This segment apparently allocates data and then deletes it in another thread. Which means that if your tool detects leak within current function only, it willl report a leak.
The code overall seems to overuse new/delete (it looks like there’s no reason to create frame/stream/etc via “new”), but I don’t see an obvious glitch there. Might be overlooking something, of course.
the memory leak its obvious, my application crash after 60 seconds running out of memory. Not use any special tool, i can see in the W10 task manager how the “Committed” go from 6gb to the limit 25GB before crash.
in the AH post there are a file with the memory profiling.
hi knack, how exacly you pass the data? opencv automanage some things like memory clean! if your project work as you say, maybe you need clean the buffer by yourself
try something like this:
uint8* data = "whatever you get";
updateyourTextureRHI(data);//this always copy the memory and pass to the render thread, but your old data keep exist
delete data;//so clean by hand
this is when you create, process the data inside a method, i am very sure something like that is happening to you!
you need be responsable some times with some buffer source, for example Mat from opencv clean the data after finish the method, or something similar with kinect color buffer
if your Tarray<FColor> Data is part of your class, i suggest intialize in the beginPlay method with know width and height, and never use Empty and Init inside your update method, maybe this raise your memory leak, anyway your code looks ok for me, try this too.
maybe your problem is not code maybe is hardware, maybe some old webcam with very slow performance, more and less 15 fps, check your code to match with the camera fps
try something like this:
const int32 frameRate = 15; //15 fps
float camTime;
void update(float deltaTime)
{
camTime += deltaTime;
if(camTime >= 1/frameRate)
{
//pass and reset time
camTime = 0;
//try to get your data
uint8* data = getData();
//always check for errors
if(data != NULL)
{
//update your texture
}
}
}
I can’t init on begin play because i need check the bitmap if its the same x/y and if not resize. (ins’t a camera).
i slow down the refresh rate a lot with the same problems. I check a lot of things hehe
and again i remove all things about updatetextureregions, and still receive the bitmaps and put it on Data without any memory leak.
That mean if i remove that line
UpdateTextureRegions(VideoTexture, (int32)0, (uint32)3, VideoUpdateTextureRegion, (uint32)(4 * VideoSize.X), (uint32)4, (uint8*)Data.GetData(), false);
now for correct implementation lets avoid use Tarray instead we use pure unsigned char*
this is the Tick:
void YourClass::Tick(float DeltaTime)
{
unsigned char* data = getData();//hypotetic get data
if (data == NULL) return;
//your new data in BGRA for your texture
unsigned char* bgra = new unsigned char[width * height *4];
//now you fix the order of your data to whatever vals of your data to BGRA texture type!!!!
for (int i = 0, i < width * height; i ++) {
bgra[4 * i + 0] = data[4 * i + 0];
bgra[4 * i + 1] = data[4 * i + 1];
bgra[4 * i + 2] = data[4 * i + 2];
bgra[4 * i + 3] = 255;
}
//now pass to your region data
// fill my Texture Region data
FUpdateTextureRegionsData* RegionData = new FUpdateTextureRegionsData;
RegionData->Texture2DResource = (FTexture2DResource*)texture->Resource;
RegionData->MipIndex = 0;
RegionData->NumRegions = 1;
RegionData->Regions = echoUpdateTextureRegion;
RegionData->SrcPitch = (uint32)(4 * width);
RegionData->SrcBpp = 4;
RegionData->SrcData = bgra;
//and finally update the texture
bool bFreeData = false;
// call the RHIUpdateTexture2D to refresh the texture with new info
ENQUEUE_UNIQUE_RENDER_COMMAND_TWOPARAMETER(
UpdateTextureRegionsData,
FUpdateTextureRegionsData*, RegionData, RegionData,
bool, bFreeData, bFreeData,
{
for (uint32 RegionIndex = 0; RegionIndex < RegionData->NumRegions; ++RegionIndex)
{
int32 CurrentFirstMip = RegionData->Texture2DResource->GetCurrentFirstMip();
if (RegionData->MipIndex >= CurrentFirstMip)
{
RHIUpdateTexture2D(
RegionData->Texture2DResource->GetTexture2DRHI(),
RegionData->MipIndex - CurrentFirstMip,
RegionData->Regions[RegionIndex],
RegionData->SrcPitch,
RegionData->SrcData
+ RegionData->Regions[RegionIndex].SrcY * RegionData->SrcPitch
+ RegionData->Regions[RegionIndex].SrcX * RegionData->SrcBpp
);
}
}
if (bFreeData)
{
FMemory::Free(RegionData->Regions);
FMemory::Free(RegionData->SrcData);
}
delete RegionData;
});
//one last thing is delete this bgra unsigned char* to avoid the leak
delete] bgra;
}
this code is functional, and must be solve your leak, at least if you dont perform in another code portion
mmm you told me about resize images and thats way you cant init your Tarray<FColor> in begin play, if you need check this every tick i afraid this approach never works, if that is the case, you need receive your data, and then resize to match with the texture size, otherway never works
I found the leak, nothing with the example code finally.
The first problem was i check my code alone and seems work fine without leak and then i assume the problem was in the example code that i c&p, especially in the update function that i not understand comptely what do and when i remove the UpdateTexture seems stop the leak. Then i focus on that. seem normal no?
Well that was another error, checking second time i see the leak continue, and finally begin again checking the code and the leak was in my “build image” code and the worst its was evident. I don’t known why first time check seems i haven’t any leak but i lost a lot of hours and your time for a stupid fault.
Thanks to all!
y Gracias ZkarmaKun y perdona las molestias, tu codigo me puede valer si decido quitar FColor que lo pense tambien.
I am using the example code , but not work. the same as you said ,“The only differences its the source of the Data, but this ins’t the problem if disable the updatetextureregions no leaks.”
