mirror of
https://github.com/bkaradzic/bgfx.git
synced 2026-10-25 18:25:29 +00:00
Reworked blit to use regions, added buffer blit and read buffer. (#3924)
This commit is contained in:
Branimir Karadžić
+209
-7
@@ -1416,14 +1416,16 @@ WGPU_IMPORT
|
||||
BX_TRACE("\tmaxComputeWorkgroupsPerDimension: %u", m_limits.maxComputeWorkgroupsPerDimension);
|
||||
BX_TRACE("\tmaxImmediateSize: %u", m_limits.maxImmediateSize);
|
||||
|
||||
g_caps.limits.maxTextureSize = m_limits.maxTextureDimension2D;
|
||||
g_caps.limits.maxTextureLayers = m_limits.maxTextureArrayLayers;
|
||||
g_caps.limits.maxTextureSamplers = bx::min(m_limits.maxSamplersPerShaderStage, BGFX_CONFIG_MAX_TEXTURE_SAMPLERS);
|
||||
g_caps.limits.maxComputeBindings = BGFX_CONFIG_MAX_TEXTURE_SAMPLERS;
|
||||
g_caps.limits.maxFBAttachments = bx::min(m_limits.maxColorAttachments, BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS);
|
||||
g_caps.limits.maxVertexStreams = bx::min(m_limits.maxVertexBuffers, BGFX_CONFIG_MAX_VERTEX_STREAMS);
|
||||
g_caps.limits.maxTextureSize = m_limits.maxTextureDimension2D;
|
||||
g_caps.limits.maxTextureLayers = m_limits.maxTextureArrayLayers;
|
||||
g_caps.limits.maxTextureSamplers = bx::min(m_limits.maxSamplersPerShaderStage, BGFX_CONFIG_MAX_TEXTURE_SAMPLERS);
|
||||
g_caps.limits.maxComputeBindings = BGFX_CONFIG_MAX_TEXTURE_SAMPLERS;
|
||||
g_caps.limits.maxFBAttachments = bx::min(m_limits.maxColorAttachments, BGFX_CONFIG_MAX_FRAME_BUFFER_ATTACHMENTS);
|
||||
g_caps.limits.maxVertexStreams = bx::min(m_limits.maxVertexBuffers, BGFX_CONFIG_MAX_VERTEX_STREAMS);
|
||||
g_caps.limits.maxVertexAttributes = m_limits.maxVertexAttributes;
|
||||
g_caps.limits.maxInstanceData = bx::min<uint32_t>(g_caps.limits.maxInstanceData, g_caps.limits.maxVertexAttributes);
|
||||
g_caps.limits.maxInstanceData = bx::min<uint32_t>(g_caps.limits.maxInstanceData, g_caps.limits.maxVertexAttributes);
|
||||
g_caps.limits.blitRowPitchAlign = 256;
|
||||
g_caps.limits.blitOffsetAlign = 256;
|
||||
|
||||
g_caps.supported = 0
|
||||
| BGFX_CAPS_ALPHA_TO_COVERAGE
|
||||
@@ -1798,6 +1800,36 @@ WGPU_IMPORT
|
||||
*(bool*)(_userdata2) = true;
|
||||
}
|
||||
|
||||
struct ReadBuffer
|
||||
{
|
||||
WGPUBuffer buffer;
|
||||
uint32_t size;
|
||||
void* data;
|
||||
};
|
||||
|
||||
static void readBufferCb(WGPUMapAsyncStatus _status, WGPUStringView _message, void* _userdata1, void* _userdata2)
|
||||
{
|
||||
BX_ASSERT(WGPUMapAsyncStatus_Success == _status, "%d", _status);
|
||||
|
||||
BX_UNUSED(_status, _message, _userdata2);
|
||||
|
||||
ReadBuffer& readBuffer = *(ReadBuffer*)_userdata1;
|
||||
|
||||
const void* result = (const void*)WGPU_CHECK(wgpuBufferGetConstMappedRange(
|
||||
readBuffer.buffer
|
||||
, 0
|
||||
, readBuffer.size
|
||||
) );
|
||||
|
||||
bx::memCopy(readBuffer.data, result, readBuffer.size);
|
||||
|
||||
WGPU_CHECK(wgpuBufferUnmap(readBuffer.buffer) );
|
||||
|
||||
wgpuRelease(readBuffer.buffer);
|
||||
|
||||
*(bool*)(_userdata2) = true;
|
||||
}
|
||||
|
||||
void readTexture(TextureHandle _handle, void* _data, uint16_t _layer, uint8_t _mip) override
|
||||
{
|
||||
const TextureWGPU& texture = m_textures[_handle.idx];
|
||||
@@ -1887,6 +1919,66 @@ WGPU_IMPORT
|
||||
}
|
||||
}
|
||||
|
||||
BufferWGPU& getBuffer(Handle _handle)
|
||||
{
|
||||
if (_handle.isIndexBuffer() )
|
||||
{
|
||||
return m_indexBuffers[_handle.idx];
|
||||
}
|
||||
|
||||
return m_vertexBuffers[_handle.idx];
|
||||
}
|
||||
|
||||
void readBuffer(Handle _handle, void* _data, uint32_t _offset, uint32_t _size) override
|
||||
{
|
||||
const BufferWGPU& buffer = getBuffer(_handle);
|
||||
|
||||
WGPUBufferDescriptor readBufferDesc =
|
||||
{
|
||||
.nextInChain = NULL,
|
||||
.label = toWGPUStringView("Read Buffer"),
|
||||
.usage = 0
|
||||
| WGPUBufferUsage_MapRead
|
||||
| WGPUBufferUsage_CopyDst
|
||||
,
|
||||
.size = _size,
|
||||
.mappedAtCreation = false,
|
||||
};
|
||||
|
||||
WGPUBuffer readBufferDst = WGPU_CHECK(wgpuDeviceCreateBuffer(m_device, &readBufferDesc) );
|
||||
|
||||
s_done = false;
|
||||
|
||||
m_cmd.copyBufferToBuffer(buffer.m_buffer, _offset, readBufferDst, 0, _size);
|
||||
|
||||
m_cmd.kick();
|
||||
|
||||
ReadBuffer readBuffer =
|
||||
{
|
||||
.buffer = readBufferDst,
|
||||
.size = _size,
|
||||
.data = _data,
|
||||
};
|
||||
|
||||
WGPU_CHECK(wgpuBufferMapAsync(
|
||||
readBufferDst
|
||||
, WGPUMapMode_Read
|
||||
, 0
|
||||
, _size
|
||||
, {
|
||||
.nextInChain = NULL,
|
||||
.mode = WGPUCallbackMode_AllowProcessEvents,
|
||||
.callback = readBufferCb,
|
||||
.userdata1 = &readBuffer,
|
||||
.userdata2 = &s_done,
|
||||
}) );
|
||||
|
||||
while (!s_done)
|
||||
{
|
||||
wgpuInstanceProcessEvents(m_instance);
|
||||
}
|
||||
}
|
||||
|
||||
void resizeTexture(TextureHandle _handle, uint16_t _width, uint16_t _height, uint8_t _numMips, uint16_t _numLayers) override
|
||||
{
|
||||
TextureWGPU& texture = m_textures[_handle.idx];
|
||||
@@ -2052,6 +2144,8 @@ WGPU_IMPORT
|
||||
|
||||
void submitBlit(BlitState& _bs, uint16_t _view);
|
||||
|
||||
void blitBufferTexture(const BlitItem& _blit);
|
||||
|
||||
void submitUniformCache(UniformCacheState& _ucs, uint16_t _view);
|
||||
|
||||
void generateMips(WGPUCommandEncoder _cmdEncoder, TextureWGPU& _texture, TextureHandle _textureHandle);
|
||||
@@ -3620,6 +3714,7 @@ WGPU_IMPORT
|
||||
.label = WGPU_STRING_VIEW_INIT,
|
||||
.usage = 0
|
||||
| (storage ? WGPUBufferUsage_Storage : 0)
|
||||
| (storage ? WGPUBufferUsage_CopySrc : 0)
|
||||
| (indirect
|
||||
? WGPUBufferUsage_Indirect
|
||||
: _vertex ? WGPUBufferUsage_Vertex : WGPUBufferUsage_Index
|
||||
@@ -5628,12 +5723,118 @@ m_resolution.formatColor = TextureFormat::BGRA8;
|
||||
}
|
||||
}
|
||||
|
||||
void RendererContextWGPU::blitBufferTexture(const BlitItem& _blit)
|
||||
{
|
||||
const bool toBuffer = _blit.m_dst.isBuffer();
|
||||
|
||||
const TextureWGPU& texture = m_textures[toBuffer ? _blit.m_src.idx : _blit.m_dst.idx];
|
||||
const BufferWGPU& buffer = getBuffer(toBuffer ? _blit.m_dst : _blit.m_src);
|
||||
|
||||
const uint32_t bufferOffset = toBuffer ? _blit.m_dstOffset : _blit.m_srcOffset;
|
||||
|
||||
const TextureFormat::Enum format = TextureFormat::Enum(texture.m_textureFormat);
|
||||
|
||||
const uint32_t rowPitch = _blit.m_rowPitch;
|
||||
const uint32_t slicePitch = _blit.m_slicePitch;
|
||||
const uint32_t numRows = calcTextureRegionNumRows(format, _blit.m_height);
|
||||
const uint32_t depth = bx::max<uint32_t>(1, _blit.m_depth);
|
||||
|
||||
const uint32_t rowsPerImage = slicePitch / rowPitch;
|
||||
|
||||
const uint8_t mip = toBuffer ? _blit.m_srcMip : _blit.m_dstMip;
|
||||
const uint16_t x = toBuffer ? _blit.m_srcX : _blit.m_dstX;
|
||||
const uint16_t y = toBuffer ? _blit.m_srcY : _blit.m_dstY;
|
||||
const uint16_t z = toBuffer ? _blit.m_srcZ : _blit.m_dstZ;
|
||||
|
||||
const bimg::ImageBlockInfo& blockInfo = bimg::getBlockInfo(bimg::TextureFormat::Enum(format) );
|
||||
|
||||
const uint32_t blockHeight = bx::max<uint32_t>(1, blockInfo.blockHeight);
|
||||
|
||||
const uint32_t offsetAlign = bimg::isDepth(bimg::TextureFormat::Enum(format) )
|
||||
? 4
|
||||
: blockInfo.blockSize
|
||||
;
|
||||
|
||||
const bool direct = (1 == numRows*depth || 0 == rowPitch % 256)
|
||||
&& 0 == bufferOffset % offsetAlign
|
||||
;
|
||||
|
||||
const uint32_t numCopies = direct ? 1 : numRows*depth;
|
||||
const uint32_t copyRows = direct ? numRows : 1;
|
||||
const uint32_t copyDepth = direct ? depth : 1;
|
||||
|
||||
for (uint32_t ii = 0; ii < numCopies; ++ii)
|
||||
{
|
||||
const uint32_t slice = direct ? 0 : ii / numRows;
|
||||
const uint32_t row = direct ? 0 : ii % numRows;
|
||||
|
||||
const WGPUTexelCopyBufferInfo bufferInfo =
|
||||
{
|
||||
.layout =
|
||||
{
|
||||
.offset = bufferOffset + slice*slicePitch + row*rowPitch,
|
||||
.bytesPerRow = 1 < copyRows*copyDepth ? rowPitch : WGPU_COPY_STRIDE_UNDEFINED,
|
||||
.rowsPerImage = 1 < copyDepth ? rowsPerImage : WGPU_COPY_STRIDE_UNDEFINED,
|
||||
},
|
||||
.buffer = buffer.m_buffer,
|
||||
};
|
||||
|
||||
const WGPUTexelCopyTextureInfo textureInfo =
|
||||
{
|
||||
.texture = texture.m_texture,
|
||||
.mipLevel = mip,
|
||||
.origin =
|
||||
{
|
||||
.x = x,
|
||||
.y = y + row*blockHeight,
|
||||
.z = z + slice,
|
||||
},
|
||||
.aspect = WGPUTextureAspect_All,
|
||||
};
|
||||
|
||||
const WGPUExtent3D copySize =
|
||||
{
|
||||
.width = _blit.m_width,
|
||||
.height = copyRows*blockHeight,
|
||||
.depthOrArrayLayers = copyDepth,
|
||||
};
|
||||
|
||||
if (toBuffer)
|
||||
{
|
||||
m_cmd.copyTextureToBuffer(textureInfo, bufferInfo, copySize);
|
||||
}
|
||||
else
|
||||
{
|
||||
m_cmd.copyBufferToTexture(bufferInfo, textureInfo, copySize);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void RendererContextWGPU::submitBlit(BlitState& _bs, uint16_t _view)
|
||||
{
|
||||
while (_bs.hasItem(_view) )
|
||||
{
|
||||
const BlitItem& blit = _bs.advance();
|
||||
|
||||
if (blit.m_src.isBuffer()
|
||||
&& blit.m_dst.isBuffer() )
|
||||
{
|
||||
const BufferWGPU& srcBuf = getBuffer(blit.m_src);
|
||||
const BufferWGPU& dstBuf = getBuffer(blit.m_dst);
|
||||
|
||||
m_cmd.copyBufferToBuffer(srcBuf.m_buffer, blit.m_srcOffset, dstBuf.m_buffer, blit.m_dstOffset, blit.m_size);
|
||||
|
||||
continue;
|
||||
}
|
||||
|
||||
if (blit.m_src.isBuffer()
|
||||
|| blit.m_dst.isBuffer() )
|
||||
{
|
||||
blitBufferTexture(blit);
|
||||
|
||||
continue;
|
||||
}
|
||||
|
||||
const TextureWGPU& src = m_textures[blit.m_src.idx];
|
||||
const TextureWGPU& dst = m_textures[blit.m_dst.idx];
|
||||
|
||||
@@ -7007,6 +7208,7 @@ m_resolution.formatColor = TextureFormat::BGRA8;
|
||||
perfStats.numDraw = statsKeyType[0];
|
||||
perfStats.numCompute = statsKeyType[1];
|
||||
perfStats.numBlit = _render->m_numBlitItems;
|
||||
perfStats.numBlitRepack = _render->m_numBlitRepack;
|
||||
|
||||
perfStats.gpuMemoryMax = -INT64_MAX;
|
||||
perfStats.gpuMemoryUsed = -INT64_MAX;
|
||||
|
||||
Reference in New Issue
Block a user