Skip to content

[ET-VK] Using uint16 for quantized linear tiling shader to reduce register pressure and improve performance. #10509

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 11 commits into from
Apr 28, 2025
Merged

[ET-VK] Using uint16 for quantized linear tiling shader to reduce register pressure and improve performance. #10509

Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
7e168c5
[ET-VK] Modify quantized linear tiling shader to linearly dispatch wo…
trivedivivek Apr 28, 2025
0942988
[ET-VK] Using uint16 for quantized linear tiling shader to reduce reg…
trivedivivek Apr 28, 2025
2f5be1d
Update base for Update on "[ET-VK] Using uint16 for quantized linear …
trivedivivek Apr 28, 2025
68e7442
Update on "[ET-VK] Using uint16 for quantized linear tiling shader to…
trivedivivek Apr 28, 2025
158ef37
Update base for Update on "[ET-VK] Using uint16 for quantized linear …
trivedivivek Apr 28, 2025
29e9aba
Update on "[ET-VK] Using uint16 for quantized linear tiling shader to…
trivedivivek Apr 28, 2025
363b90c
Update base for Update on "[ET-VK] Using uint16 for quantized linear …
trivedivivek Apr 28, 2025
97e21c6
Update on "[ET-VK] Using uint16 for quantized linear tiling shader to…
trivedivivek Apr 28, 2025
8a338c2
Update base for Update on "[ET-VK] Using uint16 for quantized linear …
trivedivivek Apr 28, 2025
5eb160d
Update on "[ET-VK] Using uint16 for quantized linear tiling shader to…
trivedivivek Apr 28, 2025
a1696a0
Merge branch 'main' into gh/trivedivivek/79/head
SS-JIA Apr 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 11 additions & 9 deletions backends/vulkan/runtime/graph/ops/glsl/q_8w_linear_tiled.glsl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,12 +40,14 @@ layout(push_constant) uniform restrict Block {

layout(local_size_x_id = 0, local_size_y_id = 1, local_size_z_id = 2) in;

#extension GL_EXT_shader_explicit_arithmetic_types_int16 : require

void main() {
const uint out_width_ntexels = divup4(out_sizes.x);
const uint out_col = (gl_GlobalInvocationID.x % out_width_ntexels) << 2;
const uint out_row = (gl_GlobalInvocationID.x / out_width_ntexels) * TILE_ROWS;
const uint16_t out_width_ntexels = uint16_t(divup4(out_sizes.x));
const uint16_t out_col = uint16_t((gl_GlobalInvocationID.x % out_width_ntexels) << 2);
const uint16_t out_row = uint16_t((gl_GlobalInvocationID.x / out_width_ntexels) * TILE_ROWS);

if (out_row >= out_sizes.y) {
if (out_row >= uint16_t(out_sizes.y)) {
return;
}

Expand All @@ -54,29 +56,29 @@ void main() {
VEC4_T c[TILE_ROWS];

$if SCALES_STORAGE == "buffer":
const VEC4_T scales = VEC4_T(t_scales[out_col >> 2]);
const VEC4_T scales = VEC4_T(t_scales[int(out_col >> 2)]);
$else:
const VEC4_T scales = VEC4_T(texelFetch(t_scales, ivec2(out_col >> 2, 0), 0));
const VEC4_T scales = VEC4_T(texelFetch(t_scales, u16vec2(out_col >> 2, 0), 0));

[[unroll]] for (int i = 0; i < TILE_ROWS; ++i) {
c[i] = VEC4_T(0.0);
}

for (int pos = 0; pos < in_sizes.x; pos += 4) {
for (uint16_t pos = uint16_t(0); pos < uint16_t(in_sizes.x); pos += uint16_t(4)) {
// Preload weight tensor
[[unroll]] for (int i = 0; i < 4; i++) {
$if WEIGHT_STORAGE == "buffer":
b[i] = t_weight[((pos + i) * out_sizes.x + out_col) >> 2];
$else:
b[i] = VEC4_T(texelFetch(t_weight, ivec2(out_col >> 2, pos + i), 0));
b[i] = VEC4_T(texelFetch(t_weight, u16vec2(out_col >> 2, pos + i), 0));
}

// Preload input tensor
[[unroll]] for (int i = 0; i < TILE_ROWS; i++) {
$if IN_STORAGE == "buffer":
a[i] = t_in[((out_row + i) * in_sizes.x + pos) >> 2];
$else:
a[i] = VEC4_T(texelFetch(t_in, ivec3(pos >> 2, out_row + i, 0), 0));
a[i] = VEC4_T(texelFetch(t_in, u16vec3(pos >> 2, out_row + i, 0), 0));
}

// Accumulate output
Expand Down
Loading