Mirrors_Framework/SDL
1
0
Fork 0
mirror of https://github.com/libsdl-org/SDL.git synced 2026-05-31 22:42:52 +08:00
Code Issues Actions 3 Packages Projects Releases Wiki Activity

[N-Gage] Simplify rendering back-end; more micro-optimization

Browse Source 
- Remove redundant null checks
- Use cached texture properties instead of API calls (GetBitmapWidth/Height/Pitch)
- Eliminate duplicate SDL_GetRenderScale() call in Copy()
- Reorder CopyEx() fast paths to check no-transform case first
- Combine operations in NGAGE_ConvertColor() to reduce intermediate steps
This commit is contained in:
Michael Fitzmayer
2026-04-16 21:09:58 +02:00
parent b53b31b74a
commit 87e356f102
1 changed files with 60 additions and 76 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/render/ngage/SDL_render_ngage.cpp
+53 -69
View File
@@ -82,7 +82,7 @@ void NGAGE_DestroyTextureData(NGAGE_TextureData *data)
void *NGAGE_GetBitmapDataAddress(NGAGE_TextureData *data) void *NGAGE_GetBitmapDataAddress(NGAGE_TextureData *data)
{ {
if (data && data->bitmap) { if (data) {
return data->bitmap->DataAddress(); return data->bitmap->DataAddress();
} }
return NULL; return NULL;
@@ -90,25 +90,24 @@ void *NGAGE_GetBitmapDataAddress(NGAGE_TextureData *data)
int NGAGE_GetBitmapPitch(NGAGE_TextureData *data) int NGAGE_GetBitmapPitch(NGAGE_TextureData *data)
{ {
if (data && data->bitmap) { if (data) {
TSize size = data->bitmap->SizeInPixels(); return data->cachedPitch;
return data->bitmap->ScanLineLength(size.iWidth, data->bitmap->DisplayMode());
} }
return 0; return 0;
} }
int NGAGE_GetBitmapWidth(NGAGE_TextureData *data) int NGAGE_GetBitmapWidth(NGAGE_TextureData *data)
{ {
if (data && data->bitmap) { if (data) {
return data->bitmap->SizeInPixels().iWidth; return data->cachedWidth;
} }
return 0; return 0;
} }
int NGAGE_GetBitmapHeight(NGAGE_TextureData *data) int NGAGE_GetBitmapHeight(NGAGE_TextureData *data)
{ {
if (data && data->bitmap) { if (data) {
return data->bitmap->SizeInPixels().iHeight; return data->cachedHeight;
} }
return 0; return 0;
} }
@@ -140,10 +139,8 @@ void NGAGE_SetClipRect(const SDL_Rect *rect)
void NGAGE_SetDrawColor(const Uint32 color) void NGAGE_SetDrawColor(const Uint32 color)
{ {
if (gRenderer) {
gRenderer->SetDrawColor(color); gRenderer->SetDrawColor(color);
} }
}
void NGAGE_PumpEventsInternal() void NGAGE_PumpEventsInternal()
{ {
@@ -177,7 +174,9 @@ CRenderer *CRenderer::NewL()
return self; return self;
} }
CRenderer::CRenderer() : iRenderer(0), iDirectScreen(0), iScreenGc(0), iWsSession(), iWsWindowGroup(), iWsWindowGroupID(0), iWsWindow(), iWsScreen(0), iWsEventStatus(), iWsEvent(), iShowFPS(EFalse), iFPS(0), iFont(0), iWorkBuffer1(0), iWorkBuffer2(0), iWorkBufferSize(0), iTempRenderBitmap(0), iTempRenderBitmapWidth(0), iTempRenderBitmapHeight(0), iLastColorR(-1), iLastColorG(-1), iLastColorB(-1), iLinePointsBuffer(0), iLinePointsBufferCapacity(0), iLastDrawColor(0xFFFFFFFF) {} CRenderer::CRenderer() : iRenderer(0), iDirectScreen(0), iScreenGc(0), iWsSession(), iWsWindowGroup(), iWsWindowGroupID(0), iWsWindow(), iWsScreen(0), iWsEventStatus(), iWsEvent(), iShowFPS(EFalse), iFPS(0), iFont(0), iWorkBuffer1(0), iWorkBuffer2(0), iWorkBufferSize(0), iTempRenderBitmap(0), iTempRenderBitmapWidth(0), iTempRenderBitmapHeight(0), iLastColorR(-1), iLastColorG(-1), iLastColorB(-1), iLinePointsBuffer(0), iLinePointsBufferCapacity(0), iLastDrawColor(0)
{
}
CRenderer::~CRenderer() CRenderer::~CRenderer()
{ {
@@ -505,13 +504,9 @@ Uint32 NGAGE_ConvertColor(float r, float g, float b, float a, float color_scale)
TFixed bf = Real2Fix(b); TFixed bf = Real2Fix(b);
TFixed af = Real2Fix(a); TFixed af = Real2Fix(a);
rf = FixMul(rf, scalef); rf = SDL_clamp(FixMul(rf, scalef), 0, ff);
gf = FixMul(gf, scalef); gf = SDL_clamp(FixMul(gf, scalef), 0, ff);
bf = FixMul(bf, scalef); bf = SDL_clamp(FixMul(bf, scalef), 0, ff);
rf = SDL_clamp(rf, 0, ff);
gf = SDL_clamp(gf, 0, ff);
bf = SDL_clamp(bf, 0, ff);
af = SDL_clamp(af, 0, ff); af = SDL_clamp(af, 0, ff);
rf = FixMul(rf, ff) >> 16; rf = FixMul(rf, ff) >> 16;
@@ -539,25 +534,18 @@ bool CRenderer::Copy(SDL_Renderer *renderer, SDL_Texture *texture, const SDL_Rec
SDL_FColor *c = &texture->color; SDL_FColor *c = &texture->color;
// Fast path 1: No transformations needed; direct BitBlt. // Get render scale once.
if (c->a == 1.f && c->r == 1.f && c->g == 1.f && c->b == 1.f) {
// Get render scale.
float sx; float sx;
float sy; float sy;
SDL_GetRenderScale(renderer, &sx, &sy); SDL_GetRenderScale(renderer, &sx, &sy);
if (sx == 1.f && sy == 1.f) { // Fast path 1: No transformations needed; direct BitBlt.
if (c->a == 1.f && c->r == 1.f && c->g == 1.f && c->b == 1.f && sx == 1.f && sy == 1.f) {
TRect aSource(TPoint(srcrect->x, srcrect->y), TSize(srcrect->w, srcrect->h)); TRect aSource(TPoint(srcrect->x, srcrect->y), TSize(srcrect->w, srcrect->h));
TPoint aDest(dstrect->x, dstrect->y); TPoint aDest(dstrect->x, dstrect->y);
iRenderer->Gc()->BitBlt(aDest, phdata->bitmap, aSource); iRenderer->Gc()->BitBlt(aDest, phdata->bitmap, aSource);
return true; return true;
} }
}
// Get render scale (moved here to avoid redundant call in fast path).
float sx;
float sy;
SDL_GetRenderScale(renderer, &sx, &sy);
// Slow path: Transformations needed. // Slow path: Transformations needed.
int w = phdata->cachedWidth; int w = phdata->cachedWidth;
@@ -577,7 +565,6 @@ bool CRenderer::Copy(SDL_Renderer *renderer, SDL_Texture *texture, const SDL_Rec
} }
dest = iWorkBuffer1; dest = iWorkBuffer1;
bool useBuffer1 = true;
if (c->a != 1.f || c->r != 1.f || c->g != 1.f || c->b != 1.f) { if (c->a != 1.f || c->r != 1.f || c->g != 1.f || c->b != 1.f) {
TFixed rf = Real2Fix(c->r); TFixed rf = Real2Fix(c->r);
@@ -591,7 +578,7 @@ bool CRenderer::Copy(SDL_Renderer *renderer, SDL_Texture *texture, const SDL_Rec
ApplyColorMod(dest, source, pitch, w, h, texture->color, iColorModLUT); ApplyColorMod(dest, source, pitch, w, h, texture->color, iColorModLUT);
source = dest; source = dest;
useBuffer1 = !useBuffer1; dest = (dest == iWorkBuffer1) ? iWorkBuffer2 : iWorkBuffer1;
} }
if (sx != 1.f || sy != 1.f) { if (sx != 1.f || sy != 1.f) {
@@ -600,10 +587,8 @@ bool CRenderer::Copy(SDL_Renderer *renderer, SDL_Texture *texture, const SDL_Rec
TFixed center_x = Int2Fix(w / 2); TFixed center_x = Int2Fix(w / 2);
TFixed center_y = Int2Fix(h / 2); TFixed center_y = Int2Fix(h / 2);
dest = useBuffer1 ? iWorkBuffer1 : iWorkBuffer2;
ApplyScale(dest, source, pitch, w, h, center_x, center_y, scale_x, scale_y); ApplyScale(dest, source, pitch, w, h, center_x, center_y, scale_x, scale_y);
source = dest; source = dest;
useBuffer1 = !useBuffer1;
} }
// Use temp bitmap to avoid destroying source texture. // Use temp bitmap to avoid destroying source texture.
@@ -637,14 +622,22 @@ bool CRenderer::CopyEx(SDL_Renderer *renderer, SDL_Texture *texture, const NGAGE
const bool isNoFlip = (!copydata->flip); const bool isNoFlip = (!copydata->flip);
const bool isNoColorMod = (c->a == 1.f && c->r == 1.f && c->g == 1.f && c->b == 1.f); const bool isNoColorMod = (c->a == 1.f && c->r == 1.f && c->g == 1.f && c->b == 1.f);
// Fast path 1: Check for cardinal rotation cache opportunity (0°, 90°, 180°, 270°). // Fast path 1: No transformations needed; direct BitBlt.
if (isNoFlip && isIdentityScale && isNoColorMod && !isNoRotation) { if (isNoFlip && isIdentityScale && isNoRotation && isNoColorMod) {
TInt angleIndex = -1; TRect aSource(TPoint(copydata->srcrect.x, copydata->srcrect.y), TSize(copydata->srcrect.w, copydata->srcrect.h));
TFixed angle = copydata->angle; TPoint aDest(copydata->dstrect.x, copydata->dstrect.y);
iRenderer->Gc()->BitBlt(aDest, phdata->bitmap, aSource);
return true;
}
// Use pre-calculated angle constants for comparison. // Fast path 2: Check for cardinal rotation cache opportunity (0°, 90°, 180°, 270°).
if (angle == kAngleZero) { if (isNoFlip && isIdentityScale && isNoColorMod && !isNoRotation) {
angleIndex = 0; TFixed angle = copydata->angle;
TInt angleIndex = -1;
// Check cardinal angles with tolerance - optimized for early exit.
if (SDL_abs(angle - kAngleZero) < kAngleTolerance) {
angleIndex = 0; // 0°
} else if (SDL_abs(angle - kAnglePi_2) < kAngleTolerance) { } else if (SDL_abs(angle - kAnglePi_2) < kAngleTolerance) {
angleIndex = 1; // 90° angleIndex = 1; // 90°
} else if (SDL_abs(angle - kAnglePi) < kAngleTolerance) { } else if (SDL_abs(angle - kAnglePi) < kAngleTolerance) {
@@ -666,14 +659,6 @@ bool CRenderer::CopyEx(SDL_Renderer *renderer, SDL_Texture *texture, const NGAGE
} }
} }
// Fast path 2: No transformations needed; direct BitBlt.
if (isNoFlip && isIdentityScale && isNoRotation && isNoColorMod) {
TRect aSource(TPoint(copydata->srcrect.x, copydata->srcrect.y), TSize(copydata->srcrect.w, copydata->srcrect.h));
TPoint aDest(copydata->dstrect.x, copydata->dstrect.y);
iRenderer->Gc()->BitBlt(aDest, phdata->bitmap, aSource);
return true;
}
// Slow path: Transformations needed. // Slow path: Transformations needed.
int w = phdata->cachedWidth; int w = phdata->cachedWidth;
int h = phdata->cachedHeight; int h = phdata->cachedHeight;
@@ -692,26 +677,23 @@ bool CRenderer::CopyEx(SDL_Renderer *renderer, SDL_Texture *texture, const NGAGE
} }
dest = iWorkBuffer1; dest = iWorkBuffer1;
bool useBuffer1 = true;
if (copydata->flip) { if (copydata->flip) {
ApplyFlip(dest, source, pitch, w, h, copydata->flip); ApplyFlip(dest, source, pitch, w, h, copydata->flip);
source = dest; source = dest;
useBuffer1 = !useBuffer1; dest = (dest == iWorkBuffer1) ? iWorkBuffer2 : iWorkBuffer1;
} }
if (!isIdentityScale) { if (!isIdentityScale) {
dest = useBuffer1 ? iWorkBuffer1 : iWorkBuffer2;
ApplyScale(dest, source, pitch, w, h, copydata->center.x, copydata->center.y, copydata->scale_x, copydata->scale_y); ApplyScale(dest, source, pitch, w, h, copydata->center.x, copydata->center.y, copydata->scale_x, copydata->scale_y);
source = dest; source = dest;
useBuffer1 = !useBuffer1; dest = (dest == iWorkBuffer1) ? iWorkBuffer2 : iWorkBuffer1;
} }
if (copydata->angle) { if (copydata->angle) {
dest = useBuffer1 ? iWorkBuffer1 : iWorkBuffer2;
ApplyRotation(dest, source, pitch, w, h, copydata->center.x, copydata->center.y, copydata->angle); ApplyRotation(dest, source, pitch, w, h, copydata->center.x, copydata->center.y, copydata->angle);
source = dest; source = dest;
useBuffer1 = !useBuffer1; dest = (dest == iWorkBuffer1) ? iWorkBuffer2 : iWorkBuffer1;
} }
if (!isNoColorMod) { if (!isNoColorMod) {
@@ -724,10 +706,8 @@ bool CRenderer::CopyEx(SDL_Renderer *renderer, SDL_Texture *texture, const NGAGE
BuildColorModLUT(rf, gf, bf); BuildColorModLUT(rf, gf, bf);
} }
dest = useBuffer1 ? iWorkBuffer1 : iWorkBuffer2;
ApplyColorMod(dest, source, pitch, w, h, texture->color, iColorModLUT); ApplyColorMod(dest, source, pitch, w, h, texture->color, iColorModLUT);
source = dest; source = dest;
useBuffer1 = !useBuffer1;
} }
// Use temp bitmap to avoid destroying source texture. // Use temp bitmap to avoid destroying source texture.
@@ -799,10 +779,12 @@ void CRenderer::DrawLines(NGAGE_Vertex *aVerts, const TInt aCount)
iLinePointsBuffer[i] = TPoint(aVerts[i].x, aVerts[i].y); iLinePointsBuffer[i] = TPoint(aVerts[i].x, aVerts[i].y);
} }
TUint32 aColor = (((TUint8)aVerts->color.a << 24) | // Pack color once - all vertices use the same color in polyline.
((TUint8)aVerts->color.b << 16) | Uint8 ca = aVerts->color.a;
((TUint8)aVerts->color.g << 8) | Uint8 cr = aVerts->color.r;
(TUint8)aVerts->color.r); Uint8 cg = aVerts->color.g;
Uint8 cb = aVerts->color.b;
TUint32 aColor = (ca << 24) | (cb << 16) | (cg << 8) | cr;
iRenderer->Gc()->SetPenColor(aColor); iRenderer->Gc()->SetPenColor(aColor);
iRenderer->Gc()->DrawPolyLineNoEndPoint(iLinePointsBuffer, aCount); iRenderer->Gc()->DrawPolyLineNoEndPoint(iLinePointsBuffer, aCount);
@@ -813,14 +795,15 @@ void CRenderer::DrawPoints(NGAGE_Vertex *aVerts, const TInt aCount)
{ {
if (iRenderer && iRenderer->Gc()) { if (iRenderer && iRenderer->Gc()) {
// Batch points by color to minimize SetPenColor calls. // Batch points by color to minimize SetPenColor calls.
TUint32 currentColor = 0xFFFFFFFF; // Invalid initial color TUint32 currentColor = 0;
bool colorSet = false; bool colorSet = false;
for (TInt i = 0; i < aCount; i++, aVerts++) { for (TInt i = 0; i < aCount; i++, aVerts++) {
TUint32 aColor = (((TUint8)aVerts->color.a << 24) | Uint8 ca = aVerts->color.a;
((TUint8)aVerts->color.b << 16) | Uint8 cr = aVerts->color.r;
((TUint8)aVerts->color.g << 8) | Uint8 cg = aVerts->color.g;
(TUint8)aVerts->color.r); Uint8 cb = aVerts->color.b;
TUint32 aColor = (ca << 24) | (cb << 16) | (cg << 8) | cr;
// Only set pen color when it changes. // Only set pen color when it changes.
if (!colorSet || aColor != currentColor) { if (!colorSet || aColor != currentColor) {
@@ -838,7 +821,7 @@ void CRenderer::FillRects(NGAGE_Vertex *aVerts, const TInt aCount)
{ {
if (iRenderer && iRenderer->Gc()) { if (iRenderer && iRenderer->Gc()) {
// Batch rectangles by color to minimize SetPenColor/SetBrushColor calls. // Batch rectangles by color to minimize SetPenColor/SetBrushColor calls.
TUint32 currentColor = 0xFFFFFFFF; // Invalid initial color TUint32 currentColor = 0;
bool colorSet = false; bool colorSet = false;
// Process rectangles (each rect uses 2 vertices: position and size). // Process rectangles (each rect uses 2 vertices: position and size).
@@ -847,10 +830,11 @@ void CRenderer::FillRects(NGAGE_Vertex *aVerts, const TInt aCount)
TSize size(aVerts[i + 1].x, aVerts[i + 1].y); TSize size(aVerts[i + 1].x, aVerts[i + 1].y);
TRect rect(pos, size); TRect rect(pos, size);
TUint32 aColor = (((TUint8)aVerts[i].color.a << 24) | Uint8 ca = aVerts[i].color.a;
((TUint8)aVerts[i].color.b << 16) | Uint8 cr = aVerts[i].color.r;
((TUint8)aVerts[i].color.g << 8) | Uint8 cg = aVerts[i].color.g;
(TUint8)aVerts[i].color.r); Uint8 cb = aVerts[i].color.b;
TUint32 aColor = (ca << 24) | (cb << 16) | (cg << 8) | cr;
// Only set colors when they change. // Only set colors when they change.
if (!colorSet || aColor != currentColor) { if (!colorSet || aColor != currentColor) {