Skip to content

Commit

Permalink
AGS 6.0
Browse files Browse the repository at this point in the history
  • Loading branch information
@rys
rys committed Nov 3, 2020
1 parent 4c48321 commit 4f2442c
Show file tree
Hide file tree
Showing 38 changed files with 513 additions and 234 deletions.
3 changes: 3 additions & 0 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,9 @@ In addition to the library itself, the AGS SDK includes several samples to demon
<a href="https://github.com/GPUOpen-LibrariesAndSDKs/AGS_SDK/releases/latest/"><img src="http://gpuopen-librariesandsdks.github.io/media/latest-release-button.svg" alt="Latest release" title="Latest release"></a>
</div>

### What's new in AGS 6.0
Version 6.0 introduces several new shader intrinsics, namely a DX12 ray tracing hit token for RDNA2 hardware for ray tracing optimisation, ReadLaneAt and explicit float conversions. There is also a change to the initialization API to make sure the AGS dll matches the header and calling code.

### What's new in AGS 5.4.2
Version 5.4.2 reinstates the sharedMemoryInBytes field which is required when calculating the memory available on APUs.

Expand Down
Binary file modified ags_lib/doc/amd_ags.chm
View file
Binary file not shown.
234 changes: 234 additions & 0 deletions ags_lib/hlsl/ags_shader_intrinsics_dx12.hlsl
View file
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,8 @@ RWByteAddressBuffer AmdExtD3DShaderIntrinsicsUAV : register(u0, AmdExtD3DShaderI
#define AmdExtD3DShaderIntrinsicsOpcode_GetWaveSize 0x19
#define AmdExtD3DShaderIntrinsicsOpcode_BaseInstance 0x1a
#define AmdExtD3DShaderIntrinsicsOpcode_BaseVertex 0x1b
#define AmdExtD3DShaderIntrinsicsOpcode_FloatConversion 0x1c
#define AmdExtD3DShaderIntrinsicsOpcode_ReadlaneAt 0x1d

/**
***********************************************************************************************************************
Expand Down Expand Up @@ -253,6 +255,17 @@ RWByteAddressBuffer AmdExtD3DShaderIntrinsicsUAV : register(u0, AmdExtD3DShaderI
#define AmdExtD3DShaderIntrinsicsAtomicOp_XchgU64 0x07
#define AmdExtD3DShaderIntrinsicsAtomicOp_CmpXchgU64 0x08

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsicsFloatConversion defines for supported rounding modes from float to float16 conversions.
* To be used as an input AmdExtD3DShaderIntrinsicsOpcode_FloatConversion instruction
***********************************************************************************************************************
*/
#define AmdExtD3DShaderIntrinsicsFloatConversionOp_FToF16Near 0x01
#define AmdExtD3DShaderIntrinsicsFloatConversionOp_FToF16NegInf 0x02
#define AmdExtD3DShaderIntrinsicsFloatConversionOp_FToF16PlusInf 0x03


/**
***********************************************************************************************************************
* MakeAmdShaderIntrinsicsInstruction
Expand Down Expand Up @@ -1315,6 +1328,133 @@ uint AmdExtD3DShaderIntrinsics_GetBaseVertex()
return retVal;
}



/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ReadlaneAt : uint
*
* The following function is available if CheckSupport(AmdExtD3DShaderIntrinsicsSupport_ReadlaneAt) returned S_OK.
*
* Returns the value of the source for the given lane index within the specified wave. The lane index
* can be non-uniform across the wave.
*
***********************************************************************************************************************
*/
uint AmdExtD3DShaderIntrinsics_ReadlaneAt(uint src, uint laneId)
{
uint retVal;

uint instruction;
instruction = MakeAmdShaderIntrinsicsInstruction(AmdExtD3DShaderIntrinsicsOpcode_ReadlaneAt,
AmdExtD3DShaderIntrinsicsOpcodePhase_0,
0);
AmdExtD3DShaderIntrinsicsUAV.InterlockedCompareExchange(instruction, src, laneId, retVal);

return retVal;
}

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ReadlaneAt : int
***********************************************************************************************************************
*/
int AmdExtD3DShaderIntrinsics_ReadlaneAt(int src, uint laneId)
{
uint retVal;

uint instruction;
instruction = MakeAmdShaderIntrinsicsInstruction(AmdExtD3DShaderIntrinsicsOpcode_ReadlaneAt,
AmdExtD3DShaderIntrinsicsOpcodePhase_0,
0);
AmdExtD3DShaderIntrinsicsUAV.InterlockedCompareExchange(instruction, asuint(src), laneId, retVal);

return asint(retVal);
}

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ReadlaneAt : float
***********************************************************************************************************************
*/
float AmdExtD3DShaderIntrinsics_ReadlaneAt(float src, uint laneId)
{
uint retVal;

uint instruction;
instruction = MakeAmdShaderIntrinsicsInstruction(AmdExtD3DShaderIntrinsicsOpcode_ReadlaneAt,
AmdExtD3DShaderIntrinsicsOpcodePhase_0,
0);
AmdExtD3DShaderIntrinsicsUAV.InterlockedCompareExchange(instruction, asuint(src), laneId, retVal);

return asfloat(retVal);
}

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ConvertF32toF16
*
* The following functions are available if CheckSupport(AmdExtD3DShaderIntrinsicsSupport_FloatConversion) returned
* S_OK.
*
* Converts 32bit floating point numbers into 16bit floating point number using a specified rounding mode
*
* Available in all shader stages.
*
***********************************************************************************************************************
*/

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ConvertF32toF16 - helper to convert f32 to f16 number
***********************************************************************************************************************
*/
uint3 AmdExtD3DShaderIntrinsics_ConvertF32toF16(in uint convOp, in float3 val)
{
uint instruction = MakeAmdShaderIntrinsicsInstruction(AmdExtD3DShaderIntrinsicsOpcode_FloatConversion,
AmdExtD3DShaderIntrinsicsOpcodePhase_0,
convOp);

uint3 retVal;
AmdExtD3DShaderIntrinsicsUAV.InterlockedCompareExchange(instruction, asuint(val.x), 0, retVal.x);
AmdExtD3DShaderIntrinsicsUAV.InterlockedCompareExchange(instruction, asuint(val.y), 0, retVal.y);
AmdExtD3DShaderIntrinsicsUAV.InterlockedCompareExchange(instruction, asuint(val.z), 0, retVal.z);

return retVal;
}

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ConvertF32toF16Near - convert f32 to f16 number using nearest rounding mode
***********************************************************************************************************************
*/
uint3 AmdExtD3DShaderIntrinsics_ConvertF32toF16Near(in float3 inVec)
{
return AmdExtD3DShaderIntrinsics_ConvertF32toF16(AmdExtD3DShaderIntrinsicsFloatConversionOp_FToF16Near, inVec);
}

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ConvertF32toF16Near - convert f32 to f16 number using -inf rounding mode
***********************************************************************************************************************
*/
uint3 AmdExtD3DShaderIntrinsics_ConvertF32toF16NegInf(in float3 inVec)
{
return AmdExtD3DShaderIntrinsics_ConvertF32toF16(AmdExtD3DShaderIntrinsicsFloatConversionOp_FToF16NegInf, inVec);
}

/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_ConvertF32toF16Near - convert f32 to f16 number using +inf rounding mode
***********************************************************************************************************************
*/
uint3 AmdExtD3DShaderIntrinsics_ConvertF32toF16PosInf(in float3 inVec)
{
return AmdExtD3DShaderIntrinsics_ConvertF32toF16(AmdExtD3DShaderIntrinsicsFloatConversionOp_FToF16PlusInf, inVec);
}



/**
***********************************************************************************************************************
* AmdExtD3DShaderIntrinsics_MakeAtomicInstructions
Expand Down Expand Up @@ -3756,4 +3896,98 @@ uint4 AmdExtD3DShaderIntrinsics_WavePostfixMax(uint4 src)
}


#if defined (AGS_RAY_HIT_TOKEN)

//=====================================================================================================================
struct AmdExtRtHitToken
{
uint dword[2];
};

/**
***********************************************************************************************************************
* @brief
* AmdExtD3DShaderIntrinsicsRT structure when included in a Ray Tracing payload will indicate to the driver
* that the dwords are already supplied in AmdExtRtHitTokenIn and only requires a call to intersect
* ray, bypassing the traversal of the acceleration structure.
***********************************************************************************************************************
*/
struct AmdExtRtHitTokenIn : AmdExtRtHitToken { };

/**
***********************************************************************************************************************
* @brief
* AmdExtD3DShaderIntrinsicsRT structure when included in a Ray Tracing payload will indicate to the driver
* that the dwords must be patched into the payload after traversal. The application can store this
* data in a buffer which can then be used for hit group sorting so shading divergence can be avoided.
***********************************************************************************************************************
*/
struct AmdExtRtHitTokenOut : AmdExtRtHitToken { };

/**
***********************************************************************************************************************
* @brief
* Group shared memory reserved for temprary storage of hit tokens. Not intended to touched by the app shader.
* Application shader must only use the extension functions defined below to access the hit tokens
*
***********************************************************************************************************************
*/
groupshared AmdExtRtHitToken AmdHitToken;

/**
***********************************************************************************************************************
* @brief
* Accessor function to obtain the hit tokens from the last call to TraceRays(). The data returned by this
* function only guarantees valid values for the last call to TraceRays() prior to calling this function.
*
***********************************************************************************************************************
*/
uint2 AmdGetLastHitToken()
{
return uint2(AmdHitToken.dword[0], AmdHitToken.dword[1]);
}

/**
***********************************************************************************************************************
* @brief
* This function initialises hit tokens for subsequent TraceRays() call. Note, any TraceRay() that intends to use
* these hit tokens must include this function call in the same basic block. Applications can use a convenience macro
* defined below to enforce that.
*
***********************************************************************************************************************
*/
void AmdSetHitToken(uint2 token)
{
AmdHitToken.dword[0] = token.x;
AmdHitToken.dword[1] = token.y;
}

/**
***********************************************************************************************************************
* @brief
* Convenience macro for calling TraceRays that uses the hit token
*
***********************************************************************************************************************
*/
#define AmdTraceRay(accelStruct, \
rayFlags, \
instanceInclusionMask, \
rayContributionToHitGroupIndex, \
geometryMultiplier, \
missShaderIndex, \
ray, \
payload, \
token) \
AmdSetHitToken(token); \
TraceRay(accelStruct, \
rayFlags, \
instanceInclusionMask, \
rayContributionToHitGroupIndex, \
geometryMultiplier, \
missShaderIndex, \
ray, \
payload); \

#endif // AGS_RAY_HIT_TOKEN

#endif // _AMDEXTD3DSHADERINTRINICS_HLSL
Loading

0 comments on commit 4f2442c

Please sign in to comment.