// V2 working images store premultiplied ACEScct/AP1 code values in float32. inline float3 dyManagedDecode(float3 c){return float3(dyPrintCCTDecode(c.r),dyPrintCCTDecode(c.g),dyPrintCCTDecode(c.b));} inline float3 dyManagedEncode(float3 c){return float3(dyPrintCCTEncode(c.r),dyPrintCCTEncode(c.g),dyPrintCCTEncode(c.b));} kernel void dyManagedCopy(texture2d a [[texture(0)]],texture2d b [[texture(1)]],uint2 p [[thread_position_in_grid]]){if(any(p>=uint2(b.get_width(),b.get_height())))return;b.write(a.read(p),p);} kernel void dyManagedBasis(texture2d a [[texture(0)]],texture2d b [[texture(1)]],constant BasisParams& settings [[buffer(0)]],constant float& amount [[buffer(1)]],uint2 p [[thread_position_in_grid]]){ if(any(p>=uint2(b.get_width(),b.get_height())))return;float4 c=a.read(p);if(!(c.a>0)||amount<=0){b.write(c,p);return;} float3 raw=c.rgb/c.a,result=basis_air(half4(half3(raw),1.h),settings).rgb; float3 mixed=amount>=1?result:dyManagedEncode(mix(dyManagedDecode(raw),dyManagedDecode(result),amount));b.write(float4(mixed*c.a,c.a),p); } kernel void dyManagedPrint(texture2d a [[texture(0)]],texture3d cube [[texture(1)]],texture2d b [[texture(2)]],constant float& amount [[buffer(0)]],uint2 p [[thread_position_in_grid]]){ if(any(p>=uint2(b.get_width(),b.get_height())))return;float4 c=a.read(p);if(!(c.a>0)||amount<=0){b.write(c,p);return;}float3 raw=c.rgb/c.a,result=dyPrintTetra(cube,raw); b.write(float4(dyManagedEncode(mix(dyManagedDecode(raw),dyManagedDecode(result),amount))*c.a,c.a),p); } kernel void dyManagedBlend(texture2d before [[texture(0)]],texture2d after [[texture(1)]],texture2d output [[texture(2)]],device const DYToolMaskEntry* masks [[buffer(0)]],constant DYToolMaskHeader& h [[buffer(1)]],uint2 p [[thread_position_in_grid]]){ if(any(p>=uint2(output.get_width(),output.get_height())))return;float4 a=before.read(p),b=after.read(p);if(!(a.a>0)||!(b.a>0)){output.write(a,p);return;} float3 ac=dyManagedDecode(a.rgb/a.a),bc=dyManagedDecode(b.rgb/b.a);float2 uv=(float2(p)+.5f)/float2(output.get_width(),output.get_height()); float w=dyToolCoverage(uv,dyToolEncode(dyPrintFromAP1(ac)),masks,h,h.section);if(w<=0){output.write(a,p);return;}if(w>=1){output.write(b,p);return;} output.write(float4(dyManagedEncode(mix(ac,bc,w))*a.a,a.a),p); } // Explicit code-value carrier for expert camera/LUT input. Host-profile mode // never uses this reinterpretation; it converts the actual source ICC instead. kernel void dyManagedCarrier(texture2d a [[texture(0)]],texture2d b [[texture(1)]],texture3d cube [[texture(2)]],constant uint& useLUT [[buffer(0)]],uint2 p [[thread_position_in_grid]]){ if(any(p>=uint2(b.get_width(),b.get_height())))return;float4 c=a.read(p);if(c.a>0){float3 v=dyEncode(c.rgb/c.a);if(useLUT)v=dyPrintTetra(cube,v);c.rgb=v*c.a;}b.write(c,p); } fragment float4 dyManagedCopyFragment(DYRaster in [[stage_in]],texture2d source [[texture(0)]]){constexpr sampler f(coord::normalized,address::clamp_to_edge,filter::linear);return source.sample(f,in.uv);} kernel void dyManagedPreview(texture2d a [[texture(0)]],device float4* out [[buffer(0)]],uint2 p [[thread_position_in_grid]]){if(any(p>=uint2(320,180)))return;constexpr sampler s(coord::normalized,address::clamp_to_edge,filter::linear);float4 c=a.sample(s,(float2(p)+.5f)/float2(320,180));out[p.y*320+p.x]=float4(c.a>0?dyEncode(c.rgb/c.a):float3(0),c.a);} kernel void dyManagedLookPreview(texture2d a [[texture(0)]],device float4* out [[buffer(0)]], constant uint4& geometry [[buffer(2)]],uint2 p [[thread_position_in_grid]]){ uint2 size=geometry.xy; if(any(p>=size))return; uint2 sample=uint2(p.x,geometry.z?size.y-1-p.y:p.y); constexpr sampler s(coord::normalized,address::clamp_to_edge,filter::linear); float4 c=a.sample(s,(float2(sample)+.5f)/float2(size)); out[p.y*size.x+p.x]=float4(c.a>0?dyEncode(c.rgb/c.a):float3(0),c.a); }