Skip to content

ggml-hexagon: fix rope failure at test-backend-ops #17565

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

Open
wants to merge 9 commits into
base: master
Choose a base branch
from
+37 −41
Open

ggml-hexagon: fix rope failure at test-backend-ops #17565

Changes from 1 commit
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
407b408
fix test failure
chraac Nov 27, 2025
4ddb8a4
fix: correct scaling calculations in rope_cache_init
chraac Nov 27, 2025
cfca78b
wip
chraac Nov 27, 2025
e9a02fd
wip
chraac Nov 28, 2025
e324bb0
fix: optimize element copying in rope_hex_f32 using memcpy
chraac Nov 28, 2025
0121291
fix: optimize loop boundaries in rope_hex_f32 for better performance
chraac Nov 28, 2025
010039a
rename
chraac Nov 28, 2025
a6ef41f
wip
chraac Nov 28, 2025
0376146
Merge branch 'master' into dev-fix-rope
chraac Nov 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
Prev Previous commit
Next Next commit
fix: correct scaling calculations in rope_cache_init
  • Loading branch information
@chraac
chraac committed Nov 27, 2025
commit 4ddb8a449cbe057d67b684b7c34192e69f743d30
11 changes: 6 additions & 5 deletions ggml/src/ggml-hexagon/htp/rope-ops.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,17 +93,18 @@ static void rope_cache_init(const float theta_base,
// Get n-d rotational scaling corrected for extrapolation
float theta_interp = freq_scale * theta_extrap;
float theta2 = theta_interp;
float mscale2 = mscale;

if (ext_factor != 0.0f) {
float ramp_mix = rope_yarn_ramp(corr_dims[0], corr_dims[1], i0) * ext_factor;
theta2 = theta_interp * (1 - ramp_mix) + theta_extrap * ramp_mix;

// Get n-d magnitude scaling corrected for interpolation
mscale *= 1.0f + 0.1f * logf(1.0f / freq_scale);
mscale2 *= 1.0f + 0.1f * logf(1.0f / freq_scale);
}

cache[i0 + 0] = cosf(theta2) * mscale;
cache[i0 + 1] = sinf(theta2) * mscale;
cache[i0 + 0] = cosf(theta2) * mscale2;
cache[i0 + 1] = sinf(theta2) * mscale2;

theta *= theta_scale;
}
Expand Down Expand Up @@ -337,8 +338,8 @@ static void rope_hex_f32(struct rope_th_ctx * rope_ctx,
}
}

src_loc += is_neox ? (rope_ctx->n_dims / 2) : 0;
dst_data_loc += is_neox ? (rope_ctx->n_dims / 2) : 0;
src_loc += (is_neox ? (rope_ctx->n_dims / 2) : 0);
dst_data_loc += (is_neox ? (rope_ctx->n_dims / 2) : 0);
for (uint32_t i0 = rope_ctx->n_dims; i0 < ne0; i0 += 2) {
dst_data_loc[0] = src_loc[0];
dst_data_loc[1] = src_loc[1];
Expand Down