mirror of
https://github.com/scratchfoundation/bgfx.git
synced 2024-11-25 09:08:22 -05:00
1427 lines
36 KiB
C++
Executable file
1427 lines
36 KiB
C++
Executable file
/*
|
|
* Copyright 2011-2013 Branimir Karadzic. All rights reserved.
|
|
* License: http://www.opensource.org/licenses/BSD-2-Clause
|
|
*/
|
|
|
|
#include "bgfx_p.h"
|
|
|
|
#if BGFX_CONFIG_USE_TINYSTL
|
|
namespace tinystl
|
|
{
|
|
void* bgfx_allocator::static_allocate(size_t _bytes)
|
|
{
|
|
return bgfx::g_realloc(NULL, _bytes);
|
|
}
|
|
|
|
void bgfx_allocator::static_deallocate(void* _ptr, size_t /*_bytes*/)
|
|
{
|
|
bgfx::g_free(_ptr);
|
|
}
|
|
|
|
} // namespace tinystl
|
|
#endif // BGFX_CONFIG_USE_TINYSTL
|
|
|
|
namespace bgfx
|
|
{
|
|
#define BGFX_MAIN_THREAD_MAGIC 0x78666762
|
|
|
|
#if BGFX_CONFIG_MULTITHREADED
|
|
# define BGFX_CHECK_MAIN_THREAD() BX_CHECK(BGFX_MAIN_THREAD_MAGIC == s_threadIndex, "Must be called from main thread.")
|
|
# define BGFX_CHECK_RENDER_THREAD() BX_CHECK(BGFX_MAIN_THREAD_MAGIC != s_threadIndex, "Must be called from render thread.")
|
|
#else
|
|
# define BGFX_CHECK_MAIN_THREAD()
|
|
# define BGFX_CHECK_RENDER_THREAD()
|
|
#endif // BGFX_CONFIG_MULTITHREADED
|
|
|
|
#if BX_PLATFORM_ANDROID
|
|
::ANativeWindow* g_bgfxAndroidWindow = NULL;
|
|
void androidSetWindow(ANativeWindow* _window)
|
|
{
|
|
g_bgfxAndroidWindow = _window;
|
|
}
|
|
#elif BX_PLATFORM_OSX
|
|
void* g_bgfxNSWindow = NULL;
|
|
|
|
void osxSetNSWindow(void* _window)
|
|
{
|
|
g_bgfxNSWindow = _window;
|
|
}
|
|
#elif BX_PLATFORM_WINDOWS
|
|
::HWND g_bgfxHwnd = NULL;
|
|
|
|
void winSetHwnd(::HWND _window)
|
|
{
|
|
g_bgfxHwnd = _window;
|
|
}
|
|
#endif // BX_PLATFORM_*
|
|
|
|
struct CallbackStub : public CallbackI
|
|
{
|
|
virtual ~CallbackStub()
|
|
{
|
|
}
|
|
|
|
virtual void fatal(Fatal::Enum _code, const char* _str) BX_OVERRIDE
|
|
{
|
|
BX_TRACE("0x%08x: %s", _code, _str);
|
|
abort();
|
|
}
|
|
|
|
virtual uint32_t cacheReadSize(uint64_t /*_id*/) BX_OVERRIDE
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
virtual bool cacheRead(uint64_t /*_id*/, void* /*_data*/, uint32_t /*_size*/) BX_OVERRIDE
|
|
{
|
|
return false;
|
|
}
|
|
|
|
virtual void cacheWrite(uint64_t /*_id*/, const void* /*_data*/, uint32_t /*_size*/) BX_OVERRIDE
|
|
{
|
|
}
|
|
|
|
virtual void screenShot(const char* _filePath, uint32_t _width, uint32_t _height, uint32_t _pitch, const void* _data, uint32_t /*_size*/, bool _yflip) BX_OVERRIDE
|
|
{
|
|
saveTga(_filePath, _width, _height, _pitch, _data, false, _yflip);
|
|
}
|
|
|
|
virtual void captureBegin(uint32_t /*_width*/, uint32_t /*_height*/, uint32_t /*_pitch*/, TextureFormat::Enum /*_format*/, bool /*_yflip*/) BX_OVERRIDE
|
|
{
|
|
BX_TRACE("Warning: using capture without callback (a.k.a. pointless).");
|
|
}
|
|
|
|
virtual void captureEnd() BX_OVERRIDE
|
|
{
|
|
}
|
|
|
|
virtual void captureFrame(const void* /*_data*/, uint32_t /*_size*/) BX_OVERRIDE
|
|
{
|
|
}
|
|
};
|
|
|
|
static CallbackStub s_callbackStub;
|
|
|
|
static void* reallocStub(void* _ptr, size_t _size)
|
|
{
|
|
void* ptr = ::realloc(_ptr, _size);
|
|
BX_CHECK(NULL != ptr, "Out of memory!");
|
|
// BX_TRACE("alloc %d, %p", _size, ptr);
|
|
return ptr;
|
|
}
|
|
|
|
static void freeStub(void* _ptr)
|
|
{
|
|
// BX_TRACE("free %p", _ptr);
|
|
::free(_ptr);
|
|
}
|
|
|
|
CallbackI* g_callback = &s_callbackStub;
|
|
ReallocFn g_realloc = reallocStub;
|
|
FreeFn g_free = freeStub;
|
|
|
|
static BX_THREAD uint32_t s_threadIndex = 0;
|
|
static Context s_ctx;
|
|
|
|
void fatal(Fatal::Enum _code, const char* _format, ...)
|
|
{
|
|
char temp[8192];
|
|
|
|
va_list argList;
|
|
va_start(argList, _format);
|
|
vsnprintf(temp, sizeof(temp), _format, argList);
|
|
va_end(argList);
|
|
|
|
temp[sizeof(temp)-1] = '\0';
|
|
|
|
g_callback->fatal(_code, temp);
|
|
}
|
|
|
|
inline void vec4MulMtx(float* __restrict _result, const float* __restrict _vec, const float* __restrict _mat)
|
|
{
|
|
_result[0] = _vec[0] * _mat[ 0] + _vec[1] * _mat[4] + _vec[2] * _mat[ 8] + _vec[3] * _mat[12];
|
|
_result[1] = _vec[0] * _mat[ 1] + _vec[1] * _mat[5] + _vec[2] * _mat[ 9] + _vec[3] * _mat[13];
|
|
_result[2] = _vec[0] * _mat[ 2] + _vec[1] * _mat[6] + _vec[2] * _mat[10] + _vec[3] * _mat[14];
|
|
_result[3] = _vec[0] * _mat[ 3] + _vec[1] * _mat[7] + _vec[2] * _mat[11] + _vec[3] * _mat[15];
|
|
}
|
|
|
|
void mtxMul(float* __restrict _result, const float* __restrict _a, const float* __restrict _b)
|
|
{
|
|
vec4MulMtx(&_result[ 0], &_a[ 0], _b);
|
|
vec4MulMtx(&_result[ 4], &_a[ 4], _b);
|
|
vec4MulMtx(&_result[ 8], &_a[ 8], _b);
|
|
vec4MulMtx(&_result[12], &_a[12], _b);
|
|
}
|
|
|
|
void mtxOrtho(float* _result, float _left, float _right, float _bottom, float _top, float _near, float _far)
|
|
{
|
|
const float aa = 2.0f/(_right - _left);
|
|
const float bb = 2.0f/(_top - _bottom);
|
|
const float cc = 1.0f/(_far - _near);
|
|
const float dd = (_left + _right)/(_left - _right);
|
|
const float ee = (_top + _bottom)/(_bottom - _top);
|
|
const float ff = _near / (_near - _far);
|
|
|
|
memset(_result, 0, sizeof(float)*16);
|
|
_result[0] = aa;
|
|
_result[5] = bb;
|
|
_result[10] = cc;
|
|
_result[12] = dd;
|
|
_result[13] = ee;
|
|
_result[14] = ff;
|
|
_result[15] = 1.0f;
|
|
}
|
|
|
|
void saveTga(const char* _filePath, uint32_t _width, uint32_t _height, uint32_t _srcPitch, const void* _src, bool _grayscale, bool _yflip)
|
|
{
|
|
FILE* file = fopen(_filePath, "wb");
|
|
if ( NULL != file )
|
|
{
|
|
uint8_t type = _grayscale ? 3 : 2;
|
|
uint8_t bpp = _grayscale ? 8 : 32;
|
|
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(type, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(0, file);
|
|
putc(_width&0xff, file);
|
|
putc( (_width>>8)&0xff, file);
|
|
putc(_height&0xff, file);
|
|
putc( (_height>>8)&0xff, file);
|
|
putc(bpp, file);
|
|
putc(32, file);
|
|
|
|
uint32_t dstPitch = _width*bpp/8;
|
|
if (_yflip)
|
|
{
|
|
uint8_t* data = (uint8_t*)_src + _srcPitch*_height - _srcPitch;
|
|
for (uint32_t yy = 0; yy < _height; ++yy)
|
|
{
|
|
fwrite(data, dstPitch, 1, file);
|
|
data -= _srcPitch;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
uint8_t* data = (uint8_t*)_src;
|
|
for (uint32_t yy = 0; yy < _height; ++yy)
|
|
{
|
|
fwrite(data, dstPitch, 1, file);
|
|
data += _srcPitch;
|
|
}
|
|
}
|
|
|
|
fclose(file);
|
|
}
|
|
}
|
|
|
|
#include "charset.h"
|
|
|
|
void charsetFillTexture(const uint8_t* _charset, uint8_t* _rgba, uint32_t _height, uint32_t _pitch, uint32_t _bpp)
|
|
{
|
|
for (uint32_t ii = 0; ii < 256; ++ii)
|
|
{
|
|
uint8_t* pix = &_rgba[ii*8*_bpp];
|
|
for (uint32_t yy = 0; yy < _height; ++yy)
|
|
{
|
|
for (uint32_t xx = 0; xx < 8; ++xx)
|
|
{
|
|
uint8_t bit = 1<<(7-xx);
|
|
memset(&pix[xx*_bpp], _charset[ii*_height+yy]&bit ? 255 : 0, _bpp);
|
|
}
|
|
|
|
pix += _pitch;
|
|
}
|
|
}
|
|
}
|
|
|
|
static const uint32_t numCharsPerBatch = 1024;
|
|
static const uint32_t numBatchVertices = numCharsPerBatch*4;
|
|
static const uint32_t numBatchIndices = numCharsPerBatch*6;
|
|
|
|
void TextVideoMemBlitter::init()
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
m_decl.begin();
|
|
m_decl.add(Attrib::Position, 3, AttribType::Float);
|
|
m_decl.add(Attrib::Color0, 4, AttribType::Uint8, true);
|
|
m_decl.add(Attrib::Color1, 4, AttribType::Uint8, true);
|
|
m_decl.add(Attrib::TexCoord0, 2, AttribType::Float);
|
|
m_decl.end();
|
|
|
|
uint16_t width = 2048;
|
|
uint16_t height = 24;
|
|
uint8_t bpp = 1;
|
|
uint32_t pitch = width*bpp;
|
|
|
|
const Memory* mem;
|
|
|
|
mem = alloc(pitch*height);
|
|
uint8_t* rgba = mem->data;
|
|
charsetFillTexture(vga8x8, rgba, 8, pitch, bpp);
|
|
charsetFillTexture(vga8x16, &rgba[8*pitch], 16, pitch, bpp);
|
|
m_texture = createTexture2D(width, height, 1, TextureFormat::L8
|
|
, BGFX_TEXTURE_MIN_POINT
|
|
| BGFX_TEXTURE_MAG_POINT
|
|
| BGFX_TEXTURE_MIP_POINT
|
|
| BGFX_TEXTURE_U_CLAMP
|
|
| BGFX_TEXTURE_V_CLAMP
|
|
, mem
|
|
);
|
|
|
|
#if BGFX_CONFIG_RENDERER_DIRECT3D9
|
|
mem = makeRef(vs_debugfont_dx9, sizeof(vs_debugfont_dx9) );
|
|
#elif BGFX_CONFIG_RENDERER_DIRECT3D11
|
|
mem = makeRef(vs_debugfont_dx11, sizeof(vs_debugfont_dx11) );
|
|
#else
|
|
mem = makeRef(vs_debugfont_glsl, sizeof(vs_debugfont_glsl) );
|
|
#endif // BGFX_CONFIG_RENDERER_
|
|
VertexShaderHandle vsh = createVertexShader(mem);
|
|
|
|
#if BGFX_CONFIG_RENDERER_DIRECT3D9
|
|
mem = makeRef(fs_debugfont_dx9, sizeof(fs_debugfont_dx9) );
|
|
#elif BGFX_CONFIG_RENDERER_DIRECT3D11
|
|
mem = makeRef(fs_debugfont_dx11, sizeof(fs_debugfont_dx11) );
|
|
#else
|
|
mem = makeRef(fs_debugfont_glsl, sizeof(fs_debugfont_glsl) );
|
|
#endif // BGFX_CONFIG_RENDERER_
|
|
FragmentShaderHandle fsh = createFragmentShader(mem);
|
|
|
|
m_program = createProgram(vsh, fsh);
|
|
destroyVertexShader(vsh);
|
|
destroyFragmentShader(fsh);
|
|
|
|
m_vb = s_ctx.createTransientVertexBuffer(numBatchVertices*m_decl.m_stride, &m_decl);
|
|
m_ib = s_ctx.createTransientIndexBuffer(numBatchIndices*2);
|
|
}
|
|
|
|
void TextVideoMemBlitter::shutdown()
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
|
|
destroyProgram(m_program);
|
|
destroyTexture(m_texture);
|
|
s_ctx.destroyTransientVertexBuffer(m_vb);
|
|
s_ctx.destroyTransientIndexBuffer(m_ib);
|
|
}
|
|
|
|
void TextVideoMemBlitter::blit(const TextVideoMem& _mem)
|
|
{
|
|
BGFX_CHECK_RENDER_THREAD();
|
|
struct Vertex
|
|
{
|
|
float m_x;
|
|
float m_y;
|
|
float m_z;
|
|
uint32_t m_fg;
|
|
uint32_t m_bg;
|
|
float m_u;
|
|
float m_v;
|
|
};
|
|
|
|
static uint32_t palette[16] =
|
|
{
|
|
0x0,
|
|
0xff0000cc,
|
|
0xff069a4e,
|
|
0xff00a0c4,
|
|
0xffa46534,
|
|
0xff7b5075,
|
|
0xff9a9806,
|
|
0xffcfd7d3,
|
|
0xff535755,
|
|
0xff2929ef,
|
|
0xff34e28a,
|
|
0xff4fe9fc,
|
|
0xffcf9f72,
|
|
0xffa87fad,
|
|
0xffe2e234,
|
|
0xffeceeee,
|
|
};
|
|
|
|
uint32_t yy = 0;
|
|
uint32_t xx = 0;
|
|
|
|
const float texelWidth = 1.0f/2048.0f;
|
|
const float texelWidthHalf = texelWidth*0.5f;
|
|
const float texelHeight = 1.0f/24.0f;
|
|
#if BGFX_CONFIG_RENDERER_DIRECT3D9
|
|
const float texelHeightHalf = texelHeight*0.5f;
|
|
#else
|
|
const float texelHeightHalf = 0.0f;
|
|
#endif // BGFX_CONFIG_RENDERER_
|
|
const float utop = (_mem.m_small ? 0.0f : 8.0f)*texelHeight + texelHeightHalf;
|
|
const float ubottom = (_mem.m_small ? 8.0f : 24.0f)*texelHeight + texelHeightHalf;
|
|
const float fontHeight = (_mem.m_small ? 8.0f : 16.0f);
|
|
|
|
setup();
|
|
|
|
for (;yy < _mem.m_height;)
|
|
{
|
|
Vertex* vertex = (Vertex*)m_vb->data;
|
|
uint16_t* indices = (uint16_t*)m_ib->data;
|
|
uint32_t startVertex = 0;
|
|
uint32_t numIndices = 0;
|
|
|
|
for (; yy < _mem.m_height && numIndices < numBatchIndices; ++yy)
|
|
{
|
|
xx = xx < _mem.m_width ? xx : 0;
|
|
const uint8_t* line = &_mem.m_mem[(yy*_mem.m_width+xx)*2];
|
|
|
|
for (; xx < _mem.m_width && numIndices < numBatchIndices; ++xx)
|
|
{
|
|
uint8_t ch = line[0];
|
|
uint8_t attr = line[1];
|
|
|
|
if (0 != (ch|attr)
|
|
&& (' ' != ch || 0 != (attr&0xf0) ) )
|
|
{
|
|
uint32_t fg = palette[attr&0xf];
|
|
uint32_t bg = palette[(attr>>4)&0xf];
|
|
|
|
Vertex vert[4] =
|
|
{
|
|
{ (xx )*8.0f, (yy )*fontHeight, 0.0f, fg, bg, (ch )*8.0f*texelWidth - texelWidthHalf, utop },
|
|
{ (xx+1)*8.0f, (yy )*fontHeight, 0.0f, fg, bg, (ch+1)*8.0f*texelWidth - texelWidthHalf, utop },
|
|
{ (xx+1)*8.0f, (yy+1)*fontHeight, 0.0f, fg, bg, (ch+1)*8.0f*texelWidth - texelWidthHalf, ubottom },
|
|
{ (xx )*8.0f, (yy+1)*fontHeight, 0.0f, fg, bg, (ch )*8.0f*texelWidth - texelWidthHalf, ubottom },
|
|
};
|
|
|
|
memcpy(vertex, vert, sizeof(vert) );
|
|
vertex += 4;
|
|
|
|
indices[0] = startVertex+0;
|
|
indices[1] = startVertex+1;
|
|
indices[2] = startVertex+2;
|
|
indices[3] = startVertex+2;
|
|
indices[4] = startVertex+3;
|
|
indices[5] = startVertex+0;
|
|
|
|
startVertex += 4;
|
|
indices += 6;
|
|
|
|
numIndices += 6;
|
|
}
|
|
|
|
line += 2;
|
|
}
|
|
|
|
if (numIndices >= numBatchIndices)
|
|
{
|
|
break;
|
|
}
|
|
}
|
|
|
|
render(numIndices);
|
|
}
|
|
}
|
|
|
|
void ClearQuad::init()
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
#if BGFX_CONFIG_CLEAR_QUAD
|
|
m_decl.begin();
|
|
m_decl.add(Attrib::Position, 3, AttribType::Float);
|
|
m_decl.add(Attrib::Color0, 4, AttribType::Uint8, true);
|
|
m_decl.end();
|
|
|
|
const Memory* mem;
|
|
|
|
# if BGFX_CONFIG_RENDERER_DIRECT3D11
|
|
mem = makeRef(vs_clear_dx11, sizeof(vs_clear_dx11) );
|
|
# elif BGFX_CONFIG_RENDERER_OPENGL
|
|
mem = makeRef(vs_clear_glsl, sizeof(vs_clear_glsl) );
|
|
# endif // BGFX_CONFIG_RENDERER_*
|
|
VertexShaderHandle vsh = createVertexShader(mem);
|
|
|
|
# if BGFX_CONFIG_RENDERER_DIRECT3D11
|
|
mem = makeRef(fs_clear_dx11, sizeof(fs_clear_dx11) );
|
|
# elif BGFX_CONFIG_RENDERER_OPENGL
|
|
mem = makeRef(fs_clear_glsl, sizeof(fs_clear_glsl) );
|
|
# endif // BGFX_CONFIG_RENDERER_*
|
|
FragmentShaderHandle fsh = createFragmentShader(mem);
|
|
|
|
m_program = createProgram(vsh, fsh);
|
|
destroyVertexShader(vsh);
|
|
destroyFragmentShader(fsh);
|
|
|
|
m_vb = s_ctx.createTransientVertexBuffer(4*m_decl.m_stride, &m_decl);
|
|
|
|
mem = alloc(6*sizeof(uint16_t) );
|
|
uint16_t* indices = (uint16_t*)mem->data;
|
|
indices[0] = 0;
|
|
indices[1] = 1;
|
|
indices[2] = 2;
|
|
indices[3] = 2;
|
|
indices[4] = 3;
|
|
indices[5] = 0;
|
|
m_ib = s_ctx.createIndexBuffer(mem);
|
|
#endif // BGFX_CONFIG_CLEAR_QUAD
|
|
}
|
|
|
|
void ClearQuad::shutdown()
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
|
|
#if BGFX_CONFIG_CLEAR_QUAD
|
|
destroyProgram(m_program);
|
|
destroyIndexBuffer(m_ib);
|
|
s_ctx.destroyTransientVertexBuffer(m_vb);
|
|
#endif // BGFX_CONFIG_CLEAR_QUAD
|
|
}
|
|
|
|
static const char* s_predefinedName[PredefinedUniform::Count] =
|
|
{
|
|
"u_viewRect",
|
|
"u_viewTexel",
|
|
"u_view",
|
|
"u_viewProj",
|
|
"u_viewProjX",
|
|
"u_model",
|
|
"u_modelView",
|
|
"u_modelViewProj",
|
|
"u_modelViewProjX",
|
|
"u_alphaRef",
|
|
};
|
|
|
|
const char* getPredefinedUniformName(PredefinedUniform::Enum _enum)
|
|
{
|
|
return s_predefinedName[_enum];
|
|
}
|
|
|
|
PredefinedUniform::Enum nameToPredefinedUniformEnum(const char* _name)
|
|
{
|
|
for (uint32_t ii = 0; ii < PredefinedUniform::Count; ++ii)
|
|
{
|
|
if (0 == strcmp(_name, s_predefinedName[ii]) )
|
|
{
|
|
return PredefinedUniform::Enum(ii);
|
|
}
|
|
}
|
|
|
|
return PredefinedUniform::Count;
|
|
}
|
|
|
|
void Frame::submit(uint8_t _id, int32_t _depth)
|
|
{
|
|
if (m_discard)
|
|
{
|
|
m_discard = false;
|
|
return;
|
|
}
|
|
|
|
if (BGFX_CONFIG_MAX_DRAW_CALLS-1 <= m_num
|
|
|| (0 == m_state.m_numVertices && 0 == m_state.m_numIndices) )
|
|
{
|
|
++m_numDropped;
|
|
return;
|
|
}
|
|
|
|
BX_WARN(invalidHandle != m_key.m_program, "Program with invalid handle");
|
|
if (invalidHandle != m_key.m_program)
|
|
{
|
|
m_key.m_depth = _depth;
|
|
m_key.m_view = _id;
|
|
m_key.m_seq = s_ctx.m_seq[_id] & s_ctx.m_seqMask[_id];
|
|
s_ctx.m_seq[_id]++;
|
|
uint64_t key = m_key.encode();
|
|
m_sortKeys[m_num] = key;
|
|
m_sortValues[m_num] = m_numRenderStates;
|
|
++m_num;
|
|
|
|
m_state.m_constEnd = m_constantBuffer->getPos();
|
|
m_state.m_flags |= m_flags;
|
|
m_renderState[m_numRenderStates] = m_state;
|
|
++m_numRenderStates;
|
|
}
|
|
|
|
m_state.clear();
|
|
m_flags = BGFX_STATE_NONE;
|
|
}
|
|
|
|
void Frame::submitMask(uint32_t _viewMask, int32_t _depth)
|
|
{
|
|
if (m_discard)
|
|
{
|
|
m_discard = false;
|
|
return;
|
|
}
|
|
|
|
if (BGFX_CONFIG_MAX_DRAW_CALLS-1 <= m_num
|
|
|| (0 == m_state.m_numVertices && 0 == m_state.m_numIndices) )
|
|
{
|
|
m_numDropped += uint32_cntbits(_viewMask);
|
|
return;
|
|
}
|
|
|
|
BX_WARN(invalidHandle != m_key.m_program, "Program with invalid handle");
|
|
if (invalidHandle != m_key.m_program)
|
|
{
|
|
m_key.m_depth = _depth;
|
|
|
|
for (uint32_t id = 0, viewMask = _viewMask, ntz = uint32_cnttz(_viewMask); 0 != viewMask; viewMask >>= 1, id += 1, ntz = uint32_cnttz(viewMask) )
|
|
{
|
|
viewMask >>= ntz;
|
|
id += ntz;
|
|
|
|
m_key.m_view = id;
|
|
m_key.m_seq = s_ctx.m_seq[id] & s_ctx.m_seqMask[id];
|
|
s_ctx.m_seq[id]++;
|
|
uint64_t key = m_key.encode();
|
|
m_sortKeys[m_num] = key;
|
|
m_sortValues[m_num] = m_numRenderStates;
|
|
++m_num;
|
|
}
|
|
|
|
m_state.m_constEnd = m_constantBuffer->getPos();
|
|
m_state.m_flags |= m_flags;
|
|
m_renderState[m_numRenderStates] = m_state;
|
|
++m_numRenderStates;
|
|
}
|
|
|
|
m_state.clear();
|
|
m_flags = BGFX_STATE_NONE;
|
|
}
|
|
|
|
void Frame::sort()
|
|
{
|
|
bx::radixSort64(m_sortKeys, s_ctx.m_tempKeys, m_sortValues, s_ctx.m_tempValues, m_num);
|
|
}
|
|
|
|
RendererType::Enum getRendererType()
|
|
{
|
|
#if BGFX_CONFIG_RENDERER_DIRECT3D9
|
|
return RendererType::Direct3D9;
|
|
#elif BGFX_CONFIG_RENDERER_DIRECT3D11
|
|
return RendererType::Direct3D11;
|
|
#elif BGFX_CONFIG_RENDERER_OPENGL
|
|
return RendererType::OpenGL;
|
|
#elif BGFX_CONFIG_RENDERER_OPENGLES2
|
|
return RendererType::OpenGLES2;
|
|
#elif BGFX_CONFIG_RENDERER_OPENGLES3
|
|
return RendererType::OpenGLES3;
|
|
#else
|
|
return RendererType::Null;
|
|
#endif // BGFX_CONFIG_RENDERER_
|
|
}
|
|
|
|
void init(CallbackI* _callback, ReallocFn _realloc, FreeFn _free)
|
|
{
|
|
if (NULL != _callback)
|
|
{
|
|
g_callback = _callback;
|
|
}
|
|
|
|
if (NULL != _realloc
|
|
&& NULL != _free)
|
|
{
|
|
g_realloc = _realloc;
|
|
g_free = _free;
|
|
}
|
|
|
|
s_threadIndex = BGFX_MAIN_THREAD_MAGIC;
|
|
|
|
// On NaCl renderer is on the main thread.
|
|
s_ctx.init(!BX_PLATFORM_NACL);
|
|
}
|
|
|
|
void shutdown()
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.shutdown();
|
|
|
|
s_threadIndex = 0;
|
|
g_callback = &s_callbackStub;
|
|
g_realloc = reallocStub;
|
|
g_free = freeStub;
|
|
}
|
|
|
|
void reset(uint32_t _width, uint32_t _height, uint32_t _flags)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.reset(_width, _height, _flags);
|
|
}
|
|
|
|
void frame()
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.frame();
|
|
}
|
|
|
|
bool renderFrame()
|
|
{
|
|
BGFX_CHECK_RENDER_THREAD();
|
|
return s_ctx.renderFrame();
|
|
}
|
|
|
|
const uint32_t g_uniformTypeSize[UniformType::Count] =
|
|
{
|
|
sizeof(int32_t),
|
|
sizeof(float),
|
|
0,
|
|
1*sizeof(int32_t),
|
|
1*sizeof(float),
|
|
2*sizeof(float),
|
|
3*sizeof(float),
|
|
4*sizeof(float),
|
|
3*3*sizeof(float),
|
|
4*4*sizeof(float),
|
|
};
|
|
|
|
void ConstantBuffer::writeUniform(UniformType::Enum _type, uint16_t _loc, const void* _value, uint16_t _num)
|
|
{
|
|
uint32_t opcode = encodeOpcode(_type, _loc, _num, true);
|
|
write(opcode);
|
|
write(_value, g_uniformTypeSize[_type]*_num);
|
|
}
|
|
|
|
void ConstantBuffer::writeUniformRef(UniformType::Enum _type, uint16_t _loc, const void* _value, uint16_t _num)
|
|
{
|
|
uint32_t opcode = encodeOpcode(_type, _loc, _num, false);
|
|
write(opcode);
|
|
write(&_value, sizeof(void*) );
|
|
}
|
|
|
|
void Context::init(bool _createRenderThread)
|
|
{
|
|
BX_CHECK(!m_rendererInitialized, "Already initialized?");
|
|
BX_TRACE("init");
|
|
|
|
m_exit = false;
|
|
m_frames = 0;
|
|
m_render = &m_frame[0];
|
|
m_submit = &m_frame[1];
|
|
m_debug = BGFX_DEBUG_NONE;
|
|
|
|
m_submit->create();
|
|
m_render->create();
|
|
|
|
#if BGFX_CONFIG_MULTITHREADED
|
|
if (_createRenderThread)
|
|
{
|
|
m_thread.init(renderThread, this);
|
|
}
|
|
#else
|
|
BX_UNUSED(_createRenderThread);
|
|
#endif // BGFX_CONFIG_MULTITHREADED
|
|
|
|
memset(m_rt, 0xff, sizeof(m_rt) );
|
|
memset(m_clear, 0, sizeof(m_clear) );
|
|
memset(m_rect, 0, sizeof(m_rect) );
|
|
memset(m_seq, 0, sizeof(m_seq) );
|
|
memset(m_seqMask, 0, sizeof(m_seqMask) );
|
|
|
|
for (uint32_t ii = 0; ii < countof(m_rect); ++ii)
|
|
{
|
|
m_rect[ii].m_width = 1;
|
|
m_rect[ii].m_height = 1;
|
|
}
|
|
|
|
m_declRef.init();
|
|
|
|
frameNoRenderWait();
|
|
|
|
getCommandBuffer(CommandBuffer::RendererInit);
|
|
|
|
m_textVideoMemBlitter.init();
|
|
m_clearQuad.init();
|
|
|
|
m_submit->m_transientVb = createTransientVertexBuffer(BGFX_CONFIG_TRANSIENT_VERTEX_BUFFER_SIZE);
|
|
m_submit->m_transientIb = createTransientIndexBuffer(BGFX_CONFIG_TRANSIENT_INDEX_BUFFER_SIZE);
|
|
frame();
|
|
m_submit->m_transientVb = createTransientVertexBuffer(BGFX_CONFIG_TRANSIENT_VERTEX_BUFFER_SIZE);
|
|
m_submit->m_transientIb = createTransientIndexBuffer(BGFX_CONFIG_TRANSIENT_INDEX_BUFFER_SIZE);
|
|
frame();
|
|
}
|
|
|
|
void Context::shutdown()
|
|
{
|
|
BX_TRACE("shutdown");
|
|
|
|
getCommandBuffer(CommandBuffer::RendererShutdownBegin);
|
|
frame();
|
|
|
|
destroyTransientVertexBuffer(m_submit->m_transientVb);
|
|
destroyTransientIndexBuffer(m_submit->m_transientIb);
|
|
m_textVideoMemBlitter.shutdown();
|
|
m_clearQuad.shutdown();
|
|
frame();
|
|
|
|
destroyTransientVertexBuffer(m_submit->m_transientVb);
|
|
destroyTransientIndexBuffer(m_submit->m_transientIb);
|
|
frame();
|
|
|
|
getCommandBuffer(CommandBuffer::RendererShutdownEnd);
|
|
frame();
|
|
|
|
m_declRef.shutdown(m_vertexDeclHandle);
|
|
|
|
#if BGFX_CONFIG_MULTITHREADED
|
|
if (m_thread.isRunning() )
|
|
{
|
|
m_thread.shutdown();
|
|
}
|
|
#endif // BGFX_CONFIG_MULTITHREADED
|
|
|
|
renderSemWait();
|
|
|
|
m_submit->destroy();
|
|
m_render->destroy();
|
|
|
|
#if BGFX_CONFIG_DEBUG
|
|
# define CHECK_HANDLE_LEAK(_handleAlloc) \
|
|
do { \
|
|
BX_WARN(0 == _handleAlloc.getNumHandles(), "LEAK: " #_handleAlloc " %d (max: %d)", _handleAlloc.getNumHandles(), _handleAlloc.getMaxHandles() ); \
|
|
} while (0)
|
|
|
|
CHECK_HANDLE_LEAK(m_dynamicIndexBufferHandle);
|
|
CHECK_HANDLE_LEAK(m_dynamicVertexBufferHandle);
|
|
CHECK_HANDLE_LEAK(m_indexBufferHandle);
|
|
CHECK_HANDLE_LEAK(m_vertexDeclHandle);
|
|
CHECK_HANDLE_LEAK(m_vertexBufferHandle);
|
|
CHECK_HANDLE_LEAK(m_vertexShaderHandle);
|
|
CHECK_HANDLE_LEAK(m_fragmentShaderHandle);
|
|
CHECK_HANDLE_LEAK(m_programHandle);
|
|
CHECK_HANDLE_LEAK(m_textureHandle);
|
|
CHECK_HANDLE_LEAK(m_renderTargetHandle);
|
|
CHECK_HANDLE_LEAK(m_uniformHandle);
|
|
|
|
# undef CHECK_HANDLE_LEAK
|
|
#endif // BGFX_CONFIG_DEBUG
|
|
}
|
|
|
|
const Memory* alloc(uint32_t _size)
|
|
{
|
|
Memory* mem = (Memory*)g_realloc(NULL, sizeof(Memory) + _size);
|
|
mem->size = _size;
|
|
mem->data = (uint8_t*)mem + sizeof(Memory);
|
|
return mem;
|
|
}
|
|
|
|
const Memory* makeRef(const void* _data, uint32_t _size)
|
|
{
|
|
Memory* mem = (Memory*)g_realloc(NULL, sizeof(Memory) );
|
|
mem->size = _size;
|
|
mem->data = (uint8_t*)_data;
|
|
return mem;
|
|
}
|
|
|
|
void release(const Memory* _mem)
|
|
{
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
g_free(const_cast<Memory*>(_mem) );
|
|
}
|
|
|
|
void setDebug(uint32_t _debug)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_debug = _debug;
|
|
}
|
|
|
|
void dbgTextClear(uint8_t _attr, bool _small)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.dbgTextClear(_attr, _small);
|
|
}
|
|
|
|
void dbgTextPrintf(uint16_t _x, uint16_t _y, uint8_t _attr, const char* _format, ...)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
va_list argList;
|
|
va_start(argList, _format);
|
|
s_ctx.dbgTextPrintfVargs(_x, _y, _attr, _format, argList);
|
|
va_end(argList);
|
|
}
|
|
|
|
IndexBufferHandle createIndexBuffer(const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.createIndexBuffer(_mem);
|
|
}
|
|
|
|
void destroyIndexBuffer(IndexBufferHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyIndexBuffer(_handle);
|
|
}
|
|
|
|
VertexBufferHandle createVertexBuffer(const Memory* _mem, const VertexDecl& _decl)
|
|
{
|
|
return s_ctx.createVertexBuffer(_mem, _decl);
|
|
}
|
|
|
|
void destroyVertexBuffer(VertexBufferHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyVertexBuffer(_handle);
|
|
}
|
|
|
|
DynamicIndexBufferHandle createDynamicIndexBuffer(uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.createDynamicIndexBuffer(_num);
|
|
}
|
|
|
|
DynamicIndexBufferHandle createDynamicIndexBuffer(const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
return s_ctx.createDynamicIndexBuffer(_mem);
|
|
}
|
|
|
|
void updateDynamicIndexBuffer(DynamicIndexBufferHandle _handle, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
s_ctx.updateDynamicIndexBuffer(_handle, _mem);
|
|
}
|
|
|
|
void destroyDynamicIndexBuffer(DynamicIndexBufferHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyDynamicIndexBuffer(_handle);
|
|
}
|
|
|
|
DynamicVertexBufferHandle createDynamicVertexBuffer(uint16_t _num, const VertexDecl& _decl)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.createDynamicVertexBuffer(_num, _decl);
|
|
}
|
|
|
|
DynamicVertexBufferHandle createDynamicVertexBuffer(const Memory* _mem, const VertexDecl& _decl)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
return s_ctx.createDynamicVertexBuffer(_mem, _decl);
|
|
}
|
|
|
|
void updateDynamicVertexBuffer(DynamicVertexBufferHandle _handle, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
s_ctx.updateDynamicVertexBuffer(_handle, _mem);
|
|
}
|
|
|
|
void destroyDynamicVertexBuffer(DynamicVertexBufferHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyDynamicVertexBuffer(_handle);
|
|
}
|
|
|
|
bool checkAvailTransientIndexBuffer(uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.m_submit->checkAvailTransientIndexBuffer(_num);
|
|
}
|
|
|
|
void allocTransientIndexBuffer(TransientIndexBuffer* _tib, uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _tib, "_tib can't be NULL");
|
|
return s_ctx.allocTransientIndexBuffer(_tib, _num);
|
|
}
|
|
|
|
bool checkAvailTransientVertexBuffer(uint16_t _num, const VertexDecl& _decl)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.m_submit->checkAvailTransientVertexBuffer(_num, _decl.m_stride);
|
|
}
|
|
|
|
void allocTransientVertexBuffer(TransientVertexBuffer* _tvb, uint16_t _num, const VertexDecl& _decl)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _tvb, "_tvb can't be NULL");
|
|
return s_ctx.allocTransientVertexBuffer(_tvb, _num, _decl);
|
|
}
|
|
|
|
const InstanceDataBuffer* allocInstanceDataBuffer(uint16_t _num, uint16_t _stride)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.allocInstanceDataBuffer(_num, _stride);
|
|
}
|
|
|
|
VertexShaderHandle createVertexShader(const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
return s_ctx.createVertexShader(_mem);
|
|
}
|
|
|
|
void destroyVertexShader(VertexShaderHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyVertexShader(_handle);
|
|
}
|
|
|
|
FragmentShaderHandle createFragmentShader(const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
return s_ctx.createFragmentShader(_mem);
|
|
}
|
|
|
|
void destroyFragmentShader(FragmentShaderHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyFragmentShader(_handle);
|
|
}
|
|
|
|
ProgramHandle createProgram(VertexShaderHandle _vsh, FragmentShaderHandle _fsh)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.createProgram(_vsh, _fsh);
|
|
}
|
|
|
|
void destroyProgram(ProgramHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyProgram(_handle);
|
|
}
|
|
|
|
static const uint32_t s_bitsPerPixel[TextureFormat::Count] =
|
|
{
|
|
4, // BC1
|
|
8, // BC2
|
|
8, // BC3
|
|
4, // BC4
|
|
8, // BC5
|
|
0, // Unknown
|
|
8, // L8
|
|
32, // BGRX8
|
|
32, // BGRA8
|
|
64, // RGBA16
|
|
64, // RGBA16F
|
|
16, // R5G6B5
|
|
16, // RGBA4
|
|
16, // RGB5A1
|
|
32, // RGB10A2
|
|
};
|
|
|
|
void calcTextureSize(TextureInfo& _info, uint16_t _width, uint16_t _height, uint16_t _depth, uint8_t _numMips, TextureFormat::Enum _format)
|
|
{
|
|
_width = uint32_max(1, _width);
|
|
_height = uint32_max(1, _height);
|
|
_depth = uint32_max(1, _depth);
|
|
_numMips = uint32_max(1, _numMips);
|
|
|
|
uint32_t width = _width;
|
|
uint32_t height = _height;
|
|
uint32_t depth = _depth;
|
|
|
|
uint32_t bpp = s_bitsPerPixel[_format];
|
|
uint32_t size = 0;
|
|
|
|
for (uint32_t lod = 0; lod < _numMips; ++lod)
|
|
{
|
|
width = uint32_max(1, width);
|
|
height = uint32_max(1, height);
|
|
depth = uint32_max(1, depth);
|
|
|
|
size += _width*_height*depth*bpp/8;
|
|
|
|
width >>= 1;
|
|
height >>= 1;
|
|
depth >>= 1;
|
|
}
|
|
|
|
_info.format = _format;
|
|
_info.storageSize = size;
|
|
_info.width = _width;
|
|
_info.height = _height;
|
|
_info.depth = _depth;
|
|
_info.numMips = _numMips;
|
|
_info.bitsPerPixel = bpp;
|
|
}
|
|
|
|
TextureHandle createTexture(const Memory* _mem, uint32_t _flags, TextureInfo* _info)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
return s_ctx.createTexture(_mem, _flags, _info);
|
|
}
|
|
|
|
TextureHandle createTexture2D(uint16_t _width, uint16_t _height, uint8_t _numMips, TextureFormat::Enum _format, uint32_t _flags, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
|
|
#if BGFX_CONFIG_DEBUG
|
|
if (NULL != _mem)
|
|
{
|
|
TextureInfo ti;
|
|
calcTextureSize(ti, _width, _height, 1, _numMips, _format);
|
|
BX_CHECK(ti.storageSize == _mem->size
|
|
, "createTexture2D: Texture storage size doesn't match passed memory size (storage size: %d, memory size: %d)"
|
|
, ti.storageSize
|
|
, _mem->size
|
|
);
|
|
}
|
|
#endif // BGFX_CONFIG_DEBUG
|
|
|
|
uint32_t size = sizeof(uint32_t)+sizeof(TextureCreate);
|
|
const Memory* mem = alloc(size);
|
|
|
|
bx::StaticMemoryBlockWriter writer(mem->data, mem->size);
|
|
uint32_t magic = BGFX_CHUNK_MAGIC_TEX;
|
|
bx::write(&writer, magic);
|
|
|
|
TextureCreate tc;
|
|
tc.m_flags = _flags;
|
|
tc.m_width = _width;
|
|
tc.m_height = _height;
|
|
tc.m_sides = 0;
|
|
tc.m_depth = 0;
|
|
tc.m_numMips = _numMips;
|
|
tc.m_format = uint8_t(_format);
|
|
tc.m_cubeMap = false;
|
|
tc.m_mem = _mem;
|
|
bx::write(&writer, tc);
|
|
|
|
return s_ctx.createTexture(mem, _flags, NULL);
|
|
}
|
|
|
|
TextureHandle createTexture3D(uint16_t _width, uint16_t _height, uint16_t _depth, uint8_t _numMips, TextureFormat::Enum _format, uint32_t _flags, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
|
|
#if BGFX_CONFIG_DEBUG
|
|
if (NULL != _mem)
|
|
{
|
|
TextureInfo ti;
|
|
calcTextureSize(ti, _width, _height, _depth, _numMips, _format);
|
|
BX_CHECK(ti.storageSize == _mem->size
|
|
, "createTexture3D: Texture storage size doesn't match passed memory size (storage size: %d, memory size: %d)"
|
|
, ti.storageSize
|
|
, _mem->size
|
|
);
|
|
}
|
|
#endif // BGFX_CONFIG_DEBUG
|
|
|
|
uint32_t size = sizeof(uint32_t)+sizeof(TextureCreate);
|
|
const Memory* mem = alloc(size);
|
|
|
|
bx::StaticMemoryBlockWriter writer(mem->data, mem->size);
|
|
uint32_t magic = BGFX_CHUNK_MAGIC_TEX;
|
|
bx::write(&writer, magic);
|
|
|
|
TextureCreate tc;
|
|
tc.m_flags = _flags;
|
|
tc.m_width = _width;
|
|
tc.m_height = _height;
|
|
tc.m_sides = 0;
|
|
tc.m_depth = _depth;
|
|
tc.m_numMips = _numMips;
|
|
tc.m_format = uint8_t(_format);
|
|
tc.m_cubeMap = false;
|
|
tc.m_mem = _mem;
|
|
bx::write(&writer, tc);
|
|
|
|
return s_ctx.createTexture(mem, _flags, NULL);
|
|
}
|
|
|
|
TextureHandle createTextureCube(uint16_t _sides, uint16_t _width, uint8_t _numMips, TextureFormat::Enum _format, uint32_t _flags, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
|
|
#if BGFX_CONFIG_DEBUG
|
|
if (NULL != _mem)
|
|
{
|
|
TextureInfo ti;
|
|
calcTextureSize(ti, _width, _width, 1, _numMips, _format);
|
|
BX_CHECK(ti.storageSize*_sides == _mem->size
|
|
, "createTextureCube: Texture storage size doesn't match passed memory size (storage size: %d, memory size: %d)"
|
|
, ti.storageSize*_sides
|
|
, _mem->size
|
|
);
|
|
}
|
|
#endif // BGFX_CONFIG_DEBUG
|
|
|
|
uint32_t size = sizeof(uint32_t)+sizeof(TextureCreate);
|
|
const Memory* mem = alloc(size);
|
|
|
|
bx::StaticMemoryBlockWriter writer(mem->data, mem->size);
|
|
uint32_t magic = BGFX_CHUNK_MAGIC_TEX;
|
|
bx::write(&writer, magic);
|
|
|
|
TextureCreate tc;
|
|
tc.m_flags = _flags;
|
|
tc.m_width = _width;
|
|
tc.m_height = _width;
|
|
tc.m_sides = _sides;
|
|
tc.m_depth = 0;
|
|
tc.m_numMips = _numMips;
|
|
tc.m_format = uint8_t(_format);
|
|
tc.m_cubeMap = true;
|
|
tc.m_mem = _mem;
|
|
bx::write(&writer, tc);
|
|
|
|
return s_ctx.createTexture(mem, _flags, NULL);
|
|
}
|
|
|
|
void destroyTexture(TextureHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyTexture(_handle);
|
|
}
|
|
|
|
void updateTexture2D(TextureHandle _handle, uint8_t _mip, uint16_t _x, uint16_t _y, uint16_t _width, uint16_t _height, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
if (_width == 0
|
|
|| _height == 0)
|
|
{
|
|
release(_mem);
|
|
}
|
|
else
|
|
{
|
|
s_ctx.updateTexture(_handle, 0, _mip, _x, _y, 0, _width, _height, 1, _mem);
|
|
}
|
|
}
|
|
|
|
void updateTexture3D(TextureHandle _handle, uint8_t _mip, uint16_t _x, uint16_t _y, uint16_t _z, uint16_t _width, uint16_t _height, uint16_t _depth, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
if (_width == 0
|
|
|| _height == 0
|
|
|| _depth == 0)
|
|
{
|
|
release(_mem);
|
|
}
|
|
else
|
|
{
|
|
s_ctx.updateTexture(_handle, 0, _mip, _x, _y, _z, _width, _height, _depth, _mem);
|
|
}
|
|
}
|
|
|
|
void updateTextureCube(TextureHandle _handle, uint8_t _side, uint8_t _mip, uint16_t _x, uint16_t _y, uint16_t _width, uint16_t _height, const Memory* _mem)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _mem, "_mem can't be NULL");
|
|
BX_CHECK(_side >= 0 && _side <= 5, "Invalid side %d.", _side);
|
|
if (_width == 0
|
|
|| _height == 0)
|
|
{
|
|
release(_mem);
|
|
}
|
|
else
|
|
{
|
|
s_ctx.updateTexture(_handle, _side, _mip, _x, _y, 0, _width, _height, 1, _mem);
|
|
}
|
|
}
|
|
|
|
RenderTargetHandle createRenderTarget(uint16_t _width, uint16_t _height, uint32_t _flags, uint32_t _textureFlags)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_WARN(0 != _width && 0 != _height, "Render target resolution width or height cannot be 0 (width %d, height %d).", _width, _height);
|
|
return s_ctx.createRenderTarget(uint16_max(1, _width), uint16_max(1, _height), _flags, _textureFlags);
|
|
}
|
|
|
|
void destroyRenderTarget(RenderTargetHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyRenderTarget(_handle);
|
|
}
|
|
|
|
UniformHandle createUniform(const char* _name, UniformType::Enum _type, uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.createUniform(_name, _type, _num);
|
|
}
|
|
|
|
void destroyUniform(UniformHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.destroyUniform(_handle);
|
|
}
|
|
|
|
void setViewRect(uint8_t _id, uint16_t _x, uint16_t _y, uint16_t _width, uint16_t _height)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewRect(_id, _x, _y, _width, _height);
|
|
}
|
|
|
|
void setViewRectMask(uint32_t _viewMask, uint16_t _x, uint16_t _y, uint16_t _width, uint16_t _height)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewRectMask(_viewMask, _x, _y, _width, _height);
|
|
}
|
|
|
|
void setViewClear(uint8_t _id, uint8_t _flags, uint32_t _rgba, float _depth, uint8_t _stencil)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(_flags != BGFX_CLEAR_NONE, "At least one clear bit flag must be used.");
|
|
s_ctx.setViewClear(_id, _flags, _rgba, _depth, _stencil);
|
|
}
|
|
|
|
void setViewClearMask(uint32_t _viewMask, uint8_t _flags, uint32_t _rgba, float _depth, uint8_t _stencil)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(_flags != BGFX_CLEAR_NONE, "At least one clear bit flag must be used.");
|
|
s_ctx.setViewClearMask(_viewMask, _flags, _rgba, _depth, _stencil);
|
|
}
|
|
|
|
void setViewSeq(uint8_t _id, bool _enabled)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewSeq(_id, _enabled);
|
|
}
|
|
|
|
void setViewSeqMask(uint32_t _viewMask, bool _enabled)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewSeqMask(_viewMask, _enabled);
|
|
}
|
|
|
|
void setViewRenderTarget(uint8_t _id, RenderTargetHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewRenderTarget(_id, _handle);
|
|
}
|
|
|
|
void setViewRenderTargetMask(uint32_t _mask, RenderTargetHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewRenderTargetMask(_mask, _handle);
|
|
}
|
|
|
|
void setViewTransform(uint8_t _id, const void* _view, const void* _proj, uint8_t _other)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewTransform(_id, _view, _proj, _other);
|
|
}
|
|
|
|
void setViewTransformMask(uint32_t _viewMask, const void* _view, const void* _proj, uint8_t _other)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setViewTransformMask(_viewMask, _view, _proj, _other);
|
|
}
|
|
|
|
void setState(uint64_t _state, uint32_t _rgba)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setState(_state, _rgba);
|
|
}
|
|
|
|
void setStencil(uint32_t _fstencil, uint32_t _bstencil)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setStencil(_fstencil, _bstencil);
|
|
}
|
|
|
|
uint32_t setTransform(const void* _mtx, uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
return s_ctx.m_submit->setTransform(_mtx, _num);
|
|
}
|
|
|
|
void setTransform(uint32_t _cache, uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setTransform(_cache, _num);
|
|
}
|
|
|
|
void setUniform(UniformHandle _handle, const void* _value, uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setUniform(_handle, _value, _num);
|
|
}
|
|
|
|
void setUniform(ProgramHandle _program, UniformHandle _handle, const void* _value)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.setUniform(_program, _handle, _value);
|
|
}
|
|
|
|
void setIndexBuffer(IndexBufferHandle _handle, uint32_t _firstIndex, uint32_t _numIndices)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setIndexBuffer(_handle, _firstIndex, _numIndices);
|
|
}
|
|
|
|
void setIndexBuffer(DynamicIndexBufferHandle _handle, uint32_t _firstIndex, uint32_t _numIndices)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setIndexBuffer(s_ctx.m_dynamicIndexBuffers[_handle.idx].m_handle, _firstIndex, _numIndices);
|
|
}
|
|
|
|
void setIndexBuffer(const TransientIndexBuffer* _tib, uint32_t _numIndices)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _tib, "_tib can't be NULL");
|
|
uint32_t numIndices = uint32_min(_numIndices, _tib->size/2);
|
|
s_ctx.m_submit->setIndexBuffer(_tib, numIndices);
|
|
}
|
|
|
|
void setVertexBuffer(VertexBufferHandle _handle, uint32_t _numVertices)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setVertexBuffer(_handle, _numVertices);
|
|
}
|
|
|
|
void setVertexBuffer(DynamicVertexBufferHandle _handle, uint32_t _numVertices)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setVertexBuffer(s_ctx.m_dynamicVertexBuffers[_handle.idx], _numVertices);
|
|
}
|
|
|
|
void setVertexBuffer(const TransientVertexBuffer* _tvb, uint32_t _numVertices)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
BX_CHECK(NULL != _tvb, "_tvb can't be NULL");
|
|
s_ctx.m_submit->setVertexBuffer(_tvb, _numVertices);
|
|
}
|
|
|
|
void setInstanceDataBuffer(const InstanceDataBuffer* _idb, uint16_t _num)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setInstanceDataBuffer(_idb, _num);
|
|
}
|
|
|
|
void setProgram(ProgramHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setProgram(_handle);
|
|
}
|
|
|
|
void setTexture(uint8_t _stage, UniformHandle _sampler, TextureHandle _handle)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setTexture(_stage, _sampler, _handle);
|
|
}
|
|
|
|
void setTexture(uint8_t _stage, UniformHandle _sampler, RenderTargetHandle _handle, bool _depth)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->setTexture(_stage, _sampler, _handle, _depth);
|
|
}
|
|
|
|
void submit(uint8_t _id, int32_t _depth)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->submit(_id, _depth);
|
|
}
|
|
|
|
void submitMask(uint32_t _viewMask, int32_t _depth)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
s_ctx.m_submit->submitMask(_viewMask, _depth);
|
|
}
|
|
|
|
void saveScreenShot(const char* _filePath)
|
|
{
|
|
BGFX_CHECK_MAIN_THREAD();
|
|
uint32_t len = (uint32_t)strlen(_filePath)+1;
|
|
const Memory* mem = alloc(len);
|
|
memcpy(mem->data, _filePath, mem->size);
|
|
return s_ctx.saveScreenShot(mem);
|
|
}
|
|
}
|