Merged
Conversation
github-actions
bot
added
Vulkan
ORippler
added a commit
to ORippler/llama.cpp
that referenced
this pull request
Feb 4, 2026
Seems memory layout is shared with Vulkan so we can port fix from ggml-org#19299
0cc4m
approved these changes
Feb 5, 2026
ORippler
added a commit
to ORippler/llama.cpp
that referenced
this pull request
Feb 6, 2026
Seems memory layout is shared with Vulkan so we can port fix from ggml-org#19299
ggerganov
added a commit
that referenced
this pull request
Feb 8, 2026
* Rename variables + fix rope_neox Seems memory layout is shared with Vulkan so we can port fix from #19299 * Fix rope_multi * Fix rope_vision * Fix rope_norm * Rename ne* to ne0* for consistent variable naming * cont : consistent stride names --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
maxious
added a commit
to maxious/llama.cpp
that referenced
this pull request
Feb 8, 2026
This fix adds support for non-contiguous (strided) source tensors and proper destination stride handling for inplace operations. Changes: - Add s3 (dimension 3 stride) parameter to all ROPE kernels - Add d1, d2, d3 (destination strides) parameters to all ROPE kernels - Properly decompose 4D coordinates (i1, i2, i3) from row_dst - Use strided indexing (i3*s3 + i2*s2 + i1*s1 + i0) for source access - Use destination strides (i3*d3 + i2*d2 + i1*d1 + i0) for destination access - Fix pos array indexing to use i2 (actual dimension 2 index) This aligns the SYCL implementation with the Vulkan fix in PR ggml-org#19299 and enables proper support for KV cache shift operations which use non-contiguous tensor views. All 288 ROPE tests now pass including: - Non-contiguous (v=1) tests - Inplace (inplace=1) tests - Combined non-contiguous + inplace tests - All modes: NORM (0), NEOX (2), MROPE (8), VISION (24), etc.
Merged
maxious
added a commit
to maxious/llama.cpp
that referenced
this pull request
Feb 14, 2026
This fix adds support for non-contiguous (strided) source tensors and proper destination stride handling for inplace operations. Changes: - Add s3 (dimension 3 stride) parameter to all ROPE kernels - Add d1, d2, d3 (destination strides) parameters to all ROPE kernels - Properly decompose 4D coordinates (i1, i2, i3) from row_dst - Use strided indexing (i3*s3 + i2*s2 + i1*s1 + i0) for source access - Use destination strides (i3*d3 + i2*d2 + i1*d1 + i0) for destination access - Fix pos array indexing to use i2 (actual dimension 2 index) This aligns the SYCL implementation with the Vulkan fix in PR ggml-org#19299 and enables proper support for KV cache shift operations which use non-contiguous tensor views. All 288 ROPE tests now pass including: - Non-contiguous (v=1) tests - Inplace (inplace=1) tests - Combined non-contiguous + inplace tests - All modes: NORM (0), NEOX (2), MROPE (8), VISION (24), etc.
ggerganov
added a commit
to ggml-org/ggml
that referenced
this pull request
Feb 14, 2026
* Rename variables + fix rope_neox Seems memory layout is shared with Vulkan so we can port fix from ggml-org/llama.cpp#19299 * Fix rope_multi * Fix rope_vision * Fix rope_norm * Rename ne* to ne0* for consistent variable naming * cont : consistent stride names --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
ggerganov
added a commit
to ggml-org/ggml
that referenced
this pull request
Feb 14, 2026
* Rename variables + fix rope_neox Seems memory layout is shared with Vulkan so we can port fix from ggml-org/llama.cpp#19299 * Fix rope_multi * Fix rope_vision * Fix rope_norm * Rename ne* to ne0* for consistent variable naming * cont : consistent stride names --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
ggerganov
added a commit
to ggml-org/whisper.cpp
that referenced
this pull request
Feb 15, 2026
* Rename variables + fix rope_neox Seems memory layout is shared with Vulkan so we can port fix from ggml-org/llama.cpp#19299 * Fix rope_multi * Fix rope_vision * Fix rope_norm * Rename ne* to ne0* for consistent variable naming * cont : consistent stride names --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
ggerganov
added a commit
to ggml-org/whisper.cpp
that referenced
this pull request
Feb 15, 2026
* Rename variables + fix rope_neox Seems memory layout is shared with Vulkan so we can port fix from ggml-org/llama.cpp#19299 * Fix rope_multi * Fix rope_vision * Fix rope_norm * Rename ne* to ne0* for consistent variable naming * cont : consistent stride names --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
liparetejas
pushed a commit
to liparetejas/llama.cpp
that referenced
this pull request
Feb 23, 2026
liparetejas
pushed a commit
to liparetejas/llama.cpp
that referenced
this pull request
Feb 23, 2026
* Rename variables + fix rope_neox Seems memory layout is shared with Vulkan so we can port fix from ggml-org#19299 * Fix rope_multi * Fix rope_vision * Fix rope_norm * Rename ne* to ne0* for consistent variable naming * cont : consistent stride names --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
maxious
added a commit
to maxious/llama.cpp
that referenced
this pull request
Mar 2, 2026
This fix adds support for non-contiguous (strided) source tensors and proper destination stride handling for inplace operations. Changes: - Add s3 (dimension 3 stride) parameter to all ROPE kernels - Add d1, d2, d3 (destination strides) parameters to all ROPE kernels - Properly decompose 4D coordinates (i1, i2, i3) from row_dst - Use strided indexing (i3*s3 + i2*s2 + i1*s1 + i0) for source access - Use destination strides (i3*d3 + i2*d2 + i1*d1 + i0) for destination access - Fix pos array indexing to use i2 (actual dimension 2 index) This aligns the SYCL implementation with the Vulkan fix in PR ggml-org#19299 and enables proper support for KV cache shift operations which use non-contiguous tensor views. All 288 ROPE tests now pass including: - Non-contiguous (v=1) tests - Inplace (inplace=1) tests - Combined non-contiguous + inplace tests - All modes: NORM (0), NEOX (2), MROPE (8), VISION (24), etc.
bartowski1182
pushed a commit
to bartowski1182/llama.cpp
that referenced
this pull request
Mar 2, 2026
* Rename variables + fix rope_neox Seems memory layout is shared with Vulkan so we can port fix from ggml-org#19299 * Fix rope_multi * Fix rope_vision * Fix rope_norm * Rename ne* to ne0* for consistent variable naming * cont : consistent stride names --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
For #19296.