From ef16086fea92b1cbb767077535bf469b518cb8c0 Mon Sep 17 00:00:00 2001 From: Jezze Date: Sat, 22 Oct 2016 22:39:14 +0200 Subject: [PATCH] Disabled clearing of render targets in several passes (HLSL) - which was quite a performance overhead and not necessary because the shaders fill every texel of a target without blending --- hlsl/distortion.fx | 16 +++++---- hlsl/post.fx | 29 ++++++++-------- src/osd/modules/render/d3d/d3dhlsl.cpp | 47 ++++++++++---------------- 3 files changed, 40 insertions(+), 52 deletions(-) diff --git a/hlsl/distortion.fx b/hlsl/distortion.fx index 7624d962a58..3dd1dbd44b6 100644 --- a/hlsl/distortion.fx +++ b/hlsl/distortion.fx @@ -120,6 +120,7 @@ uniform float RoundCornerAmount = 0.0f; uniform float SmoothBorderAmount = 0.0f; uniform float VignettingAmount = 0.0f; uniform float ReflectionAmount = 0.0f; +uniform float3 LightReflectionColor = float3(1.0f, 0.90f, 0.80f); // color temperature 5.000 Kelvin uniform bool SwapXY = false; @@ -280,7 +281,12 @@ float4 ps_main(PS_INPUT Input) : COLOR float2 QuadCoord = GetQuadCoords(Input.TexCoord, BaseTargetQuadScale, distortCornerAmount, 0.0f); // clip border - clip(BaseCoord < 0.0f - TexelDims || BaseCoord > 1.0f + TexelDims ? -1 : 1); + if (BaseCoord.x < 0.0f - TexelDims.x || BaseCoord.y < 0.0f - TexelDims.y || + BaseCoord.x > 1.0f + TexelDims.x || BaseCoord.y > 1.0f + TexelDims.y) + { + // we don't use the clip function, because we don't clear the render target before + return float4(0.0f, 0.0f, 0.0f, 1.0f); + } // Color float4 BaseColor = tex2D(DiffuseSampler, BaseCoord); @@ -293,14 +299,10 @@ float4 ps_main(PS_INPUT Input) : COLOR BaseColor.rgb *= VignetteFactor; // Light Reflection Simulation - float3 LightColor = float3(1.0f, 0.90f, 0.80f); // color temperature 5.000 Kelvin - float2 SpotCoord = QuadCoord; - float2 NoiseCoord = QuadCoord; - float SpotAddend = GetSpotAddend(SpotCoord, ReflectionAmount); - float NoiseFactor = GetNoiseFactor(SpotAddend, random(NoiseCoord)); - BaseColor.rgb += SpotAddend * NoiseFactor * LightColor; + float SpotAddend = GetSpotAddend(SpotCoord, ReflectionAmount) * LightReflectionColor; + BaseColor.rgb += SpotAddend * GetNoiseFactor(SpotAddend, random(SpotCoord)); // Round Corners Simulation float2 RoundCornerCoord = QuadCoord; diff --git a/hlsl/post.fx b/hlsl/post.fx index e6dda5a9474..dcc4ee3b194 100644 --- a/hlsl/post.fx +++ b/hlsl/post.fx @@ -49,17 +49,15 @@ struct VS_OUTPUT { float4 Position : POSITION; float4 Color : COLOR0; - float2 SourceCoord : TEXCOORD0; - float2 TexCoord : TEXCOORD1; - float2 ScreenCoord : TEXCOORD2; + float2 TexCoord : TEXCOORD0; + float2 ScreenCoord : TEXCOORD1; }; struct PS_INPUT { float4 Color : COLOR0; - float2 SourceCoord : TEXCOORD0; - float2 TexCoord : TEXCOORD1; - float2 ScreenCoord : TEXCOORD2; + float2 TexCoord : TEXCOORD0; + float2 ScreenCoord : TEXCOORD1; }; //----------------------------------------------------------------------------- @@ -104,9 +102,6 @@ VS_OUTPUT vs_main(VS_INPUT Input) Output.ScreenCoord = Input.Position.xy / ScreenDims; - Output.SourceCoord = Input.TexCoord; - Output.SourceCoord += 0.5f / TargetDims; // fix half texel offset (DX9) - Output.Color = Input.Color; return Output; @@ -209,19 +204,23 @@ float4 ps_main(PS_INPUT Input) : COLOR { float2 ScreenCoord = Input.ScreenCoord; float2 BaseCoord = GetAdjustedCoords(Input.TexCoord); - float2 SourceCoord = GetAdjustedCoords(Input.SourceCoord); // Color float4 BaseColor = tex2D(DiffuseSampler, BaseCoord); BaseColor.a = 1.0f; // clip border - clip(BaseCoord < 0.0f || BaseCoord > 1.0f ? -1 : 1); + if (BaseCoord.x < 0.0f || BaseCoord.y < 0.0f || + BaseCoord.x > 1.0f || BaseCoord.y > 1.0f) + { + // we don't use the clip function, because we don't clear the render target before + return float4(0.0f, 0.0f, 0.0f, 1.0f); + } // Mask Simulation (may not affect bloom) if (!PrepareBloom && ShadowAlpha > 0.0f) { - float2 ShadowCoord = GetShadowCoord(ScreenCoord, SourceCoord); + float2 ShadowCoord = GetShadowCoord(ScreenCoord, BaseCoord); float4 ShadowColor = tex2D(ShadowSampler, ShadowCoord); float3 ShadowMaskColor = lerp(1.0f, ShadowColor.rgb, ShadowAlpha); @@ -256,7 +255,7 @@ float4 ps_main(PS_INPUT Input) : COLOR float ColorBrightness = 0.299f * BaseColor.r + 0.587f * BaseColor.g + 0.114 * BaseColor.b; - float ScanlineCoord = SourceCoord.y; + float ScanlineCoord = BaseCoord.y; ScanlineCoord += SwapXY ? QuadDims.x <= SourceDims.x * 2.0f ? 0.5f / QuadDims.x // uncenter scanlines if the quad is less than twice the size of the source @@ -280,7 +279,7 @@ float4 ps_main(PS_INPUT Input) : COLOR if (!VectorScreen && HumBarAlpha > 0.0f) { float HumBarStep = frac(TimeMilliseconds * HumBarDesync); - float HumBarBrightness = 1.0 - frac(SourceCoord.y + HumBarStep) * HumBarAlpha; + float HumBarBrightness = 1.0 - frac(BaseCoord.y + HumBarStep) * HumBarAlpha; BaseColor.rgb *= HumBarBrightness; } } @@ -301,4 +300,4 @@ technique DefaultTechnique VertexShader = compile vs_3_0 vs_main(); PixelShader = compile ps_3_0 ps_main(); } -} \ No newline at end of file +} diff --git a/src/osd/modules/render/d3d/d3dhlsl.cpp b/src/osd/modules/render/d3d/d3dhlsl.cpp index db39deb0aad..3ea23ba5b04 100644 --- a/src/osd/modules/render/d3d/d3dhlsl.cpp +++ b/src/osd/modules/render/d3d/d3dhlsl.cpp @@ -841,10 +841,10 @@ void shaders::begin_draw() downsample_effect->set_technique("DefaultTechnique"); vector_effect->set_technique("DefaultTechnique"); - HRESULT result = d3d->get_device()->GetRenderTarget(0, &backbuffer); + HRESULT result = d3d->get_device()->SetRenderTarget(0, backbuffer); if (FAILED(result)) { - osd_printf_verbose("Direct3D: Error %08lX during device GetRenderTarget call\n", result); + osd_printf_verbose("Direct3D: Error %08lX during device SetRenderTarget call\n", result); } } @@ -940,7 +940,7 @@ int shaders::ntsc_pass(d3d_render_target *rt, int source_index, poly_info *poly, curr_effect->set_float("SignalOffset", signal_offset); next_index = rt->next_index(next_index); - blit(rt->source_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->source_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); color_effect->set_texture("Diffuse", rt->source_texture[next_index]); @@ -997,7 +997,7 @@ int shaders::color_convolution_pass(d3d_render_target *rt, int source_index, pol // initial "Diffuse" texture is set in shaders::set_texture() or the result of shaders::ntsc_pass() next_index = rt->next_index(next_index); - blit(rt->source_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->source_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1011,7 +1011,7 @@ int shaders::prescale_pass(d3d_render_target *rt, int source_index, poly_info *p curr_effect->set_texture("Diffuse", rt->source_texture[next_index]); next_index = rt->next_index(next_index); - blit(rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->target_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1034,7 +1034,7 @@ int shaders::deconverge_pass(d3d_render_target *rt, int source_index, poly_info curr_effect->set_texture("Diffuse", rt->target_texture[next_index]); next_index = rt->next_index(next_index); - blit(rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->target_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1054,7 +1054,7 @@ int shaders::defocus_pass(d3d_render_target *rt, int source_index, poly_info *po curr_effect->set_texture("Diffuse", rt->target_texture[next_index]); next_index = rt->next_index(next_index); - blit(rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->target_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1076,7 +1076,7 @@ int shaders::phosphor_pass(d3d_render_target *rt, int source_index, poly_info *p curr_effect->set_bool("Passthrough", false); next_index = rt->next_index(next_index); - blit(rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->target_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); // Pass along our phosphor'd screen curr_effect->update_uniforms(); @@ -1084,8 +1084,7 @@ int shaders::phosphor_pass(d3d_render_target *rt, int source_index, poly_info *p curr_effect->set_texture("LastPass", rt->target_texture[next_index]); curr_effect->set_bool("Passthrough", true); - // Avoid changing targets due to page flipping - blit(rt->cache_surface, true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->cache_surface, false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1130,7 +1129,7 @@ int shaders::post_pass(d3d_render_target *rt, int source_index, poly_info *poly, curr_effect->set_bool("PrepareBloom", prepare_bloom); next_index = rt->next_index(next_index); - blit(prepare_bloom ? rt->source_surface[next_index] : rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(prepare_bloom ? rt->source_surface[next_index] : rt->target_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1156,7 +1155,7 @@ int shaders::downsample_pass(d3d_render_target *rt, int source_index, poly_info ? rt->source_texture[next_index] : rt->bloom_texture[bloom_index - 1]); - blit(rt->bloom_surface[bloom_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->bloom_surface[bloom_index], false, D3DPT_TRIANGLELIST, 0, 2); } return next_index; @@ -1204,7 +1203,7 @@ int shaders::bloom_pass(d3d_render_target *rt, int source_index, poly_info *poly } next_index = rt->next_index(next_index); - blit(rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->target_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1230,7 +1229,7 @@ int shaders::distortion_pass(d3d_render_target *rt, int source_index, poly_info curr_effect->set_texture("DiffuseTexture", rt->target_texture[next_index]); next_index = rt->next_index(next_index); - blit(rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); + blit(rt->target_surface[next_index], false, D3DPT_TRIANGLELIST, 0, 2); return next_index; } @@ -1245,6 +1244,7 @@ int shaders::vector_pass(d3d_render_target *rt, int source_index, poly_info *pol curr_effect->set_float("LengthScale", options->vector_length_scale); curr_effect->set_float("BeamSmooth", options->vector_beam_smooth); + // we need to clear the vector render target here blit(rt->target_surface[next_index], true, poly->type(), vertnum, poly->count()); return next_index; @@ -1260,6 +1260,7 @@ int shaders::vector_buffer_pass(d3d_render_target *rt, int source_index, poly_in curr_effect->set_texture("Diffuse", rt->target_texture[next_index]); + // we need to clear the vector render target here next_index = rt->next_index(next_index); blit(rt->target_surface[next_index], true, D3DPT_TRIANGLELIST, 0, 2); @@ -1276,27 +1277,19 @@ int shaders::screen_pass(d3d_render_target *rt, int source_index, poly_info *pol curr_effect->set_texture("Diffuse", rt->target_texture[next_index]); - // we do not clear the backbuffer here because multiple screens might be rendered into blit(backbuffer, false, poly->type(), vertnum, poly->count()); if (recording_movie) { blit(recorder->target_surface(), false, poly->type(), vertnum, poly->count()); - HRESULT result = d3d->get_device()->SetRenderTarget(0, backbuffer); - if (FAILED(result)) - osd_printf_verbose("Direct3D: Error %08lX during device SetRenderTarget call\n", result); - recorder->save_frame(); } if (render_snap) { - blit(snap_target, false, poly->type(), vertnum, poly->count()); - - HRESULT result = d3d->get_device()->SetRenderTarget(0, backbuffer); - if (FAILED(result)) - osd_printf_verbose("Direct3D: Error %08lX during device SetRenderTarget call\n", result); + // we need to clear the snap render target here + blit(snap_target, true, poly->type(), vertnum, poly->count()); render_snapshot(snap_target); @@ -1458,12 +1451,6 @@ void shaders::render_quad(poly_info *poly, int vertnum) next_index = screen_pass(rt, next_index, poly, vertnum); d3d->set_wrap(PRIMFLAG_GET_TEXWRAP(curr_texture->get_flags()) ? D3DTADDRESS_WRAP : D3DTADDRESS_CLAMP); - HRESULT result = d3d->get_device()->SetRenderTarget(0, backbuffer); - if (FAILED(result)) - { - osd_printf_verbose("Direct3D: Error %08lX during device SetRenderTarget call\n", result); - } - curr_screen++; } else