You've already forked UnrealEngineUWP
mirror of
https://github.com/izzy2lost/UnrealEngineUWP.git
synced 2026-03-26 18:15:20 -07:00
e4bae00099
Requires two additional compute passes post-culling to compact the draw command instances. This feature is not currently supported on mobile platforms; more work is required to make instance compaction work with mobile vertex buffers. Updated ISM to use order preservation for meshes that render with a translucent material #rb ola.olsson, dmitriy.dyomin #jira none #preflight 61e98458c92021e535994cd7 #ROBOMERGE-AUTHOR: jamie.hayes #ROBOMERGE-SOURCE: CL 18676076 in //UE5/Release-5.0/... via CL 18676089 via CL 18676097 #ROBOMERGE-BOT: UE5 (Release-Engine-Test -> Main) (v902-18672795) [CL 18679861 by jamie hayes in ue5-main branch]
67 lines
2.6 KiB
Plaintext
67 lines
2.6 KiB
Plaintext
// Copyright Epic Games, Inc. All Rights Reserved.
|
|
|
|
#pragma once
|
|
|
|
/**
|
|
* ThreadGroupPrefixSum(Value, GroupThreadIndex [, GroupSumOut])
|
|
*
|
|
* Calculates the prefix sum (and optionally group sum) of a given value across threads of a thread group.
|
|
*
|
|
* EXAMPLE USES:
|
|
* float GroupSum;
|
|
* float PrefixSum = ThreadGroupPrefixSum(ValueFloat, GroupThreadIndex, GroupSum);
|
|
*
|
|
* int PrefixSum = ThreadGroupPrefixSum(ValueInt, GroupThreadIndex);
|
|
*
|
|
* NOTES:
|
|
* - PrefixSum means the sum of the value for all threads whose group thread index is LESS than the current
|
|
* - Only scalar types are currently supported
|
|
* - All threads in the group must be active when calling this method; it cannot be used in a branch (or after an early return)
|
|
* that doesn't include the entire group.
|
|
*/
|
|
|
|
#ifndef NUM_THREADS_PER_GROUP
|
|
#error NUM_THREADS_PER_GROUP must be defined, and must be equal to the thread group size of the caller of ThreadGroupPrefixSum
|
|
#endif
|
|
|
|
// re-use the same groupshared memory, in case the caller utilizes multiple overloads
|
|
groupshared uint ThreadGroupPrefixSumWorkspace[2][NUM_THREADS_PER_GROUP];
|
|
|
|
#define DECLARE_THREAD_GROUP_PREFIX_SUM(ValType, CastToUint, CastFromUint) \
|
|
ValType ThreadGroupPrefixSum(ValType Value, uint GroupThreadIndex, inout ValType GroupSum) \
|
|
{ \
|
|
uint Curr = 0; \
|
|
ThreadGroupPrefixSumWorkspace[Curr][GroupThreadIndex] = CastToUint(Value); \
|
|
GroupMemoryBarrierWithGroupSync(); \
|
|
for (uint i = 1U; i <= (NUM_THREADS_PER_GROUP / 2U); i *= 2U) \
|
|
{ \
|
|
const uint Next = 1U - Curr; \
|
|
if (GroupThreadIndex < i) \
|
|
{ \
|
|
ThreadGroupPrefixSumWorkspace[Next][GroupThreadIndex] = \
|
|
ThreadGroupPrefixSumWorkspace[Curr][GroupThreadIndex]; \
|
|
} \
|
|
else \
|
|
{ \
|
|
ThreadGroupPrefixSumWorkspace[Next][GroupThreadIndex] = CastToUint( \
|
|
CastFromUint(ThreadGroupPrefixSumWorkspace[Curr][GroupThreadIndex]) + \
|
|
CastFromUint(ThreadGroupPrefixSumWorkspace[Curr][GroupThreadIndex - i]) \
|
|
); \
|
|
} \
|
|
Curr = Next; \
|
|
GroupMemoryBarrierWithGroupSync(); \
|
|
} \
|
|
GroupSum = CastFromUint(ThreadGroupPrefixSumWorkspace[Curr][NUM_THREADS_PER_GROUP - 1]); \
|
|
return (GroupThreadIndex == 0U) ? CastFromUint(0) : CastFromUint(ThreadGroupPrefixSumWorkspace[Curr][GroupThreadIndex - 1]); \
|
|
} \
|
|
ValType ThreadGroupPrefixSum(ValType Value, uint GroupThreadIndex) \
|
|
{ \
|
|
ValType GroupSum; \
|
|
return ThreadGroupPrefixSum(Value, GroupThreadIndex, GroupSum); \
|
|
}
|
|
|
|
DECLARE_THREAD_GROUP_PREFIX_SUM(float, asuint, asfloat)
|
|
DECLARE_THREAD_GROUP_PREFIX_SUM(int, uint, int)
|
|
DECLARE_THREAD_GROUP_PREFIX_SUM(uint, uint, uint)
|
|
|
|
#undef DECLARE_THREAD_GROUP_PREFIX_SUM |