No mipmap support for Dynamic Texture 2D

I’m trying to set material onto object at run-time from texture, normal map, displacement map etc. located in the hard disk. To achieve this i need to create dynamic material instance and set the parameter to the material by loading the image file at run-time.

I’m creating material from texture & displacement map image file. These image files are loading from folder using Victory Plugin’s “Victory Load Texture 2D from File” Node.

I’m creating the material instance dynamic of the “test material”. Now I’m setting the texture and displacement parameter as shown in figure 4.
After creating material instance dynamic, I’m setting this material to the object.

figure[1] test material:

figure[2] texture:

figure[3] displacement map:

figure[4] Method:

Now here is the problem.

When I directly create material in material editor from texture and displacement map and set it onto object. It doesn’t generate** moiré pattern.**

But when I run-time set the material onto object using figure [4]. It will generate moiré pattern.

Here are the sample screen shots of both.

figure[5]: Material Editor result:

figure[6]: Runtime result: Imgur: The magic of the Internet

I think the “Load Texture 2D from File” nodes is not calculating the Mip Map correctly.

I tried with following nodes but the result is same.
1> “Victory Load Texture 2D from File” (Created by rama)
2> “Load Texture 2D from File by Extension” (Created by you)
3> “Victory Load Texture 2D from File Pixels” (Created by rama)

For MipMap Calculation I modified the plugin file VictoryBPFunctionLibrary.cpp.

Here is the function that I’m using without any success.

UTexture2D* UVictoryBPFunctionLibrary::Victory_LoadTexture2D_FromFile(const FString& FullFilePath,EJoyImageFormats ImageFormat, bool& IsValid,int32& Width, int32& Height)
{

IsValid = false;
UTexture2D* LoadedT2D = NULL;

IImageWrapperModule& ImageWrapperModule = FModuleManager::LoadModuleChecked<IImageWrapperModule>(FName("ImageWrapper"));
IImageWrapperPtr ImageWrapper = ImageWrapperModule.CreateImageWrapper(GetJoyImageFormat(ImageFormat));

//Load From File
TArray<uint8> RawFileData;
if (!FFileHelper::LoadFileToArray(RawFileData, *FullFilePath)) return NULL;
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  
//Create T2D!
if (ImageWrapper.IsValid() && ImageWrapper->SetCompressed(RawFileData.GetData(), RawFileData.Num()))
{ 
	const TArray<uint8>* UncompressedBGRA = NULL;
	if (ImageWrapper->GetRaw(ERGBFormat::BGRA, 8, UncompressedBGRA))
	{
		LoadedT2D = UTexture2D::CreateTransient(ImageWrapper->GetWidth(), ImageWrapper->GetHeight(), PF_B8G8R8A8);
		LoadedT2D->MipGenSettings=TMGS_FromTextureGroup;
		
		//Valid?
		if(!LoadedT2D) return NULL;
		//~~~~~~~~~~~~~~
		
		//Out!
		Width = ImageWrapper->GetWidth();
		Height = ImageWrapper->GetHeight();
		 
		//Copy!
		void* TextureData = LoadedT2D->PlatformData->Mips[0].BulkData.Lock(LOCK_READ_WRITE);
		FMemory::Memcpy(TextureData, UncompressedBGRA->GetData(), UncompressedBGRA->Num());
		LoadedT2D->PlatformData->Mips[0].BulkData.Unlock();

		//Update!
		LoadedT2D->UpdateResource();
	}
}
 
// Success!
IsValid = true;
return LoadedT2D;

}

I also tried with

LoadedT2D->MipGenSettings=TMGS_Blur1;

to

LoadedT2D->MipGenSettings=TMGS_Blur5;

LoadedT2D->MipGenSettings=TMGS_Sharpen0;

to

LoadedT2D->MipGenSettings=TMGS_Sharpen10;

LoadedT2D->MipGenSettings=TMGS_SimpleAverage;

After so many attempts result is same.

While playing the game when event occur, it closes the editor without any prompt or crash report.

I think engine is not able to set the mipmap at runtime.

Hi ,
Can you reproduce this without the Victory Plugin or nodes from this plugin? If so I can look into this further. We do not support other users plugin’s and bugs that may be generated because of those plugins. MipMaps cannot be set at runtime.

If you can reproduce please with the base editor please include the repro steps here, If not this question would likely be better handled in the C++ section where the community can offer assistance.

Thank you!

Tim

Hi Tim,

Thank you for the reply. Actually my project is blueprint project. I’m novice in workflow of UE4 for c++ programming. So I have moved the question to C++ section. I hope someone can help here.

you have mentioned that “MipMaps cannot be set at runtime”. Do you mean that it can not be possible to generate MipMap using C++ or Blueprint Programming ?

Regards,

The texture mipmap settings cannot be adjusted at runtime. These are non-accessible for BP or C++ out of the box. However, I’m sure if you have some C++ savvy that can be adjusted, but it’s not an area I’m familiar with in regards to programming.

Its possible to set mips at runtime but when I looked last year it wasn’t possible to use the engine mip generation code as this was all private and not available during texture loading. Perhaps this has changed?

Even a simple averaged pixel algorithm will give you better results than no mips though - I’ve pasted one below. Note that you’ll want square textures.

auto workFunc = [texPtr, bThreaded]()
	{
		if (!texPtr.IsValid())
			return;

		auto *pNewTexture = texPtr.Get();

		int mipsToAdd = 3;// pNewTexture->RequestedMips - 1;

						  //Declaring buffers here to reduce reallocs
						  //We double buffer mips, using the prior buffer to build the next buffer
		TArray<uint8> _mipRGBAs;
		TArray<uint8> _mipRGBBs;

		//Access source data
		auto* priorData = (const uint8 *)pNewTexture->PlatformData->Mips[0].BulkData.Lock(LOCK_READ_WRITE);
		int priorwidth = pNewTexture->PlatformData->Mips[0].SizeX;
		int priorheight = pNewTexture->PlatformData->Mips[0].SizeY;

		while (mipsToAdd > 0)
		{
			auto *mipRGBAs = mipsToAdd & 1 ? &_mipRGBAs : &_mipRGBBs;

			int mipwidth = priorwidth >> 1;
			int mipheight = priorheight >> 1;
			if ((mipwidth == 0) || (mipheight == 0))
			{
				break;
			}
			//++pNewTexture->RequestedMips;

			mipRGBAs->Reset();
			mipRGBAs->AddUninitialized(mipwidth * mipheight * BYTES_PER_PIXEL);

			int dataPerRow = priorwidth * BYTES_PER_PIXEL;

			//Average out the values
			auto *dataOut = mipRGBAs->GetData();
			for (int y = 0; y < mipheight; y++)
			{
				auto *dataInRow0 = priorData + (dataPerRow * y * 2);
				auto *dataInRow1 = dataInRow0 + dataPerRow;
				for (int x = 0; x < mipwidth; x++)
				{
					int totalB = *dataInRow0++;
					int totalG = *dataInRow0++;
					int totalR = *dataInRow0++;
					int totalA = *dataInRow0++;
					totalB += *dataInRow0++;
					totalG += *dataInRow0++;
					totalR += *dataInRow0++;
					totalA += *dataInRow0++;

					totalB += *dataInRow1++;
					totalG += *dataInRow1++;
					totalR += *dataInRow1++;
					totalA += *dataInRow1++;
					totalB += *dataInRow1++;
					totalG += *dataInRow1++;
					totalR += *dataInRow1++;
					totalA += *dataInRow1++;

					totalB >>= 2;
					totalG >>= 2;
					totalR >>= 2;
					totalA >>= 2;

					*dataOut++ = (uint8)totalB;
					*dataOut++ = (uint8)totalG;
					*dataOut++ = (uint8)totalR;
					*dataOut++ = (uint8)totalA;
				}
				dataInRow0 += priorwidth * 2;
				dataInRow1 += priorwidth * 2;
			}

			// Allocate next mipmap.
			FTexture2DMipMap* Mip = new(pNewTexture->PlatformData->Mips) FTexture2DMipMap();
			Mip->SizeX = mipwidth;
			Mip->SizeY = mipheight;
			Mip->BulkData.Lock(LOCK_READ_WRITE);
			void* mipData = Mip->BulkData.Realloc(mipRGBAs->Num());
			FMemory::Memcpy(mipData, mipRGBAs->GetData(), mipRGBAs->Num());
			Mip->BulkData.Unlock();

			priorData = mipRGBAs->GetData();
			priorwidth = mipwidth;
			priorheight = mipheight;
			--mipsToAdd;
		}

		pNewTexture->PlatformData->Mips[0].BulkData.Unlock();
		if (bThreaded)
		{
			AppData::CallOnMainThread([texPtr] {
				if (texPtr.IsValid())
				{
					texPtr->UpdateResource();
				}
			});
		}
		else
		{
			pNewTexture->UpdateResource();
		}
	};
3 Likes

(CallOnMainThread is custom in my codebase, if you are not threading anything you don’t need to worry about that its just that you need to call UpdateResource from the main thread and I shift my mip generation onto a background worker thread).

Hi , have you solved your problem?

That’s what the engine code does - my assumption is that there is some sort of custom new handling but I didn’t step into it. It’s a good question though.

Where you call this for each mip:

FTexture2DMipMap* Mip = new(pNewTexture->PlatformData->Mips) FTexture2DMipMap();

Aren’t you erasing the prior mip with each placement new in the loop? Shouldn’t it be indexing into PlatformData->Mips? I’m not quite sure how TIndirectArray works, so maybe something else is changing it during the loop.

ah thanks I’ll check into that, TIndirectArray does seem to overload placement new

FWIW I was able to resolve a similar issue by using UCanvasRenderTarget2D’s “bAutoGenerateMips = true”

Unfortunately this only seems to work on some platforms - Mac and Android are a no-go. Looks like I will have to stick to CPU-side mip generation - what a waste of resources!