Improve typing in render.

This commit is contained in:
Roz K 2023-01-04 09:33:49 +01:00
parent ebc6ededf3
commit b46b4bddba
Signed by: roz
GPG Key ID: 51FBF4E483E1C822
5 changed files with 197 additions and 212 deletions

View File

@ -18,15 +18,12 @@
#include "types.hpp" #include "types.hpp"
template<typename _type> template<typename _small>
bool _rk_cmp_memcpy_small( bool _rk_cmp_memcpy_small(
void * const __restrict _dst, _small * __restrict dst,
void const * const __restrict _src, _small const * __restrict src,
unsigned const size) { unsigned count) {
unsigned count = (size / sizeof(_type)); _small cmp = 0;
_type * dst = reinterpret_cast<_type *>(_dst);
_type const * src = reinterpret_cast<_type const *>(_src);
_type cmp = 0;
do { do {
cmp |= *dst ^ *src; cmp |= *dst ^ *src;
*dst++ = *src++; *dst++ = *src++;
@ -36,134 +33,67 @@ bool _rk_cmp_memcpy_small(
template<typename _big, typename _small> template<typename _big, typename _small>
bool _rk_cmp_memcpy_big( bool _rk_cmp_memcpy_big(
void * const __restrict _dst, _small * const __restrict _dst,
void const * const __restrict _src, _small const * const __restrict _src,
unsigned const size) { unsigned const _count) {
unsigned count = size / sizeof(_big); unsigned const ratio = sizeof(_big) / sizeof(_small);
unsigned const remain = size % sizeof(_big); unsigned big_count = _count / ratio;
unsigned const small_count = _count % ratio;
_big * dst = reinterpret_cast<_big *>(_dst); _big * dst = reinterpret_cast<_big *>(_dst);
_big const * src = reinterpret_cast<_big const *>(_src); _big const * src = reinterpret_cast<_big const *>(_src);
_big cmp = 0; _big cmp = 0;
do { do {
cmp |= *dst ^ *src; cmp |= *dst ^ *src;
*dst++ = *src++; *dst++ = *src++;
} while(--count > 0); } while(--big_count > 0);
bool modified = (cmp != 0); bool modified = (cmp != 0);
if (remain) { if (small_count) {
modified |= _rk_cmp_memcpy_small<_small>(dst, src, remain); modified |= _rk_cmp_memcpy_small<_small>(
reinterpret_cast<_small *>(dst), reinterpret_cast<_small const *>(src), small_count);
} }
return modified; return modified;
} }
#ifdef RK_CMP_MEMCPY_UNALIGNED #ifdef RK_CMP_MEMCPY_UNALIGNED
#define _rk_size_and_alignment(_t) (size >= sizeof(_t)) #define _rk_count_and_alignment(_t) (count >= (sizeof(_t) / sizeof(_small)))
#else #else
#define _rk_size_and_alignment(_t) (size >= sizeof(_t) && !(alignment & (sizeof(_t) - 1))) #define _rk_count_and_alignment(_t) ((count >= (sizeof(_t) / sizeof(_small))) && !(alignment % sizeof(_t)))
#endif #endif
template<unsigned _s> template<typename _small>
bool rk_cmp_memcpy( bool rk_cmp_memcpy(
void * const __restrict _dst, _small * const __restrict _dst,
void const * const __restrict _src, _small const * const __restrict _src,
unsigned const size); unsigned const count) {
template<>
bool rk_cmp_memcpy<sizeof(rk_ubyte)>(
void * const __restrict _dst,
void const * const __restrict _src,
unsigned const size) {
#ifndef RK_CMP_MEMCPY_UNALIGNED #ifndef RK_CMP_MEMCPY_UNALIGNED
unsigned const alignment = reinterpret_cast<uintptr_t>(_dst) | reinterpret_cast<uintptr_t const>(_src); unsigned const alignment = reinterpret_cast<uintptr_t>(_dst) | reinterpret_cast<uintptr_t const>(_src);
#endif #endif
if (_rk_size_and_alignment(rk_ullong)) { if (sizeof(_small) < sizeof(rk_ullong)) {
return _rk_cmp_memcpy_big<rk_ullong, rk_ubyte>(_dst, _src, size); if (_rk_count_and_alignment(rk_ullong)) {
return _rk_cmp_memcpy_big<rk_ullong, _small>(_dst, _src, count);
} }
if (_rk_size_and_alignment(rk_ulong)) {
return _rk_cmp_memcpy_big<rk_ulong, rk_ubyte>(_dst, _src, size);
} }
if (_rk_size_and_alignment(rk_uint)) { if (sizeof(_small) < sizeof(rk_ulong)) {
return _rk_cmp_memcpy_big<rk_uint, rk_ubyte>(_dst, _src, size); if (_rk_count_and_alignment(rk_ulong)) {
return _rk_cmp_memcpy_big<rk_ulong, _small>(_dst, _src, count);
} }
if (_rk_size_and_alignment(rk_ushort)) {
return _rk_cmp_memcpy_big<rk_ushort, rk_ubyte>(_dst, _src, size);
} }
if (size > 0) { if (sizeof(_small) < sizeof(rk_uint)) {
return _rk_cmp_memcpy_small<rk_ubyte>(_dst, _src, size); if (_rk_count_and_alignment(rk_uint)) {
return _rk_cmp_memcpy_big<rk_uint, _small>(_dst, _src, count);
}
}
if (sizeof(_small) < sizeof(rk_ushort)) {
if (_rk_count_and_alignment(rk_ushort)) {
return _rk_cmp_memcpy_big<rk_ushort, _small>(_dst, _src, count);
}
}
if (count) {
return _rk_cmp_memcpy_small<_small>(_dst, _src, count);
} }
return false; return false;
} }
template<> #undef _rk_count_and_alignment
bool rk_cmp_memcpy<sizeof(rk_ushort)>(
void * const __restrict _dst,
void const * const __restrict _src,
unsigned const size) {
#ifndef RK_CMP_MEMCPY_UNALIGNED
unsigned const alignment = reinterpret_cast<uintptr_t>(_dst) | reinterpret_cast<uintptr_t const>(_src);
#endif
if (_rk_size_and_alignment(rk_ullong)) {
return _rk_cmp_memcpy_big<rk_ullong, rk_ushort>(_dst, _src, size);
}
if (_rk_size_and_alignment(rk_ulong)) {
return _rk_cmp_memcpy_big<rk_ulong, rk_ushort>(_dst, _src, size);
}
if (_rk_size_and_alignment(rk_uint)) {
return _rk_cmp_memcpy_big<rk_uint, rk_ushort>(_dst, _src, size);
}
if (size > 0) {
return _rk_cmp_memcpy_small<rk_ushort>(_dst, _src, size);
}
return false;
}
template<>
bool rk_cmp_memcpy<sizeof(rk_uint)>(
void * const __restrict _dst,
void const * const __restrict _src,
unsigned const size) {
#ifndef RK_CMP_MEMCPY_UNALIGNED
unsigned const alignment = reinterpret_cast<uintptr_t>(_dst) | reinterpret_cast<uintptr_t const>(_src);
#endif
if (_rk_size_and_alignment(rk_ullong)) {
return _rk_cmp_memcpy_big<rk_ullong, rk_uint>(_dst, _src, size);
}
if (_rk_size_and_alignment(rk_ulong)) {
return _rk_cmp_memcpy_big<rk_ulong, rk_uint>(_dst, _src, size);
}
if (size > 0) {
return _rk_cmp_memcpy_small<rk_uint>(_dst, _src, size);
}
return false;
}
template<>
bool rk_cmp_memcpy<sizeof(rk_ulong)>(
void * const __restrict _dst,
void const * const __restrict _src,
unsigned const size) {
#ifndef RK_CMP_MEMCPY_UNALIGNED
unsigned const alignment = reinterpret_cast<uintptr_t>(_dst) | reinterpret_cast<uintptr_t const>(_src);
#endif
if (_rk_size_and_alignment(rk_ullong)) {
return _rk_cmp_memcpy_big<rk_ullong, rk_ulong>(_dst, _src, size);
}
if (size > 0) {
return _rk_cmp_memcpy_small<rk_ulong>(_dst, _src, size);
}
return false;
}
template<>
bool rk_cmp_memcpy<sizeof(rk_ullong)>(
void * const __restrict _dst,
void const * const __restrict _src,
unsigned const size) {
if (size > 0) {
return _rk_cmp_memcpy_small<rk_ullong>(_dst, _src, size);
}
return false;
}
#undef _rk_size_and_alignment
#endif // RK_ENGINE_CMP_MEMCPY_H #endif // RK_ENGINE_CMP_MEMCPY_H

View File

@ -26,6 +26,14 @@ typedef glm::vec4 rk_vec4;
typedef glm::mat3 rk_mat3; typedef glm::mat3 rk_mat3;
typedef glm::mat4 rk_mat4; typedef glm::mat4 rk_mat4;
#define RK_CHECK_MATH_TYPE(_t, _e, _c) static_assert(sizeof(_t) == sizeof(_e) * (_c))
RK_CHECK_MATH_TYPE(rk_vec2, float, 2);
RK_CHECK_MATH_TYPE(rk_vec3, float, 3);
RK_CHECK_MATH_TYPE(rk_vec4, float, 4);
RK_CHECK_MATH_TYPE(rk_mat3, rk_vec3, 3);
RK_CHECK_MATH_TYPE(rk_mat4, rk_vec4, 4);
#define vec3_right (rk_vec3(1.f, 0.f, 0.f)) #define vec3_right (rk_vec3(1.f, 0.f, 0.f))
#define vec3_forward (rk_vec3(0.f, 1.f, 0.f)) #define vec3_forward (rk_vec3(0.f, 1.f, 0.f))
#define vec3_up (rk_vec3(0.f, 0.f, 1.f)) #define vec3_up (rk_vec3(0.f, 0.f, 1.f))

View File

@ -27,14 +27,18 @@ typedef rk_handle_t rk_triangles_t;
typedef rk_handle_t rk_vertices_t; typedef rk_handle_t rk_vertices_t;
typedef rk_handle_t rk_batch_t; typedef rk_handle_t rk_batch_t;
enum rk_texture_format : rk_uint { typedef rk_uint rk_texture_format;
enum : rk_uint {
RK_TEXTURE_FORMAT_SRGB8_A8 = 0, RK_TEXTURE_FORMAT_SRGB8_A8 = 0,
RK_TEXTURE_FORMAT_RGBA8 = 1, RK_TEXTURE_FORMAT_RGBA8 = 1,
RK_TEXTURE_FORMAT_RGB10_A2 = 2, RK_TEXTURE_FORMAT_RGB10_A2 = 2,
RK_TEXTURE_FORMAT_FLOAT_32 = 3 RK_TEXTURE_FORMAT_FLOAT_32 = 3
}; };
enum rk_texture_flags : rk_uint { typedef rk_uint rk_texture_flags;
enum : rk_uint {
RK_TEXTURE_FLAG_3D = RK_FLAG(0), RK_TEXTURE_FLAG_3D = RK_FLAG(0),
RK_TEXTURE_FLAG_MIPMAPS = RK_FLAG(1), RK_TEXTURE_FLAG_MIPMAPS = RK_FLAG(1),
RK_TEXTURE_FLAG_MIN_NEAREST = 0, RK_TEXTURE_FLAG_MIN_NEAREST = 0,
@ -43,34 +47,49 @@ enum rk_texture_flags : rk_uint {
RK_TEXTURE_FLAG_MAG_LINEAR = RK_FLAG(3), RK_TEXTURE_FLAG_MAG_LINEAR = RK_FLAG(3),
}; };
enum rk_vertex_format : rk_ubyte { typedef rk_ubyte rk_vertex_format;
enum : rk_ubyte {
RK_VERTEX_FORMAT_VEC3_FLOAT = 1, RK_VERTEX_FORMAT_VEC3_FLOAT = 1,
RK_VERTEX_FORMAT_VEC3_INT10 = 2, RK_VERTEX_FORMAT_VEC3_INT10 = 2,
RK_VERTEX_FORMAT_VEC3_UINT10 = 3 RK_VERTEX_FORMAT_VEC3_UINT10 = 3,
RK_VERTEX_FORMAT_NORMALIZE = RK_FLAG(7),
RK_VERTEX_FORMAT_MASK = RK_VERTEX_FORMAT_NORMALIZE - 1
}; };
enum : rk_ubyte { RK_VERTEX_FORMAT_NORMALIZE = RK_FLAG(7) }; typedef rk_ubyte rk_param_format;
enum : rk_ubyte { RK_VERTEX_FORMAT_MASK = RK_VERTEX_FORMAT_NORMALIZE - 1 };
enum rk_param_format : rk_ubyte { enum : rk_ubyte {
RK_PARAM_FORMAT_VEC3_FLOAT = 1, RK_PARAM_FORMAT_VEC3_FLOAT = 1,
RK_PARAM_FORMAT_VEC3_SHORT = 2, RK_PARAM_FORMAT_VEC3_SHORT = 2,
RK_PARAM_FORMAT_VEC3_INT10 = 3, RK_PARAM_FORMAT_VEC3_INT10 = 3,
RK_PARAM_FORMAT_MAT3_FLOAT = 4, RK_PARAM_FORMAT_MAT3_FLOAT = 4,
RK_PARAM_FORMAT_MAT3_INT10 = 5 RK_PARAM_FORMAT_MAT3_INT10 = 5,
RK_PARAM_FORMAT_NORMALIZE = RK_FLAG(7),
RK_PARAM_FORMAT_MASK = RK_PARAM_FORMAT_NORMALIZE - 1
}; };
enum : rk_ubyte { RK_PARAM_FORMAT_NORMALIZE = RK_FLAG(7) }; typedef rk_ubyte rk_instance_flags;
enum : rk_ubyte { RK_PARAM_FORMAT_MASK = RK_PARAM_FORMAT_NORMALIZE - 1 };
enum rk_instance_flags : rk_ubyte { enum : rk_ubyte {
RK_INSTANCE_FLAG_SPAWNED = RK_FLAG(0), RK_INSTANCE_FLAG_SPAWNED = RK_FLAG(0),
RK_INSTANCE_FLAG_VISIBLE = RK_FLAG(1) RK_INSTANCE_FLAG_VISIBLE = RK_FLAG(1),
RK_INSTANCE_FLAGS_SPAWNED_VISIBLE = RK_INSTANCE_FLAG_SPAWNED | RK_INSTANCE_FLAG_VISIBLE
}; };
enum : rk_ubyte { RK_INSTANCE_FLAGS_SPAWNED_VISIBLE = RK_INSTANCE_FLAG_SPAWNED | RK_INSTANCE_FLAG_VISIBLE }; typedef rk_ushort rk_instance_index;
enum : rk_uint { RK_BATCH_MAX_SIZE = 65536 }; enum : rk_uint {
RK_BATCH_MAX_SIZE = 1 << (sizeof(rk_instance_index) * 8)
};
typedef rk_ushort rk_vertex_index;
typedef rk_ushort rk_mesh_index;
// param input types must be size compatible with an array of rk_param_input
typedef rk_uint rk_param_input;
#define RK_CHECK_PARAM_INPUT_TYPE(_t) static_assert(!(sizeof(_t) % sizeof(rk_param_input)))
struct rk_mesh { struct rk_mesh {
rk_uint base_index; rk_uint base_index;
@ -113,7 +132,7 @@ RK_EXPORT rk_vertices_t rk_create_vertices(
rk_uint nvertices, rk_uint nvertices,
rk_ubyte const * vertices, rk_ubyte const * vertices,
rk_uint nindices, rk_uint nindices,
rk_ushort const * indices, rk_vertex_index const * indices,
rk_uint nmeshes, rk_uint nmeshes,
rk_mesh const * meshes); rk_mesh const * meshes);
@ -126,8 +145,8 @@ RK_EXPORT void rk_fill_batch(
rk_batch_t batch, rk_batch_t batch,
rk_uint count, rk_uint count,
rk_instance_flags const * flags, rk_instance_flags const * flags,
rk_ushort const * meshes, rk_mesh_index const * meshes,
rk_ubyte const * const * params); rk_param_input const * const * params);
RK_EXPORT void rk_clear_buffer( RK_EXPORT void rk_clear_buffer(
rk_bool pixels, rk_bool pixels,

View File

@ -27,12 +27,6 @@ typedef void (*rk_MultiDrawElementsIndirectFunc)(rk_uint, rk_uint, const void *,
static rk_DrawElementsInstancedBaseInstanceFunc rk_DrawElementsInstancedBaseInstance = nullptr; static rk_DrawElementsInstancedBaseInstanceFunc rk_DrawElementsInstancedBaseInstance = nullptr;
static rk_MultiDrawElementsIndirectFunc rk_MultiDrawElementsIndirect = nullptr; static rk_MultiDrawElementsIndirectFunc rk_MultiDrawElementsIndirect = nullptr;
struct rk_bucket {
unsigned size;
unsigned count;
rk_ushort * indices;
};
static unsigned rk_nbuckets = 0; static unsigned rk_nbuckets = 0;
static rk_bucket * rk_buckets = nullptr; static rk_bucket * rk_buckets = nullptr;
@ -325,7 +319,7 @@ rk_vertices_t rk_create_vertices(
rk_uint nvertices, rk_uint nvertices,
rk_ubyte const * _vertices, rk_ubyte const * _vertices,
rk_uint nindices, rk_uint nindices,
rk_ushort const * indices, rk_vertex_index const * indices,
rk_uint nmeshes, rk_uint nmeshes,
rk_mesh const * meshes) { rk_mesh const * meshes) {
if (!format || !nvertices || !_vertices || !nindices || !indices) { if (!format || !nvertices || !_vertices || !nindices || !indices) {
@ -363,8 +357,8 @@ rk_vertices_t rk_create_vertices(
memcpy(vertices->format, format, (format_size + 1) * sizeof(rk_vertex_format)); memcpy(vertices->format, format, (format_size + 1) * sizeof(rk_vertex_format));
vertices->vertices = new rk_ubyte[nvertices * vertex_size]; vertices->vertices = new rk_ubyte[nvertices * vertex_size];
memcpy(vertices->vertices, _vertices, nvertices * vertex_size); memcpy(vertices->vertices, _vertices, nvertices * vertex_size);
vertices->indices = new rk_ushort[nindices]; vertices->indices = new rk_vertex_index[nindices];
memcpy(vertices->indices, indices, nindices * sizeof(rk_ushort)); memcpy(vertices->indices, indices, nindices * sizeof(rk_vertex_index));
vertices->meshes = new rk_mesh[nmeshes]; vertices->meshes = new rk_mesh[nmeshes];
memcpy(vertices->meshes, meshes, nmeshes * sizeof(rk_mesh)); memcpy(vertices->meshes, meshes, nmeshes * sizeof(rk_mesh));
vertices->vertices_buffer = 0; vertices->vertices_buffer = 0;
@ -383,7 +377,7 @@ static void rk_buckets_alloc(
for (unsigned index = 0; index < count; ++index) { for (unsigned index = 0; index < count; ++index) {
rk_bucket & bucket = rk_buckets[index]; rk_bucket & bucket = rk_buckets[index];
bucket.size = size; bucket.size = size;
bucket.indices = reinterpret_cast<rk_ushort *>(malloc(size * sizeof(rk_ushort))); bucket.indices = reinterpret_cast<rk_instance_index *>(malloc(size * sizeof(rk_instance_index)));
} }
reallocated = true; reallocated = true;
} }
@ -392,7 +386,8 @@ static void rk_buckets_alloc(
rk_bucket & bucket = rk_buckets[index]; rk_bucket & bucket = rk_buckets[index];
if (bucket.size < size) { if (bucket.size < size) {
bucket.size = size; bucket.size = size;
bucket.indices = reinterpret_cast<rk_ushort *>(realloc(bucket.indices, size * sizeof(rk_ushort))); bucket.indices = reinterpret_cast<rk_instance_index *>(
realloc(bucket.indices, size * sizeof(rk_instance_index)));
reallocated = true; reallocated = true;
} }
} }
@ -403,13 +398,15 @@ static void rk_buckets_alloc(
rk_bucket & bucket = rk_buckets[index]; rk_bucket & bucket = rk_buckets[index];
if (bucket.size < size) { if (bucket.size < size) {
bucket.size = size; bucket.size = size;
bucket.indices = reinterpret_cast<rk_ushort *>(realloc(bucket.indices, size * sizeof(rk_ushort))); bucket.indices = reinterpret_cast<rk_instance_index *>(
realloc(bucket.indices, size * sizeof(rk_instance_index)));
} }
} }
for (unsigned index = rk_nbuckets; index < count; ++index) { for (unsigned index = rk_nbuckets; index < count; ++index) {
rk_bucket & bucket = rk_buckets[index]; rk_bucket & bucket = rk_buckets[index];
bucket.size = size; bucket.size = size;
bucket.indices = reinterpret_cast<rk_ushort *>(malloc(size * sizeof(rk_ushort))); bucket.indices = reinterpret_cast<rk_instance_index *>(
malloc(size * sizeof(rk_instance_index)));
} }
rk_nbuckets = count; rk_nbuckets = count;
reallocated = true; reallocated = true;
@ -418,7 +415,7 @@ static void rk_buckets_alloc(
unsigned total_size = rk_nbuckets * sizeof(rk_bucket); unsigned total_size = rk_nbuckets * sizeof(rk_bucket);
for (unsigned index = 0; index < rk_nbuckets; ++index) { for (unsigned index = 0; index < rk_nbuckets; ++index) {
rk_bucket const & bucket = rk_buckets[index]; rk_bucket const & bucket = rk_buckets[index];
total_size += bucket.size * sizeof(rk_ushort); total_size += bucket.size * sizeof(rk_instance_index);
} }
printf("[RK] rk_buckets_alloc() -> %d KiB\n", total_size / 1024); printf("[RK] rk_buckets_alloc() -> %d KiB\n", total_size / 1024);
} }
@ -426,26 +423,26 @@ static void rk_buckets_alloc(
static void rk_pack_vec3_float( static void rk_pack_vec3_float(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_vec3_float * __restrict dst = reinterpret_cast<rk_vec3_float *>(_dst); rk_vec3_float * __restrict dst = reinterpret_cast<rk_vec3_float *>(_dst);
rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src); rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src);
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
*dst = src[*index]; *dst = src[*index];
} }
} }
static void rk_pack_vec3_short( static void rk_pack_vec3_short(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_vec3_short * __restrict dst = reinterpret_cast<rk_vec3_short *>(_dst); rk_vec3_short * __restrict dst = reinterpret_cast<rk_vec3_short *>(_dst);
rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src); rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src);
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
rk_vec3_float const & input = src[*index]; rk_vec3_float const & input = src[*index];
dst->x = static_cast<rk_short>(input.x); dst->x = static_cast<rk_short>(input.x);
dst->y = static_cast<rk_short>(input.y); dst->y = static_cast<rk_short>(input.y);
@ -456,14 +453,14 @@ static void rk_pack_vec3_short(
static void rk_pack_vec3_short_norm( static void rk_pack_vec3_short_norm(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_vec3_short * __restrict dst = reinterpret_cast<rk_vec3_short *>(_dst); rk_vec3_short * __restrict dst = reinterpret_cast<rk_vec3_short *>(_dst);
rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src); rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src);
#define _convert(s) (static_cast<rk_short>((s) * ((s) < 0.f ? 32768.f : 32767.f))) #define _convert(s) (static_cast<rk_short>((s) * ((s) < 0.f ? 32768.f : 32767.f)))
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
rk_vec3_float const & input = src[*index]; rk_vec3_float const & input = src[*index];
dst->x = _convert(input.x); dst->x = _convert(input.x);
dst->y = _convert(input.y); dst->y = _convert(input.y);
@ -475,14 +472,14 @@ static void rk_pack_vec3_short_norm(
static void rk_pack_vec3_int10( static void rk_pack_vec3_int10(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_vec3_int10 * __restrict dst = reinterpret_cast<rk_vec3_int10 *>(_dst); rk_vec3_int10 * __restrict dst = reinterpret_cast<rk_vec3_int10 *>(_dst);
rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src); rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src);
#define _convert(s) (static_cast<rk_int>((s)) & 1023) #define _convert(s) (static_cast<rk_int>((s)) & 1023)
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
rk_vec3_float const & input = src[*index]; rk_vec3_float const & input = src[*index];
*dst = _convert(input.x) | (_convert(input.y) << 10) | (_convert(input.z) << 20); *dst = _convert(input.x) | (_convert(input.y) << 10) | (_convert(input.z) << 20);
} }
@ -491,14 +488,14 @@ static void rk_pack_vec3_int10(
static void rk_pack_vec3_int10_norm( static void rk_pack_vec3_int10_norm(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_vec3_int10 * __restrict dst = reinterpret_cast<rk_vec3_int10 *>(_dst); rk_vec3_int10 * __restrict dst = reinterpret_cast<rk_vec3_int10 *>(_dst);
rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src); rk_vec3_float const * const __restrict src = reinterpret_cast<rk_vec3_float const *>(_src);
#define _convert(s) (static_cast<rk_int>((s) * ((s) < 0.f ? 512.f : 511.f)) & 1023) #define _convert(s) (static_cast<rk_int>((s) * ((s) < 0.f ? 512.f : 511.f)) & 1023)
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
rk_vec3_float const & input = src[*index]; rk_vec3_float const & input = src[*index];
*dst = _convert(input.x) | (_convert(input.y) << 10) | (_convert(input.z) << 20); *dst = _convert(input.x) | (_convert(input.y) << 10) | (_convert(input.z) << 20);
} }
@ -507,13 +504,13 @@ static void rk_pack_vec3_int10_norm(
static void rk_pack_mat3_float( static void rk_pack_mat3_float(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_mat3_float * __restrict dst = reinterpret_cast<rk_mat3_float *>(_dst); rk_mat3_float * __restrict dst = reinterpret_cast<rk_mat3_float *>(_dst);
rk_mat3_float const * const __restrict src = reinterpret_cast<rk_mat3_float const *>(_src); rk_mat3_float const * const __restrict src = reinterpret_cast<rk_mat3_float const *>(_src);
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
*dst = src[*index]; *dst = src[*index];
} }
#undef _convert #undef _convert
@ -521,14 +518,14 @@ static void rk_pack_mat3_float(
static void rk_pack_mat3_int10( static void rk_pack_mat3_int10(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_mat3_int10 * __restrict dst = reinterpret_cast<rk_mat3_int10 *>(_dst); rk_mat3_int10 * __restrict dst = reinterpret_cast<rk_mat3_int10 *>(_dst);
rk_mat3_float const * const __restrict src = reinterpret_cast<rk_mat3_float const *>(_src); rk_mat3_float const * const __restrict src = reinterpret_cast<rk_mat3_float const *>(_src);
#define _convert(s) (static_cast<rk_int>((s)) & 1023) #define _convert(s) (static_cast<rk_int>((s)) & 1023)
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
rk_mat3_float const & input = src[*index]; rk_mat3_float const & input = src[*index];
dst->x = _convert(input.x.x) | (_convert(input.x.y) << 10) | (_convert(input.x.z) << 20); dst->x = _convert(input.x.x) | (_convert(input.x.y) << 10) | (_convert(input.x.z) << 20);
dst->y = _convert(input.y.x) | (_convert(input.y.y) << 10) | (_convert(input.y.z) << 20); dst->y = _convert(input.y.x) | (_convert(input.y.y) << 10) | (_convert(input.y.z) << 20);
@ -539,14 +536,14 @@ static void rk_pack_mat3_int10(
static void rk_pack_mat3_int10_norm( static void rk_pack_mat3_int10_norm(
unsigned const count, unsigned const count,
rk_ushort const * const __restrict indices, rk_instance_index const * const __restrict indices,
rk_ubyte * __restrict _dst, rk_param_output * __restrict _dst,
rk_ubyte const * const __restrict _src) { rk_param_input const * const __restrict _src) {
rk_ushort const * const last_index = indices + count; rk_instance_index const * const last_index = indices + count;
rk_mat3_int10 * __restrict dst = reinterpret_cast<rk_mat3_int10 *>(_dst); rk_mat3_int10 * __restrict dst = reinterpret_cast<rk_mat3_int10 *>(_dst);
rk_mat3_float const * const __restrict src = reinterpret_cast<rk_mat3_float const *>(_src); rk_mat3_float const * const __restrict src = reinterpret_cast<rk_mat3_float const *>(_src);
#define _convert(s) (static_cast<rk_int>((s) * ((s) < 0.f ? 512.f : 511.f)) & 1023) #define _convert(s) (static_cast<rk_int>((s) * ((s) < 0.f ? 512.f : 511.f)) & 1023)
for (rk_ushort const * __restrict index = indices; index < last_index; ++index, ++dst) { for (rk_instance_index const * __restrict index = indices; index < last_index; ++index, ++dst) {
rk_mat3_float const & input = src[*index]; rk_mat3_float const & input = src[*index];
dst->x = _convert(input.x.x) | (_convert(input.x.y) << 10) | (_convert(input.x.z) << 20); dst->x = _convert(input.x.x) | (_convert(input.x.y) << 10) | (_convert(input.x.z) << 20);
dst->y = _convert(input.y.x) | (_convert(input.y.y) << 10) | (_convert(input.y.z) << 20); dst->y = _convert(input.y.x) | (_convert(input.y.y) << 10) | (_convert(input.y.z) << 20);
@ -614,9 +611,11 @@ rk_batch_t rk_create_batch(
batch->nparams = nparams; batch->nparams = nparams;
batch->vertices = vertices; batch->vertices = vertices;
batch->flags = new rk_instance_flags[max_size]; batch->flags = new rk_instance_flags[max_size];
batch->meshes = new rk_ushort[max_size]; memset(batch->flags, 0xFF, max_size * sizeof(rk_instance_flags));
batch->indices = new rk_ushort[max_size]; batch->meshes = new rk_mesh_index[max_size];
memset(batch->indices, 0xFF, max_size * sizeof(rk_ushort)); memset(batch->meshes, 0xFF, max_size * sizeof(rk_mesh_index));
batch->indices = new rk_instance_index[max_size];
memset(batch->indices, 0, max_size * sizeof(rk_instance_index));
batch->commands = new rk_command[vertices->nmeshes]; batch->commands = new rk_command[vertices->nmeshes];
memset(batch->commands, 0, vertices->nmeshes * sizeof(rk_command)); memset(batch->commands, 0, vertices->nmeshes * sizeof(rk_command));
if (nparams) { if (nparams) {
@ -637,7 +636,8 @@ rk_batch_t rk_create_batch(
} else { } else {
glGenBuffers(1, &vertices->indices_buffer); glGenBuffers(1, &vertices->indices_buffer);
glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, vertices->indices_buffer); glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, vertices->indices_buffer);
glBufferData(GL_ELEMENT_ARRAY_BUFFER, vertices->nindices * sizeof(rk_ushort), vertices->indices, GL_STATIC_DRAW); glBufferData(GL_ELEMENT_ARRAY_BUFFER,
vertices->nindices * sizeof(rk_vertex_index), vertices->indices, GL_STATIC_DRAW);
} }
if (rk_MultiDrawElementsIndirect) { if (rk_MultiDrawElementsIndirect) {
glGenBuffers(1, &batch->commands_buffer); glGenBuffers(1, &batch->commands_buffer);
@ -747,7 +747,10 @@ rk_batch_t rk_create_batch(
break; break;
} }
glVertexBindingDivisor(binding, 1); glVertexBindingDivisor(binding, 1);
param->source = new rk_ubyte[max_size * param->src_size]; param->src_len = param->src_size / sizeof(rk_param_input);
param->dst_len = param->dst_size / sizeof(rk_param_output);
param->source = new rk_param_input[max_size * param->src_len];
memset(param->source, 0xFF, max_size * param->src_size);
offset += max_size * param->dst_size; offset += max_size * param->dst_size;
} }
} }
@ -763,20 +766,20 @@ static void rk_sort_batch(
bucket->count = 0; bucket->count = 0;
} }
rk_instance_flags const * __restrict flags = batch.flags; rk_instance_flags const * __restrict flags = batch.flags;
rk_ushort const * __restrict mesh_index = batch.meshes; rk_mesh_index const * __restrict mesh_index = batch.meshes;
for (unsigned index = 0; index < batch.count; ++index, ++flags, ++mesh_index) { for (unsigned index = 0; index < batch.count; ++index, ++flags, ++mesh_index) {
if ((*flags & RK_INSTANCE_FLAGS_SPAWNED_VISIBLE) == RK_INSTANCE_FLAGS_SPAWNED_VISIBLE) { if ((*flags & RK_INSTANCE_FLAGS_SPAWNED_VISIBLE) == RK_INSTANCE_FLAGS_SPAWNED_VISIBLE) {
rk_bucket & __restrict bucket = rk_buckets[*mesh_index]; rk_bucket & __restrict bucket = rk_buckets[*mesh_index];
bucket.indices[bucket.count++] = index; bucket.indices[bucket.count++] = index;
} }
} }
rk_ushort * __restrict indices = batch.indices; rk_instance_index * __restrict indices = batch.indices;
rk_command * __restrict command = batch.commands; rk_command * __restrict command = batch.commands;
rk_mesh const * __restrict mesh = batch.vertices->meshes; rk_mesh const * __restrict mesh = batch.vertices->meshes;
for (rk_bucket const * __restrict bucket = rk_buckets; bucket < last_bucket; ++bucket, ++mesh) { for (rk_bucket const * __restrict bucket = rk_buckets; bucket < last_bucket; ++bucket, ++mesh) {
if (bucket->count) { if (bucket->count) {
memcpy(indices, bucket->indices, bucket->count * sizeof(rk_ushort)); memcpy(indices, bucket->indices, bucket->count * sizeof(rk_instance_index));
command->nvertices = static_cast<GLuint>(mesh->ntriangles) * 3; command->nvertices = mesh->ntriangles * 3;
command->ninstances = bucket->count; command->ninstances = bucket->count;
command->base_index = mesh->base_index; command->base_index = mesh->base_index;
command->base_instance = indices - batch.indices; command->base_instance = indices - batch.indices;
@ -800,7 +803,7 @@ static void rk_pack_batch(
if (param->dirty) { if (param->dirty) {
param->dirty = false; param->dirty = false;
if (batch.ninstances) { if (batch.ninstances) {
rk_ubyte * const dst = reinterpret_cast<rk_ubyte *>( rk_param_output * const dst = reinterpret_cast<rk_param_output *>(
glMapBufferRange(GL_ARRAY_BUFFER, param->offset, batch.ninstances * param->dst_size, glMapBufferRange(GL_ARRAY_BUFFER, param->offset, batch.ninstances * param->dst_size,
GL_MAP_WRITE_BIT | GL_MAP_INVALIDATE_RANGE_BIT | GL_MAP_UNSYNCHRONIZED_BIT)); GL_MAP_WRITE_BIT | GL_MAP_INVALIDATE_RANGE_BIT | GL_MAP_UNSYNCHRONIZED_BIT));
if (dst) { if (dst) {
@ -819,8 +822,8 @@ void rk_fill_batch(
rk_batch_t _batch, rk_batch_t _batch,
rk_uint count, rk_uint count,
rk_instance_flags const * flags, rk_instance_flags const * flags,
rk_ushort const * meshes, rk_mesh_index const * meshes,
rk_ubyte const * const * params) { rk_param_input const * const * params) {
rk_batch const * const batch = reinterpret_cast<rk_batch const *>(_batch); rk_batch const * const batch = reinterpret_cast<rk_batch const *>(_batch);
if (!batch || !count || count > batch->max_size) { if (!batch || !count || count > batch->max_size) {
rk_printf("rk_fill_batch(): invalid params."); rk_printf("rk_fill_batch(): invalid params.");
@ -831,7 +834,7 @@ void rk_fill_batch(
if (batch->nparams) { if (batch->nparams) {
got_all_params = (params != nullptr); got_all_params = (params != nullptr);
if (params) { if (params) {
for (rk_ubyte const * const * param = params; param < params + batch->nparams; ++param) { for (rk_param_input const * const * param = params; param < params + batch->nparams; ++param) {
bool const got_param = (*param != nullptr); bool const got_param = (*param != nullptr);
got_any_params |= got_param; got_any_params |= got_param;
got_all_params &= got_param; got_all_params &= got_param;
@ -849,23 +852,20 @@ void rk_fill_batch(
return; return;
} }
batch->count = count; batch->count = count;
bool const cmp_flags = (flags && bool const cmp_flags = (flags && rk_cmp_memcpy(batch->flags, flags, batch->count));
rk_cmp_memcpy<sizeof(rk_ubyte)>(batch->flags, flags, batch->count * sizeof(rk_instance_flags))); bool const cmp_meshes = (meshes && rk_cmp_memcpy(batch->meshes, meshes, batch->count));
bool const cmp_meshes = (meshes &&
rk_cmp_memcpy<sizeof(rk_ushort)>(batch->meshes, meshes, batch->count * sizeof(rk_mesh)));
bool const need_sorting = (cmp_flags || cmp_meshes || resized); bool const need_sorting = (cmp_flags || cmp_meshes || resized);
if (batch->nparams) { if (batch->nparams) {
rk_parameter const * const last_param = batch->params + batch->nparams; rk_parameter const * const last_param = batch->params + batch->nparams;
if (got_any_params) { if (got_any_params) {
rk_ubyte const * const * src = params; rk_param_input const * const * src = params;
for (rk_parameter const * dst = batch->params; dst < last_param; ++dst, ++src) { for (rk_parameter const * param = batch->params; param < last_param; ++param, ++src) {
dst->dirty = ((*src && param->dirty =
rk_cmp_memcpy<sizeof(rk_uint)>(dst->source, *src, batch->count * dst->src_size)) ((*src && rk_cmp_memcpy(param->source, *src, batch->count * param->src_len)) || need_sorting);
|| need_sorting);
} }
} else if (need_sorting) { } else if (need_sorting) {
for (rk_parameter const * dst = batch->params; dst < last_param; ++dst) { for (rk_parameter const * param = batch->params; param < last_param; ++param) {
dst->dirty = true; param->dirty = true;
} }
} }
} }

View File

@ -17,6 +17,7 @@
#define _RK_ENGINE_RENDER_OPENGLES_H #define _RK_ENGINE_RENDER_OPENGLES_H
#include "../types.hpp" #include "../types.hpp"
#include "../math.hpp"
#include <GLES3/gl32.h> #include <GLES3/gl32.h>
#include <GLES3/gl3ext.h> #include <GLES3/gl3ext.h>
#include <GLES3/gl3platform.h> #include <GLES3/gl3platform.h>
@ -44,7 +45,7 @@ struct rk_vertices {
unsigned nmeshes; unsigned nmeshes;
rk_vertex_format * format; rk_vertex_format * format;
rk_ubyte * vertices; rk_ubyte * vertices;
rk_ushort * indices; rk_vertex_index * indices;
rk_mesh * meshes; rk_mesh * meshes;
GLuint vertices_buffer; GLuint vertices_buffer;
GLuint indices_buffer; GLuint indices_buffer;
@ -58,12 +59,19 @@ struct rk_command {
GLuint base_instance; GLuint base_instance;
}; };
// param output types must be size compatible with an array of rk_param_output
typedef rk_uint rk_param_output;
#define RK_CHECK_PARAM_OUTPUT_TYPE(_t) static_assert(!(sizeof(_t) % sizeof(rk_param_output)))
struct rk_vec3_float { struct rk_vec3_float {
float x; float x;
float y; float y;
float z; float z;
}; };
static_assert(sizeof(rk_vec3_float) == sizeof(rk_vec3));
struct rk_vec3_short { struct rk_vec3_short {
rk_short x; rk_short x;
rk_short y; rk_short y;
@ -80,17 +88,29 @@ struct rk_mat3_float {
rk_vec3_float z; rk_vec3_float z;
}; };
static_assert(sizeof(rk_mat3_float) == sizeof(rk_mat3));
struct rk_mat3_int10 { struct rk_mat3_int10 {
rk_vec3_int10 x; rk_vec3_int10 x;
rk_vec3_int10 y; rk_vec3_int10 y;
rk_vec3_int10 z; rk_vec3_int10 z;
}; };
RK_CHECK_PARAM_INPUT_TYPE(rk_vec3_float);
RK_CHECK_PARAM_INPUT_TYPE(rk_mat3_float);
RK_CHECK_PARAM_OUTPUT_TYPE(rk_vec3_float);
RK_CHECK_PARAM_OUTPUT_TYPE(rk_vec3_short);
RK_CHECK_PARAM_OUTPUT_TYPE(rk_vec3_int10);
RK_CHECK_PARAM_OUTPUT_TYPE(rk_vec3_uint10);
RK_CHECK_PARAM_OUTPUT_TYPE(rk_mat3_float);
RK_CHECK_PARAM_OUTPUT_TYPE(rk_mat3_int10);
typedef void (*rk_packer)( typedef void (*rk_packer)(
unsigned const, // count unsigned const, // count
rk_ushort const * const, // indices rk_instance_index const * const, // indices
rk_ubyte *, // dst rk_param_output *, // dst
rk_ubyte const * const); // src rk_param_input const * const); // src
struct rk_parameter { struct rk_parameter {
mutable bool dirty; mutable bool dirty;
@ -98,10 +118,18 @@ struct rk_parameter {
unsigned offset; unsigned offset;
unsigned src_size; unsigned src_size;
unsigned dst_size; unsigned dst_size;
rk_ubyte * source; unsigned src_len;
unsigned dst_len;
rk_param_input * source;
rk_packer packer; rk_packer packer;
}; };
struct rk_bucket {
unsigned size;
unsigned count;
rk_instance_index * indices;
};
enum rk_batch_state { enum rk_batch_state {
RK_BATCH_STATE_EMPTY = 0, RK_BATCH_STATE_EMPTY = 0,
RK_BATCH_STATE_FILLED = 1, RK_BATCH_STATE_FILLED = 1,
@ -118,8 +146,8 @@ struct rk_batch {
unsigned nparams; unsigned nparams;
rk_vertices const * vertices; rk_vertices const * vertices;
rk_instance_flags * flags; rk_instance_flags * flags;
rk_ushort * meshes; rk_mesh_index * meshes;
rk_ushort * indices; rk_instance_index * indices;
rk_command * commands; rk_command * commands;
rk_parameter * params; rk_parameter * params;
GLuint vertex_array; GLuint vertex_array;