| struct S { |
| int before; |
| matrix<float16_t, 2, 2> m; |
| int after; |
| }; |
| |
| cbuffer cbuffer_u : register(b0, space0) { |
| uint4 u[32]; |
| }; |
| static S p[4] = (S[4])0; |
| |
| matrix<float16_t, 2, 2> u_load_3(uint offset) { |
| const uint scalar_offset = ((offset + 0u)) / 4; |
| uint ubo_load = u[scalar_offset / 4][scalar_offset % 4]; |
| const uint scalar_offset_1 = ((offset + 4u)) / 4; |
| uint ubo_load_1 = u[scalar_offset_1 / 4][scalar_offset_1 % 4]; |
| return matrix<float16_t, 2, 2>(vector<float16_t, 2>(float16_t(f16tof32(ubo_load & 0xFFFF)), float16_t(f16tof32(ubo_load >> 16))), vector<float16_t, 2>(float16_t(f16tof32(ubo_load_1 & 0xFFFF)), float16_t(f16tof32(ubo_load_1 >> 16)))); |
| } |
| |
| S u_load_1(uint offset) { |
| const uint scalar_offset_2 = ((offset + 0u)) / 4; |
| const uint scalar_offset_3 = ((offset + 64u)) / 4; |
| const S tint_symbol = {asint(u[scalar_offset_2 / 4][scalar_offset_2 % 4]), u_load_3((offset + 4u)), asint(u[scalar_offset_3 / 4][scalar_offset_3 % 4])}; |
| return tint_symbol; |
| } |
| |
| typedef S u_load_ret[4]; |
| u_load_ret u_load(uint offset) { |
| S arr[4] = (S[4])0; |
| { |
| for(uint i = 0u; (i < 4u); i = (i + 1u)) { |
| arr[i] = u_load_1((offset + (i * 128u))); |
| } |
| } |
| return arr; |
| } |
| |
| [numthreads(1, 1, 1)] |
| void f() { |
| p = u_load(0u); |
| p[1] = u_load_1(256u); |
| p[3].m = u_load_3(260u); |
| uint ubo_load_2 = u[0].z; |
| p[1].m[0] = vector<float16_t, 2>(float16_t(f16tof32(ubo_load_2 & 0xFFFF)), float16_t(f16tof32(ubo_load_2 >> 16))).yx; |
| return; |
| } |