d1479f871a
This avoids a lot of huge code changes. Arrays generally cannot be copied in and out of buffers, at least no compiler frontend seems to do it. Also avoids a lot of issues surrounding packed vectors and matrices.
47 lines
825 B
Plaintext
47 lines
825 B
Plaintext
#include <metal_stdlib>
|
|
#include <simd/simd.h>
|
|
|
|
using namespace metal;
|
|
|
|
struct A
|
|
{
|
|
int a;
|
|
int b;
|
|
};
|
|
|
|
struct A_1
|
|
{
|
|
A Data[1];
|
|
};
|
|
|
|
struct A_2
|
|
{
|
|
int a;
|
|
int b;
|
|
char _m0_final_padding[8];
|
|
};
|
|
|
|
struct A_3
|
|
{
|
|
A_2 Data[1024];
|
|
};
|
|
|
|
struct B
|
|
{
|
|
A Data[1];
|
|
};
|
|
|
|
struct B_1
|
|
{
|
|
A_2 Data[1024];
|
|
};
|
|
|
|
kernel void main0(device A_1& C1 [[buffer(0)]], constant A_3& C2 [[buffer(1)]], device B& C3 [[buffer(2)]], constant B_1& C4 [[buffer(3)]], uint3 gl_GlobalInvocationID [[thread_position_in_grid]])
|
|
{
|
|
C1.Data[gl_GlobalInvocationID.x].a = C2.Data[gl_GlobalInvocationID.x].a;
|
|
C1.Data[gl_GlobalInvocationID.x].b = C2.Data[gl_GlobalInvocationID.x].b;
|
|
C3.Data[gl_GlobalInvocationID.x].a = C4.Data[gl_GlobalInvocationID.x].a;
|
|
C3.Data[gl_GlobalInvocationID.x].b = C4.Data[gl_GlobalInvocationID.x].b;
|
|
}
|
|
|