ec054dad7f
Emit synthetic functions before function constants. Support use of spvQuantizeToF16() in function constants for numerical behavior consistency with the op code. Ensure subnormal results from OpQuantizeToF16 are flushed to zero per SPIR-V spec. Adjust SPIRV-Cross unit test reference shaders to accommodate these changes. Any MSL reference shader that inclues a synthetic function is affected, since the location it is emitted has changed.
39 lines
1.6 KiB
Plaintext
39 lines
1.6 KiB
Plaintext
#pragma clang diagnostic ignored "-Wmissing-prototypes"
|
|
#pragma clang diagnostic ignored "-Wunused-variable"
|
|
|
|
#include <metal_stdlib>
|
|
#include <simd/simd.h>
|
|
#include <metal_atomic>
|
|
|
|
using namespace metal;
|
|
|
|
// The required alignment of a linear texture of R32Uint format.
|
|
constant uint spvLinearTextureAlignmentOverride [[function_constant(65535)]];
|
|
constant uint spvLinearTextureAlignment = is_function_constant_defined(spvLinearTextureAlignmentOverride) ? spvLinearTextureAlignmentOverride : 4;
|
|
// Returns buffer coords corresponding to 2D texture coords for emulating 2D texture atomics
|
|
#define spvImage2DAtomicCoord(tc, tex) (((((tex).get_width() + spvLinearTextureAlignment / 4 - 1) & ~( spvLinearTextureAlignment / 4 - 1)) * (tc).y) + (tc).x)
|
|
|
|
struct SSBO
|
|
{
|
|
float4 outdata;
|
|
};
|
|
|
|
constant uint3 gl_WorkGroupSize [[maybe_unused]] = uint3(1u);
|
|
|
|
struct spvDescriptorSetBuffer0
|
|
{
|
|
texture2d<uint> uImage [[id(0)]];
|
|
device atomic_uint* uImage_atomic [[id(1)]];
|
|
device SSBO* m_31 [[id(2)]];
|
|
texture2d<float> uTexture [[id(3)]];
|
|
sampler uTextureSmplr [[id(4)]];
|
|
};
|
|
|
|
kernel void main0(constant spvDescriptorSetBuffer0& spvDescriptorSet0 [[buffer(0)]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]])
|
|
{
|
|
uint _26 = atomic_fetch_add_explicit((device atomic_uint*)&spvDescriptorSet0.uImage_atomic[spvImage2DAtomicCoord(int2(gl_GlobalInvocationID.xy), spvDescriptorSet0.uImage)], 10u, memory_order_relaxed);
|
|
uint ret = _26;
|
|
(*spvDescriptorSet0.m_31).outdata = spvDescriptorSet0.uTexture.sample(spvDescriptorSet0.uTextureSmplr, float2(gl_GlobalInvocationID.xy), level(0.0)) + float4(float(ret));
|
|
}
|
|
|