Merge pull request #6815 from unknownbrackets/gpu-minor

Eat cycles during GE block transfers

GPU/Directx9/GPU_DX9.cpp

@@ -1478,6 +1478,9 @@ void DIRECTX9_GPU::DoBlockTransfer() {
 			dstStride == 512 && height == 272) {
 		framebufferManager_.DrawPixels(Memory::GetPointerUnchecked(dstBasePtr), GE_FORMAT_8888, 512);
 	}
+
+	// TODO: Correct timing appears to be 1.9, but erring a bit low since some of our other timing is inaccurate.
+	cyclesExecuted += ((height * width * bpp) * 16) / 10;
 }
 
 void DIRECTX9_GPU::InvalidateCache(u32 addr, int size, GPUInvalidationType type) {

GPU/GLES/GLES_GPU.cpp

@@ -2000,6 +2000,9 @@ void GLES_GPU::DoBlockTransfer() {
 	CBreakPoints::ExecMemCheck(srcBasePtr + (srcY * srcStride + srcX) * bpp, false, height * srcStride * bpp, currentMIPS->pc);
 	CBreakPoints::ExecMemCheck(dstBasePtr + (srcY * dstStride + srcX) * bpp, true, height * dstStride * bpp, currentMIPS->pc);
 #endif
+
+	// TODO: Correct timing appears to be 1.9, but erring a bit low since some of our other timing is inaccurate.
+	cyclesExecuted += ((height * width * bpp) * 16) / 10;
 }
 
 void GLES_GPU::InvalidateCache(u32 addr, int size, GPUInvalidationType type) {

GPU/GLES/SoftwareTransform.cpp

@@ -548,9 +548,9 @@ void TransformDrawEngine::SoftwareTransformAndDraw(
 	glBindBuffer(GL_ARRAY_BUFFER, 0);
 	glVertexAttribPointer(ATTR_POSITION, 4, GL_FLOAT, GL_FALSE, vertexSize, drawBuffer);
 	int attrMask = program->attrMask;
-	if (attrMask & (1 << ATTR_TEXCOORD)) glVertexAttribPointer(ATTR_TEXCOORD, doTextureProjection ? 3 : 2, GL_FLOAT, GL_FALSE, vertexSize, ((uint8_t*)drawBuffer) + 4 * 4);
-	if (attrMask & (1 << ATTR_COLOR0)) glVertexAttribPointer(ATTR_COLOR0, 4, GL_UNSIGNED_BYTE, GL_TRUE, vertexSize, ((uint8_t*)drawBuffer) + 7 * 4);
-	if (attrMask & (1 << ATTR_COLOR1)) glVertexAttribPointer(ATTR_COLOR1, 3, GL_UNSIGNED_BYTE, GL_TRUE, vertexSize, ((uint8_t*)drawBuffer) + 8 * 4);
+	if (attrMask & (1 << ATTR_TEXCOORD)) glVertexAttribPointer(ATTR_TEXCOORD, doTextureProjection ? 3 : 2, GL_FLOAT, GL_FALSE, vertexSize, ((uint8_t*)drawBuffer) + offsetof(TransformedVertex, u));
+	if (attrMask & (1 << ATTR_COLOR0)) glVertexAttribPointer(ATTR_COLOR0, 4, GL_UNSIGNED_BYTE, GL_TRUE, vertexSize, ((uint8_t*)drawBuffer) + offsetof(TransformedVertex, color0));
+	if (attrMask & (1 << ATTR_COLOR1)) glVertexAttribPointer(ATTR_COLOR1, 3, GL_UNSIGNED_BYTE, GL_TRUE, vertexSize, ((uint8_t*)drawBuffer) + offsetof(TransformedVertex, color1));
 	if (drawIndexed) {
 #if 1  // USING_GLES2
 		glDrawElements(glprim[prim], numTrans, GL_UNSIGNED_SHORT, inds);

GPU/Null/NullGpu.cpp

@@ -16,12 +16,14 @@
 // https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
 
 
-#include "NullGpu.h"
-#include "../GPUState.h"
-#include "../ge_constants.h"
-#include "../../Core/MemMap.h"
-#include "../../Core/HLE/sceKernelInterrupt.h"
-#include "../../Core/HLE/sceGe.h"
+#include "GPU/Null/NullGpu.h"
+#include "GPU/GPUState.h"
+#include "GPU/ge_constants.h"
+#include "Core/Debugger/Breakpoints.h"
+#include "Core/MemMap.h"
+#include "Core/MIPS/MIPS.h"
+#include "Core/HLE/sceKernelInterrupt.h"
+#include "Core/HLE/sceGe.h"
 
 NullGPU::NullGPU() { }
 NullGPU::~NullGPU() { }
@@ -306,10 +308,38 @@ void NullGPU::ExecuteOp(u32 op, u32 diff) {
 
 	case GE_CMD_TRANSFERSTART:
 		{
-			DEBUG_LOG(G3D, "DL Texture Transfer Start: PixFormat %i", data);
-			// TODO: Here we should check if the transfer overlaps a framebuffer or any textures,
-			// and take appropriate action. If not, this should just be a block transfer within
-			// GPU memory which could be implemented by a copy loop.
+			u32 srcBasePtr = gstate.getTransferSrcAddress();
+			u32 srcStride = gstate.getTransferSrcStride();
+
+			u32 dstBasePtr = gstate.getTransferDstAddress();
+			u32 dstStride = gstate.getTransferDstStride();
+
+			int srcX = gstate.getTransferSrcX();
+			int srcY = gstate.getTransferSrcY();
+
+			int dstX = gstate.getTransferDstX();
+			int dstY = gstate.getTransferDstY();
+
+			int width = gstate.getTransferWidth();
+			int height = gstate.getTransferHeight();
+
+			int bpp = gstate.getTransferBpp();
+
+			DEBUG_LOG(G3D, "Block transfer: %08x/%x -> %08x/%x, %ix%ix%i (%i,%i)->(%i,%i)", srcBasePtr, srcStride, dstBasePtr, dstStride, width, height, bpp, srcX, srcY, dstX, dstY);
+
+			for (int y = 0; y < height; y++) {
+				const u8 *src = Memory::GetPointer(srcBasePtr + ((y + srcY) * srcStride + srcX) * bpp);
+				u8 *dst = Memory::GetPointer(dstBasePtr + ((y + dstY) * dstStride + dstX) * bpp);
+				memcpy(dst, src, width * bpp);
+			}
+
+#ifndef MOBILE_DEVICE
+			CBreakPoints::ExecMemCheck(srcBasePtr + (srcY * srcStride + srcX) * bpp, false, height * srcStride * bpp, currentMIPS->pc);
+			CBreakPoints::ExecMemCheck(dstBasePtr + (srcY * dstStride + srcX) * bpp, true, height * dstStride * bpp, currentMIPS->pc);
+#endif
+
+			// TODO: Correct timing appears to be 1.9, but erring a bit low since some of our other timing is inaccurate.
+			cyclesExecuted += ((height * width * bpp) * 16) / 10;
 			break;
 		}
 

GPU/Software/SoftGpu.cpp

@@ -626,6 +626,9 @@ void SoftGPU::ExecuteOp(u32 op, u32 diff)
 			CBreakPoints::ExecMemCheck(dstBasePtr + (srcY * dstStride + srcX) * bpp, true, height * dstStride * bpp, currentMIPS->pc);
 #endif
 
+			// TODO: Correct timing appears to be 1.9, but erring a bit low since some of our other timing is inaccurate.
+			cyclesExecuted += ((height * width * bpp) * 16) / 10;
+
 			// Could theoretically dirty the framebuffer.
 			framebufferDirty_ = true;
 			break;

UI/GameInfoCache.cpp

@@ -609,7 +609,7 @@ again:
 void GameInfoCache::SetupTexture(GameInfo *info, std::string &textureData, Thin3DContext *thin3d, Thin3DTexture *&tex, double &loadTime) {
 	if (textureData.size()) {
 		if (!tex) {
-			tex = thin3d->CreateTextureFromFileData(textureData.data(), textureData.size(), T3DFileType::PNG);
+			tex = thin3d->CreateTextureFromFileData(textureData.data(), (int)textureData.size(), T3DFileType::PNG);
 			if (tex) {
 				loadTime = time_now_d();
 			} else {