GLMULTIDRAWELEMENTSI(3G)
π GLMULTIDRAWELEMENTSI(3G) π
okay bestie, so this one is glMultiDrawElementsIndirect from the 3G manual, and it is literally the "draw a million things with one function call" icon we all deserve. sit down, hydrate, let's go π
β¨ NAME β¨
glMultiDrawElementsIndirect β render indexed primitives from array data, taking parameters from memory. yes the whole vibe is "the params live in memory, not in your twelve million arguments" π§
πΈ C SPECIFICATION πΈ
void glMultiDrawElementsIndirect(GLenum mode, GLenum type,
const void *indirect,
GLsizei drawcount, GLsizei stride);
π PARAMETERS π
modeβ tells the GPU what kind of primitives to render. accepted values are the constantsGL_POINTS,GL_LINE_STRIP,GL_LINE_LOOP,GL_LINES,GL_LINE_STRIP_ADJACENCY,GL_LINES_ADJACENCY,GL_TRIANGLE_STRIP,GL_TRIANGLE_FAN,GL_TRIANGLES,GL_TRIANGLE_STRIP_ADJACENCY,GL_TRIANGLES_ADJACENCY, andGL_PATCHES. that's the whole menu, no substitutions πtypeβ the type of data in the buffer bound to theGL_ELEMENT_ARRAY_BUFFERbinding. this matters, okay? πindirectβ the address of a structure containing an array of draw parameters. translation: the receipts for every draw call, all lined up in memory πdrawcountβ how many elements are in the array thatindirectpoints at.strideβ the distance in basic machine units between elements of the draw parameter array. think of it as the gap between one receipt and the next β¨
π¦ DESCRIPTION π¦
glMultiDrawElementsIndirect specifies multiple indexed geometric primitives with very few subroutine calls. like, very few. barely any. we love efficiency π
it behaves similarly to a multitude of calls to glDrawElementsInstancedBaseVertexBaseInstance(), except β key detail, no cap β the parameters to glDrawElementsInstancedBaseVertexBaseInstance() are stored in an array in memory at the address given by indirect, separated by the stride, in basic machine units, specified by stride. and if stride is zero, the array is assumed to be tightly packed in memory. zero means "no gap, we're cozy" π«Ά
the parameters addressed by indirect are packed into a structure that takes this form (in C):
typedef struct {
uint count;
uint instanceCount;
uint firstIndex;
uint baseVertex;
uint baseInstance;
} DrawElementsIndirectCommand;
a single call to glMultiDrawElementsIndirect is equivalent β assuming no errors are generated β to this, and bestie this loop is iconic:
GLsizei n;
for (n = 0; n < drawcount; n++) {
const DrawElementsIndirectCommand *cmd;
if (stride != 0) {
cmd = (const DrawElementsIndirectCommand *)((uintptr)indirect + n * stride);
} else {
cmd = (const DrawElementsIndirectCommand *)indirect + n;
}
glDrawElementsInstancedBaseVertexBaseInstance(mode,
cmd->count,
type,
cmd->firstIndex * size-of-type,
cmd->instanceCount,
cmd->baseVertex,
cmd->baseInstance);
}
okay so if a buffer is bound to the GL_DRAW_INDIRECT_BUFFER binding at the time of a call to glDrawElementsIndirect, then indirect is interpreted as an offset, in basic machine units, into that buffer, and the parameter data is read from the buffer rather than from client memory. buffer-based, client-memory-free, immaculate β¨
β οΈ heads up: indices stored in client memory are not supported. if no buffer is bound to the GL_ELEMENT_ARRAY_BUFFER binding, an error will be generated. no exceptions, no workarounds, no vibes π
the results of the operation are undefined if the reservedMustBeZero member of the parameter structure is non-zero. however β and this is the sneaky part β no error is generated in this case. so like, it won't yell at you, it'll just do undefined things. love that for us π
vertex attributes that are modified by glDrawElementsIndirect have an unspecified value after glDrawElementsIndirect returns. attributes that aren't modified remain well defined. so the ones you didn't touch are safe, the ones you did are in their little mystery era π
β NOTES β
the baseInstance member of the DrawElementsIndirectCommand structure is defined only if the GL version is 4.2 or greater. for versions of the GL less than 4.2, this parameter is present but is reserved and should be set to zero. on earlier versions of the GL, behavior is undefined if it is non-zero. so check your version, bestie, please π«Ά
π₯ ERRORS π₯
GL_INVALID_ENUMis generated ifmodeis not an accepted value.GL_INVALID_VALUEis generated ifstrideis not a multiple of four.GL_INVALID_VALUEis generated ifdrawcountis negative.GL_INVALID_OPERATIONis generated if no buffer is bound to theGL_ELEMENT_ARRAY_BUFFERbinding, or if such a buffer's data store is currently mapped.GL_INVALID_OPERATIONis generated if a non-zero buffer object name is bound to an enabled array or to theGL_DRAW_INDIRECT_BUFFERbinding and the buffer object's data store is currently mapped.GL_INVALID_OPERATIONis generated if a geometry shader is active andmodeis incompatible with the input primitive type of the geometry shader in the currently installed program object.GL_INVALID_OPERATIONis generated ifmodeisGL_PATCHESand no tessellation control shader is active.
that's a lot of GL_INVALID_OPERATION and honestly? same energy π€
π VERSION SUPPORT π
the table in the original is giving β¨spreadsheetβ¨, so here's the tea in list form. the function in question is glMultiDrawElementsIndirect, and across the OpenGL versions it looks like this:
- OpenGL 2.0 β
- - OpenGL 2.1 β
- - OpenGL 3.0 β
- - OpenGL 3.1 β
- - OpenGL 3.2 β
- - OpenGL 3.3 β
- - OpenGL 4.0 β
- - OpenGL 4.1 β
- - OpenGL 4.2 β
- - OpenGL 4.3 β
β - OpenGL 4.4 β
β - OpenGL 4.5 β
β
so it's a hard no until 4.3, and then it's a full slay π the column header was "OpenGL Version" and the row label was "Function / Feature Name", in case you were wondering where the table went. i gotchu π«Ά
π SEE ALSO π
glDrawArrays(), glDrawArraysInstanced(), glDrawArraysIndirect(), glDrawElements(), glDrawRangeElements(), glDrawElementsIndirect(), glMultiDrawArraysIndirect()
go read those next, bestie, the whole friend group is there π§
π COPYRIGHT π
Copyright Β© 2010-2014 Khronos Group. This material may be distributed subject to the terms and conditions set forth in the Open Publication License, v 1.0, 8 June 1999. http://opencontent.org/openpub/.
π COPYRIGHT (again, yes really) π
Copyright Β© 2010-2014 Khronos Group.
and that's the whole page, bestie π mwah π β sourced from [FIXME: source], dated 08/09/2021, and forever enshrined as GLMULTIDRAWELEMENTSI(3G) β¨π¦π
π¦ where this page comes from
- centos-stream 9: gl-manpages 1.1-24.20190306.el9 Β· MIT and Open Publication
- π girlified by deepseek-ai/DeepSeek-V4.1-Flash-1017 on : a language model's rewrite of the original, under its terms