5b952d2cbf
We were passing arrays by value which the compiler fails to optimize, causing abyssal performance. To fix this, we need to consider that descriptors can be in constant or const device address spaces. Also, lone descriptors are passed by value, so we explicitly remove address space qualifiers. One failure case is when shader passes a texture/sampler array as an argument. It's all UniformConstant in SPIR-V, but in MSL it might be thread, const device or constant, so that won't work ... Global variable use works fine though, and that should cover 99.9999999% of use cases.
49 lines
1.8 KiB
GLSL
49 lines
1.8 KiB
GLSL
#pragma clang diagnostic ignored "-Wmissing-prototypes"
|
|
|
|
#include <metal_stdlib>
|
|
#include <simd/simd.h>
|
|
|
|
using namespace metal;
|
|
|
|
struct main0_out
|
|
{
|
|
float4 FragColor [[color(0)]];
|
|
};
|
|
|
|
struct main0_in
|
|
{
|
|
float2 vTex [[user(locn0), flat]];
|
|
int vIndex [[user(locn1)]];
|
|
};
|
|
|
|
static inline __attribute__((always_inline))
|
|
float4 sample_from_global(thread int& vIndex, thread float2& vTex, thread const array<texture2d<float>, 4>& uSampler, thread const array<sampler, 4>& uSamplerSmplr)
|
|
{
|
|
return uSampler[vIndex].sample(uSamplerSmplr[vIndex], (vTex + float2(0.100000001490116119384765625)));
|
|
}
|
|
|
|
static inline __attribute__((always_inline))
|
|
float4 sample_from_argument(thread const array<texture2d<float>, 4>& samplers, thread const array<sampler, 4>& samplersSmplr, thread int& vIndex, thread float2& vTex)
|
|
{
|
|
return samplers[vIndex].sample(samplersSmplr[vIndex], (vTex + float2(0.20000000298023223876953125)));
|
|
}
|
|
|
|
static inline __attribute__((always_inline))
|
|
float4 sample_single_from_argument(texture2d<float> samp, sampler sampSmplr, thread float2& vTex)
|
|
{
|
|
return samp.sample(sampSmplr, (vTex + float2(0.300000011920928955078125)));
|
|
}
|
|
|
|
fragment main0_out main0(main0_in in [[stage_in]], array<texture2d<float>, 4> uSampler [[texture(0)]], array<texture2d<float>, 4> uTextures [[texture(4)]], array<sampler, 4> uSamplerSmplr [[sampler(0)]], array<sampler, 4> uSamplers [[sampler(4)]])
|
|
{
|
|
main0_out out = {};
|
|
out.FragColor = float4(0.0);
|
|
out.FragColor += uTextures[2].sample(uSamplers[1], in.vTex);
|
|
out.FragColor += uSampler[in.vIndex].sample(uSamplerSmplr[in.vIndex], in.vTex);
|
|
out.FragColor += sample_from_global(in.vIndex, in.vTex, uSampler, uSamplerSmplr);
|
|
out.FragColor += sample_from_argument(uSampler, uSamplerSmplr, in.vIndex, in.vTex);
|
|
out.FragColor += sample_single_from_argument(uSampler[3], uSamplerSmplr[3], in.vTex);
|
|
return out;
|
|
}
|
|
|