Merge pull request #13466 from hrydgard/misc-optimization

Misc minor optimizations

Core/MemMap.h

@@ -261,6 +261,10 @@ inline bool IsVRAMAddress(const u32 address) {
 inline bool IsDepthTexVRAMAddress(const u32 address) {
 	return ((address & 0x3FE00000) == 0x04200000) || ((address & 0x3FE00000) == 0x04600000);
 }
+inline bool IsKernelAddress(const u32 address) {
+	return ((address & 0x3F800000) == 0x08000000);
+}
+
 bool IsScratchpadAddress(const u32 address);
 
 // Used for auto-converted char * parameters, which can sometimes legitimately be null -

GPU/Common/TextureCacheCommon.cpp

@@ -506,7 +506,10 @@ TexCacheEntry *TextureCacheCommon::SetTexture() {
 		}
 
 		entry = entryNew;
-		if (g_Config.bTextureBackoffCache) {
+		if (Memory::IsKernelAddress(texaddr)) {
+			// It's the builtin font texture.
+			entry->status = TexCacheEntry::STATUS_RELIABLE;
+		} else if (g_Config.bTextureBackoffCache) {
 			entry->status = TexCacheEntry::STATUS_HASHING;
 		} else {
 			entry->status = TexCacheEntry::STATUS_UNRELIABLE;
@@ -568,10 +571,9 @@ std::vector<AttachCandidate> TextureCacheCommon::GetFramebufferCandidates(const
 
 	FramebufferNotificationChannel channel = Memory::IsDepthTexVRAMAddress(entry.addr) ? FramebufferNotificationChannel::NOTIFY_FB_DEPTH : FramebufferNotificationChannel::NOTIFY_FB_COLOR;
 
-	auto framebuffers = framebufferManager_->Framebuffers();
+	const std::vector<VirtualFramebuffer *> &framebuffers = framebufferManager_->Framebuffers();
 
-	for (size_t i = 0, n = framebuffers.size(); i < n; ++i) {
-		auto framebuffer = framebuffers[i];
+	for (VirtualFramebuffer *framebuffer : framebuffers) {
 		FramebufferMatchInfo match = MatchFramebuffer(entry, framebuffer, texAddrOffset, channel);
 		switch (match.match) {
 		case FramebufferMatch::VALID:
@@ -707,7 +709,8 @@ void TextureCacheCommon::HandleTextureChange(TexCacheEntry *const entry, const c
 		ReleaseTexture(entry, true);
 		entry->status &= ~TexCacheEntry::STATUS_IS_SCALED;
 	}
-	// Clear the reliable bit if set.
+
+	// Mark as hashing, if marked as reliable.
 	if (entry->GetHashStatus() == TexCacheEntry::STATUS_RELIABLE) {
 		entry->SetHashStatus(TexCacheEntry::STATUS_HASHING);
 	}
@@ -723,7 +726,6 @@ void TextureCacheCommon::HandleTextureChange(TexCacheEntry *const entry, const c
 		}
 	}
 
-	entry->status |= TexCacheEntry::STATUS_UNRELIABLE;
 	if (entry->numFrames < TEXCACHE_FRAME_CHANGE_FREQUENT) {
 		if (entry->status & TexCacheEntry::STATUS_FREE_CHANGE) {
 			entry->status &= ~TexCacheEntry::STATUS_FREE_CHANGE;
@@ -1073,9 +1075,8 @@ void TextureCacheCommon::LoadClut(u32 clutAddr, u32 loadBytes) {
 			static const u32 MAX_CLUT_OFFSET = 4096;
 
 			clutRenderOffset_ = MAX_CLUT_OFFSET;
-			auto framebuffers = framebufferManager_->Framebuffers();
-			for (size_t i = 0, n = framebuffers.size(); i < n; ++i) {
-				auto framebuffer = framebuffers[i];
+			const std::vector<VirtualFramebuffer *> &framebuffers = framebufferManager_->Framebuffers();
+			for (VirtualFramebuffer *framebuffer : framebuffers) {
 				const u32 fb_address = framebuffer->fb_address & 0x3FFFFFFF;
 				const u32 bpp = framebuffer->drawnFormat == GE_FORMAT_8888 ? 4 : 2;
 				u32 offset = clutFramebufAddr - fb_address;

GPU/GPUCommon.cpp

@@ -1231,7 +1231,7 @@ void GPUCommon::Execute_Origin(u32 op, u32 diff) {
 void GPUCommon::Execute_Jump(u32 op, u32 diff) {
 	const u32 target = gstate_c.getRelativeAddress(op & 0x00FFFFFC);
 	if (!Memory::IsValidAddress(target)) {
-		ERROR_LOG_REPORT(G3D, "JUMP to illegal address %08x - ignoring! data=%06x", target, op & 0x00FFFFFF);
+		ERROR_LOG(G3D, "JUMP to illegal address %08x - ignoring! data=%06x", target, op & 0x00FFFFFF);
 		UpdateState(GPUSTATE_ERROR);
 		return;
 	}
@@ -1253,7 +1253,7 @@ void GPUCommon::Execute_BJump(u32 op, u32 diff) {
 			UpdatePC(currentList->pc, target - 4);
 			currentList->pc = target - 4; // pc will be increased after we return, counteract that
 		} else {
-			ERROR_LOG_REPORT(G3D, "BJUMP to illegal address %08x - ignoring! data=%06x", target, op & 0x00FFFFFF);
+			ERROR_LOG(G3D, "BJUMP to illegal address %08x - ignoring! data=%06x", target, op & 0x00FFFFFF);
 			UpdateState(GPUSTATE_ERROR);
 		}
 	}
@@ -1264,7 +1264,7 @@ void GPUCommon::Execute_Call(u32 op, u32 diff) {
 
 	const u32 target = gstate_c.getRelativeAddress(op & 0x00FFFFFC);
 	if (!Memory::IsValidAddress(target)) {
-		ERROR_LOG_REPORT(G3D, "CALL to illegal address %08x - ignoring! data=%06x", target, op & 0x00FFFFFF);
+		ERROR_LOG(G3D, "CALL to illegal address %08x - ignoring! data=%06x", target, op & 0x00FFFFFF);
 		UpdateState(GPUSTATE_ERROR);
 		return;
 	}
@@ -1297,7 +1297,7 @@ void GPUCommon::DoExecuteCall(u32 target) {
 	}
 
 	if (currentList->stackptr == ARRAY_SIZE(currentList->stack)) {
-		ERROR_LOG_REPORT(G3D, "CALL: Stack full!");
+		ERROR_LOG(G3D, "CALL: Stack full!");
 	} else {
 		auto &stackEntry = currentList->stack[currentList->stackptr++];
 		stackEntry.pc = retval;
@@ -1310,7 +1310,7 @@ void GPUCommon::DoExecuteCall(u32 target) {
 
 void GPUCommon::Execute_Ret(u32 op, u32 diff) {
 	if (currentList->stackptr == 0) {
-		DEBUG_LOG_REPORT(G3D, "RET: Stack empty!");
+		DEBUG_LOG(G3D, "RET: Stack empty!");
 	} else {
 		auto &stackEntry = currentList->stack[--currentList->stackptr];
 		gstate_c.offsetAddr = stackEntry.offsetAddr;

ext/native/util/text/wrap_text.cpp

@@ -130,8 +130,8 @@ void WordWrapper::AppendWord(int endIndex, bool addNewline) {
 		scanForNewline_ = false;
 	} else {
 		// We may have appended a newline - check.
-		size_t pos = out_.substr(lastLineStart_).find_last_of("\n");
-		if (pos != out_.npos) {
+		size_t pos = out_.find_last_of("\n");
+		if (pos != out_.npos && pos >= lastLineStart_) {
 			lastLineStart_ += pos;
 		}
 	}
@@ -143,6 +143,10 @@ void WordWrapper::Wrap() {
 
 	// First, let's check if it fits as-is.
 	size_t len = strlen(str_);
+
+	// We know it'll be approximately this size. It's fine if the guess is a little off.
+	out_.reserve(len + len / 16);
+
 	if (MeasureWidth(str_, len) <= maxW_) {
 		// If it fits, we don't need to go through each character.
 		out_ = str_;