if (SORTING)
	{
		device->EventBegin("SortEmittedParticles", threadID);

		// initialize sorting arguments:
		device->BindCS(kickoffSortCS, threadID);
		device->Dispatch(1, 1, 1, threadID);

		// initial sorting:
		bool bDone = true;

		// calculate how many threads we'll require:
		//   we'll sort 512 elements per CU (threadgroupsize 256)
		//     maybe need to optimize this or make it changeable during init
		//     TGS=256 is a good intermediate value

		unsigned int numThreadGroups = ((MAX_PARTICLES - 1) >> 9) + 1;

		if (numThreadGroups>1) bDone = false;

		// sort all buffers of size 512 (and presort bigger ones)
		device->BindCS(sortCS, threadID);
		device->DispatchIndirect(indirectBuffers, ARGUMENTBUFFER_OFFSET_DISPATCHSORT, threadID);

		int presorted = 512;
		while (!bDone)
		{
			bDone = true;
			device->BindCS(sortStepCS, threadID);

			// prepare thread group description data
			unsigned int numThreadGroups = 0;

			if (MAX_PARTICLES > (uint32_t)presorted)
			{
				if (MAX_PARTICLES>(uint32_t)presorted * 2)
					bDone = false;

				unsigned int pow2 = presorted;
				while (pow2<MAX_PARTICLES)
					pow2 *= 2;
				numThreadGroups = pow2 >> 9;
			}

			unsigned int nMergeSize = presorted * 2;
			for (unsigned int nMergeSubSize = nMergeSize >> 1; nMergeSubSize>256; nMergeSubSize = nMergeSubSize >> 1)
			//	for( int nMergeSubSize=nMergeSize>>1; nMergeSubSize>0; nMergeSubSize=nMergeSubSize>>1 ) 
			{
				SortConstants sc;
				sc.job_params.x = nMergeSubSize;
				if (nMergeSubSize == nMergeSize >> 1)
				{
					sc.job_params.y = (2 * nMergeSubSize - 1);
					sc.job_params.z = -1;
				}
				else
				{
					sc.job_params.y = nMergeSubSize;
					sc.job_params.z = 1;
				}
				sc.job_params.w = 0;

				device->UpdateBuffer(sortCB, &sc, threadID);
				
				device->Dispatch(numThreadGroups, 1, 1, threadID);
			}

			device->BindCS(sortInnerCS, threadID);
			device->Dispatch(numThreadGroups, 1, 1, threadID);


			presorted *= 2;
		}