e29acf74fe
* vulkan (DRAFT): split shader generation by GLSL source file, to improve incremental build times * support dep-files so shaders are recompiled if their included files change * rename shader files which are used as "headers" to use .glsl extension * move glslc extension detection shaders to separate folders * the above is to prevent them from getting glob'd with the actual compute shaders that need to be compiled * vulkan : only write embedded shader .hpp/.cpp when they change * avoid recompiling ggml-vulkan.cpp when editing shaders * pass single --source argument instead of --input-dir & --filter to shader gen * check for source file match earlier * fix hang in vulkan-shaders-gen when there are compilation errors * early out did not decrement compile_count * clean up * fix glslc integer dot product test * unconditionally write the embedded shader cpp output * replace output filepath in generated dep-files to match output in CMakeLists --------- Co-authored-by: Jeff Bolz <jbolz@nvidia.com>
61 lines
1.4 KiB
Plaintext
61 lines
1.4 KiB
Plaintext
#version 450
|
|
|
|
#include "generic_head.glsl"
|
|
#include "types.glsl"
|
|
|
|
#extension GL_EXT_control_flow_attributes : enable
|
|
|
|
#define FLT_MAX 3.402823466e+38F
|
|
|
|
layout(local_size_x_id = 0, 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[];};
|
|
|
|
layout (constant_id = 0) const uint BLOCK_SIZE = 32;
|
|
|
|
shared FLOAT_TYPE tmpmax[BLOCK_SIZE];
|
|
shared uint tmp[BLOCK_SIZE];
|
|
|
|
void main() {
|
|
const uint row = gl_WorkGroupID.z * 262144 + gl_WorkGroupID.y * 512 + gl_WorkGroupID.x;
|
|
const uint col = gl_LocalInvocationID.x;
|
|
|
|
if (row >= p.KY) {
|
|
return;
|
|
}
|
|
|
|
A_TYPE amax = -FLT_MAX;
|
|
uint acol = col;
|
|
|
|
if (col < p.KX) {
|
|
amax = data_a[row*p.KX + col];
|
|
}
|
|
|
|
for (uint i = col + BLOCK_SIZE; i < p.KX; i += BLOCK_SIZE) {
|
|
A_TYPE val = data_a[row*p.KX + i];
|
|
if (val > amax) {
|
|
amax = val;
|
|
acol = i;
|
|
}
|
|
}
|
|
|
|
tmp[col] = acol;
|
|
tmpmax[col] = amax;
|
|
|
|
barrier();
|
|
[[unroll]] for (int s = int(BLOCK_SIZE) / 2; s > 0; s >>= 1) {
|
|
if (col < s && col + s < p.KX) {
|
|
if (tmpmax[col] < tmpmax[col + s]) {
|
|
tmpmax[col] = tmpmax[col + s];
|
|
tmp[col] = tmp[col + s];
|
|
}
|
|
}
|
|
barrier();
|
|
}
|
|
|
|
if (col == 0) {
|
|
data_d[row] = D_TYPE(tmp[0]);
|
|
}
|
|
}
|