &used);
mapping.readWriteResources[i].used = (used != 0);
mapping.readWriteResources[i].arraySize = 1;
}
}
}
else
{
mapping.readWriteResources[i].bindset = -1;
mapping.readWriteResources[i].bind = -1;
mapping.readWriteResources[i].used = false;
mapping.readWriteResources[i].arraySize = 1;
}
}
mapping.constantBlocks.resize(refl->constantBlocks.size());
for(size_t i = 0; i < refl->constantBlocks.size(); i++)
{
if(refl->constantBlocks[i].bufferBacked)
{
GLint loc = GL.glGetUniformBlockIndex(curProg, refl->constantBlocks[i].name.c_str());
if(loc >= 0)
{
GL.glGetActiveUniformBlockiv(curProg, loc, eGL_UNIFORM_BLOCK_BINDING, dummyReadback);
mapping.constantBlocks[i].bindset = 0;
mapping.constantBlocks[i].bind = dummyReadback[0];
mapping.constantBlocks[i].arraySize = 1;
}
}
else
{
mapping.constantBlocks[i].bindset = -1;
mapping.constantBlocks[i].bind = -1;
mapping.constantBlocks[i].arraySize = 1;
}
if(!refl->constantBlocks[i].bufferBacked)
{
mapping.constantBlocks[i].used = true;
}
else
{
GLuint idx = GL.glGetProgramResourceIndex(curProg, eGL_UNIFORM_BLOCK,
refl->constantBlocks[i].name.c_str());
if(idx == GL_INVALID_INDEX)
{
mapping.constantBlocks[i].used = false;
}
else
{
GLint used = 0;
GL.glGetProgramResourceiv(curProg, eGL_UNIFORM_BLOCK, idx, 1, &refEnum[shadIdx], 1, NULL,
&used);
mapping.constantBlocks[i].used = (used != 0);
}
}
}
GLint numVAttribBindings = 16;
GL.glGetIntegerv(eGL_MAX_VERTEX_ATTRIBS, &numVAttribBindings);
mapping.inputAttributes.resize(numVAttribBindings);
for(int32_t i = 0; i < numVAttribBindings; i++)
mapping.inputAttributes[i] = -1;
// override identity map with bindings
if(shadIdx == 0)
{
for(int32_t i = 0; i < refl->inputSignature.count(); i++)
{
// skip system inputs, as some drivers will return a location for them
if(refl->inputSignature[i].systemValue != ShaderBuiltin::Undefined)
continue;
int32_t matrixRow = 0;
std::string varName = refl->inputSignature[i].varName;
size_t offs = varName.find(":row");
if(offs != std::string::npos)
{
matrixRow = varName[offs + 4] - '0';
varName.erase(offs);
}
GLint loc = GL.glGetAttribLocation(curProg, varName.c_str());
if(loc >= 0 && loc < numVAttribBindings)
{
mapping.inputAttributes[loc + matrixRow] = i;
}
}
}
#if ENABLED(RDOC_DEVEL)
for(size_t i = 1; i < ARRAY_COUNT(dummyReadback); i++)
if(dummyReadback[i] != 0x6c7b8a9d)
RDCERR("Invalid uniform readback - data beyond first element modified!");
#endif
}
void EvaluateSPIRVBindpointMapping(GLuint curProg, int shadIdx, const ShaderReflection *refl,
ShaderBindpointMapping &mapping)
{
// this is similar in principle to GetBindpointMapping - we want to look up the actual uniform
// values right now and replace the bindpoint mapping list. However for SPIR-V we can't call
// glGetUniformLocation. Instead we assume the *current* bind value is a location, and we
// overwrite it with the read uniform value.
// in case of bugs, we readback into this array instead of a single int
GLint dummyReadback[32];
#if ENABLED(RDOC_DEVEL)
for(size_t i = 1; i < ARRAY_COUNT(dummyReadback); i++)
dummyReadback[i] = 0x6c7b8a9d;
#endif
// GL_ARB_gl_spirv spec says that glBindAttribLocation does nothing on SPIR-V, so we don't have to
// remap inputAttributes.
// It's fuzzy on whether UBOs can be remapped with glUniformBlockBinding so for now we hope that
// anyone using UBOs and SPIR-V will at least specify immutable bindings in the SPIR-V.
for(size_t i = 0; i < mapping.constantBlocks.size(); i++)
{
Bindpoint &bind = mapping.constantBlocks[i];
if(!bind.used)
continue;
if(bind.bind < 0)
{
RDCERR("Invalid constant block binding found: '%s' = %d",
refl->constantBlocks[i].name.c_str(), bind.bind);
bind.bind = 0;
}
}
// shouldn't have any separate samplers - this is GL
RDCASSERT(mapping.samplers.size() == 0);
// for other resources we handle textures only, other resource types are assumed to have valid fix
// binds already. Any negative inputs are locations, so get the uniform value and assign it as the
// binding index.
for(size_t i = 0; i < refl->readOnlyResources.size(); i++)
{
if(!mapping.readOnlyResources[i].used)
continue;
if(refl->readOnlyResources[i].isTexture && mapping.readOnlyResources[i].bind < 0)
{
GL.glGetUniformiv(curProg, -mapping.readOnlyResources[i].bind, dummyReadback);
mapping.readOnlyResources[i].bind = dummyReadback[0];
if(mapping.readOnlyResources[i].bind < 0)
{
RDCERR("Invalid uniform value retrieved: '%s' = %d",
refl->readOnlyResources[i].name.c_str(), mapping.readOnlyResources[i].bind);
mapping.readOnlyResources[i].bind = 0;
}
}
else
{
if(mapping.readOnlyResources[i].bind < 0)
{
RDCERR("Invalid read-only resource binding found: '%s' = %d",
refl->readOnlyResources[i].name.c_str(), mapping.readOnlyResources[i].bind);
mapping.readOnlyResources[i].bind = 0;
}
}
}
for(size_t i = 0; i < refl->readWriteResources.size(); i++)
{
if(!mapping.readWriteResources[i].used)
continue;
if(refl->readWriteResources[i].isTexture && mapping.readWriteResources[i].bind < 0)
{
GL.glGetUniformiv(curProg, -mapping.readWriteResources[i].bind, dummyReadback);
mapping.readWriteResources[i].bind = dummyReadback[0];
if(mapping.readWriteResources[i].bind < 0)
{
RDCERR("Invalid uniform value retrieved: '%s' = %d",
refl->readWriteResources[i].name.c_str(), mapping.readWriteResources[i].bind);
mapping.readWriteResources[i].bind = 0;
}
}
else
{
if(mapping.readWriteResources[i].bind < 0)
{
RDCERR("Invalid read-only resource binding found: '%s' = %d",
refl->readWriteResources[i].name.c_str(), mapping.readWriteResources[i].bind);
mapping.readWriteResources[i].bind = 0;
}
}
}
for(size_t i = 0; i < mapping.inputAttributes.size(); i++)
if(mapping.inputAttributes[i] < 0)
mapping.inputAttributes[i] = -1;
#if ENABLED(RDOC_DEVEL)
for(size_t i = 1; i < ARRAY_COUNT(dummyReadback); i++)
if(dummyReadback[i] != 0x6c7b8a9d)
RDCERR("Invalid uniform readback - data beyond first element modified!");
#endif
}
// first int - the mapping index, second int - the binding
typedef std::vector<rdcpair<size_t, int> > Permutation;
// copy permutation by value since we mutate it to track the algorithm
static void ApplyPermutation(Permutation permutation, std::function<void(size_t, size_t)> DoSwap)
{
// permutations can always be decomposed into a series of disjoint cycles (one or more). Think
// of
// e.g:
//
// 0 1 2 3 4 5 6 7 8 9 10 11 12
// 8 0 4 5 2 11 1 12 6 3 10 9 7
//
// this is multiple cycles: 0 -> 8, 8 -> 6, 6 -> 1, 1 -> 0
// 2 -> 4, 4 -> 2
// 3 -> 5, 5 -> 11, 11 -> 9, 9 -> 3
// 7 -> 12, 12 -> 7
// 10 -> 10
//
// The general case is we just iterate along the permutation, find the first element that
// isn't in
// the right place, and then follow the cycle along - swapping the first element along into
// place
// until we eventually find that the cycle closes and we've swapped the first element into the
// right place. As we go we set the permutation values to an invalid marker so we know that
// they've been processed by a previous cycle when we continue with the iteration.
//
// This boils down to nothing in the case where the cycle is 2 long, it's just one swap.
size_t processedIdx = permutation.size();
for(size_t i = 0; i < permutation.size(); i++)
{
size_t dst = permutation[i].first;
// check if i is already in place or is already processed
if(i == dst || dst == processedIdx)
continue;
size_t src = i;
do
{
// do this swap
DoSwap(src, dst);
// mark this permutation as processed
permutation[src].first = processedIdx;
// move onto the next link in the cycle
src = dst;
dst = permutation[src].first;
// stop when we reach the start again - we've already done the swap to put this into place
} while(dst != i);
// close the cycle marking the last one as processed
permutation[src].first = processedIdx;
}
}
void ResortBindings(ShaderReflection *refl, ShaderBindpointMapping *mapping)
{
// In addition to the annoyance with texture unit handling in GL below, there's also an additional
// problem with the way bindings are handled. Nominally we have a set of bindings reflected out
// from the shader - these may come in alphabetical, declaration, location, or some other
// implementation defined order. We want a single fixed set of bindings so that the same shader
// always presents the same set of bindings to the user. However we can't use this reflected order
// as the set, because the mapping from these binds to the actual API slots used is *mutable*. If
// it were fixed at shader compile/specification time then we could sort it once after reflection
// and then go on with our lives, however because it's mutable we have to do the sort here based
// on the latest uniform values.
//
// Other alternatives would be to never sort, but then we land ourselves in a quagmire where the
// bindings could be "diffuse, normals, shadow, depthbuffer" in reflected order, but then
// depthbuffer could be assigned to slot 0 - then we are listing the textures in an arbitrary
// unsorted order. It's not possible to sort by current binding anywhere above this level because
// we don't want to do this in any of the generic code. No matter how we represent the bindings,
// this fundamentally comes down to two competing orders: The order that actually makes sense (but
// is mutable), and the order that is fixed at reflection time (but is useless).
//
// In general the hope is that no-one actually makes use of this ability to remap uniform values
// at runtime, and in practice everyone either uses the layout qualifiers in shaders to fix the
// bindings are shader compile time anyway (hah), or they reflect the samplers and set them one
// time, then leave them fixed. In the worst case, if an application does actually remap the
// uniforms from draw to draw, they will end up seeing the bindings re-order themselves in the UI.
// This might be confusing, but it's a) technically what the application is actually doing, from a
// certain perspective, and b) limited to a very small niche of people that are doing something
// kind of ridiculous.
//
// So here we re-sort the actual reflection data and bindpoint mapping so that the 'bind' is in
// ascending order. It looks ugly because it is ugly.
if(!refl || !mapping)
return;
Permutation permutation;
// sort by the binding
struct permutation_sort
{
bool operator()(const rdcpair<size_t, int> &a, const rdcpair<size_t, int> &b) const
{
return a.second < b.second;
}
};
permutation.resize(mapping->readOnlyResources.size());
for(size_t i = 0; i < mapping->readOnlyResources.size(); i++)
permutation[i] = make_rdcpair(i, mapping->readOnlyResources[i].bind);
std::sort(permutation.begin(), permutation.end(), permutation_sort());
// apply the permutation to the mapping array, and update the bindPoint values in the shader
// reflection to match, so that the re-order is applied
ApplyPermutation(permutation, [mapping](size_t a, size_t b) {
std::swap(mapping->readOnlyResources[a], mapping->readOnlyResources[b]);
});
for(size_t i = 0; i < permutation.size(); i++)
refl->readOnlyResources[permutation[i].first].bindPoint = (int)i;
permutation.resize(mapping->readWriteResources.size());
for(size_t i = 0; i < mapping->readWriteResources.size(); i++)
permutation[i] = make_rdcpair(i, mapping->readWriteResources[i].bind);
std::sort(permutation.begin(), permutation.end(), permutation_sort());
ApplyPermutation(permutation, [mapping](size_t a, size_t b) {
std::swap(mapping->readWriteResources[a], mapping->readWriteResources[b]);
});
for(size_t i = 0; i < permutation.size(); i++)
refl->readWriteResources[permutation[i].first].bindPoint = (int)i;
permutation.resize(mapping->constantBlocks.size());
for(size_t i = 0; i < mapping->constantBlocks.size(); i++)
permutation[i] = make_rdcpair(i, mapping->constantBlocks[i].bind);
std::sort(permutation.begin(), permutation.end(), permutation_sort());
ApplyPermutation(permutation, [mapping](size_t a, size_t b) {
std::swap(mapping->constantBlocks[a], mapping->constantBlocks[b]);
});
for(size_t i = 0; i < permutation.size(); i++)
refl->constantBlocks[permutation[i].first].bindPoint = (int)i;
}