piwalker/dolphin
1
0
Fork 0
mirror of https://github.com/dolphin-emu/dolphin.git synced 2024-11-15 13:57:57 -07:00
Code Issues Actions Packages Projects Releases Wiki Activity
d53ab32506
dolphin/Source/Core/VideoCommon/TextureConverterShaderGen.cpp

332 lines
11 KiB
C++
Raw Normal View History

VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
// Copyright 2017 Dolphin Emulator Project
// Licensed under GPLv2+
// Refer to the license.txt file included.
VideoCommon: Remove unnecessary memset on ShaderUid instances. Zero-initialization zeroes out all members and padding bits, so this is safe to do. While we're at it, also add static assertions that enforce the necessary requirements of a UID type explicitly within the ShaderUid class. This way, we can remove several memset calls around the shader generation code that makes sure the underlying UID data is zeroed out. Now our ShaderUid class enforces this for us, so we don't need to care about it at the usage sites.
2019-05-29 23:05:06 -06:00
#include "VideoCommon/TextureConverterShaderGen.h"
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
#include "Common/Assert.h"
#include "Common/CommonTypes.h"
#include "VideoCommon/BPMemory.h"
#include "VideoCommon/VideoCommon.h"
#include "VideoCommon/VideoConfig.h"
VideoCommon: Create a namespace for TextureConversionShaderGen.
2017-11-25 03:07:14 -07:00
namespace TextureConversionShaderGen
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
{
VideoCommon: Create a namespace for TextureConversionShaderGen.
2017-11-25 03:07:14 -07:00
TCShaderUid GetShaderUid(EFBCopyFormat dst_format, bool is_depth_copy, bool is_intensity,
TextureConversionShader: Don't sample from adjacent rows when not needed
2018-05-02 22:09:32 -06:00
bool scale_by_half, bool copy_filter)
VideoCommon: Create a namespace for TextureConversionShaderGen.
2017-11-25 03:07:14 -07:00
{
TCShaderUid out;
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
VideoCommon: Remove unnecessary memset on ShaderUid instances. Zero-initialization zeroes out all members and padding bits, so this is safe to do. While we're at it, also add static assertions that enforce the necessary requirements of a UID type explicitly within the ShaderUid class. This way, we can remove several memset calls around the shader generation code that makes sure the underlying UID data is zeroed out. Now our ShaderUid class enforces this for us, so we don't need to care about it at the usage sites.
2019-05-29 23:05:06 -06:00
UidData* const uid_data = out.GetUidData();
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
uid_data->dst_format = dst_format;
Convert BPMemory to BitField and enum class Additional changes: - For TevStageCombiner's ColorCombiner and AlphaCombiner, op/comparison and scale/compare_mode have been split as there are different meanings and enums if bias is set to compare. (Shift has also been renamed to scale) - In TexMode0, min_filter has been split into min_mip and min_filter. - In TexImage1, image_type is now cache_manually_managed. - The unused bit in GenMode is now exposed. - LPSize's lineaspect is now named adjust_for_aspect_ratio.
2021-02-10 19:11:31 -07:00
uid_data->efb_has_alpha = bpmem.zcontrol.pixel_format == PixelFormat::RGBA6_Z24;
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
uid_data->is_depth_copy = is_depth_copy;
uid_data->is_intensity = is_intensity;
uid_data->scale_by_half = scale_by_half;
TextureConversionShader: Don't sample from adjacent rows when not needed
2018-05-02 22:09:32 -06:00
uid_data->copy_filter = copy_filter;
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
return out;
}
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
static void WriteHeader(APIType api_type, ShaderCode& out)
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
{
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
if (api_type == APIType::D3D)
TextureConverterShaderGen: Make generated shaders HLSL-compatible
2017-12-05 10:00:40 -07:00
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("cbuffer PSBlock : register(b0) {{\n"
" float2 src_offset, src_size;\n"
" float3 filter_coefficients;\n"
" float gamma_rcp;\n"
" float2 clamp_tb;\n"
" float pixel_height;\n"
"}};\n\n");
TextureConverterShaderGen: Make generated shaders HLSL-compatible
2017-12-05 10:00:40 -07:00
}
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
else if (api_type == APIType::OpenGL || api_type == APIType::Vulkan)
TextureConverterShaderGen: Make generated shaders HLSL-compatible
2017-12-05 10:00:40 -07:00
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("UBO_BINDING(std140, 1) uniform PSBlock {{\n"
" float2 src_offset, src_size;\n"
" float3 filter_coefficients;\n"
" float gamma_rcp;\n"
" float2 clamp_tb;\n"
" float pixel_height;\n"
"}};\n");
TextureConverterShaderGen: Make generated shaders HLSL-compatible
2017-12-05 10:00:40 -07:00
}
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
}
ShaderCode GenerateVertexShader(APIType api_type)
{
ShaderCode out;
WriteHeader(api_type, out);
if (api_type == APIType::D3D)
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("void main(in uint id : SV_VertexID, out float3 v_tex0 : TEXCOORD0,\n"
" out float4 opos : SV_Position) {{\n");
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
}
else if (api_type == APIType::OpenGL || api_type == APIType::Vulkan)
{
ShaderGen: Use interface blocks when geometry shaders are supported We don't use explicit locations in OpenGL currently, so this breaks when we use alternative names in the geometry shaders.
2019-04-20 07:44:33 -06:00
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("VARYING_LOCATION(0) out VertexData {{\n"
" float3 v_tex0;\n"
"}};\n");
ShaderGen: Use interface blocks when geometry shaders are supported We don't use explicit locations in OpenGL currently, so this breaks when we use alternative names in the geometry shaders.
2019-04-20 07:44:33 -06:00
}
else
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("VARYING_LOCATION(0) out float3 v_tex0;\n");
ShaderGen: Use interface blocks when geometry shaders are supported We don't use explicit locations in OpenGL currently, so this breaks when we use alternative names in the geometry shaders.
2019-04-20 07:44:33 -06:00
}
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("#define id gl_VertexID\n"
"#define opos gl_Position\n"
"void main() {{\n");
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
}
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" v_tex0 = float3(float((id << 1) & 2), float(id & 2), 0.0f);\n");
out.Write(
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
" opos = float4(v_tex0.xy * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);\n");
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" v_tex0 = float3(src_offset + (src_size * v_tex0.xy), 0.0f);\n");
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
// NDC space is flipped in Vulkan
if (api_type == APIType::Vulkan)
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" opos.y = -opos.y;\n");
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("}}\n");
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
return out;
}
ShaderCode GeneratePixelShader(APIType api_type, const UidData* uid_data)
{
const bool mono_depth = uid_data->is_depth_copy && g_ActiveConfig.bStereoEFBMonoDepth;
ShaderCode out;
WriteHeader(api_type, out);
if (api_type == APIType::D3D)
TextureConverterShaderGen: Make generated shaders HLSL-compatible
2017-12-05 10:00:40 -07:00
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("Texture2DArray tex0 : register(t0);\n"
"SamplerState samp0 : register(s0);\n"
"float4 SampleEFB(float3 uv, float y_offset) {{\n"
" return tex0.Sample(samp0, float3(uv.x, clamp(uv.y + (y_offset * pixel_height), "
"clamp_tb.x, clamp_tb.y), {}));\n"
"}}\n\n",
mono_depth ? "0.0" : "uv.z");
out.Write("void main(in float3 v_tex0 : TEXCOORD0, out float4 ocol0 : SV_Target)\n{{\n");
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
}
else if (api_type == APIType::OpenGL || api_type == APIType::Vulkan)
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("SAMPLER_BINDING(0) uniform sampler2DArray samp0;\n");
out.Write("float4 SampleEFB(float3 uv, float y_offset) {{\n"
" return texture(samp0, float3(uv.x, clamp(uv.y + (y_offset * pixel_height), "
"clamp_tb.x, clamp_tb.y), {}));\n"
"}}\n",
mono_depth ? "0.0" : "uv.z");
ShaderGen: Use interface blocks when geometry shaders are supported We don't use explicit locations in OpenGL currently, so this breaks when we use alternative names in the geometry shaders.
2019-04-20 07:44:33 -06:00
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("VARYING_LOCATION(0) in VertexData {{\n"
" float3 v_tex0;\n"
"}};\n");
ShaderGen: Use interface blocks when geometry shaders are supported We don't use explicit locations in OpenGL currently, so this breaks when we use alternative names in the geometry shaders.
2019-04-20 07:44:33 -06:00
}
else
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("VARYING_LOCATION(0) in vec3 v_tex0;\n");
ShaderGen: Use interface blocks when geometry shaders are supported We don't use explicit locations in OpenGL currently, so this breaks when we use alternative names in the geometry shaders.
2019-04-20 07:44:33 -06:00
}
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("FRAGMENT_OUTPUT_LOCATION(0) out vec4 ocol0;"
"void main()\n{{\n");
TextureConverterShaderGen: Make generated shaders HLSL-compatible
2017-12-05 10:00:40 -07:00
}
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
Implement EFB copy filter and gamma in hardware backends Also makes y_scale a dynamic parameter for EFB copies, as it doesn't make sense to keep it as part of the uid, otherwise we're generating redundant shaders.
2018-04-29 02:52:30 -06:00
// The copy filter applies to both color and depth copies. This has been verified on hardware.
// The filter is only applied to the RGB channels, the alpha channel is left intact.
TextureConversionShader: Don't sample from adjacent rows when not needed
2018-05-02 22:09:32 -06:00
if (uid_data->copy_filter)
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" float4 prev_row = SampleEFB(v_tex0, -1.0f);\n"
" float4 current_row = SampleEFB(v_tex0, 0.0f);\n"
" float4 next_row = SampleEFB(v_tex0, 1.0f);\n"
" float4 texcol = float4(min(prev_row.rgb * filter_coefficients[0] +\n"
" current_row.rgb * filter_coefficients[1] +\n"
" next_row.rgb * filter_coefficients[2], \n"
" float3(1, 1, 1)), current_row.a);\n");
TextureConversionShader: Don't sample from adjacent rows when not needed
2018-05-02 22:09:32 -06:00
}
else
{
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
" float4 current_row = SampleEFB(v_tex0, 0.0f);\n"
EFB2RAM: Apply copy filter as a float coefficient after sampling Using 8-bit integer math here lead to precision loss for depth copies, which broke various effects in games, e.g. lens flare in MK:DD. It's unlikely the console implements this as a floating-point multiply (fixed-point perhaps), but since we have the float round trip in our EFB2RAM shaders anyway, it's not going to make things any worse. If we do rewrite our shaders to use integer math completely, then it might be worth switching this conversion back to integers. However, the range of the values (format) should be known, or we should expand all values out to 24-bits first.
2018-05-21 20:14:48 -06:00
" float4 texcol = float4(min(current_row.rgb * filter_coefficients[1], float3(1, 1, 1)),\n"
" current_row.a);\n");
TextureConversionShader: Don't sample from adjacent rows when not needed
2018-05-02 22:09:32 -06:00
}
Implement EFB copy filter and gamma in hardware backends Also makes y_scale a dynamic parameter for EFB copies, as it doesn't make sense to keep it as part of the uid, otherwise we're generating redundant shaders.
2018-04-29 02:52:30 -06:00
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
if (uid_data->is_depth_copy)
{
Move most backend functionality to VideoCommon
2019-02-14 18:59:50 -07:00
if (!g_ActiveConfig.backend_info.bSupportsReversedDepthRange)
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("texcol.x = 1.0 - texcol.x;\n");
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" int depth = int(texcol.x * 16777216.0);\n"
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
// Convert to Z24 format
" int4 workspace;\n"
" workspace.r = (depth >> 16) & 255;\n"
" workspace.g = (depth >> 8) & 255;\n"
" workspace.b = depth & 255;\n"
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
// Convert to Z4 format
" workspace.a = (depth >> 16) & 0xF0;\n"
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
// Normalize components to [0.0..1.0]
" texcol = float4(workspace) / 255.0;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
switch (uid_data->dst_format)
{
case EFBCopyFormat::R4: // Z4
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.aaaa;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
case EFBCopyFormat::R8_0x1: // Z8
case EFBCopyFormat::R8: // Z8H
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.rrrr;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::RA8: // Z16
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.gggr;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::RG8: // Z16 (reverse order)
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.rrrg;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::RGBA8: // Z24X8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = float4(texcol.rgb, 1.0);\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::G8: // Z8M
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.gggg;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::B8: // Z8L
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.bbbb;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::GB8: // Z16L - copy lower 16 depth bits
// expected to be used as an IA8 texture (upper 8 bits stored as intensity, lower 8 bits
// stored as alpha)
// Used e.g. in Zelda: Skyward Sword
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.gggb;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
default:
VideoCommon: Migrate over to fmt Migrates off the printf-based formatting where applicable.
2020-11-13 20:33:26 -07:00
ERROR_LOG_FMT(VIDEO, "Unknown copy zbuf format: {:#X}",
static_cast<int>(uid_data->dst_format));
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = float4(texcol.bgr, 0.0);\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
}
}
else if (uid_data->is_intensity)
{
Reimplement EFB copy for intensity formats This is the direct continuation of #5887, and removes the last usage of the colmat matrix in EFB copy conversion shaders.
2017-12-02 16:10:00 -07:00
bool has_four_bits =
(uid_data->dst_format == EFBCopyFormat::R4 || uid_data->dst_format == EFBCopyFormat::RA4);
bool has_alpha =
(uid_data->dst_format == EFBCopyFormat::RA4 || uid_data->dst_format == EFBCopyFormat::RA8);
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
switch (uid_data->dst_format)
{
case EFBCopyFormat::R4: // I4
case EFBCopyFormat::R8_0x1: // I8
case EFBCopyFormat::R8: // I8
case EFBCopyFormat::RA4: // IA4
case EFBCopyFormat::RA8: // IA8
Reimplement EFB copy for intensity formats This is the direct continuation of #5887, and removes the last usage of the colmat matrix in EFB copy conversion shaders.
2017-12-02 16:10:00 -07:00
if (has_four_bits)
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" texcol = float4(int4(texcol * 255.0) & 0xF0) * (1.0 / 240.0);\n");
Reimplement EFB copy for intensity formats This is the direct continuation of #5887, and removes the last usage of the colmat matrix in EFB copy conversion shaders.
2017-12-02 16:10:00 -07:00
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
// TODO - verify these coefficients
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" const float3 coefficients = float3(0.257, 0.504, 0.098);\n"
" float intensity = dot(texcol.rgb, coefficients) + 16.0 / 255.0;\n"
" ocol0 = float4(intensity, intensity, intensity, {});\n",
has_alpha ? "texcol.a" : "intensity");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
default:
VideoCommon: Migrate over to fmt Migrates off the printf-based formatting where applicable.
2020-11-13 20:33:26 -07:00
ERROR_LOG_FMT(VIDEO, "Unknown copy intensity format: {:#X}",
static_cast<int>(uid_data->dst_format));
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
}
}
else
{
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
if (!uid_data->efb_has_alpha)
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" texcol.a = 1.0;\n");
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
switch (uid_data->dst_format)
{
case EFBCopyFormat::R4: // R4
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" float red = float(int(texcol.r * 255.0) & 0xF0) * (1.0 / 240.0);\n"
" ocol0 = float4(red, red, red, red);\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
case EFBCopyFormat::R8_0x1: // R8
case EFBCopyFormat::R8: // R8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.rrrr;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::RA4: // RA4
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" float2 red_alpha = float2(int2(texcol.ra * 255.0) & 0xF0) * (1.0 / 240.0);\n"
" ocol0 = red_alpha.rrrg;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
case EFBCopyFormat::RA8: // RA8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.rrra;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::A8: // A8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.aaaa;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::G8: // G8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.gggg;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
case EFBCopyFormat::B8: // B8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.bbbb;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::RG8: // RG8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.rrrg;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::GB8: // GB8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol.gggb;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::RGB565: // RGB565
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" float2 red_blue = float2(int2(texcol.rb * 255.0) & 0xF8) * (1.0 / 248.0);\n"
" float green = float(int(texcol.g * 255.0) & 0xFC) * (1.0 / 252.0);\n"
" ocol0 = float4(red_blue.r, green, red_blue.g, 1.0);\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
case EFBCopyFormat::RGB5A3: // RGB5A3
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
// TODO: The MSB controls whether we have RGB5 or RGB4A3, this selection
// will need to be implemented once we move away from floats.
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" float3 color = float3(int3(texcol.rgb * 255.0) & 0xF8) * (1.0 / 248.0);\n"
" float alpha = float(int(texcol.a * 255.0) & 0xE0) * (1.0 / 224.0);\n"
" ocol0 = float4(color, alpha);\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
VideoCommon: Reimplement texture format conversion This fixes the rendering of the scan visor in Metroid Prime 2: Echoes, as seen in https://fifoci.dolphin-emu.org/dff/mp2-scanner/ The alpha channel was off-by-one on Ivy Bridge due to the rounding after multiplication with colmat. This commit removes this matrix altogether in most cases, making them simple GLSL swizzles.
2017-08-04 12:14:14 -06:00
case EFBCopyFormat::RGBA8: // RGBA8
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
Implement EFB copy filter and gamma in hardware backends Also makes y_scale a dynamic parameter for EFB copies, as it doesn't make sense to keep it as part of the uid, otherwise we're generating redundant shaders.
2018-04-29 02:52:30 -06:00
case EFBCopyFormat::XFB:
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(
TextureConverterShader: Write EFB2Tex XFB copies with alpha value of 1 This way we don't end up with artifacts of the EFB's alpha values in frame dumps. XFB copies loaded from RAM also set the alpha to 1, so this will match.
2019-02-01 21:14:58 -07:00
" ocol0 = float4(pow(texcol.rgb, float3(gamma_rcp, gamma_rcp, gamma_rcp)), 1.0f);\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
default:
VideoCommon: Migrate over to fmt Migrates off the printf-based formatting where applicable.
2020-11-13 20:33:26 -07:00
ERROR_LOG_FMT(VIDEO, "Unknown copy color format: {:#X}",
static_cast<int>(uid_data->dst_format));
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write(" ocol0 = texcol;\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
break;
}
}
ShaderGenCommon: Rename WriteFmt() to Write() Now that we've converted all of the shader generators over to using fmt, we can drop the old Write() member function and perform a rename operation on the WriteFmt() to turn it into the new Write() function. All changes within this are the removal of a <cstdarg> header, since the previous printf-based Write() required it, and renaming. No functional changes are made at all.
2020-11-09 00:23:32 -07:00
out.Write("}}\n");
VideoBackends: Use VideoCommon shader generators for efb2tex copies. This will generate one shader per copy format. For now, it is the same shader with the colmat hard coded. So it should already improve the GPU performance a bit, but a rewrite of the shader generator is suggested. Half of the patch is done by linkmauve1: VideoCommon: Reorganise the shader writes.
2017-08-04 09:56:24 -06:00
return out;
}
VideoCommon: Create a namespace for TextureConversionShaderGen.
2017-11-25 03:07:14 -07:00
} // namespace TextureConversionShaderGen
Reference in New Issue Copy Permalink