| cbuffer cbuffer_m : register(b0, space0) { |
| uint4 m[1]; |
| }; |
| static int counter = 0; |
| |
| int i() { |
| counter = (counter + 1); |
| return counter; |
| } |
| |
| matrix<float16_t, 2, 3> m_load(uint offset) { |
| const uint scalar_offset = ((offset + 0u)) / 4; |
| uint4 ubo_load_1 = m[scalar_offset / 4]; |
| uint2 ubo_load = ((scalar_offset & 2) ? ubo_load_1.zw : ubo_load_1.xy); |
| vector<float16_t, 2> ubo_load_xz = vector<float16_t, 2>(f16tof32(ubo_load & 0xFFFF)); |
| float16_t ubo_load_y = f16tof32(ubo_load[0] >> 16); |
| const uint scalar_offset_1 = ((offset + 8u)) / 4; |
| uint4 ubo_load_3 = m[scalar_offset_1 / 4]; |
| uint2 ubo_load_2 = ((scalar_offset_1 & 2) ? ubo_load_3.zw : ubo_load_3.xy); |
| vector<float16_t, 2> ubo_load_2_xz = vector<float16_t, 2>(f16tof32(ubo_load_2 & 0xFFFF)); |
| float16_t ubo_load_2_y = f16tof32(ubo_load_2[0] >> 16); |
| return matrix<float16_t, 2, 3>(vector<float16_t, 3>(ubo_load_xz[0], ubo_load_y, ubo_load_xz[1]), vector<float16_t, 3>(ubo_load_2_xz[0], ubo_load_2_y, ubo_load_2_xz[1])); |
| } |
| |
| [numthreads(1, 1, 1)] |
| void f() { |
| const int p_m_i_save = i(); |
| const matrix<float16_t, 2, 3> l_m = m_load(0u); |
| const uint scalar_offset_2 = ((8u * uint(p_m_i_save))) / 4; |
| uint4 ubo_load_5 = m[scalar_offset_2 / 4]; |
| uint2 ubo_load_4 = ((scalar_offset_2 & 2) ? ubo_load_5.zw : ubo_load_5.xy); |
| vector<float16_t, 2> ubo_load_4_xz = vector<float16_t, 2>(f16tof32(ubo_load_4 & 0xFFFF)); |
| float16_t ubo_load_4_y = f16tof32(ubo_load_4[0] >> 16); |
| const vector<float16_t, 3> l_m_i = vector<float16_t, 3>(ubo_load_4_xz[0], ubo_load_4_y, ubo_load_4_xz[1]); |
| return; |
| } |