// Recursively figure out how many bytes of xfb buffer are used by the given type.
// Return the size of type, in bytes.
// Sets containsDouble to true if the type contains a double.
// N.B. Caller must set containsDouble to false before calling.
unsigned int TIntermediate::computeTypeXfbSize(const TType& type, bool& containsDouble) const
{
// "...if applied to an aggregate containing a double, the offset must also be a multiple of 8,
// and the space taken in the buffer will be a multiple of 8.
// ...within the qualified entity, subsequent components are each
// assigned, in order, to the next available offset aligned to a multiple of
// that component's size. Aggregate types are flattened down to the component
// level to get this sequence of components."
if (type.isArray()) {
// TODO: perf: this can be flattened by using getCumulativeArraySize(), and a deref that discards all arrayness
assert(type.isExplicitlySizedArray());
TType elementType(type, 0);
return type.getOuterArraySize() * computeTypeXfbSize(elementType, containsDouble);
}
if (type.isStruct()) {
unsigned int size = 0;
bool structContainsDouble = false;
for (int member = 0; member < (int)type.getStruct()->size(); ++member) {
TType memberType(type, member);
// "... if applied to
// an aggregate containing a double, the offset must also be a multiple of 8,
// and the space taken in the buffer will be a multiple of 8."
bool memberContainsDouble = false;
int memberSize = computeTypeXfbSize(memberType, memberContainsDouble);
if (memberContainsDouble) {
structContainsDouble = true;
RoundToPow2(size, 8);
}
size += memberSize;
}
if (structContainsDouble) {
containsDouble = true;
RoundToPow2(size, 8);
}
return size;
}
int numComponents;
if (type.isScalar())
numComponents = 1;
else if (type.isVector())
numComponents = type.getVectorSize();
else if (type.isMatrix())
numComponents = type.getMatrixCols() * type.getMatrixRows();
else {
assert(0);
numComponents = 1;
}
if (type.getBasicType() == EbtDouble) {
containsDouble = true;
return 8 * numComponents;
} else
return 4 * numComponents;
}
// Accumulate locations used for inputs, outputs, and uniforms, and check for collisions
// as the accumulation is done.
//
// Returns < 0 if no collision, >= 0 if collision and the value returned is a colliding value.
//
// typeCollision is set to true if there is no direct collision, but the types in the same location
// are different.
//
int TIntermediate::addUsedLocation(const TQualifier& qualifier, const TType& type, bool& typeCollision)
{
typeCollision = false;
int set;
if (qualifier.isPipeInput())
set = 0;
else if (qualifier.isPipeOutput())
set = 1;
else if (qualifier.storage == EvqUniform)
set = 2;
else if (qualifier.storage == EvqBuffer)
set = 3;
else
return -1;
int size;
if (qualifier.isUniformOrBuffer()) {
if (type.isArray())
size = type.getCumulativeArraySize();
else
size = 1;
} else {
// Strip off the outer array dimension for those having an extra one.
if (type.isArray() && qualifier.isArrayedIo(language)) {
TType elementType(type, 0);
size = computeTypeLocationSize(elementType);
} else
size = computeTypeLocationSize(type);
}
TRange locationRange(qualifier.layoutLocation, qualifier.layoutLocation + size - 1);
TRange componentRange(0, 3);
if (qualifier.hasComponent()) {
componentRange.start = qualifier.layoutComponent;
componentRange.last = componentRange.start + type.getVectorSize() - 1;
}
TIoRange range(locationRange, componentRange, type.getBasicType(), qualifier.hasIndex() ? qualifier.layoutIndex : 0);
// check for collisions, except for vertex inputs on desktop
if (! (profile != EEsProfile && language == EShLangVertex && qualifier.isPipeInput())) {
for (size_t r = 0; r < usedIo[set].size(); ++r) {
if (range.overlap(usedIo[set][r])) {
// there is a collision; pick one
return std::max(locationRange.start, usedIo[set][r].location.start);
} else if (locationRange.overlap(usedIo[set][r].location) && type.getBasicType() != usedIo[set][r].basicType) {
// aliased-type mismatch
typeCollision = true;
return std::max(locationRange.start, usedIo[set][r].location.start);
}
}
}
usedIo[set].push_back(range);
return -1; // no collision
}
// Recursively figure out how many locations are used up by an input or output type.
// Return the size of type, as measured by "locations".
int TIntermediate::computeTypeLocationSize(const TType& type) const
{
// "If the declared input is an array of size n and each element takes m locations, it will be assigned m * n
// consecutive locations..."
if (type.isArray()) {
// TODO: perf: this can be flattened by using getCumulativeArraySize(), and a deref that discards all arrayness
TType elementType(type, 0);
if (type.isImplicitlySizedArray()) {
// TODO: are there valid cases of having an implicitly-sized array with a location? If so, running this code too early.
return computeTypeLocationSize(elementType);
} else
return type.getOuterArraySize() * computeTypeLocationSize(elementType);
}
// "The locations consumed by block and structure members are determined by applying the rules above
// recursively..."
if (type.isStruct()) {
int size = 0;
for (int member = 0; member < (int)type.getStruct()->size(); ++member) {
TType memberType(type, member);
size += computeTypeLocationSize(memberType);
}
return size;
}
// ES: "If a shader input is any scalar or vector type, it will consume a single location."
// Desktop: "If a vertex shader input is any scalar or vector type, it will consume a single location. If a non-vertex
// shader input is a scalar or vector type other than dvec3 or dvec4, it will consume a single location, while
// types dvec3 or dvec4 will consume two consecutive locations. Inputs of type double and dvec2 will
// consume only a single location, in all stages."
if (type.isScalar())
return 1;
if (type.isVector()) {
if (language == EShLangVertex && type.getQualifier().isPipeInput())
return 1;
if (type.getBasicType() == EbtDouble && type.getVectorSize() > 2)
return 2;
else
return 1;
}
// "If the declared input is an n x m single- or double-precision matrix, ...
// The number of locations assigned for each matrix will be the same as
// for an n-element array of m-component vectors..."
if (type.isMatrix()) {
TType columnType(type, 0);
return type.getMatrixCols() * computeTypeLocationSize(columnType);
}
assert(0);
return 1;
}
// Ensure index is in bounds, correct if necessary.
// Give an error if not.
void TParseContextBase::checkIndex(const TSourceLoc& loc, const TType& type, int& index)
{
if (index < 0) {
error(loc, "", "[", "index out of range '%d'", index);
index = 0;
} else if (type.isArray()) {
if (type.isSizedArray() && index >= type.getOuterArraySize()) {
error(loc, "", "[", "array index out of range '%d'", index);
index = type.getOuterArraySize() - 1;
}
} else if (type.isVector()) {
if (index >= type.getVectorSize()) {
error(loc, "", "[", "vector index out of range '%d'", index);
index = type.getVectorSize() - 1;
}
} else if (type.isMatrix()) {
if (index >= type.getMatrixCols()) {
error(loc, "", "[", "matrix index out of range '%d'", index);
index = type.getMatrixCols() - 1;
}
}
}
// Implement base-alignment and size rules from section 7.6.2.2 Standard Uniform Block Layout
// Operates recursively.
//
// If std140 is true, it does the rounding up to vec4 size required by std140,
// otherwise it does not, yielding std430 rules.
//
// The size is returned in the 'size' parameter
//
// The stride is only non-0 for arrays or matrices, and is the stride of the
// top-level object nested within the type. E.g., for an array of matrices,
// it is the distances needed between matrices, despite the rules saying the
// stride comes from the flattening down to vectors.
//
// Return value is the alignment of the type.
int TIntermediate::getBaseAlignment(const TType& type, int& size, int& stride, bool std140, bool rowMajor)
{
int alignment;
// When using the std140 storage layout, structures will be laid out in buffer
// storage with its members stored in monotonically increasing order based on their
// location in the declaration. A structure and each structure member have a base
// offset and a base alignment, from which an aligned offset is computed by rounding
// the base offset up to a multiple of the base alignment. The base offset of the first
// member of a structure is taken from the aligned offset of the structure itself. The
// base offset of all other structure members is derived by taking the offset of the
// last basic machine unit consumed by the previous member and adding one. Each
// structure member is stored in memory at its aligned offset. The members of a top-
// level uniform block are laid out in buffer storage by treating the uniform block as
// a structure with a base offset of zero.
//
// 1. If the member is a scalar consuming N basic machine units, the base alignment is N.
//
// 2. If the member is a two- or four-component vector with components consuming N basic
// machine units, the base alignment is 2N or 4N, respectively.
//
// 3. If the member is a three-component vector with components consuming N
// basic machine units, the base alignment is 4N.
//
// 4. If the member is an array of scalars or vectors, the base alignment and array
// stride are set to match the base alignment of a single array element, according
// to rules (1), (2), and (3), and rounded up to the base alignment of a vec4. The
// array may have padding at the end; the base offset of the member following
// the array is rounded up to the next multiple of the base alignment.
//
// 5. If the member is a column-major matrix with C columns and R rows, the
// matrix is stored identically to an array of C column vectors with R
// components each, according to rule (4).
//
// 6. If the member is an array of S column-major matrices with C columns and
// R rows, the matrix is stored identically to a row of S C column vectors
// with R components each, according to rule (4).
//
// 7. If the member is a row-major matrix with C columns and R rows, the matrix
// is stored identically to an array of R row vectors with C components each,
// according to rule (4).
//
// 8. If the member is an array of S row-major matrices with C columns and R
// rows, the matrix is stored identically to a row of S R row vectors with C
// components each, according to rule (4).
//
// 9. If the member is a structure, the base alignment of the structure is N , where
// N is the largest base alignment value of any of its members, and rounded
// up to the base alignment of a vec4. The individual members of this substructure
// are then assigned offsets by applying this set of rules recursively,
// where the base offset of the first member of the sub-structure is equal to the
// aligned offset of the structure. The structure may have padding at the end;
// the base offset of the member following the sub-structure is rounded up to
// the next multiple of the base alignment of the structure.
//
// 10. If the member is an array of S structures, the S elements of the array are laid
// out in order, according to rule (9).
//
// Assuming, for rule 10: The stride is the same as the size of an element.
stride = 0;
int dummyStride;
// rules 4, 6, 8, and 10
if (type.isArray()) {
// TODO: perf: this might be flattened by using getCumulativeArraySize(), and a deref that discards all arrayness
TType derefType(type, 0);
alignment = getBaseAlignment(derefType, size, dummyStride, std140, rowMajor);
if (std140)
alignment = std::max(baseAlignmentVec4Std140, alignment);
RoundToPow2(size, alignment);
stride = size; // uses full matrix size for stride of an array of matrices (not quite what rule 6/8, but what's expected)
// uses the assumption for rule 10 in the comment above
size = stride * type.getOuterArraySize();
return alignment;
}
// rule 9
if (type.getBasicType() == EbtStruct) {
const TTypeList& memberList = *type.getStruct();
size = 0;
int maxAlignment = std140 ? baseAlignmentVec4Std140 : 0;
for (size_t m = 0; m < memberList.size(); ++m) {
int memberSize;
// modify just the children's view of matrix layout, if there is one for this member
TLayoutMatrix subMatrixLayout = memberList[m].type->getQualifier().layoutMatrix;
int memberAlignment = getBaseAlignment(*memberList[m].type, memberSize, dummyStride, std140,
(subMatrixLayout != ElmNone) ? (subMatrixLayout == ElmRowMajor) : rowMajor);
maxAlignment = std::max(maxAlignment, memberAlignment);
RoundToPow2(size, memberAlignment);
size += memberSize;
}
// The structure may have padding at the end; the base offset of
// the member following the sub-structure is rounded up to the next
// multiple of the base alignment of the structure.
RoundToPow2(size, maxAlignment);
return maxAlignment;
}
// rule 1
if (type.isScalar())
return getBaseAlignmentScalar(type, size);
// rules 2 and 3
if (type.isVector()) {
int scalarAlign = getBaseAlignmentScalar(type, size);
switch (type.getVectorSize()) {
case 2:
size *= 2;
return 2 * scalarAlign;
default:
size *= type.getVectorSize();
return 4 * scalarAlign;
}
}
// rules 5 and 7
if (type.isMatrix()) {
// rule 5: deref to row, not to column, meaning the size of vector is num columns instead of num rows
TType derefType(type, 0, rowMajor);
alignment = getBaseAlignment(derefType, size, dummyStride, std140, rowMajor);
if (std140)
alignment = std::max(baseAlignmentVec4Std140, alignment);
RoundToPow2(size, alignment);
stride = size; // use intra-matrix stride for stride of a just a matrix
if (rowMajor)
size = stride * type.getMatrixRows();
else
size = stride * type.getMatrixCols();
return alignment;
}
assert(0); // all cases should be covered above
size = baseAlignmentVec4Std140;
return baseAlignmentVec4Std140;
}
