D3D9: Fixed 16-bit texture formats.

src/image.cpp

@@ -1252,6 +1252,26 @@ namespace bgfx
 		_dst[3] = float( ( (packed>>12) & 0xf) ) / 15.0f;
 	}
 
+	// RGBA4
+	void packBgra4(void* _dst, const float* _src)
+	{
+		*( (uint16_t*)_dst) = 0
+			| uint16_t(toUnorm(_src[0], 15.0f)<< 8)
+			| uint16_t(toUnorm(_src[1], 15.0f)<< 4)
+			| uint16_t(toUnorm(_src[2], 15.0f)    )
+			| uint16_t(toUnorm(_src[3], 15.0f)<<12)
+			;
+	}
+
+	void unpackBgra4(float* _dst, const void* _src)
+	{
+		uint16_t packed = *( (const uint16_t*)_src);
+		_dst[0] = float( ( (packed>> 8) & 0xf) ) / 15.0f;
+		_dst[1] = float( ( (packed>> 4) & 0xf) ) / 15.0f;
+		_dst[2] = float( ( (packed    ) & 0xf) ) / 15.0f;
+		_dst[3] = float( ( (packed>>12) & 0xf) ) / 15.0f;
+	}
+
 	// RGB5A1
 	void packRgb5a1(void* _dst, const float* _src)
 	{
@@ -1272,6 +1292,26 @@ namespace bgfx
 		_dst[3] = float( ( (packed>>14) &  0x1) );
 	}
 
+	// BGR5A1
+	void packBgr5a1(void* _dst, const float* _src)
+	{
+		*( (uint16_t*)_dst) = 0
+			| uint16_t(toUnorm(_src[0], 31.0f)<<10)
+			| uint16_t(toUnorm(_src[1], 31.0f)<< 5)
+			| uint16_t(toUnorm(_src[2], 31.0f)    )
+			| uint16_t(toUnorm(_src[3],  1.0f)<<15)
+			;
+	}
+
+	void unpackBgr5a1(float* _dst, const void* _src)
+	{
+		uint16_t packed = *( (const uint16_t*)_src);
+		_dst[0] = float( ( (packed>>10) & 0x1f) ) / 31.0f;
+		_dst[1] = float( ( (packed>> 5) & 0x1f) ) / 31.0f;
+		_dst[2] = float( ( (packed    ) & 0x1f) ) / 31.0f;
+		_dst[3] = float( ( (packed>>14) &  0x1) );
+	}
+
 	// RGB10A2
 	void packRgb10A2(void* _dst, const float* _src)
 	{
@@ -1310,9 +1350,6 @@ namespace bgfx
 		_dst[2] = bx::halfToFloat( (packed>>17) & 0x7fe0);
 	}
 
-	typedef void (*PackFn)(void*, const float*);
-	typedef void (*UnpackFn)(float*, const void*);
-
 	struct PackUnpack
 	{
 		PackFn pack;
@@ -1405,6 +1442,40 @@ namespace bgfx
 			;
 	}
 
+	void imageConvert(void* _dst, uint32_t _bpp, PackFn _pack, const void* _src, UnpackFn _unpack, uint32_t _size)
+	{
+		const uint8_t* src = (uint8_t*)_src;
+		uint8_t* dst = (uint8_t*)_dst;
+
+		const uint32_t size = _size * 8 / _bpp;
+
+		for (uint32_t ii = 0; ii < size; ++ii)
+		{
+			float rgba[4];
+			_unpack(rgba, &src[ii*_bpp/8]);
+			_pack(&dst[ii*_bpp/8], rgba);
+		}
+	}
+
+	void imageConvert(void* _dst, uint32_t _dstBpp, PackFn _pack, const void* _src, uint32_t _srcBpp, UnpackFn _unpack, uint32_t _width, uint32_t _height)
+	{
+		const uint8_t* src = (uint8_t*)_src;
+		uint8_t* dst = (uint8_t*)_dst;
+
+		const uint32_t srcPitch = _width * _srcBpp / 8;
+		const uint32_t dstPitch = _width * _dstBpp / 8;
+
+		for (uint32_t yy = 0; yy < _height; ++yy, src += srcPitch, dst += dstPitch)
+		{
+			for (uint32_t xx = 0; xx < _width; ++xx)
+			{
+				float rgba[4];
+				_unpack(rgba, &src[xx*_srcBpp/8]);
+				_pack(&dst[xx*_dstBpp/8], rgba);
+			}
+		}
+	}
+
 	bool imageConvert(void* _dst, TextureFormat::Enum _dstFormat, const void* _src, TextureFormat::Enum _srcFormat, uint32_t _width, uint32_t _height)
 	{
 		UnpackFn unpack = s_packUnpack[_srcFormat].unpack;
@@ -1415,23 +1486,9 @@ namespace bgfx
 			return false;
 		}
 
-		const uint8_t* src = (uint8_t*)_src;
-		uint8_t* dst = (uint8_t*)_dst;
-
 		const uint32_t srcBpp = s_imageBlockInfo[_srcFormat].bitsPerPixel;
 		const uint32_t dstBpp = s_imageBlockInfo[_dstFormat].bitsPerPixel;
-		const uint32_t srcPitch = _width * srcBpp / 8;
+		imageConvert(_dst, dstBpp, pack, _src, srcBpp, unpack, _width, _height);
-		const uint32_t dstPitch = _width * dstBpp / 8;
-
-		for (uint32_t yy = 0; yy < _height; ++yy, src += srcPitch, dst += dstPitch)
-		{
-			for (uint32_t xx = 0; xx < _width; ++xx)
-			{
-				float rgba[4];
-				unpack(rgba, &src[xx*srcBpp/8]);
-				pack(&dst[xx*dstBpp/8], rgba);
-			}
-		}
 
 		return true;
 	}

src/image.h

@@ -66,6 +66,189 @@ namespace bgfx
 		uint8_t encoding;
 	};
 
+	typedef void (*PackFn)(void*, const float*);
+	typedef void (*UnpackFn)(float*, const void*);
+
+	// R8
+	void packR8(void* _dst, const float* _src);
+	void unpackR8(float* _dst, const void* _src);
+
+	// R8S
+	void packR8S(void* _dst, const float* _src);
+	void unpackR8S(float* _dst, const void* _src);
+
+	// R8I
+	void packR8I(void* _dst, const float* _src);
+	void unpackR8I(float* _dst, const void* _src);
+
+	// R8U
+	void packR8U(void* _dst, const float* _src);
+	void unpackR8U(float* _dst, const void* _src);
+
+	// RG8
+	void packRg8(void* _dst, const float* _src);
+	void unpackRg8(float* _dst, const void* _src);
+
+	// RG8S
+	void packRg8S(void* _dst, const float* _src);
+	void unpackRg8S(float* _dst, const void* _src);
+
+	// RG8I
+	void packRg8I(void* _dst, const float* _src);
+	void unpackRg8I(float* _dst, const void* _src);
+
+	// RG8U
+	void packRg8U(void* _dst, const float* _src);
+	void unpackRg8U(float* _dst, const void* _src);
+
+	// RGBA8
+	void packRgba8(void* _dst, const float* _src);
+	void unpackRgba8(float* _dst, const void* _src);
+
+	// BGRA8
+	void packBgra8(void* _dst, const float* _src);
+	void unpackBgra8(float* _dst, const void* _src);
+
+	// RGBA8S
+	void packRgba8S(void* _dst, const float* _src);
+	void unpackRgba8S(float* _dst, const void* _src);
+
+	// RGBA8I
+	void packRgba8I(void* _dst, const float* _src);
+	void unpackRgba8I(float* _dst, const void* _src);
+
+	// RGBA8U
+	void packRgba8U(void* _dst, const float* _src);
+	void unpackRgba8U(float* _dst, const void* _src);
+
+	// R16
+	void packR16(void* _dst, const float* _src);
+	void unpackR16(float* _dst, const void* _src);
+
+	// R16S
+	void packR16S(void* _dst, const float* _src);
+	void unpackR16S(float* _dst, const void* _src);
+
+	// R16I
+	void packR16I(void* _dst, const float* _src);
+	void unpackR16I(float* _dst, const void* _src);
+
+	// R16U
+	void packR16U(void* _dst, const float* _src);
+	void unpackR16U(float* _dst, const void* _src);
+
+	// R16F
+	void packR16F(void* _dst, const float* _src);
+	void unpackR16F(float* _dst, const void* _src);
+
+	// RG16
+	void packRg16(void* _dst, const float* _src);
+	void unpackRg16(float* _dst, const void* _src);
+
+	// RG16S
+	void packRg16S(void* _dst, const float* _src);
+	void unpackRg16S(float* _dst, const void* _src);
+
+	// RG16I
+	void packRg16I(void* _dst, const float* _src);
+	void unpackRg16I(float* _dst, const void* _src);
+
+	// RG16U
+	void packRg16U(void* _dst, const float* _src);
+	void unpackRg16U(float* _dst, const void* _src);
+
+	// RG16F
+	void packRg16F(void* _dst, const float* _src);
+	void unpackRg16F(float* _dst, const void* _src);
+
+	// RGBA16
+	void packRgba16(void* _dst, const float* _src);
+	void unpackRgba16(float* _dst, const void* _src);
+
+	// RGBA16S
+	void packRgba16S(void* _dst, const float* _src);
+	void unpackRgba16S(float* _dst, const void* _src);
+
+	// RGBA16I
+	void packRgba16I(void* _dst, const float* _src);
+	void unpackRgba16I(float* _dst, const void* _src);
+
+	// RGBA16U
+	void packRgba16U(void* _dst, const float* _src);
+	void unpackRgba16U(float* _dst, const void* _src);
+
+	// RGBA16F
+	void packRgba16F(void* _dst, const float* _src);
+	void unpackRgba16F(float* _dst, const void* _src);
+
+	// R32I
+	void packR32I(void* _dst, const float* _src);
+	void unpackR32I(float* _dst, const void* _src);
+
+	// R32U
+	void packR32U(void* _dst, const float* _src);
+	void unpackR32U(float* _dst, const void* _src);
+
+	// R32F
+	void packR32F(void* _dst, const float* _src);
+	void unpackR32F(float* _dst, const void* _src);
+
+	// RG32I
+	void packRg32I(void* _dst, const float* _src);
+	void unpackRg32I(float* _dst, const void* _src);
+
+	// RG32U
+	void packRg32U(void* _dst, const float* _src);
+	void unpackRg32U(float* _dst, const void* _src);
+
+	// RGB9E5F
+	void packRgb9E5F(void* _dst, const float* _src);
+	void unpackRgb9E5F(float* _dst, const void* _src);
+
+	// RGBA32I
+	void packRgba32I(void* _dst, const float* _src);
+	void unpackRgba32I(float* _dst, const void* _src);
+
+	// RGBA32U
+	void packRgba32U(void* _dst, const float* _src);
+	void unpackRgba32U(float* _dst, const void* _src);
+
+	// RGBA32F
+	void packRgba32F(void* _dst, const float* _src);
+	void unpackRgba32F(float* _dst, const void* _src);
+
+	// R5G6B5
+	void packR5G6B5(void* _dst, const float* _src);
+	void unpackR5G6B5(float* _dst, const void* _src);
+
+	// RGBA4
+	void packRgba4(void* _dst, const float* _src);
+	void unpackRgba4(float* _dst, const void* _src);
+
+	// RGBA4
+	void packBgra4(void* _dst, const float* _src);
+	void unpackBgra4(float* _dst, const void* _src);
+
+	// RGB5A1
+	void packRgb5a1(void* _dst, const float* _src);
+	void unpackRgb5a1(float* _dst, const void* _src);
+
+	// BGR5A1
+	void packBgr5a1(void* _dst, const float* _src);
+	void unpackBgr5a1(float* _dst, const void* _src);
+
+	// RGB10A2
+	void packRgb10A2(void* _dst, const float* _src);
+	void unpackRgb10A2(float* _dst, const void* _src);
+
+	// R11G11B10F
+	void packR11G11B10F(void* _dst, const float* _src);
+	void unpackR11G11B10F(float* _dst, const void* _src);
+
+	// RG32F
+	void packRg32F(void* _dst, const float* _src);
+	void unpackRg32F(float* _dst, const void* _src);
+
 	/// Returns true if texture format is compressed.
 	bool isCompressed(TextureFormat::Enum _format);
 
@@ -120,6 +303,12 @@ namespace bgfx
 	///
 	bool imageConvert(TextureFormat::Enum _dstFormat, TextureFormat::Enum _srcFormat);
 
+	///
+	void imageConvert(void* _dst, uint32_t _bpp, PackFn _pack, const void* _src, UnpackFn _unpack, uint32_t _size);
+
+	///
+	void imageConvert(void* _dst, uint32_t _dstBpp, PackFn _pack, const void* _src, uint32_t _srcBpp, UnpackFn _unpack, uint32_t _width, uint32_t _height);
+
 	///
 	bool imageConvert(void* _dst, TextureFormat::Enum _dstFormat, const void* _src, TextureFormat::Enum _srcFormat, uint32_t _width, uint32_t _height);
 

src/renderer_d3d9.cpp

@@ -2927,7 +2927,20 @@ namespace bgfx { namespace d3d9
 						else
 						{
 							uint32_t size = useMipSize ? mip.m_size : mipSize;
+							switch (m_textureFormat)
+							{
+							case TextureFormat::RGB5A1:
+								imageConvert(bits, 16, packBgr5a1, mip.m_data, unpackRgb5a1, size);
+								break;
+
+							case TextureFormat::RGBA4:
+								imageConvert(bits, 16, packBgra4, mip.m_data, unpackRgba4, size);
+								break;
+
+							default:
 								memcpy(bits, mip.m_data, size);
+								break;
+							}
 						}
 
 						unlock(side, lod);
@@ -2976,7 +2989,20 @@ namespace bgfx { namespace d3d9
 			uint8_t* dst = bits;
 			for (uint32_t yy = 0, height = _rect.m_height; yy < height; ++yy)
 			{
+				switch (m_textureFormat)
+				{
+				case TextureFormat::RGB5A1:
+					imageConvert(dst, 16, packBgr5a1, src, unpackRgb5a1, rectpitch);
+					break;
+
+				case TextureFormat::RGBA4:
+					imageConvert(dst, 16, packBgra4, src, unpackRgba4, rectpitch);
+					break;
+
+				default:
 					memcpy(dst, src, rectpitch);
+					break;
+				}
 				src += srcpitch;
 				dst += dstpitch;
 			}