Skip to content

SIMD optimisation in Hudwarp code #21

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
barnabwhy wants to merge 7 commits into
base: master
Choose a base branch
from
Open

SIMD optimisation in Hudwarp code #21

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
0f09896
Hudwarp border size define in shader
barnabwhy Nov 30, 2024
0a1f7b4
Scale HUD render by 1.333x for clarity
barnabwhy Nov 30, 2024
aea2194
Lazily update hudwarp constant buffer
barnabwhy Nov 30, 2024
ffceecb
Reorder contents of MiscRenderHooks to better match TFORevive
barnabwhy Nov 30, 2024
0af2cc6
SIMD hudwarp (for gpu & disabled)
barnabwhy Nov 30, 2024
c8d87b7
use _mm_fmadd_ps and _mm_fmsub_ps
barnabwhy Nov 30, 2024
3c734c2
hudwarp process 2 verts at once
barnabwhy Nov 30, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
52 changes: 38 additions & 14 deletions bmedll/Hudwarp.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,9 @@ ID3D11PixelShader* pPixelShader;
#define MAINVP 0
D3D11_VIEWPORT viewport{ 0 };

// render hud at 1440p on 1080p and similar scale
#define HUD_RENDER_RES 1.333f

HudwarpProcess::HudwarpProcess(ID3D11Device* pDevice, ID3D11DeviceContext** ppID3D11DeviceContext) : m_pDevice(pDevice), m_pContext(*ppID3D11DeviceContext)
{
// Initialize shaders
Expand Down Expand Up @@ -70,8 +73,8 @@ HudwarpProcess::HudwarpProcess(ID3D11Device* pDevice, ID3D11DeviceContext** ppID
m_height = *reinterpret_cast<unsigned int*>(vguimatsurfacedllBaseAddress + 0x290DD8 + 4);

// We add a border to the texture so that the HUD can't get cut off by the texture boundaries
unsigned int widthWithBorder = m_width * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f);
unsigned int heightWithBorder = m_height * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f);
unsigned int widthWithBorder = m_width * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f) * HUD_RENDER_RES;
unsigned int heightWithBorder = m_height * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f) * HUD_RENDER_RES;

// Setup the texture descriptor
D3D11_TEXTURE2D_DESC textureDesc;
Expand Down Expand Up @@ -125,6 +128,7 @@ HudwarpProcess::HudwarpProcess(ID3D11Device* pDevice, ID3D11DeviceContext** ppID
mOrtho = XMMatrixOrthographicLH(1.0f, 1.0f, 0.0f, 1.0f);
ConstantBuffer cb;
cb.mProjection = mOrtho;
cb.aspectRatio = (float)m_width / (float)m_height;
cb.xWarp = 0.0f;
cb.xScale = 1.0f;
cb.yWarp = 0.0f;
Expand Down Expand Up @@ -246,8 +250,8 @@ void HudwarpProcess::Resize(unsigned int w, unsigned int h)
m_height = *reinterpret_cast<unsigned int*>(vguimatsurfacedllBaseAddress + 0x290DD8 + 4);

// We add a border to the texture so that the HUD can't get cut off by the texture boundaries
unsigned int widthWithBorder = m_width * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f);
unsigned int heightWithBorder = m_height * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f);
unsigned int widthWithBorder = m_width * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f) * HUD_RENDER_RES;
unsigned int heightWithBorder = m_height * (1.0f + HUD_TEX_BORDER_SIZE * 2.0f) * HUD_RENDER_RES;

// Setup the texture descriptor
D3D11_TEXTURE2D_DESC textureDesc;
Expand All @@ -263,6 +267,20 @@ void HudwarpProcess::Resize(unsigned int w, unsigned int h)
textureDesc.CPUAccessFlags = 0;
textureDesc.MiscFlags = 0;

// Update the constant buffer
ConstantBuffer cb;
cb.mProjection = mOrtho;
cb.aspectRatio = (float)m_width / (float)m_height;
cb.xWarp = m_hudwarpSettings.xWarp;
cb.xScale = m_hudwarpSettings.xScale;
cb.yWarp = m_hudwarpSettings.yWarp;
cb.yScale = m_hudwarpSettings.yScale;
cb.viewDist = m_hudwarpSettings.viewDist;

m_pContext->UpdateSubresource(m_pConstantBuffer, 0, 0, &cb, 0, 0);

m_shouldUpdateConstantBuffer = false;

// Create the render texture
m_pDevice->CreateTexture2D(&textureDesc, NULL, &m_pRenderTexture);

Expand All @@ -289,6 +307,7 @@ void HudwarpProcess::Resize(unsigned int w, unsigned int h)
void HudwarpProcess::UpdateSettings(HudwarpSettings* hudwarpSettings)
{
m_hudwarpSettings = *hudwarpSettings;
m_shouldUpdateConstantBuffer = true;
}

void HudwarpProcess::Begin()
Expand Down Expand Up @@ -357,16 +376,21 @@ void HudwarpProcess::Finish()
m_pContext->IASetPrimitiveTopology(D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST);

// Update the constant buffer
ConstantBuffer cb;
cb.mProjection = mOrtho;
cb.aspectRatio = (float)m_width / (float)m_height;
cb.xWarp = m_hudwarpSettings.xWarp;
cb.xScale = m_hudwarpSettings.xScale;
cb.yWarp = m_hudwarpSettings.yWarp;
cb.yScale = m_hudwarpSettings.yScale;
cb.viewDist = m_hudwarpSettings.viewDist;

m_pContext->UpdateSubresource(m_pConstantBuffer, 0, 0, &cb, 0, 0);
if (m_shouldUpdateConstantBuffer)
{
ConstantBuffer cb;
cb.mProjection = mOrtho;
cb.aspectRatio = (float)m_width / (float)m_height;
cb.xWarp = m_hudwarpSettings.xWarp;
cb.xScale = m_hudwarpSettings.xScale;
cb.yWarp = m_hudwarpSettings.yWarp;
cb.yScale = m_hudwarpSettings.yScale;
cb.viewDist = m_hudwarpSettings.viewDist;

m_pContext->UpdateSubresource(m_pConstantBuffer, 0, 0, &cb, 0, 0);

m_shouldUpdateConstantBuffer = false;
}

// Set shader resources
m_pContext->VSSetConstantBuffers(0, 1, &m_pConstantBuffer);
Expand Down
7 changes: 5 additions & 2 deletions bmedll/Hudwarp.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,7 @@ class HudwarpProcess
ID3D11InputLayout* m_pVertexLayout = nullptr;
ID3D11Buffer* m_pIndexBuffer = nullptr;
ID3D11Buffer* m_pConstantBuffer = nullptr;
bool m_shouldUpdateConstantBuffer = false;

ID3D11RasterizerState* m_pCWcullMode = NULL;

Expand Down Expand Up @@ -77,8 +78,10 @@ struct Vertex
XMFLOAT2 texCoord;
};

// Must match hudScale in shader
// Used for hudScale in shader, prevents cut off HUD
// These values must match or the HUD will end up scaled wrong
#define HUD_TEX_BORDER_SIZE 0.025f
#define HUD_TEX_BORDER_SIZE_STR "0.025f"

constexpr const char* hudwarpShader = R"(
cbuffer ConstantBuffer : register(b0)
Expand Down Expand Up @@ -116,7 +119,7 @@ sampler Sampler : register(s0);
float2 UndoHudTexBorder(float2 texCoord)
{
// IMPORTANT: must match value of HUD_TEX_BORDER_SIZE
float hudTexBorderSize = 0.025f;
float hudTexBorderSize = )" HUD_TEX_BORDER_SIZE_STR R"(;

float hudScale = 1.0f + 2.0f * hudTexBorderSize;
float hudOffset = 0.5f - (0.5f / hudScale);
Expand Down
Loading