mirror of
https://github.com/ggerganov/llama.cpp.git
synced 2024-11-14 23:09:53 +00:00
7c7836d9d4
* Refactor shaders, extract GLSL code from ggml_vk_generate_shaders.py into vulkan-shaders directory * Improve debug log code * Add memory debug output option * Fix flake8 * Fix unnecessary high llama-3 VRAM use
36 lines
1.4 KiB
Plaintext
36 lines
1.4 KiB
Plaintext
#extension GL_EXT_shader_16bit_storage : require
|
|
|
|
layout (push_constant) uniform parameter
|
|
{
|
|
uint ne;
|
|
uint ne00; uint ne01; uint ne02; uint ne03; uint nb00; uint nb01; uint nb02; uint nb03;
|
|
uint ne10; uint ne11; uint ne12; uint ne13; uint nb10; uint nb11; uint nb12; uint nb13;
|
|
uint d_offset;
|
|
float param1; float param2;
|
|
} p;
|
|
|
|
layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
|
|
|
|
layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
|
|
layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
|
|
|
|
uint src0_idx(uint idx) {
|
|
const uint i03 = idx / (p.ne02*p.ne01*p.ne00);
|
|
const uint i03_offset = i03 * p.ne02*p.ne01*p.ne00;
|
|
const uint i02 = (idx - i03_offset) / (p.ne01*p.ne00);
|
|
const uint i02_offset = i02*p.ne01*p.ne00;
|
|
const uint i01 = (idx - i03_offset - i02_offset) / p.ne00;
|
|
const uint i00 = idx - i03_offset - i02_offset - i01*p.ne00;
|
|
return i03*p.nb03 + i02*p.nb02 + i01*p.nb01 + i00*p.nb00;
|
|
}
|
|
|
|
uint dst_idx(uint idx) {
|
|
const uint i13 = idx / (p.ne12*p.ne11*p.ne10);
|
|
const uint i13_offset = i13 * p.ne12*p.ne11*p.ne10;
|
|
const uint i12 = (idx - i13_offset) / (p.ne11*p.ne10);
|
|
const uint i12_offset = i12*p.ne11*p.ne10;
|
|
const uint i11 = (idx - i13_offset - i12_offset) / p.ne10;
|
|
const uint i10 = idx - i13_offset - i12_offset - i11*p.ne10;
|
|
return i13*p.nb13 + i12*p.nb12 + i11*p.nb11 + i10*p.nb10;
|
|
}
|