* vulkan: use memory budget extension to read memory usage
* fix: formatting and names
* formatting
* fix: detect and cache memory budget extension availability on init
* fix: read `budgetprops.heapBudget` instead of `heap.size` when memory budget extension is available
* style: lints
* SVE support for exponential functions
Add const notation to variable pg
* Update ggml/src/ggml-cpu/vec.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add const
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* vulkan: Add Integer Dot Product mul_mat_vec shader for legacy quants
* vulkan: use subgroup operations for quantize_q8_1 shader
* vulkan: add q8_1_x4 type with 128-bit alignment, use in mul_mat_vecq shader
* vulkan: use q8_1_x4 blocks in mul_mmq shader
* vulkan: do 8 calculations per invocation instead of 32 in mul_mat_vecq, similar to mul_mat_vec
* vulkan: tune mul_mat_vecq performance for Intel
* vulkan: fix quantizing issue when tensor is not divisible by 128
* vulkan: adapt integer dot mmv to mmv small m optimization (llama/15355)
* vulkan: allow all subgroup modes for mmv and mmvq
* vulkan: use prealloc intermediate reuse for mmvq path
* vulkan: tune mmvq for Intel, AMD GCN and Nvidia RTX 3090
* vulkan: adapt mmv quantize_y path to conditional sync logic
* vulkan: disable q8_0 mmvq on Nvidia
* vulkan: enable q8_0 on Nvidia pre-turing
* fix prealloc sync condition
* fix llvmpipe subgroup 8 issue
* ggml : WebGPU add TRANSPOSE and RESHAPE to supported ops
This commit adds support for the TRANSPOSE and RESHAPE operations in the
ggml webgpu backend.
Co-authored-by: Diego Devesa <slarengh@gmail.com>
Co-authored-by: Sigbjørn Skjæret <sigbjorn.skjaeret@scala.com>
* CUDA: fix build error from ambiguous __half conversions in conv2d
Building conv2d with half precision failed because `__half` defines
multiple implicit conversion operators (to float, int, short, etc.),
causing ambiguous overload resolution when multiplying with float.
Introduce a templated `to_float` helper that explicitly converts
`__half` via `__half2float`, while passing through float unchanged.
Use this helper in conv2d accumulation to ensure unambiguous and
correct promotion to float.
Fixes some build errors with half-precision kernels on CUDA.
ggml-ci
* CUDA: Replace custom to_float helper with unified ggml_cuda_cast and add half‑>float conversion
* CUDA: Add missing convert.cuh header
* CUDA: remove unnecessary extension in ggml_cuda_cast
* CUDA: Address review comment, remove second type template argument
* CANN: fix RoPE cache issue on multi-device
RoPE cache only needs to be computed once per token.
However, in multi-device scenarios, not every device starts
computation from layer 0, which may lead to unallocated memory
issues and precision errors.
This commit records the first layer of each device to avoid
the above issues.
* CANN: Optimize first-layer detection method
* CANN: Remove trailing whitespace
* CANN: Only cache the data that can be determined as unchanged through the parameters.
* CANN: Update function comment
Exposes ggml_backend_sched_split_graph() to allow splitting the graph without allocating compute buffers and uses it to split the graph for the automatic Flash Attention check.
* vulkan: mul_mat_id coopmat2 optimizations
Add a path for when the tile fits in BN/2, similar to what we have for mul_mat.
Only call fetch_scales/store_scales once per QUANT_K block, and once at the
beginning in case start_k is not aligned.
* Also add a path for BN/4 - worth a couple more percent
This commit removes the portability_enumeration_ext variable from the
ggml_vk_instance_portability_enumeration_ext_available function as it
is initialized to false but never modified, making it redundant.
Prior to this change, we faced undefined cublasLt references when
attempting to compile 'llama-cli' with GGML_STATIC=ON on Linux.
We add linking with CUDA::cublasLt_static when CUDA version is greater
than 10.1.
* CANN(flash-attn): refactor mask handling and improve performance
1. Refactored the mask computation in Flash Attention, unified the logic without separating prefill and decode.
2. Optimized performance in non-alibi scenarios by reducing one repeat operation.
3. Updated operator management to explicitly mark unsupported cases on 310P devices and when dim is not divisible by 16.
Signed-off-by: noemotiovon <757486878@qq.com>
* [CANN]: fix review
Signed-off-by: noemotiovon <757486878@qq.com>
* [CANN]: Optimization FA BNSD to BSND
Signed-off-by: noemotiovon <757486878@qq.com>
---------
Signed-off-by: noemotiovon <757486878@qq.com>
The original implementation unconditionally returned true for this operation, leading to a failure when the tensor's first dimension (ne[0]) was not a multiple of WARP_SIZE. This caused an GGML_ASSERT(ncols % WARP_SIZE == 0) failure in ggml-sycl/norm.cpp.
This change updates the ggml_backend_sycl_device_supports_op check to correctly return true for GGML_OP_RMS_NORM only when the first dimension of the tensor is a multiple of WARP_SIZE, ensuring the operation can be performed without error.
This patch improves GEMM for FP32 Data Type on PowerPC
Implements GEMM on large blocks with configurable block size mc, nc, kc
(default: 256, 256, 256).
Packing Function optimized to access blocks as per memory layout.
GEMM Optimized to work on larger blocks.
Isolated Packing from GEMM Operations for better MMA utilization.
Verified functionality and correctness uing llama-cli and stand alone
test case (performs matmul and compares final mattrix C result with base).
Minor code refactoring changes:
Replace macro with inline function
Code Indent made consistent with 4 spaces
Performance Testing:
Observed 50% ~ 70% improvement in Prompt Processing Speed mesured using
llama-bench with Meta-Llama3-8B FP32 Model. Similar gains observed with
Mistral-7b-Instruct-v0.3 Model.
model Size Params Backend Threads Test Patch Base
llama 8B all F32 29.92 GiB 8.03 B CPU 20 pp512 98.58 60.3
llama 8B all F32 29.92 GiB 8.03 B CPU 20 pp1024 95.88 57.36
llama 8B all F32 29.92 GiB 8.03 B CPU 20 pp2048 85.46 53.26
llama 8B all F32 29.92 GiB 8.03 B CPU 20 pp4096 68.66 45.78
llama 8B all F32 29.92 GiB 8.03 B CPU 20 pp6144 57.35 40.44
25 ~ 30% improvement in llama-batched-bench with Metla-Llama3-8B in
Prompt Processing Speed for large prompts (256, 512, 1024, 2048, 4096)tokens with various batch
sizes ( 1, 2, 4, 8, 16)
Signed-off-by: Shalini Salomi Bodapati <Shalini.Salomi.Bodapati@ibm.com>
* metal : optmize FA vec for large heads and sequences
* metal : adjust small-batch mul mv kernels
ggml-ci
* batched-bench : fix total speed computation
ggml-ci
* cont : add comments
ggml-ci
* metal : mul_mm_id remove hdst
* metal : remove mul_mm_id hsrc1
* metal : mul_mm_id simplify + add test
* metal : opt mul_mm_id map0
* metal : optimize mul_mm_id id gathering
* metal : mul/div opt
* metal : optimize mul_mm_id_map0
ggml-ci
* CUDA: optimize get_int_from_table_16
* CUDA: use v_perm_b32 to replace byte_perm on AMD GPUs
* revise documentation
---------
Co-authored-by: xix <xiapc@outlook.com>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* vulkan: use subgroup function for mul_mat_id shader even without coopmat
* vulkan: fix compile warnings
* vulkan: properly check for subgroup size control and require full subgroups for subgroup mul_mat_id
* vulkan: disable subgroup mul_mat_id on devices with subgroups < 16
The scalar FA shader already handled multiples of 8. The coopmat1 FA
shader assumed 16x16x16 and the shared memory allocations need the HSK
dimensions padded to a multiple of 16. NVIDIA's coopmat2 implementation
requires multiples of 16 for N and K, and needs the matrix dimensions
padded and loads clamped.
Store the FA pipelines in a map, indexed by the pipeline state.
* vulkan: optimize rms_norm, and allow the work to spread across multiple SMs
There are really two parts to this change:
(1) Some optimizations similar to what we have in soft_max, to unroll with
different numbers of iterations.
(2) A fusion optimization where we detect add followed by rms_norm, and make
the add shader atomically accumulate the values^2 into memory. Then the
rms_norm shader can just load that sum. This allows the rms_norm to be
parallelized across multiple workgroups, it just becomes a simple per-element
multiply.
The fusion optimization is currently only applied when the rms_norm is on a
single vector. This previously always ran on a single SM. It could apply more
broadly, but when there are other dimensions the work can already spread across
SMs, and there would be some complexity to tracking multiple atomic sums.
* Change add+rms_norm optimization to write out an array of partial sums
rather than using atomic add, to make it deterministic. The rms_norm
shader fetches a subgroup's worth in parallel and uses subgroupAdd to
add them up.
* complete rebase against fused adds - multi_add shader can also compute partial sums
* fix validation errors
* disable add_rms_fusion for Intel due to possible driver bug
* resolve against #15489, sync after clearing partial sums
Track a list of nodes that need synchronization, and only sync if the new node
depends on them (or overwrites them). This allows some overlap which can
improve performance, and centralizes a big chunk of the synchronization logic.
The remaining synchronization logic involves writes to memory other than the
nodes, e.g. for dequantization or split_k. Each of these allocations has a bool
indicating whether they were in use and need to be synced. This should be
checked before they are written to, and set to true after they are done being
consumed.
* vulkan : support ggml_mean
* vulkan : support sum, sum_rows and mean with non-contiguous tensors
* vulkan : fix subbuffer size not accounting for misalign offset
* tests : add backend-op tests for non-contiguous sum_rows
* cuda : require contiguous src for SUM_ROWS, MEAN support
* sycl : require contiguous src for SUM, SUM_ROWS, ARGSORT support
* require ggml_contiguous_rows in supports_op and expect nb00=1 in the shader
- Spread the work across the whole workgroup. Using more threads seems to
far outweigh the synchronization overhead.
- Specialize the code for when the division is by a power of two.
* Begin work on set_rows
* Work on set rows
* Add error buffers for reporting unsupported SET_ROWS indices
* Remove extra comments
* Work on templating for different types in shaders
* Work on shader type generation
* Working q4_0 mul_mat and some templating for different types
* Add q4_0_f16 matmul and fix device init
* Add matmul support for basic quantization types
* Add q2_k and q3_k quantization
* Add rest of k-quants
* Get firt i-quant working
* Closer to supporting all i-quants
* Support rest of i-quants
* Cleanup code
* Fix python formatting
* debug
* Bugfix for memset
* Add padding to end of buffers on creation
* Simplify bit-shifting
* Update usage of StringView
* Add Pad Reflect 1D CUDA support
* Update ggml/src/ggml-cuda/pad_reflect_1d.cu
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* vulkan: Reuse conversion results in prealloc_y
Cache the pipeline and tensor that were most recently used to fill prealloc_y,
and skip the conversion if the current pipeline/tensor match.
* don't use shared pointer for prealloc_y_last_pipeline_used
* musa: fix build warnings
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* fix warning: comparison of integers of different signs: 'const int' and 'unsigned int' [-Wsign-compare]
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
---------
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
While working on the [whisper-cpp](https://conan.io/center/recipes/whisper-cpp) Conan package for ConanCenter, I noticed that enabling the `with_blas` option fails to build due to an issue in the _MKL_ detection logic.
The problem is that the CMake condition currently expands `BLAS_INCLUDE_DIRS` without quotes:
```cmake
if (${BLAS_INCLUDE_DIRS} MATCHES "mkl" AND (${GGML_BLAS_VENDOR} MATCHES "Generic" OR ${GGML_BLAS_VENDOR} MATCHES "Intel"))
```
When `BLAS_INCLUDE_DIRS` is a list (as Conan provides it), the `if()` command receives multiple arguments and produces a CMake error:
```bash
...
-- BLAS found, Includes: /root/.conan2/p/b/openb034c5a6ca927b/p/include;/root/.conan2/p/b/openb034c5a6ca927b/p/include/openblas
CMake Error at ggml/src/ggml-blas/CMakeLists.txt:77 (if):
if given arguments:
"/root/.conan2/p/b/openb034c5a6ca927b/p/include" "/root/.conan2/p/b/openb034c5a6ca927b/p/include/openblas" "MATCHES" "mkl" "AND" "(" "OpenBLAS" "MATCHES" "Generic" "OR" "OpenBLAS" "MATCHES" "Intel" ")"
Unknown arguments specified
...
```
This PR fixes the issue by quoting the variable:
```cmake
if ("${BLAS_INCLUDE_DIRS}" MATCHES "mkl" AND (${GGML_BLAS_VENDOR} MATCHES "Generic" OR ${GGML_BLAS_VENDOR} MATCHES "Intel"))
```
With this change, the whole list is treated as a single string and the regex still works correctly.
* vulkan: Use larger workgroups for mul_mat_vec when M is small
Also use subgroup instructions for (part of) the reduction when supported.
Without this, the more expensive reductions would eat into the benefits of
the larger workgroups.
* update heuristic for amd/intel
Co-authored-by: 0cc4m <picard12@live.de>
---------
Co-authored-by: 0cc4m <picard12@live.de>
- Launch an appropriate number of invocations (next larger power of two).
32 invocations is common and the barrier is much cheaper there.
- Specialize for "needs bounds checking" vs not.
- Make the code less branchy and [[unroll]] the loops. In the final code,
I see no branches inside the main loop (only predicated stores) when
needs_bounds_check is false.
- Always sort ascending, then apply the ascending vs descending option when
doing the final stores to memory.
- Copy the values into shared memory, makes them slightly cheaper to access.
* vulkan: fuse adds
Fuse adds that have the same shape, which are common in MoE models.
It will currently fuse up to 6 adds, because we assume no more than
8 descriptors per dispatch. But this could be changed.
* check runtimeDescriptorArray feature
* disable multi_add for Intel due to likely driver bug
* vulkan: Add missing bounds checking to scalar/coopmat1 mul_mat_id
* vulkan: Support mul_mat_id with f32 accumulators, but they are not hooked up
- There's no explicit way to request f32 precision for mul_mat_id, but there
probably should be, and this gets the code in place for that.
- A couple fixes to check_results.
- Remove casts to fp16 in coopmat1 FA shader (found by inspection).
* add F16/F16 fa support
* fix kernel init
* use mad instead of fma
* use inline function
* mark FA with sinks as unsupported for now
* add pragma unroll to loops
add expicit conversion operator to support older versions of rocm
Switch over to hip_bf16 from legacy hip_bfloat16
Simplify RDNA3 define
Reduce swap over of new hipblas api to rocm 6.5 as this version is used for rocm 7.0 previews
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* vulkan: perf_logger improvements
- Account for batch dimension in flops calculation.
- Fix how "_VEC" is detected for mat_mul_id.
- Fix "n" dimension for mat_mul_id (in case of broadcasting).
- Include a->type in name.
* use <=mul_mat_vec_max_cols rather than ==1
* examples/finetune -opt SGD (stochastic gradient descent) memory opt
add unit tested GGML_OPT_OPTIMIZER_SGD to ggml - avoids allocating
m, v tensors.
support finetune.cpp arg -opt SGD (or sgd). (default adamw as before)
llama 3.2-1b-F32 result: observed 11gb gpu ram (41 sec/epoch)
when using SGD instead of 19gb (55 sec/epoch) using adamw.
(wikipedia 100 lines finetune)
(
using the same GPU memory, adamw can only do before OOM 512
batch/context, reaching:
train: [███████▉] data=0000140/0000140 loss=0.02575±0.00099 acc=99.52±0.03% t=00:00:47 ETA=00:00:00
val: [███████▉] data=0000008/0000008 loss=4.76565±0.28810 acc=41.46±0.77% t=00:00:00 ETA=00:00:00
SGD is superior, though it converges slower, with max before OOM 1728
batch/context (esp see the better validation perf):
train: [███████▉] data=0000039/0000039 loss=0.00371±0.00010 acc=99.96±0.01% t=00:00:41 ETA=00:00:00
val: [███████▉] data=0000003/0000003 loss=5.11406±0.76034 acc=48.01±0.69% t=00:00:01 ETA=00:00:00
)
note: when finetuning long enough (or w/ enough -lr),
validation accuracy *eventually* drops ('catastrophic forgetting')
-lr-half (halflife) option useful for SGD to avoid oscillation or
super slow underdamped learning (makes setting -lr more forgiving).
terminal -lr for now is set by lr-halvings i.e. if you want at most
1/8 the inital -lr you set -lr-halvings 3.
note: objective loss not directly comparable between adamw, sgd? -
check perplexity or accuracy or consider relative improvements
for convergence
new finetune args -wd 1e-9 to enable weight decay in sgd or adamw,
and max -epochs N (default 2 as before)
cache (1 - wd*alpha) in 'adamw' opt struct -
no noticeable perf benefit, disabled (still done
for new SGD though)
since opt. memory is pre-allocated, the ggml_opt_get_optimizer_params
would probably be able to change between SGD and AdamW with each epoch
but would need to use adamw for the first (unconfirmed - no cmdline arg
to set such a policy yet)
test-opt checks adamw as before and now sgd (except for a few disabled
tests for sgd only; probably just needs logging values and adding
alternate reference values); tolerance on the 'regression'
test is broader for sgd (so we don't need many more epochs)
* Vulkan: Implement GGML_OP_OPT_STEP_SGD
* tests: Fix OPT_STEP_SGD test-backend-ops
* SGD op param store weight-decay and not 1-alpha*wd
* minor + cosmetic changes
* fix vulkan sgd
* try CI fix
---------
Co-authored-by: 0cc4m <picard12@live.de>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* Factor out `reduce_rows_f32` from common.cuh
This increases iteration cycle speed by not having to recompile
every kernel all the time
* Hide memory-latency by loop unrolling in reduce_rows_f32
* Further optimizations to `reduce_rows_f32`
1. Increase threadblock size to better hide latency of memory requests.
As a consequence of bigger threadblocks, do 2-step summation, using
shared memory to communicate results between invocations
2. Use sum_temp array to reduce waits on sum
3. Adjust num_unroll to reflext bigger threadblock
4. Improve default block_dims, increase support for more block_dims
* Add perf tests for `reduce_rows_f32` kernel
* Add heuristic to toggle 128/512 threads based on sm count
Break even point was the minimum of the following multiples.
| GPU Model | Nrow SM Count Multiple |
| ----------- | ----------- |
| RTX 4000 SFF ADA | 2.0x |
| RTX 6000 ADA | 2.5x |
| RTX PRO 6000 Blackwell Max-Q | 3.04x |
| RTX PRO 4500 Blackwell | 3.15x |
* Ensure perf gains also for small ncols and large nrows
Alternative to this, one could have also made the number of unrollings
template-able, but that would require compiling the kernel multiple
times, increasing binary size unnecessarily
* Modify perf and unit-tests
* Apply auto-formatting by clang
* Fix CI build failure
See https://github.com/ggml-org/llama.cpp/actions/runs/16798370266/job/47573716079?pr=15132#step:7:486
Building with VS generator worked though.
* Remove sm_count property from `ggml_backend_cuda_context`
Requested by @JohannesGaessler, and should fix remaining CI issues as a
side-effect
* Add CUB-based implementation for GGML_OP_MEAN
Currently this branch is only executed for nrows==1
* Add heuristics to execute CUB branch only when it brings perf
Heuristics were determined on the following HW:
* RTX 4000 SFF ADA
* RTX 6000 ADA
* RTX PRO 6000 Blackwell Max-Q
* RTX PRO 4500 Blackwell
* Add unit-test for CUB-based mean
Tests should run with CUDA Graphs enabled per default on NVGPUs
* Rename `USE_CUB` to `GGML_CUDA_USE_CUB`
Suggested by @JohannesGaessler
* Unindent Preprocessor directives
See
https://github.com/ggml-org/llama.cpp/pull/15132#discussion_r2269213506
* ggml-rpc: chunk send()/recv() to avoid EINVAL for very large tensors over RPC (macOS & others). Fixes#15055
* ggml-rpc: rename RPC_IO_CHUNK->MAX_CHUNK_SIZE, use std::min() for cap, switch to GGML_LOG_ERROR, handle 0-length send/recv
* rpc: drop n==0 special case in send_data(); retry in loop per review
* rpc: remove trailing whitespace in send_data()
---------
Co-authored-by: Shinnosuke Takagi <nosuke@nosukenoMacBook-Pro.local>
* musa: fix failures in test-backend-ops for mul_mat_id op
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* Address review comments
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
---------
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* cuda: refactored ssm_scan to use CUB
* fixed compilation error when when not using CUB
* assign L to constant and use size_t instead of int
* deduplicated functions
* change min blocks per mp to 1
* Use cub load and store warp transpose
* suppress clang warning
* feat(cann): add optional support for ACL Graph execution
This commit adds support for executing ggml computational graphs using
Huawei's ACL graph mode via the USE_CANN_GRAPH flag. The support can be
enabled at compile time using the CMake option:
-DUSE_CANN_GRAPH=ON
By default, ACL graph execution is **disabled**, and the fallback path
uses node-by-node execution.
Key additions:
- CMake option to toggle graph mode
- Graph capture and execution logic using
- Tensor property matching to determine whether graph update is required
- Safe fallback and logging if the environment variable LLAMA_SET_ROWS
is unset or invalid
This prepares the backend for performance improvements in repetitive graph
execution scenarios on Ascend devices.
Signed-off-by: noemotiovon <757486878@qq.com>
* Fix review comments
Signed-off-by: noemotiovon <757486878@qq.com>
* remane USE_CANN_GRAPH to USE_ACL_GRAPH
Signed-off-by: noemotiovon <757486878@qq.com>
* fix typo
Signed-off-by: noemotiovon <757486878@qq.com>
---------
Signed-off-by: noemotiovon <757486878@qq.com>
* cmake: Add GGML_BACKEND_DIR option
This can be used by distributions to specify where to look for backends
when ggml is built with GGML_BACKEND_DL=ON.
* Fix phrasing
- Increase tile size for k-quants, to match non-k-quants
- Choose more carefully between large and medium tiles, considering how it
interacts with split_k
- Allow larger/non-power of two split_k, and make the splits a multiple of 256
- Use split_k==3 to when >1/2 and <=2/3 of the SMs would hae been used
* vulkan: optimizations for direct convolution
- Empirically choose a better tile size. Reducing BS_K/BS_NPQ helps fill
the GPU. The new size should be amenable to using coopmat, too.
- Fix shmem bank conflicts. 16B padding should work with coopmat.
- Some explicit loop unrolling.
- Skip math/stores work for parts of the tile that are OOB.
- Apply fastdiv opt.
- Disable shuffles for NV.
* Three tiles sizes for CONV_2D, and a heuristic to choose
* reallow collectives for pre-Turing
* make SHMEM_PAD a spec constant
* fixes for intel perf - no shmem padding, placeholder shader core count
* shader variants with/without unrolling
* 0cc4m's fixes for AMD perf
Co-authored-by: 0cc4m <picard12@live.de>
---------
Co-authored-by: 0cc4m <picard12@live.de>
* Initial Q2_K Block Interleaving Implementation
* Addressed review comments and clean up of the code
* Post rebase fixes
* Initial CI/CD fixes
* Update declarations in arch-fallback.h
* Changes for GEMV Q2_K in arch-fallback.h
* Enable repacking only on AVX-512 machines
* Update comments in repack.cpp
* Address q2k comments
---------
Co-authored-by: Manogna-Sree <elisetti.manognasree@multicorewareinc.com>
This is useful for testing for regressions on GCN with CDNA hardware.
With GGML_HIP_MMQ_MFMA=Off and GGML_CUDA_FORCE_MMQ=On we can conveniently test the GCN code path on CDNA. As CDNA is just GCN renamed with MFMA added and limited use ACC registers, this provides a good alternative for regression testing when GCN hardware is not available.
llvm with the amdgcn target dose not support unrolling loops with conditional break statements, when those statements can not be resolved at compile time. Similar to other places in GGML lets simply ignore this warning.
* remove redundant code in riscv
* remove redundant code in arm
* remove redundant code in loongarch
* remove redundant code in ppc
* remove redundant code in s390
* remove redundant code in wasm
* remove redundant code in x86
* remove fallback headers
* fix x86 ggml_vec_dot_q8_0_q8_0
* SYCL: Add set_rows support for quantized types
This commit adds support for GGML_OP_SET_ROWS operation for various
quantized tensor types (Q8_0, Q5_1, Q5_0, Q4_1, Q4_0, IQ4_NL) and BF16
type in the SYCL backend.
The quantization/dequantization copy kernels were moved from cpy.cpp
to cpy.hpp to make them available for set_rows.cpp.
This addresses part of the TODOs mentioned in the code.
* Use get_global_linear_id() instead
ggml-ci
* Fix formatting
ggml-ci
* Use const for ne11 and size_t variables in set_rows_sycl_q
ggml-ci
* Increase block size for q kernel to 256
ggml-ci
* Cleanup imports
* Add float.h to cpy.hpp
The pipeline member can be cast to VkPipeline.
This is a VkPipeline_T* on 64 bit but a uint64_t on 32 bit.
Cf. VK_DEFINE_NON_DISPATCHABLE_HANDLE documentation.
This commit adds support for MFMA instructions to MMQ. CDNA1/GFX908 CDNA2/GFX90a and CDNA3/GFX942 are supported by the MFMA-enabled code path added by this commit. The code path and stream-k is only enabled on CDNA3 for now as it fails to outperform blas in all cases on the other devices.
Blas is currently only consistently outperformed on CDNA3 due to issues in the amd-provided blas libraries.
This commit also improves the awareness of MMQ towards different warp sizes and as a side effect improves the performance of all quant formats besides q4_0 and q4_1, which regress slightly, on GCN gpus.
* feat: Add s_off as a parameter in the args struct
This may not be necessary, but it more closely mirrors the CUDA kernel
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* perf: Parallelize mamba2 SSM_SCAN metal kernel over d_state
This is a first attempt at optimizing the metal kernel. The changes here
are:
- Launch the kernel with a thread group of size d_state
- Use simd groups and shared memory to do the summation for the y
computation
When tested with G4 tiny preview, this shows roughly a 3x speedup on
prefill and 15% speedup on decode.
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* fix: Update logic to correctly do the multi-layer parallel sum
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* fix: Correctly size the shared memory bufer and assert expected size relationships
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* refactor: Compute block offsets once rather than once per token
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* feat: Use local variable for state recursion
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* feat: Use a secondary simd_sum instead of a for loop
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* feat: Add assertion and comment about relationship between simd size and num simd groups
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* feat: Parallelize of d_state for mamba-1
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* feat: Parallel sum in SSM_CONV
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* Revert "feat: Parallel sum in SSM_CONV"
After discussion with @compilade, the size of the parallelism here is
not worth the cost in complexity or overhead of the parallel for.
https://github.com/ggml-org/llama.cpp/pull/14743#discussion_r2223395357
This reverts commit 16bc059660c1c59e566628201c0ca2c20c9f4bc3.
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
* refactor: Simplify shared memory sizing
Branch: GraniteFourPerf
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
Co-Authored-By: Georgi Gerganov <ggerganov@gmail.com>
---------
Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Neither "g" nor "x" are valid portPos specifiers per the official
[graphviz documents](https://graphviz.org/docs/attr-types/portPos/):
> If a compass point is used, it must have the form "n","ne","e","se","s","sw","w","nw","c","_".
I tested locally for it to fall back to default portPos specifier if an
invalid portPos is specified. As a consequence, we can remove associated
code.
* musa: apply mublas API changes
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: update musa version to 4.2.0
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: restore MUSA graph settings in CMakeLists.txt
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: disable mudnnMemcpyAsync by default
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: switch back to non-mudnn images
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* minor changes
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: restore rc in docker image tag
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
---------
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* weight format to nz for 310p
* remove quant weight format to nz
* clean code
* fix
* make the conditions for converting weights to NZ format consistent
* clean code
* ggml/ggml-vulkan/test-backend-ops: adds CONV_2D for Vulkan
* ggml-vulkan: adds f32 scalar shader to compute 2D convolution directly
with gemm (no need for im2col),
* test-backend-ops: adds test_case_ref to check the validity/performance of ops
against reference implementations having different graphs, adds tests
* * Performance fixes: minimized branch divergence, uses collectives to
eliminate redundant calculation, macros removed.
* Kernel shared memory size check
* Updates test-backend-ops to support graphs for performance
measurement.
* * Apple/Win32 compile errors fixed
* Subgroup size used to determine tile size -> fixes llvmpipe errors.
* Collectives disabled by default.
* Intel support is disabled as the performance is poor.
* Conv2d enabled for Intel with disabled collectives, disabled for Apple
* test-backend-ops modifications are reverted
* Trailing spaces and missing override fixed.
* Triggering pipeline relaunch.
* Code formatted with .clang-format.
* CMake config: Create target only once
Fix error on repeated find_package(ggml).
For simplicity, check only for the top-level ggml::ggml.
* CMake config: Add CUDA link libs
* CMake config: Add OpenCL link libs
* CMake config: Use canonical find_dependency
Use set and append to control link lib variables.
Apply more $<LINK_ONLY...>.
* CMake config: Wire OpenMP dependency
This commit removes the inclusion of `<cstdlib>`.
The motivation for this change is that this source file does not seem to
use any functions from this header and the comment about `qsort` is a
little misleading/confusing.
* Fix Gemma3n not executed as CUDA_GRAPH on NVGPUs
Gemma3n uses Matrix-Matrix addition as part of their input processing,
wrongly triggering CUDA_GRAPH disablement on NVGPUs even when batch-size
of 1 is used.
* Exclude `project_per_layer_input` by matching node names
This ensures that all other graphs which don't exhibit this pattern do
not have their behavior changed.
* Revert unnecessary formatting changes
* Minimal setup of webgpu backend with dawn. Just prints out the adapter and segfaults
* Initialize webgpu device
* Making progress on setting up the backend
* Finish more boilerplate/utility functions
* Organize file and work on alloc buffer
* Add webgpu_context to prepare for actually running some shaders
* Work on memset and add shader loading
* Work on memset polyfill
* Implement set_tensor as webgpu WriteBuffer, remove host_buffer stubs since webgpu doesn't support it
* Implement get_tensor and buffer_clear
* Finish rest of setup
* Start work on compute graph
* Basic mat mul working
* Work on emscripten build
* Basic WebGPU backend instructions
* Use EMSCRIPTEN flag
* Work on passing ci, implement 4d tensor multiplication
* Pass thread safety test
* Implement permuting for mul_mat and cpy
* minor cleanups
* Address feedback
* Remove division by type size in cpy op
* Fix formatting and add github action workflows for vulkan and metal (m-series) webgpu backends
* Fix name
* Fix macos dawn prefix path
Remove un-necessary templates from class definition and packing functions
Reduce deeply nested conditionals, if-else switching in mnapck function
Replace repetitive code with inline functions in Packing functions
2 ~ 7% improvement in Q8 Model
15 ~ 50% improvement in Q4 Model
Signed-off-by: Shalini Salomi Bodapati <Shalini.Salomi.Bodapati@ibm.com>
* CUDA: add set rows for f32 and f16
* Review: change kernel params, use strides from host
* Use 1-d kernel
* Review: use int64_t for blockDim.x, rename nb->s for clarity
* vulkan: support SET_ROWS
Add variants of the copy_to_quant shader that do the SET_ROWS operation.
Change these shaders to spread the work across the workgroup.
The memory access pattern is probably not great (one thread per quant block),
but should be fine for now.
* vulkan: optimize set_rows
Larger workgroups for non-quant types.
Set "norepeat" (there is manual repeat logic).
Use fastmod.
* vulkan: allow unclamped loads in coopmat2 mul_mat_id shader
* vulkan: increase coopmat2 mul_mat_id tile size
* vulkan: optimize mat_mul_id row_ids search to batch loads, and port to coopmat1 path
* vulkan: use smaller FA row size when head size is large. applies to both scalar and CM2 paths (CM1 isn't used due to shared memory limits)
* ggml : add ggml_scale_bias
* ggml_vec_mad1_f32
* add more simd
* add CUDA
* sycl
* vulkan
* cann (placeholder)
* opencl
* will this fix cpu?
* fix cuda
* suggestions from coderabbit
* fix cann compile error
* vDSP_vsmsa
* rm __ARM_FEATURE_SVE
* use memcpy for op params
* make code looks more consistent
* use scalar for __ARM_FEATURE_SVE
* add x param to ggml_vec_mad1_f32
* vulkan: allow FA split_k with smaller KV values
* vulkan: spread split_k_reduce work across more threads
k_num can get rather large. Use the whole workgroup to reduce the M/L values.
Launch a thread for each element in the HSV dimension of the output. Helps a
lot for large HSV (like deepseek).
The fused operation was grabbing the epsilon value from the wrong place.
Add an env var to disable fusion.
Add some missing checks for supported shapes/types.
Handle fused rms_norm+mul in check_results.
* vulkan: Handle updated FA dim2/3 definition
Pack mask boolean and n_head_log2 into a single dword to keep the push
constant block under the 128B limit.
* handle null mask for gqa
* allow gqa with dim3>1
* kv-cache : use ggml_set_rows
ggml-ci
* graph : separate k and v indices
ggml-ci
* cont : remove redundant ifs
ggml-ci
* kv-cache : improve find_slot impl
* kv-cache : bounds-check when accessing slot_info indices
* kv-cache : add comments
ggml-ci
* ggml : add TODOs for adding GGML_OP_SET_ROWS support in the backends
ggml-ci
* llama : initial Mamba-2 support
* ggml : SIMD ggml_ssm_scan for Mamba-2
* ggml : improve ggml_mul speed when masking recurrent states
* llama : support running Mamba-Codestral-7B-v0.1
* llama : fix Mamba-2 conv state saving
* ggml : make the ggml_mul fast broadcast path more consistently formatted
* llama : remove unused variable
* llama : add missing break
* convert_hf : prefer SentencePiece tokenizer for Mamba-2 when present
The tokenzier.json of Mamba-Codestral-7B-v0.1 otherwise requires
workarounds to work correctly.
* llama : avoid redundant state copy for Mamba 1 and 2
* metal : attempt to adapt SSM_SCAN for Mamba-2
* metal : fix SSM_SCAN pipeline scope
* metal : use log and exp instead of log1pf and expf in SSM_SCAN
* metal : remove unused arguments for SSM_SCAN
The max index is 31, so trimming the arguments is necessary.
* metal : add back n_seqs to SSM_SCAN args
Whoops, this is needed for the offset in the concatenated output.
* metal : fix SSM_SCAN state head offset
* metal : fix wrong number of tokens per sequence in SSM_SCAN
* ggml : remove unused fast broadcast path in GGML_MUL
This was initially added because states were masked with ggml_mul,
but this is no longer done and so this "optimisation" is no longer
necessary, or at least not worth the additional code complexity.
* ggml : avoid multiply by D in GGML_OP_SSM_SCAN
This makes the weight buft detection in src/llama.cpp simpler.
* convert : transpose Mamba-2 A, D and reshape SSM_NORM
This breaks existing conversions of Mamba-2 models
to avoid some reshapes.
Not sure if it's a good idea,
but it makes the graph slightly cleaner.
* llama : more appropriate SSM_SCAN and SSM_CONV buft support checks
* convert : fix flake8 lint
* metal : fix confusion between ; and ,
* metal : add missing args for nb references in ssm_scan_f32_group
* metal : single-user mamba2 inference works
* kv-cache : remove const_cast when setting inputs for s_copy
And also fix multi-user inference for recurrent models
by using cell_id instead of i as the kv cell index
when populating s_copy.
* convert : avoid AutoConfig for Mamba and Mamba2 hparams
* kv-cache : allow context shift for recurrent models
* graph : fix recurrent state copies when avoiding copies
Works, but using lambda functions might not be that clean.
* ggml : fix mamba2 ssm scan when compiled with SVE
* ggml-cpu : reorder SVE FMA for consistency with other SIMD arches
* cuda : implement ssm scan for Mamba2
There is still room for improvement, but it works!
* cuda : adapt Mamba1 ssm scan to shape changes from Mamba2
* mamba : fix mismatched new and delete size for llm_build_mamba
Subclasses of llm_graph_context cannot have extra fields,
because the called destructor is not the one from the subclass.
This otherwise would cause problems when runnning Mamba-(1|2) inference
when compiled -DGGML_SANITIZE_ADDRESS=ON
* cuda : graceful fallback for Mamba-1 models with weird embd size
* CUDA: add softmax broadcast
* Pass by const ref
* Review: Use blockDims for indexing, remove designated initializers
* Add TODO for noncontigous input/output
* Add a callback that will be called just before abort. This allows apps without a console to display a message to the user and save data if needed.
* Return previous callback to allow callback chaining
* style fixes
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* ggml : add version function to get lib version
This commit adds a function `ggml_version()` to the ggml library that
returns the version of the library as a string.
The motivation for this is that it can be useful to be able to
programmatically check the version of the ggml library being used.
Usage:
```c
printf("GGML version: %s\n", ggml_version());
```
Output:
```console
GGML version: 0.0.2219
```
* ggml : add ggml_commit()
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Conv2D: Add CPU version
* Half decent
* Tiled approach for F32
* remove file
* Fix tests
* Support F16 operations
* add assert about size
* Review: further formatting fixes, add assert and use CPU version of fp32->fp16
* Update docker.yml
修改docker.yml文件中的内容使其停止周期性的运行该workflow，如果想要运行该workflow可以手动启动
* Remove redundant include path in CMakeLists.txt
The parent directory '..' was removed from the include directories for the ggml-cpu-feats target, to avoid unnecessary include paths.
* Enable scheduled Docker image builds
Uncomments the workflow schedule to trigger daily Docker image rebuilds at 04:12 UTC, improving automation and keeping images up to date.
* SYCL: disable faulty fp16 CPU exponent for now
* Revert "SYCL: disable faulty fp16 CPU exponent for now"
This reverts commit ed0aab1ec31b4eb4b0f275dd7acd41d96a375202.
* SYCL: disable faulty fp16 CPU exponent for now
* Fix logic of disabling exponent kernel
* implement unary REGLU/GEGLU/SWIGLU cpu ops
* relax constraints
* duplicate shape of source
* fix ggml_vec_geglu_f16
* special case gated ops
* implement unary REGLU/GEGLU/SWIGLU cuda ops
* tighten constraints again
* refactor into GGML_GLU_OP
* metal : add glu kernels
ggml-ci
* add CUDA_GLU_BLOCK_SIZE [no ci]
* more constraints and use 64bit ints
ggml-ci
* 64bit multiplication [no ci]
* implement swapped variants (cpu/cuda)
* update comment [no ci]
ggml-ci
* Vulkan: Add GLU ops and shaders
* SYCL: Implement fused kernel GEGLU, SWIGLU and REGLU for single up+gate
* ggml : implement GLU for split up/gate (llama/14181)
* implement GLU for split up/gate
* add tests for ggml_glu_split
* Vulkan: Implement glu_split logic and shader support
* add split to logging [no ci]
* SYCL: refactor element_size ops and add split up and gate support to gated kernels
* SYCL: switch GEGLU to use tanh approximation
---------
Co-authored-by: 0cc4m <picard12@live.de>
Co-authored-by: Akarshan <akarshan@menlo.ai>
* GGML: increase OP count in assertion
* Refactor: Optimize SYCL element-wise operations with unary function inlining
This commit refactors the SYCL element-wise operations to improve performance by:
- Inlining unary operations (sgn, abs, elu, gelu, silu, etc.) to reduce kernel launch overhead.
- Introducing helper functions `op_xxx` for each unary operation to encapsulate the logic.
- Replacing direct kernel calls with calls to these inlined functions.
- Using `__dpct_inline__` to encourage compiler inlining.
- Minor code cleanup and consistency improvements.
The changes aim to reduce kernel launch overhead and improve the overall efficiency of element-wise operations on SYCL devices.
* vulkan: Increase workgroup size for GLU, for performance (llama/14345)
* vulkan: Increase workgroup size for GLU, for performance
* vulkan: change GLU shaders to do one element per invocation rather than one row per workgroup
* merge fix
* metal : add support for split and swap
ggml-ci
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Co-authored-by: 0cc4m <picard12@live.de>
Co-authored-by: Akarshan <akarshan@menlo.ai>
Co-authored-by: Jeff Bolz <jbolz@nvidia.com>
* vulkan: Add fusion support for RMS_NORM+MUL
- Add a use_count to ggml_tensor, so we can detect if an output is used more than once.
- Change the ggml-vulkan rms_norm shader to optionally multiply by another tensor.
- Add detection logic and basic fusion logic in ggml-vulkan.
- Add some testing support for fusion. Rather than computing one node at a time, allow
for computing the whole graph and just testing one node's results. Add rms_norm_mul tests
and enable a llama test.
* extract some common fusion logic
* fix -Winconsistent-missing-override
* move ggml_can_fuse to a common function
* build fix
* C and C++ versions of can_fuse
* move use count to the graph to avoid data races and double increments when used in multiple threads
* use hash table lookup to find node index
* change use_counts to be indexed by hash table slot
* minimize hash lookups
style fixes
* last node doesn't need single use.
fix type.
handle mul operands being swapped.
* remove redundant parameter
---------
Co-authored-by: slaren <slarengh@gmail.com>
* CUDA: add bf16 and f32 support to cublas_mul_mat_batched
* Review: add type traits and make function more generic
* Review: make check more explicit, add back comments, and fix formatting
* Review: fix formatting, remove useless type conversion, fix naming for bools
* ggml : add ggml_set_rows
Add ggml_set_rows(a, b, c) which copies rows from 'b' into 'a' using
indices from 'c'.
ref: #8366
* use I64 for indices
* ggml : add repeat impl for i64
* ggml : add ggml_is_contiguous_rows
* ggml : ggml_set_rows support broadcast
* ggml : ggml_set_rows support quantized dst
ggml-ci
* ggml : support GGML_TYPE_F32 ".from_float" trait
* ggml : ggml_set_rows update comment + better index name
* tests : add ggml_set_rows
* metal : add ggml_set_rows implementation
ggml-ci
* ggml : simplify forward_dup_f32
* ggml : fix supports_op
* tests : add comment to set_rows
* ggml : leave the repeat_i64 for a separate PR
ggml-ci
* ggml : set_rows use std::min instead of MIN
* ggml : better error message for set_rows unsupported type
* metal : perform op->type check only once
* tests : more consistent implementation + more tests
ggml-ci
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add support for VK_EXT_debug_utils to add labels to Vulkan objects. In step 1 compute pipelines are getting labeled.
* remove #ifdef for debug utils and add queue marker.
* add "align corners" mode for bilinear upscale, and allow downscaling
* add ggml_interpolate, deprecate ggml_upscale_ext, pass in align-corners as bit-flag
* test-backend-ops: replace ggml_upscale_ext with ggml_interpolate, add test cases for downscale and align-corners
This commit renames the variable `best_mad` to `best_error` in the
`make_qkx2_quants` function.
The motivation for this is that the name `best_mad` can be somewhat
confusing if mean absolute deviation (MAD) is not in use.
* Add header and namespace to use enqueue_functions extension
* Convert submit and parallel_for to use new extension in convert.cpp
* Convert submit and parallel_for to use extension in ggml-sycl.cpp
* Convert submit and parallel_for to use extension in gla.cpp
* Convert submit and parallel_for in mmq.cpp
* Convert submit and parallel_for in mmvq.cpp
* Convert submit and parallel_for in remaining files
* Convert all simple parallel_for to nd_launch from enqueue_functions
extension
* Wrapping extension in general function
Create a general function that enable the enqueue_functions extension if
it is enable in the compiler, otherwise call the general SYCL function
to launch kernels.
---------
Signed-off-by: nscipione <nicolo.scipione@codeplay.com>
* Add PowerPC feature detection and scoring
* ggml-cpu: Implement GGML_CPU_ALL_VARIANTS for PowerPC
* ggml-cpu: Delay some initializations until function is called
When using GGML_BACKEND_DL=ON, these initializations might use
instructions that are not supported by the current CPU.
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* llama : add thread safety test
* llamafile : remove global state
* llama : better LLAMA_SPLIT_MODE_NONE logic
when main_gpu < 0 GPU devices are not used
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Update oneMath commit to merged PR https://github.com/uxlfoundation/oneMath/pull/669
which adds SYCL-Graph support for recording CUDA BLAS commands.
With this change the `MUL_MAT` tests now pass on DPC++ CUDA backends with SYCL-Graph
enabled. Prior to this change, an error would be thrown.
```
$ GGML_SYCL_DISABLE_GRAPH=0 ./bin/test-backend-ops -b SYCL0 -o MUL_MAT -p type_a=f16,type_b=f32,m=16,n=1,k=256,bs=\\[1,1\\],nr=\\[2
UR CUDA ERROR:
Value: 700
Name: CUDA_ERROR_ILLEGAL_ADDRESS
Description: an illegal memory access was encountered
Function: operator()
Source Location: $HOME/dpcpp/unified-runtime/source/adapters/cuda/queue.cpp:154
Native API failed. Native API returns: 2147483646 (UR_RESULT_ERROR_UNKNOWN)
Exception caught at file:$HOME/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp, line:3598, func:operator()
SYCL error: CHECK_TRY_ERROR((stream)->wait()): Meet error in this line code!
in function ggml_backend_sycl_synchronize at $HOME/llama.cpp/ggml/src/ggml-sycl/ggml-sycl.cpp:3598
$HOME/llama.cpp/ggml/src/ggml-sycl/../ggml-sycl/common.hpp:118: SYCL error
Could not attach to process. If your uid matches the uid of the target
process, check the setting of /proc/sys/kernel/yama/ptrace_scope, or try
again as the root user. For more details, see /etc/sysctl.d/10-ptrace.conf
ptrace: Operation not permitted.
No stack.
The program is not being run.
```
* ggml-cpu: Factor out feature detection build from x86
* ggml-cpu: Add ARM feature detection and scoring
This is analogous to cpu-feats-x86.cpp. However, to detect compile-time
activation of features, we rely on GGML_USE_<FEAT> which need to be set
in cmake, instead of GGML_<FEAT> that users would set for x86.
This is because on ARM, users specify features with GGML_CPU_ARM_ARCH,
rather than with individual flags.
* ggml-cpu: Implement GGML_CPU_ALL_VARIANTS for ARM
Like x86, however to pass around arch flags within cmake, we use
GGML_INTERNAL_<FEAT> as we don't have GGML_<FEAT>.
Some features are optional, so we may need to build multiple backends
per arch version (armv8.2_1, armv8.2_2, ...), and let the scoring
function sort out which one can be used.
* ggml-cpu: Limit ARM GGML_CPU_ALL_VARIANTS to Linux for now
The other platforms will need their own specific variants.
This also fixes the bug that the the variant-building branch was always
being executed as the else-branch of GGML_NATIVE=OFF. The branch is
moved to an elseif-branch which restores the previous behavior.
This change moves the command pool/buffer tracking into a vk_command_pool
structure. There are two instances per context (for compute+transfer) and
two instances per device for operations that don't go through a context.
This should prevent separate contexts from stomping on each other.
Use the same descriptor set layout for all pipelines (MAX_PARAMETER_COUNT == 8)
and move it to the vk_device. Move all the descriptor pool and set tracking to
the context - none of it is specific to pipelines anymore. It has a single vector
of pools and vector of sets, and a single counter to track requests and a single
counter to track use.
* ggml : disable warnings for tests when using MSVC
This commit disables warnings for tests on windows when using MSVC.
The motivation for this is that this brings the build output more
inline with what Linux/MacOS systems produce.
There is still one warning generated for the tests which is:
```console
Building Custom Rule C:/ggml/tests/CMakeLists.txt
cl : command line warning D9025: overriding '/DNDEBUG' with '/UNDEBUG'
[C:\ggml\build\tests\test-arange.vcxproj]
test-arange.cpp
test-arange.vcxproj -> C:\ggml\build\bin\Release\test-arange.exe
```
* ggml : fix typo in tests disable list
This commit removes the unused `ggml_context_container` structure from
the ggml library. It looks like the usage of this struct was removed in
Commit 4757fe18d56ec11bf9c07feaca6e9d5b5357e7f4 ("ggml : alloc
ggml_contexts on the heap (whisper/2525)").
The motivation for this changes is to improve code clarity/readability.
This commit adds the examples in the "list" of targets to ignore MSVC
warnings.
The motivation for this is that currently the examples generate a number
of warnings that are ignore/disabled for the core ggml project. This
makes for a cleaner output when building.
* Simplify the environment variable setting to specify the memory pool type.
* Adjust the GGML_CANN_ASYNC_MODE setting to accept yes, enable, 1, or on (case-insensitive) as valid options.
* update
* fix CI
* update
* delete whitespace
* fix according to review
* update CANN.md
* update CANN.md
* Add Reorder to Q6_K mmvq implementation
* Address PR comments: clean up comments
* Remove unused parameter after refactoring q4_k
* Adding inline to function and removing unnecessary reference to int
---------
Signed-off-by: nscipione <nicolo.scipione@codeplay.com>
* SYCL: Implement few same quantized type copy kernels
* Use memcpy for copying contiguous tensors
ggml-ci
* feat(sycl): add contiguous tensor copy support and device checks
Adds a memcpy path for contiguous tensors of the same type to optimize data transfer. Updates device support checks to recognize contiguous tensor operations, improving compatibility and performance.
* refactor: replace specific block copy functions with template
The changes replace multiple redundant block copy functions (e.g., cpy_block_q8_0_q8_0, cpy_block_q5_0_q5_0) with a single templated function cpy_blck_q_q. This reduces code duplication by using a generic template that works for any block type, improving maintainability while preserving the same functionality. The template is instantiated with specific block types (e.g., block_q8_0) where needed.
* Exclude BF16 support for COPY tensors for now
ggml-ci
* perf: adjust SYCL copy kernel block sizes for efficiency
Use ceil_div to ensure full element coverage and update nd_range parameters to better align with SYCL block sizes, improving parallelism and device utilization in copy operations.
* * ggml-vulkan: adds op CONV_TRANSPOSE_1D
* test-backend-ops: adds more spohisticated tests for CONV_TRANSPOSE_1D
* Missing barrier added to shader.
Number of additional tests reduced to 108.
* * Fixes typo in variable name.
* Removes extra whitespaces.
* Adds int64->int32 casts to prevent possible warnings.
* Problem size reduced in tests to pass tests with llvmpipe.
* supports_op condition moved from unintended position
* This is not needed by the normal use where the result is read
using `tensor_get`, but it allows perf mode of `test-backend-ops`
to properly measure performance.
Some systems report the CPU implementation as "Power11" instead of "POWER11".
The existing CMake logic uses a case-sensitive regular expression to extract
the CPU generation, which fails when the casing doesn't exactly match "POWER".
This patch provides a fix by first converting the string to uppercase before applying the regex.
Signed-off-by: root <root@rheldb2v.pperf.tadn.ibm.com>
Co-authored-by: root <root@rheldb2v.pperf.tadn.ibm.com>
* threading: support for GGML_SCHED_PRIO_LOW, update thread info on Windows to avoid throttling
We talked about adding LOW priority for GGML threads in the original threadpool PR.
It might be useful for some cases to avoid contention.
Latest Windows ARM64 releases started parking (offlining) the CPU cores
more aggresively which results in suboptimal performance with n_threads > 4.
To deal with that we now disable Power Throttling for our threads for the NORMAL
and higher priorities.
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* threading: disable SetThreadInfo() calls for older Windows versions
* Update tools/llama-bench/llama-bench.cpp
Co-authored-by: Diego Devesa <slarengh@gmail.com>
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* 1. add "integrated" in ggml_cuda_device_info for distinguish whether it is Intergrate_gpu or discrete_gpu
2. Adjust the func:"ggml_backend_cuda_device_supports_buft" for this new feature
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Adjusted code indentation
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Fixed incorrect setting of variable types
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Adjusted the judgment logic
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* add a host_buft assert in case of integrated_cuda_device with func:'evaluate_and_capture_cuda_graph()'
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Add a defensive security assert
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Adjusted the support judgment logic.
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* revoke the suggest commit changes due to it's not applicable in jetson_device
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Add parentheses to enforce operator precedence
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* Update ggml/src/ggml-cuda/ggml-cuda.cu
Fix ci bug: add a spaces
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---------
Co-authored-by: yangxiao <yang_xl@tju.edu.cn>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
Co-authored-by: yangxiao <yangxl_zz@qq.com>
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* SYCL: Add mrope kernel
* feat: Optimize rope operations with vectorization
Uses `sycl::vec` to load and store two elements at a time,
significantly improving performance in `rope_norm`,
`rope_neox`, and `rope_multi`. This reduces the number of memory
accesses and leverages SIMD instructions for faster execution.
* Use ceil_div
* cmake: Define function for querying architecture
The tests and results match exactly those of src/CMakeLists.txt
* Switch arch detection over to new function
* SYCL: Add non contiguous input support to norm kernel
* refactor and add RMS_NORM non contiguous input support
ggml-ci
* restore subgroup reduction for multi-subgroup thread blocks in norm kernels
* Swap grid dims of nsamples and nrows
ggml-ci
* Revert "Swap grid dims of nsamples and nrows"
This reverts commit 43be2d657fec7f7fba54e2cd154106bc0fc45adf.
* restore not required changes
ggml-ci
* address review comments: change it to more like SYCL
* Use a common function to calculate offset
* remove wrap around logic for handling broadcasts
* remove static from calculate_offset fn and use ceil_div
* cann: add the basic FA support
* cann: update the readme
* cann: update the FlashAttention with PSEShift
* cann: update the input parameters in FA
* cann: update the alibi with max_bias
* cann: add the constrints of softcap
* cann: update the docs CANN.md
* cann: update the docs CANN.md
* cann: fix typo of CANN.md
* cann: add some comments and update the CANN.md
* cann: update the CANN.md
* cann: update the inner precise for fusedInferAttention
* cann: update the constraints of flash_attn_ext on ggml-cann.cpp
* cann: clean the whitespace
* cann: clean the whitespace
* cann: add a new endline
* opencl: Add support for multiple devices
... but limited to one platform. A platform with a GPU will be preferred.
Additionally:
* Filter out devices that lack capabilities needed by the backend
implementation (half support, OpenCL 2.0+, etc).
* Make ggml_backend_opencl_reg() thread-safe.
* fixup: fix an error in sync_with_other_backends
... when there is only one OpenCL device available.
* opencl: fix couple crashes
* fix kernel launches failed on devices which do not support
non-uniform work-groups. When non-uniform work-groups are not
supported, set `local_work_size` to NULL (= let driver choose the
work-group sizes). This patch does not cover everything - just the
cases tested by test-backend-ops.
* fix sub-buffer creation failed due to `cl_buffer_region::origin` not
being aligned to `CL_DEVICE_MEM_BASE_ADDR_ALIGN`.
* OpenCL: query non-uniform WG sizes only on OpenCL 3.0+
* musa: fix build warning (unused parameter)
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: upgrade MUSA SDK version to rc4.0.1
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: use mudnn::Unary::IDENTITY op to accelerate D2D memory copy
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* Update ggml/src/ggml-cuda/cpy.cu
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* musa: remove MUDNN_CHECK_GEN and use CUDA_CHECK_GEN instead in MUDNN_CHECK
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
---------
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* Remove mmap workaround on windows
After some testing I found that mmap is supported on windows and for
many GPUs on Linux. Therefore I remove the workaround for windows since
it is not necessary.
* Update llama-bench README
SYCL backend introduced a workaround that allows execution of
llama-bench also without specifying `--mmp 0` flag
This shader uses coopmat1 to do the Q*K^T multiply. The P*V multiply is more
difficult for various reasons so I haven't done it. Performance for this
shader is around 2.5x better than for the scalar shader when doing prompt
processing. Some of the benefit may be from other optimizations like staging
through shared memory, or splitting by rows.
* batched-bench : fix pp batch contents
* metal : optimize multi-sequence FA vec kernel
ggml-ci
* metal : use FA-vec kernel up to batch size 20
ggml-ci
* llama/ggml: add LLM training support
more compact progress bar
llama_save_model_to_file
llama_opt_param_filter
ggml_graph_dup force_grads
refactor ggml_opt, fix test-opt
* remove logits_all
* refactor CUDA implementation for ACC
* reset graph at beginning of opt period
* ggml-cpu: Integrate fp32=bf16xbf16 SME KleidiAI kernel
Signed-off-by: Dan Johansson <dan.johansson@arm.com>
* * code review fixes
Signed-off-by: Dan Johansson <dan.johansson@arm.com>
* * adds a comment that clarifies barrier usage
Signed-off-by: Dan Johansson <dan.johansson@arm.com>
---------
Signed-off-by: Dan Johansson <dan.johansson@arm.com>
Co-authored-by: Charles Xu <charles.xu@arm.com>
* vulkan: scalar flash attention implementation
* vulkan: always use fp32 for scalar flash attention
* vulkan: use vector loads in scalar flash attention shader
* vulkan: remove PV matrix, helps with register usage
* vulkan: reduce register usage in scalar FA, but perf may be slightly worse
* vulkan: load each Q value once. optimize O reduction. more tuning
* vulkan: support q4_0/q8_0 KV in scalar FA
* CI: increase timeout to accommodate newly-supported tests
* vulkan: for scalar FA, select between 1 and 8 rows
* vulkan: avoid using Float16 capability in scalar FA
* sycl : Implemented reorder Q4_0 mmvq
Signed-off-by: Alberto Cabrera <alberto.cabrera@codeplay.com>
* sycl : Fixed mmvq being called when reorder is disabled
* sycl : Improved comments in the quants header
Signed-off-by: Alberto Cabrera <alberto.cabrera@codeplay.com>
* Use static_assert
* safe_div -> ceil_div
* Clarify qi comment
* change the reorder tensor from init to execute OP
* dbg
* Undo changes to test-backend-ops
* Refactor changes on top of q4_0 reorder fix
* Missing Reverts
* Refactored opt_for_reorder logic to simplify code path
* Explicit inlining and unroll
* Renamed mul_mat_algo enum for consistency
---------
Signed-off-by: Alberto Cabrera <alberto.cabrera@codeplay.com>
Co-authored-by: romain.biessy <romain.biessy@codeplay.com>
This assert fired running Qwen_Qwen3-30B-A3B-Q2_K.gguf:
GGML_ASSERT(nei0 * nei1 <= 3072);
The tensor is 8 x 512. Increase this array size to accommodate.
This patch upstreams llamafile's cpu matrix multiplication kernels for ppc64le using MMA builtins for BF16 data type.
This change results in 9x - 40x gains
in total speed S t/s (ie all tokens/total time), across various batch sizes tested using llama-batched-bench benchmark.
The patch is tested with Meta-Lllama-3-8B,
and Mistral-7B models (BF16 models generated by using llama-quantize from corresponding FP32 models) on an IBM POWER10 machine.
Signed-off-by: Shalini Salomi Bodapati <Shalini.Salomi.Bodapati@ibm.com>
The following scenario will cause an assertion failure in the graph
allocator:
- Build and allocate a graph containing a tensor with a non-NULL data
pointer
- Build and allocate a new graph where that data is NULL
Result:
ggml-alloc.c:819: GGML_ASSERT(talloc->buffer_id >= 0) failed
This happens during revalidation because we think that memory should
have been previously allocated based on the current graph but in
reality the previous graph was different. In this situation, we
should do a full reallocation pass.
* vulkan: Add bfloat16 support
This adds bfloat16 matrix multiply support based on VK_KHR_shader_bfloat16.
The extension is required for coopmat multiply support, but matrix-vector
multiply trivially promotes bf16 to fp32 and doesn't require the extension.
The copy/get_rows shaders also don't require the extension.
It's probably possible to fall back to non-coopmat and promote to fp32 when
the extension isn't supported, but this change doesn't do that.
The coopmat support also requires a glslc that supports the extension, which
currently requires a custom build.
* vulkan: Support bf16 tensors without the bf16 extension or coopmat support
Compile a variant of the scalar mul_mm shader that will promote the bf16
values to float, and use that when either the bf16 extension or the coopmat
extensions aren't available.
* vulkan: bfloat16 fixes (really works without bfloat16 support now)
* vulkan: fix spirv-val failure and reenable -O
* ggml : remove MSVC warnings pragmas
This commit removes the MSVC-specific pragmas as these are now handled
in CMakeLists.txt.
* whisper : remove MSVC warning pragmas
This commit removes the MSVC-specific pragmas. These are now handled in
the CMakeLists.txt file.
Build fails with compilation error on power pc.
This patch fixes the same.
Tested with unit tests run via
--build <build_dir> && cd <build_dir> && make test
Signed-off-by: Shalini Salomi Bodapati <Shalini.Salomi.Bodapati@ibm.com>
* fix(rpc): Improve input validation and error handling
The `rpc-server` was vulnerable to Denial of Service attacks via
several RPC commands (`SET_TENSOR`, `GRAPH_COMPUTE`, etc.). Malformed
messages could trigger failed assertions (e.g., invalid `ggml_type`)
or out-of-bounds reads/writes leading to `GGML_ABORT` calls,
crashing the server process.
This PR introduces robust input validation and replaces `abort()`
calls with graceful error handling:
- **Type Validation:** `deserialize_tensor` now checks if the
`tensor->type` is within the valid `GGML_TYPE_COUNT` range
*before* calling `ggml_new_tensor_4d`. Returns `nullptr` on
invalid type.
- **Bounds Checks:** Replaced `GGML_ABORT` in `set_tensor`,
`set_tensor_hash`, and `get_tensor` handlers with error
logging and returning `false` when data/offset parameters
are out of buffer bounds.
- **Size Checks:** Added safe arithmetic checks (for overflow) in
`graph_compute` when calculating required message sizes based
on client-provided `n_nodes` and `n_tensors`. Returns early
if the reported sizes conflict with the actual message size or
would lead to overflow.
- **Error Propagation:**
- `create_node` now checks for `nullptr` return values from
`deserialize_tensor` and its recursive calls, propagating
`nullptr` upwards on failure. Uses `find` instead of `at`
for safer map access.
- `copy_tensor` now checks for `nullptr` from `deserialize_tensor`
and sets the response status to failure if deserialization
or bounds checks fail.
- `graph_compute` now checks for `nullptr` return from
`create_node` and returns failure status correctly. The final
return value now reflects the actual computation status.
These changes improve the RPC server's resilience
against malformed client requests, preventing crashes and ensuring
errors are handled more gracefully.
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* refactor(rpc): address pr comments
removed comments and unnecessary returns
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* refactor(rpc): ambiguous nullptr from create_node
rpc_server::create_node could previously return nullptr if the input ID
was 0 (valid) or if an internal error (deserialization, recursion
failure) occurred (invalid). This ambiguity made error handling
difficult for the caller (`graph_compute`).
This commit clarifies the meaning of nullptr:
- `graph_compute` now checks if the input 'id' was non-zero when
`create_node` returns nullptr, correctly identifying failures
versus intentional null links.
- `create_node` avoids recursive calls for zero IDs and propagates
nullptr unambiguously on failure during recursion.
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* refactor(rpc): initial zero check in create_node
The caller (`graph_compute`) already checks `id != 0` when handling
a `nullptr` return from `create_node`, correctly distinguishing
intentional null links from actual errors. This makes the initial
`if (id == 0)` check redundant.
Also removes the log message when a tensor ID is not found in the
provided map which was added in this branch.
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* fix(rpc): Handle get_alloc_size failure in server
Check the return value of `server.get_alloc_size` in the RPC server
loop. If the call fails, return early to close the connection.
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* refactor(rpc): input size validation in graph_compute
Removes detailed, step-by-step size calculations and overflow
checks in favor of simpler direct comparisons, assuming 64-bit
overflow is unlikely.
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* refactor(rpc): remove extra status code setting
Removes the explicit setting of `response.result = GGML_STATUS_FAILED`
when `create_node` returns `nullptr` within `graph_compute`.
Primary signal is the `false` return value in case of failure.
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* refactor(rpc): remove redundant check for tensor->type
Breaks CI on ubuntu-cpu-make. Tensor type is uint32_t, thus
the check is not needed.
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
---------
Signed-off-by: Ville Vesilehto <ville@vesilehto.fi>
* SYCL: Add all missing unary kernels
ggml-ci
* decouple kernel launch range from data size using strided loop
* use ciel_div helper for num_blocks
ggml-ci
* clean auto imported header files
RPC_CMD_SET_TENSOR always returns an empty response and we send this 4
times per token. We can improve TG speed if we don't wait for this empty
response.
The performance impact of this change depends on the network latency.
This commit adds a check to makes sure that the target exists before
trying to add compile options to ignore warnings when using MSVC.
The motivation for this is currently the build is broken depending on
the cmake options provided. With this fix it should be possible to build
even if the targets are not actually available.
Refs: https://github.com/ggml-org/whisper.cpp/pull/3090#issuecomment-2842760104
* whisper: suppress Windows compiler warnings
This commit disables compiler warnings on window using MSVC.
The motivation for these changes is that some compilers generate
warnings for these conversion, for example Windows MSVC, and
there are quite a few of them. This makes it a little difficult to
spot new warnings that may be introduced and also can be difficult
for users/embedders of ggml where these warnings are hard to separate
from their own warnings.
* squash! whisper: suppress Windows compiler warnings
Move ggml related warnings into ggml. This commit also fixes the
indentation and adds a missing whitespace to the if statement.
* tune matmul for gcn
* this one is more power efficient
* Update ggml/src/ggml-vulkan/ggml-vulkan.cpp
Co-authored-by: 0cc4m <picard12@live.de>
* disable this tune for the proprietary driver
---------
Co-authored-by: 0cc4m <picard12@live.de>
Add RPC_CMD_HELLO for getting the version of the protocol implemend by
the server. Follow the semantic versioning rules at https://semver.org
Hopefully this bring better user experience when we make breaking
changes at the protocol level and avoid issues like #12465
* graph : make mla compatible with FA
* metal : add exp FA kernels for DeepSeek models
ggml-ci
* llama : minor naming updates
ggml-ci
* ggml : disable FA for DS head sizes
* tests : add FA tests for MLA shapes
ggml-ci
Submit operators using asynchronous threads to improve performance.
Use the environment variable GGML_CANN_ASYNC_MODE to control whether
asynchronous submission is enabled. It is disabled by default.
Testing shows a 10%–20% performance improvement in scenarios with
small parameter sizes, especially in quantized models.
The grouped query attention optmization doesn't require a power of two ratio,
the only thing relying on it was the modulo operation written as bitwise &.
split_k need not depend on gqa_ratio - enable it any time there's only one
workgroup in the X dimension. The shader gets the split index from the x coord,
and multiple workgroups in the X dimension (pre-split) indicates a larger
FA operation that wouldn't need splitting.
Replace compile-time `GGML_HIP_UMA` with environment variable `GGML_CUDA_ENABLE_UNIFIED_MEMORY`. This unifies the usage on NVIDIA and AMD GPUs, and allows a single binary to be shared between integrated and dedicated GPUs.
Multiple optional memory pools are provided for CANN, including VMM,
priority queue-based, and traditional memory pools.
1.When the memory pool is available and GGML_CANN_DISABLE_VMM_POOL
is not defined, the VMM pool is selected by default.
2.Otherwise, if GGML_CANN_ENABLE_BUF_PRIO_POOL is defined,
the priority queue-based memory pool is used.
3.If neither condition is met, the default memory pool is used.
The current usage of the SYCL-Graph extension checks for
the `sycl_ext_oneapi_graph` device aspect. However, it is also
possible to support `sycl_ext_oneapi_limied_graph` devices that
don't support update
* SYCL: Add fp16 support to some elementwise OP kernels
* remove comment
ggml-ci
* Use static_cast directly
* remove not needed cast from tanh
* Use static cast and remove unneeded castings
* Adjust device_support_op for unary OPs
* Use cast_data and typed_data struct to deduplicate casting code
* [CANN] Support ELU and CONV_TRANSPOSE_1D
* [CANN]Modification review comments
* [CANN]Modification review comments
* [CANN]name adjustment
* [CANN]remove lambda used in template
* [CANN]Use std::func instead of template
* [CANN]Modify the code according to the review comments
---------
Signed-off-by: noemotiovon <noemotiovon@gmail.com>
q4_k and q5_k had a lot of redundant global loads where the same 16B of
scale information is repeatedly loaded and decoded during each loop iteration.
This change restructures the loops to more explicitly iterate over whole
blocks in the outer loop (with unrolled inner loop) and to copy/decode the
scale data into shared memory once at the start of each outer loop. The copy
is pipelined so the scale load from global memory is relatively cheap.
This improves q4_k/q5_k model prompt processing performance by around 5-7%.
I briefly tried applying this to q6_k and q4_0, and it didn't help for q6_k
and hurt for q4_0.
The big "else" path in mul_mm_cm2.comp that had all the clamped/unclamped
variants isn't used as often as it originally was (e.g. due to the padded_N
change), so I trimmed it down to offset some of the new complexity of the
semi-manual loop unrolling.
* ggml : FA supports F32 V
* graph : cast KV to F16 when the KV cache is not used
ggml-ci
* server : add test that exercises embeddings with FA enabled
ggml-ci
nem1 must be a multiple of GGML_KQ_MASK_PAD, and GGML_KQ_MASK_PAD is a multiple
of the number of rows in the matrix. The KV dim is a multiple of the number of
columns for the aligned shader.
There seems to be a bubble waking up from waitForFences, which costs a few
percent performance and also increased variance in performance. This change
inserts an "almost_ready" fence when the graph is about 80% complete and we
waitForFences for the almost_ready fence and then spin (with _mm_pauses) waiting
for the final fence to be signaled.
* Prefer vector flash decoding kernel for Gemma models
Vector flash decoding kernel was not being picked for models with head dimension 256. Gemma models are in this category.
Removing this limit improves e2e performance by upto 12% in gen phase throughput for Gemm models.
* Update ggml/src/ggml-cuda/fattn.cu
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* CUDA: Simplify and improve CUDA graphs through use of indirect copy pointers
Previously there was complexity in the CUDA graphs implementation due
frequently changing parameters to copy kernels associated with K and V
cache pointers. This patch simplifies by using indirection to avoid
such parameters frequently changing, avoiding the need for frequent
graph updates.
Fixes#12152
* Addressed comments
* fix HIP builds
* properly sync to stream
* removed ggml_cuda_cpy_fn_ptrs
* move stream sync before free
* guard to only use indirection with graphs
* style fixes
* check for errors
---------
Co-authored-by: slaren <slarengh@gmail.com>
When using group query attention, we have one workgroup per KV batch and this
can be very few workgroups (e.g. just 8 in some models). Enable split_k to
spread the work across SMs. This helps a lot when the KV cache is large.
When adjacent batches of Q share the same batches of K/V, batch them into
the same workgroup. For example, when:
dst(128,32,1,1) = FA(q(128,1,32,1), k(128,16640,8,1), v(128,16640,8,1))
previously we would run 32 workgroups computing 1 result each, now we will
run 8 workgroups computing 4 results each.
This doesn't directly translate to better performance (at least when you have
>=32 SMs), but in a subsequent change I'll enable split_k which will scale much
better with 4x fewer workgroups.
* add bf16 support
* use convert_from_bf16_cuda instead of convert_unary_cuda for f32
* revert 7ec5085
* move functionality into convert_unary with constexpr
* cpu: refactor SIMD mappings and vectorized op functions into separate files
* Fix warning for ggml_float to float
* Fix warnings
* cpu: move all the operations (except mul_mat) to a separate c++ file
* fix whitespace
* Update ggml/src/ggml-cpu/vec.h
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* Fix PR comments - use GGML_UNUSED, use cassert in ops.cpp
* Reverse the order of import for ops.h and vec.h, to match what was present in ggml-cpu.c previously
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* Rename oneMKL Interface to oneMath
* Use oneMath for Intel vendor
* Rename occurences to mkl
* clang-format
* Silence verbose warnings
* Set oneMath HIP_TARGETS
* Fix silence warnings
* Remove step to build oneMath from build instructions
* Use fixed oneMath version
* Remove INTEL_CPU
* Fold CMake oneDNN conditions
* Use Intel oneMKL for Intel devices
* Improve CMake message
* Link against MKL::MKL_SYCL::BLAS only
* Move oneMath documentation to Nvidia and AMD sections
If users already set CMAKE_C_COMPILER_LAUNCHER globally, setting it in
cmake again will lead to conflict and compile fail.
Signed-off-by: Jay <BusyJay@users.noreply.github.com>
* ggml : FA with different K, V head sizes (CPU)
ggml-ci
* metal : add FA with HS=192
* metal : extend FA to support different K and V head sizes
ggml-ci
* metal : add FA vector kernels for heads K 192 and V 128
ggml-ci
* ggml : restrict op on other backends to equal head sizes
ggml-ci
* metal : optimize FA-vec kernel
ggml-ci
* metal : FA remove mq registers
* metal : improve MoE mul_mat_id condition
ggml-ci
* metal : fix comments + remove unnecessary addition
ggml-ci
* metal : avoid too much shared memory usage with mul_mat_id
ggml-ci
* vulkan: fix coopmat shader generation when cross-compiling
Previously the status of coopmat{,2} support isn't passed to the
vulkan-shaders-gen project building on the host, which leads to build
failure because of the cross-compiling code expecting coopmat{,2}
shaders that didn't get generated.
Fix this by passing the coopmat{,2} support status to vulkan-shaders
subproject.
Signed-off-by: Icenowy Zheng <uwu@icenowy.me>
* Only call coop-mat shaders once
* Fix whitespace
---------
Signed-off-by: Icenowy Zheng <uwu@icenowy.me>
Co-authored-by: bandoti <141645996+bandoti@users.noreply.github.com>
This patch enables usage of MMA when one of the
dimensions of the matrix(ie either M or N) is 1. This
is useful in case of token generation where N < 2.
The concept of 'GEMV Forwarding' is used where when one
of the matrix has a single row/column, the elements are
broadcasted, instead of using packing routine to prepack
the matrix elements.
This change results in 5% - 15% improvement in total
speed(ie all tokens/total time), across various batch
sizes. This is in comparision with the corresponding
dot product implementation.
The patch is tested with FP32 models of Meta-Lllama-3-8B,
Mistral-7B, Llama-2-7B-chat-hf on a IBM POWER10 machine.
Signed-off-by: Amrita H S <amritahs@linux.vnet.ibm.com>
* rpc : send hash when tensor data is above some fixed threshold
ref #10095
* rpc : put cache under $HOME/.cache/llama.cpp
* try to fix win32 build
* another try to fix win32 build
* remove llama as dependency
This change upstreams llamafile's cpu matrix
multiplication kernels for ppc64le ISA using MMA
builtins. This patch handles matrix multiplication
between quantised datatypes, block_q4_0 and
block_q8_0.
This change results in 5% - 50% improvement
in total speed(ie all tokens/total time), across
various batch sizes.
The patch is tested with Meta-Lllama-3-8B,
Mistral-7B, Llama-2-7B-chat-hf models on a
IBM POWER10 machine.
Signed-off-by: Amrita H S <amritahs@linux.vnet.ibm.com>
The OOB calculation could be wrong if the last iteration was during one of
the unrolled loops. Adjust the unrolling counts to avoid this. Add a couple
new backend tests that hit this failure on NVIDIA GPUs.
* tests: add mul_mat perf/functional tests for p021/nc vulkan shaders
* vulkan: Optimize mul_mat_vec p021 and nc shaders.
These shaders are used in attention calculations, and when the KV cache grows
large they start to dominate the run time. For the nc shader (which is called
with large 'k' dimension), use unrolling and vector loads. For the p021 shader
(which is called with large 'm' and small 'k' dimensions), take advantage of
grouped query attention to reuse loads from the A matrix for the whole group,
and reduce the number of workgroups (too much overhead from tiny dispatches).
Using subgroupAdd in the p021 shader also helps, use that conditionally.
* [SYCL] Fix build on Windows when ccache enabled (llama/9954)
* take effect only on windows and force it to icl
---------
Co-authored-by: Romain Biessy <romain.biessy@codeplay.com>
* Add block interleaving support for Q4_K quantization
* Remove whitespaces and fix CI/CD issues
* Update pointer of bsums from int16_t to const int16_t
* Add vector version of quantize_q8_K_4x8 function
* Update code formatting based on review comments
- Find out active blocks per SM using cudaOccupancyMaxActiveBlocksPerMultiprocessor API. Use this value to determine the optimal parallel_blocks value.
- Prefer vector flash attention kernels over MMA kernel for BS=1
Fixes Issue: #12182
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* ci: add visionOS build workflow
Add a new GitHub Actions workflow for building on visionOS with CMake and Xcode.
* ggml: Define _DARWIN_C_SOURCE for visionOS to fix missing u_xxx typedefs
* ci: remove define hacks for u_xxx system types
---------
Co-authored-by: Giovanni Petrantoni <7008900+sinkingsugar@users.noreply.github.com>
I've been seeing significantly worse performance for tg with flash attention
enabled vs disabled, and it seems to be related to the submit heuristic.
Change the heuristic to check how many bytes worth of weight matrix are
used and flush every 100MB, and ramp up after the first few submits.
This seems to resolve the issue, and also increases perf for non-FA a bit.
* opencl: more profiling timing
* opencl: generate trace for profiling
* opencl: reduce profiling overhead
* Populate profiling timing info at the end rather than after each
kernel run
* opencl: fix for chrome tracing
* Enable CUDA Graph on CTK < 12.x
`cudaGraphExecUpdate` API was changed on 12.x. For this reason CUDA graph support was disabled on older CUDA toolkit. This change enables CUDA support in CTK version < 12.x by using older API if CTK < 12.x.
* Fix compilation errors with MUSA
* Disable CUDA Graph for MUSA
* cmake: Factor out compiler flag function from ggml
llama.cpps's build requires it, too, and we may want to make use of it
without add_subdirectory(ggml).
* cmake: Enable building against system ggml
This facilitates package maintenance for Linux distributions, where the
libggml library most likely will be shipped as an individual package
upon which a llama.cpp package depends.
When fattn-wmma was ported over to warp64 various bits that also touch fattn-vec where converted to
selectable warp size, however the fattn-vec kernels dont work with 64 wide warps for now, so we need
to avoid launching them with parameters for warp64
refactor mmqv to unify the calculation of nwarps and rows per block between host and device code.
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This patch nudges the llama.cpp a bit to be supported on PoCL which
doesn't support OpenCL C CL2.0. The issue is solved by querying the
device for the supported OpenCL C versions and using the highest one
available.
This commit updates the compilation of default.metallib to skip the
intermediate .air (Apple Intermediate Representation) file.
The motivation for this change is to simplify the custom command a
little and avoid generating and then removing the .air file.
Nothing installs to it yet, so when attempting to use the cmake package,
set_and_check() triggers an error if the directory doesn't already exist
for other reasons.
This commit adds debug level logging for the native build options and
variables to ggml/CMakeLists.txt.
The motivation for this is that it can be useful to see the effective
result of `GGML_NATIVE`, `GGML_NATIVE_DEFAULT`, and `INS_ENB` for a
cmake build. I've found myself adding similar logging a few times now,
so I thought it might be a good idea to add this.
Example output, specifying `-DCMAKE_MESSAGE_LOG_LEVEL=DEBUG` when
running cmake produces the following output:
```console
-- GGML_NATIVE : OFF
-- GGML_NATIVE_DEFAULT : OFF
-- INS_ENB : OFF
```
This commit updates the command.wasm example by adding a server.py script to make it easy to start a local http server to try out the example, updates the build instructions, and also addresses some of the compiler warnings that were being generated.
* emscripten : fix TOTAL_STACK for wasm
This commit moves the TOTAL_STACK setting from the compile flags to the
linker flags. This is because the TOTAL_STACK setting is a linker
setting.
The motivation for this change is that currently the following warnings
are generated when building:
```console
em++: warning: linker setting ignored during compilation: 'TOTAL_STACK' [-Wunused-command-line-argument]
em++: warning: linker setting ignored during compilation: 'TOTAL_STACK' [-Wunused-command-line-argument]
em++: warning: linker setting ignored during compilation: 'TOTAL_STACK' [-Wunused-command-line-argument]
em++: warning: linker setting ignored during compilation: 'TOTAL_STACK' [-Wunused-command-line-argument]
em++: warning: linker setting ignored during compilation: 'TOTAL_STACK' [-Wunused-command-line-argument]
em++: warning: linker setting ignored during compilation: 'TOTAL_STACK' [-Wunused-command-line-argument]
```
* examples : suppress C++17 deprecation warning for std::codecvt_utf8
This commit suppresses the C++17 deprecation warning for
std::codecvt_utf8 similar to what is done in
examples/talk-llama/unicode.cpp.
The motivation for this change is to suppress these warnings:
```console
/Users/danbev/work/ai/whisper-work/examples/common.cpp:251:31: warning: 'codecvt_utf8<wchar_t>' is deprecated [-Wdeprecated-declarations]
251 | std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/codecvt:193:28: note: 'codecvt_utf8<wchar_t>' has been explicitly marked deprecated here
193 | class _LIBCPP_TEMPLATE_VIS _LIBCPP_DEPRECATED_IN_CXX17 codecvt_utf8 : public __codecvt_utf8<_Elem> {
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:723:41: note: expanded from macro '_LIBCPP_DEPRECATED_IN_CXX17'
723 | # define _LIBCPP_DEPRECATED_IN_CXX17 _LIBCPP_DEPRECATED
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:688:49: note: expanded from macro '_LIBCPP_DEPRECATED'
688 | # define _LIBCPP_DEPRECATED __attribute__((__deprecated__))
| ^
/Users/danbev/work/ai/whisper-work/examples/common.cpp:251:10: warning: 'wstring_convert<std::codecvt_utf8<wchar_t>>' is deprecated [-Wdeprecated-declarations]
251 | std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/locale:3145:28: note: 'wstring_convert<std::codecvt_utf8<wchar_t>>' has been explicitly marked deprecated here
3145 | class _LIBCPP_TEMPLATE_VIS _LIBCPP_DEPRECATED_IN_CXX17 wstring_convert {
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:723:41: note: expanded from macro '_LIBCPP_DEPRECATED_IN_CXX17'
723 | # define _LIBCPP_DEPRECATED_IN_CXX17 _LIBCPP_DEPRECATED
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:688:49: note: expanded from macro '_LIBCPP_DEPRECATED'
688 | # define _LIBCPP_DEPRECATED __attribute__((__deprecated__))
| ^
/Users/danbev/work/ai/whisper-work/examples/common.cpp:257:31: warning: 'codecvt_utf8<wchar_t>' is deprecated [-Wdeprecated-declarations]
257 | std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/codecvt:193:28: note: 'codecvt_utf8<wchar_t>' has been explicitly marked deprecated here
193 | class _LIBCPP_TEMPLATE_VIS _LIBCPP_DEPRECATED_IN_CXX17 codecvt_utf8 : public __codecvt_utf8<_Elem> {
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:723:41: note: expanded from macro '_LIBCPP_DEPRECATED_IN_CXX17'
723 | # define _LIBCPP_DEPRECATED_IN_CXX17 _LIBCPP_DEPRECATED
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:688:49: note: expanded from macro '_LIBCPP_DEPRECATED'
688 | # define _LIBCPP_DEPRECATED __attribute__((__deprecated__))
| ^
/Users/danbev/work/ai/whisper-work/examples/common.cpp:257:10: warning: 'wstring_convert<std::codecvt_utf8<wchar_t>>' is deprecated [-Wdeprecated-declarations]
257 | std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/locale:3145:28: note: 'wstring_convert<std::codecvt_utf8<wchar_t>>' has been explicitly marked deprecated here
3145 | class _LIBCPP_TEMPLATE_VIS _LIBCPP_DEPRECATED_IN_CXX17 wstring_convert {
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:723:41: note: expanded from macro '_LIBCPP_DEPRECATED_IN_CXX17'
723 | # define _LIBCPP_DEPRECATED_IN_CXX17 _LIBCPP_DEPRECATED
| ^
/Users/danbev/work/wasm/emsdk/upstream/emscripten/cache/sysroot/include/c++/v1/__config:688:49: note: expanded from macro '_LIBCPP_DEPRECATED'
688 | # define _LIBCPP_DEPRECATED __attribute__((__deprecated__))
| ^
4 warnings generated.
```
* ggml : suppress double-promotion warning in GGML_F16x4_REDUCE
This commit adds a cast to `ggml_float` in the `GGML_F16x4_REDUCE` macro
to suppress a double-promotion warning.
Currently the following warning is generated when compiling the
command.wasm example:
```console
/whisper-work/ggml/src/ggml-cpu/ggml-cpu.c:1592:5: warning: implicit conversion increases floating-point precision: 'float' to 'ggml_float' (aka 'double') [-Wdouble-promotion]
1592 | GGML_F16_VEC_REDUCE(sumf, sum);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/danbev/work/ai/whisper-work/ggml/src/ggml-cpu/ggml-cpu.c:932:37: note: expanded from macro 'GGML_F16_VEC_REDUCE'
932 | #define GGML_F16_VEC_REDUCE GGML_F16x4_REDUCE
| ^
/Users/danbev/work/ai/whisper-work/ggml/src/ggml-cpu/ggml-cpu.c:920:44: note: expanded from macro 'GGML_F16x4_REDUCE'
918 | res = wasm_f32x4_extract_lane(x[0], 0) + \
| ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
919 | wasm_f32x4_extract_lane(x[0], 1) + \
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
920 | wasm_f32x4_extract_lane(x[0], 2) + \
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~
921 | wasm_f32x4_extract_lane(x[0], 3); \
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/whisper-work/ggml/src/ggml-cpu/ggml-cpu.c:1640:9: warning: implicit conversion increases floating-point precision: 'float' to 'ggml_float' (aka 'double') [-Wdouble-promotion]
1640 | GGML_F16_VEC_REDUCE(sumf[k], sum[k]);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/danbev/work/ai/whisper-work/ggml/src/ggml-cpu/ggml-cpu.c:932:37: note: expanded from macro 'GGML_F16_VEC_REDUCE'
932 | #define GGML_F16_VEC_REDUCE GGML_F16x4_REDUCE
| ^
/Users/danbev/work/ai/whisper-work/ggml/src/ggml-cpu/ggml-cpu.c:920:44: note: expanded from macro 'GGML_F16x4_REDUCE'
918 | res = wasm_f32x4_extract_lane(x[0], 0) + \
| ~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
919 | wasm_f32x4_extract_lane(x[0], 1) + \
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
920 | wasm_f32x4_extract_lane(x[0], 2) + \
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~
921 | wasm_f32x4_extract_lane(x[0], 3); \
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2 warnings generated.
```
wasm_f32x4_extract_lane returns a 32-bit float and this is what the
addition is performed on. But there is an implicit conversion from
32-bit float to 64-bit double when the result is assigned to `res`,
which is of type `ggml_float`. My understanding here is that this is
intentional and adding a cast to `ggml_float` should suppress the
warning.
* emscripten : add -Wno-deprecated to for emscripten
This commit adds -Wno-deprecated to the CMAKE_CXX_FLAGS for emscripten
builds.
The motivation for this is that currently there a number of warnings
generated like the following:
```console
warning: JS library symbol '$print' is deprecated. Please open a bug if you have a continuing need for this symbol [-Wdeprecated]
warning: JS library symbol '$printErr' is deprecated. Please open a bug if you have a continuing need for this symbol [-Wdeprecated]
em++: warning: warnings in JS library compilation [-Wjs-compiler]
em++: warning: linker setting ignored during compilation: 'ENVIRONMENT' [-Wunused-command-line-argument]
warning: JS library symbol '$print' is deprecated. Please open a bug if you have a continuing need for this symbol [-Wdeprecated]
warning: JS library symbol '$printErr' is deprecated. Please open a bug if you have a continuing need for this symbol [-Wdeprecated]
em++: warning: warnings in JS library compilation [-Wjs-compiler]
warning: JS library symbol '$print' is deprecated. Please open a bug if you have a continuing need for this symbol [-Wdeprecated]
warning: JS library symbol '$printErr' is deprecated. Please open a bug if you have a continuing need for this symbol [-Wdeprecated]
em++: warning: warnings in JS library compilation [-Wjs-compiler]
em++: warning: linker setting ignored during compilation: 'ENVIRONMENT' [-Wunused-command-line-argument]
em++: warning: linker setting ignored during compilation: 'ENVIRONMENT' [-Wunused-command-line-argument]
```
The downside of this is that we might miss other deprecation warnings
in the future so I'm not sure if this is acceptable. But it make the
wasm examples cleaner without the warnings.
* examples : fix tautological-compare warning in stb_vorbis.c [no ci]
This commit applies a fix to address a tautological-compare warning
in stb_vorbis.c.
The motivation for this is that currently the following warning is
generated when compiling the commmand-wasm example:
```console
/Users/danbev/work/ai/whisper-work/examples/stb_vorbis.c:1404:75: warning: pointer comparison always evaluates to false [-Wtautological-compare]
1404 | if (f->stream_start + loc >= f->stream_end || f->stream_start + loc < f->stream_start) {
| ^
1 warning generated.
```
This fix was taken from an open pull request on the stb repository
that addreses this issue:
https://github.com/nothings/stb/pull/1746
* squash! examples : update command.wasm instructions [no ci]
This commit adds a Python script to serve the the wasm examples build
in the `build-em` directory. Initially I thought that it would be enough
to start a simple python server but I did not notice that there was an
error in the browser console when I did that:
```console
command.js:1 Uncaught (in promise) DataCloneError: Failed to execute 'postMessage' on 'Worker': SharedArrayBuffer transfer requires self.crossOriginIsolated.
at command.js:1:1206224
at new Promise (<anonymous>)
at loadWasmModuleToWorker (command.js:1:1204981)
at Array.map (<anonymous>)
at Object.loadWasmModuleToAllWorkers (command.js:1:1206428)
at command.js:1:1204318
at callRuntimeCallbacks (command.js:1:1202062)
at preRun (command.js:1:6136)
at run (command.js:1:1294094)
at removeRunDependency (command.js:1:7046)
```
We need a few CORS headers to be set and in order hopefully make this
easy for users a Python script is added to the examples directory.
This should be able to server all the wasm examples provided they have
been built. command.wasm's README.md is updated to reflect this change.
* examples : remove unused functions
This commit removed the unused functions convert_to_utf8 and
convert_to_wstring from examples/common.cpp.
* Revert "examples : fix tautological-compare warning in stb_vorbis.c [no ci]"
This reverts commit 8e3c47d961.
We should not make this change here and instead when the upstream PR is
merged we can sync with it.
Refs: https://github.com/ggerganov/whisper.cpp/issues/2784
* metal : refactor im2col parameters into a struct
* metal: Change im2col offset types from int32_t to uint64_t to support larger memory offsets
* metal : refactor sum_rows parameters into a struct
* metal : refactor soft_max parameters into a struct
* metal : refactor diag_mask_inf parameters into a struct
* metal : refactor ssm_conv parameters into a struct
* metal : refactor ssm_scan parameters into a struct
* metal : refactor get_rows parameters into a struct
* metal : refactor group_norm parameters into a struct
* metal : refactor conv_transpose_1d parameters into a struct
* metal : refactor upscale parameters into a struct
* metal : refactor pad parameters into a struct
* metal : refactor pad_reflect_1d parameters into a struct
* metal : refactor arange parameters into a struct
* metal : refactor timestep_embedding parameters into a struct
* metal : refactor argsort parameters into a struct
* metal : refactor leaky_relu parameters into a struct
* metal : refactor pool_2d parameters into a struct
* metal : fix trailing whitespace
---------
Co-authored-by: alexju <alexju@tencent.com>
This commit updates the custom command to build the default.metallib
file to use the correct path to ../ggml-common.h by using the variable
METALLIB_COMMON.
The motivation for this change is that currently when building and
specifying GGML_METAL_EMBED_LIBRARY=OFF the following error is
generated:
```console
[ 11%] Linking CXX shared library ../../bin/libggml.dylib
[ 11%] Built target ggml
make[2]: *** No rule to make target `ggml/src/ggml-metal/ggml-common.h', needed by `bin/default.metallib'. Stop.
make[1]: *** [ggml/src/ggml-metal/CMakeFiles/ggml-metal-lib.dir/all] Error 2
```
With the above change the build could progress but there was a follow
on error about not being able to find the ggml-common.h file in
ggml-metal.metal where is was included as a relative path:
```console
[ 11%] Compiling Metal kernels
/Users/danbev/work/llama.cpp/build/bin/ggml-metal.metal:6:10: error: '../ggml-common.h' file not found, did you mean 'ggml-common.h'?
^~~~~~~~~~~~~~~~~~
"ggml-common.h"
1 error generated.
```
Removing the relative path then allowed the build to complete
successfully.
Fix the following error:
```
ggml-alloc.c:99: not enough space in the buffer
ggml_tallocr_alloc: not enough space in the buffer to allocate blk.17.ffn_down.weight (needed 27525120, available 27521024)
```
which occurs when `ggml_backend_opencl_context::alignment` is larger
than `cl_ptr_base` (hard-coded to `0x1000`).
Also, fix `ggml_backend_opencl_context::alignment` was set to
`CL_DEVICE_MEM_BASE_ADDR_ALIGN` which was treated as bytes but the
value is reported in bits.
* ggml-cpu: Faster IQ1 mul_mat_vec on AVX2 using BMI2 instructions
* cmake: Add GGML_BMI2 build option
* ggml: enable BMI2 on relevant CPU variants
* ggml-cpu: include BMI2 in backend score
* ggml-cpu: register BMI2 in ggml_backend_cpu_get_features
* ggml-cpu: add __BMI2__ define when using MSVC
-- it might happen if ggml is loaded from 2 separate libraries since each one of them will expose the class. This is more of a guard since we want to use only Metal as embedded library and don't care about the other case.
* ggml_compute_forward_concat() for arbitrary tensor type
* Check that tensors' type match
* ggml-cpu.c: check type of source tensors
* ggml-cpu.c: move tensor type check to ggml_compute_forward_concat()
* ggml.c: check concatenated tensor type
* Remove tensor type check from ggml_compute_forward_concat() in ggml-cpu.c
..., as it was moved to ggml.c.
* Add include files for std::min/max and std::toupper/tolower
* win32: move _USE_MATH_DEFINES before includes to ensure M_PI is defined
* Use GGML_RESTRICT instead of "restrict" keyword everywhere, and use "__restrict" in MSVC plain C mode
* win32: only use __restrict in MSVC if C11/C17 support is not enabled
---------
Co-authored-by: Marcus Groeber <Marcus.Groeber@cerence.com>
* Upgrade init_tensor API to return a ggml_status
To prepare for an 'abort-free' ggml
(ggml not to abort on OOMs but return a OOM status),
as agreeed with Diego in the ggml repo,
upgrade the init_tensor() and view_init() APIs
to return a ggml_status.
* misc fixes
---------
Co-authored-by: slaren <slarengh@gmail.com>
* vulkan: implement specialized MMV kernels for IQ2 quantizations
* vulkan: add MMV kernels for IQ3 quants
* vulkan: Increase MMV batch size and unroll IQ LUT setup
* vulkan: fix init_iq_shmem for WG sizes larger than tables
* vulkan: common batch size for all I-quants
* Added SVE Support for Q2_K Quantized Models
* Use 4-space indentation in the switch cases
* removed comments lines
* Remove the loop Retain the curly bracess for better understanding of code
* Remove the comment like added for q3_k_q8_k kernel
---------
Co-authored-by: vithulep <p.m.vithule1517@gmail.com>
* Fix dependencies between ggml and backends
ggml backends link only to ggml-base and ggml links to all backends.
* Fix installation of ggml backends
Set up GNUInstallDirs before setting the installation directory of ggml backends
* cuda: restrict SILU_BACK to fp32, since fp16 exceeds the desired test threshold
* vulkan: specify fp32-only support for certain ops (that are now tested for fp16 as well)
* f32 sigmoid in vulkan supports op
* Revert "f32 sigmoid in vulkan supports op"
This reverts commit c6f04b3c19bf4504c2776149c6d8cd84e0b48acb.
* Support fp16 unary operations in the CUDA backend
* cpu: increase fp16 support for unary operators in the CPU backend
* cuda: increase fp16 support for unary operators in the CUDA backend
* Add test cases for fp16 unary operators
* metal: update supports_op for unary operators that don't support fp16, to prevent test-backend-ops from failing
* metal: fix PR comments for unary op support after fp16 unary tests
* Support float16-to-float16 add/sub/mul/div operations in the CUDA backend
* Add fp16 support for add/sub/mul/div on the CPU backend
* Add test cases for fp16 add/sub/mul/div
* opt performance by reorder for Intel GPU
* detect hw type and save opt feature, and print opt feature
* correct name
* support optimize graph once when compute graph, record the opt status in tensor->extra, make CI passed
* add env variable GGML_SYCL_DISABLE_OPT for debug
* use syclex::architecture replace the custom hw define, update the guide for GGML_SYCL_DISABLE_OPT
* add performance data
* mv getrows functions to separeted files
* fix global variables
---------
Co-authored-by: arthw <14088817+arthw@users.noreply.github.com>
* MUSA: support ARM64 and enable __dp4a .etc
* fix cross entropy loss op for musa
* update
* add cc info log for musa
* add comment for the MUSA .cc calculation block
---------
Co-authored-by: Bodhi Hu <huaishun.hu@mthreads.com>
* ggml-cpu: Add CPU backend support for KleidiAI library
* Add environmental variable GGML_KLEIDIAI_SME
* Add support for multithread LHS conversion
* Switch kernel selection order to dotprod and i8mm
* updates for review comments
* More updates for review comments
* Reorganize and rename KleidiAI files
* Move ggml-cpu-traits.h to source file
* Update cmake for SME build and add alignment for SME
* Remove append GGML_USE_CPU_KLEIDIAI to the GGML_CDEF_PUBLIC list
* vulkan: initial support for IQ1_S and IQ1_M quantizations
* vulkan: define MMV kernels for IQ1 quantizations
* devops: increase timeout of Vulkan tests again
* vulkan: simplify ifdef for init_iq_shmem
* musa: Update MUSA SDK version to rc3.1.1
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* musa: Remove workaround in PR #10042
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
---------
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* ggml-cpu : add chunking support to mul_mat_id
* allocate chunk counter in wdata
parallelize src1 quantization by column to allows parallelization even when there is only one row
* disable for arm
* cleanup
* better way to disable for arm
* fix uninitialized counter when using 1 thread only
* revert test-backend-ops changes
* Bug fix for clamp_f32
When using tensors larger than 1d clamp operation does not work due to the restriction of returning if ith is not 0.
* Bug fix for clamp_f32
* Bug fix for clamp_f32
After the barrier in last iteration is executed, still the loop termination
condition will be executed. However main thread can destroy the cgraph object
and its nodes already, then another thread will access it, but the thing is already gone.
Also trouble can happen when n_nodes == 0 or abort is called, but I'm not sure if the
prior situation is possible.
Last syncronization should be done after the loop to ensure the cgraph/cplan won't be
accessed after the main thread exits from the function.
* ggml : optimize convert f32<->f16 for loongarch_asx
* ggml : optimize loongarch_asx extend i16,i8,u8 to i32,i16
* ggml : Fix warnings when run cpu CI locally on LoongArch
Add bounds checking in `rpc_server::copy_tensor` to prevent out-of-bounds writes
+ Check if `(uint8_t *)dst->data + ggml_nbytes(src)` remains within the destination buffer’s allocated region.
This makes git as a dependency optional, and is useful in the case where
ggml is built not from git, but from a tarball, or a distribution source
package.
This conditional also affects GGML_BUILD_COMMIT. Nothing seems to be
using it, though, so there doesn't seem much value factor it out, or
even require it.
* CUDA: use mma PTX instructions for FlashAttention
* __shfl_sync workaround for movmatrix
* add __shfl_sync to HIP
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* vulkan: initial support for IQ3_S
* vulkan: initial support for IQ3_XXS
* vulkan: initial support for IQ2_XXS
* vulkan: initial support for IQ2_XS
* vulkan: optimize Q3_K by removing branches
* vulkan: implement dequantize variants for coopmat2
* vulkan: initial support for IQ2_S
* vulkan: vertically realign code
* port failing dequant callbacks from mul_mm
* Fix array length mismatches
* vulkan: avoid using workgroup size before it is referenced
* tests: increase timeout for Vulkan llvmpipe backend
---------
Co-authored-by: Jeff Bolz <jbolz@nvidia.com>
loops with bounds not known at compile time can not be unrolled.
when ncols_template == 0, the bounds of the loop are not constexpr, thus llvm cant unroll the loops here.
This disables the workaround on rocblas fixed versions (>=4.0.0) to eliminate the runtime cost and unnecessary VRAM allocation of loading all tensile objects.
Implemented ggml_sycl_op_soft_max() F16 src1(mask) support for which a pragma deprecation warning was added during #5021.
To do this, had to decouple it from ggml_sycl_op_flatten which always considered src1 to be of fp32 type(many OP functions are dependent on it).
* SYCL: SOFTMAX F16 mask support and other fixes
* test-backend-ops: Add F16 mask test cases
This fixes segmentation fault error when running tests when no metal
devices are available (for example, when not linked with Core Graphics
framework or otherwise).
With robustbufferaccess disabled, this shader was showing OOB stores. There
is a bounds check in the code, but the workgrouop dimensions were reversed vs
CUDA and it was running the wrong number of threads. So fix the workgroup
dimensions and disable robustness for this pipeline.
mul mat and flash attention shaders were loading f32 types directly into
A/B matrices, which happens to work but is technically invalid usage.
For FA, we can load it as an Accumulator matrix and convert and this
is not in the inner loop and is cheap enough. For mul mat, it's more
efficient to do this conversion in a separate pass and have the input(s)
be f16.
coopmat2 requires SPIR-V 1.6 (related using to LocalSizeId). LocalSizeId
requires maintenance4 be enabled, and SPIR-V 1.6 requires Vulkan 1.3.
* Implement host pool for matrix_info
Creating a new memory pool on the host to store memory location for
matrix_info needed to launch gemm_batch from oneMKL/oneMath.
Removing complex support in gemm_batch since it is not used in llama.cpp
* Remove unnecessary headers and cast
* Reorder member variable to avoid warning on initialization
* Formatting
* Remove unused variable
* Address PR review feedback - remove warning
---------
Signed-off-by: nscipione <nicolo.scipione@codeplay.com>
Add code similar to mul_mm_cm2 to force alignment of strides, to avoid
a performance regression.
Add noncontiguous FA tests in test-backend-ops.
Fixes#11268.
* vulkan: support copy from f32 to q4_0/q4_1/q5_0/q5_1/q8_0/iq4_nl
Shaders are based on cpy.cu.
* vulkan: support copy from q4_0/q4_1/q5_0/q5_1/q8_0/iq4_nl to f32
* ggml: copy q->f32 assumes some contiguity in the destination
* Add SVE support for q4_K_q8_K
* Update ggml/src/ggml-cpu/ggml-cpu-quants.c
change to use K_SCALE_SIZE
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* q6_k scale caching
* 16 bit unpack
* q4_k test (slow)
* revert it
* q3_k
* q2_k
* little stuff
* try precalculating products of a and q2_k scales
* Revert "try precalculating products of a and q2_k scales"
This reverts commit 65110b81f23f66331a50c6e889a7c1ab9470a86b.
* unpack should be u16, add vim swap to gitignore (about time)
* better q4_k scales
* q5_k
* better q6_k with separate paths for all threads and partial threads in use, plus some more optimizations
* q2_k better dequant
* q3_k optimizations
* q3_k use hmask simd from cpu avx version
* make the caches happy
* q3_k separate out calculation
* q2_k separate out
* little stuff
* use calc_superblock everywhere
* q2_k optimize scale calculation
* more barriers
* fix: ggml: fix vulkan-shaders-gen build
The vulkan-shaders-gen target was not being built correctly
in case of cross-compilation.
Other outputs need to be built for the cross compile target,
but vulkan-shaders-gen needs to be built for the host.
* refactor: ggml: Improve vulkan-shaders-gen toolchain setup
- Add GGML_SHADERS_GEN_TOOLCHAIN CMake option.
- Auto-detect host toolchain if not set.
* refactor: ggml: Improve vulkan-shaders-gen toolchain setup
Use configure_file to generate host_toolchain.cmake from template
* fix: ggml: Fix compile error
Fix compile error not finding vulkan-shaders-gen
* fix: vulkan-shaders-gen build and path handling
Fix build issues with vulkan-shaders-gen:
- Add target dependency for correct build order
- Use CMAKE_HOST_SYSTEM_NAME for executable suffix
- Fix MSVC output directory in host toolchain
- Normalize path handling for cross-compilation
* fix: improve host compiler detection in vulkan shader build
Improve host compiler detection for vulkan shader generation:
- Add NO_CMAKE_FIND_ROOT_PATH to all compiler searches
- Consolidate compiler detection logic
- Fix Windows-specific MSVC detection
- Ensure correct compiler search in cross-compilation
* refactor: Simplify CMake function for detecting host compiler
Simplified the CMake function to improve the process of detecting the host compiler.
* fix: Remove unnecessary Vulkan library linkage in CMakeLists.txt
Since `vulkan-shader-gen.cpp` only requires the `glslc` executable
and not the Vulkan headers or libraries, CMakeLists.txt needs to
be corrected.
(See: ecc93d0558fc3ecb8a5af69d2ece02fae4710ade)
* refactor: Rename host_toolchain.cmake.in
- Rename host_toolchain.cmake.in to cmake/host-toolchain.cmake.in
* refactor: GGML_VULKAN_SHADERS_GEN_TOOLCHAIN
Rename the macro GGML_SHADERS_GEN_TOOLCHAIN to GGML_VULKAN_SHADERS_GEN_TOOLCHAIN
* Add option to not print stack on abort
Add option/envvar to disable stack printing on abort.
Also link some unittests with Threads to fix link errors on
ubuntu/g++11.
* Update ggml/src/ggml.c
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* Refactor: Moves cuda graph executable update step to separate function.
* Refactor: Moves cuda graph update check to separate function.
* Refactor: Moves cuda graph maintenance (update or adjusting copy parameters) to separate function for improved readability.
* Fix: Adds missing reference to maintain_cuda_graph() definition.
* Refactor: Improves structure and abstractions by moving CUDA graph evaluation and capture to its own function.
* Refactor: Moves node graph checks and copy ops into individual function for improved readability.
* Refactor: Removes code permanently excluded from compilation to increase readability.
* Style: Adds missing newline
* Style: Consolidates several neighboring '#ifdef USE_CUDA_GRAPH' into a single one
* Refactor: Makes 'cuda_graph_update_required' a local variable
* remove double lines between functions
---------
Co-authored-by: slaren <slarengh@gmail.com>
Build fails when using HIP and GGML_BACKEND_DL:
```
/usr/bin/ld: ../ggml/src/libggml.so: undefined reference to `ggml_backend_cuda_reg'
collect2: error: ld returned 1 exit status
```
This patch fixes this.
* SYCL: refactor ggml_sycl_compute_forward
* SYCL: add back GGML_USED(dst) to ggml_sycl_cpy
* SYCL: add function name to noop debug
* SYCL: Some device info print refactoring and add details of XMX availability
This change upstreams llamafile's cpu matrix
multiplication kernels for ppc64le using MMA
builtins for quantised int8 datatype.
This change results in 10% - 70% improvement
in total speed(ie all tokens/total time), across
various batch sizes.
The patch is tested with Meta-Lllama-3-8B,
Mistral-7B, Llama-2-7B-chat-hf models on a
IBM POWER10 machine.
Signed-off-by: Amrita H S <amritahs@linux.vnet.ibm.com>
* Disable GL_KHR_cooperative_matrix Vulkan extension if not available.
* Perform Vulkan extensions checks in a more sensible order
* Remove unnecessary #ifdef directive
* GGUF: C++ refactor, backend support, misc fixes
remove ggml_tensor.backend
update CODEOWNERS [no ci]
remove gguf_get_data from API
revise GGUF API data types
* SYCL: Use get_multi_ptr instead of deprecated get_pointer in wkv6
* Revert "SYCL: Use get_multi_ptr instead of deprecated get_pointer in wkv6"
This reverts commit f62dc45f318e48d375e7734b34cbddee81deed52.
* Reland: Use get_multi_ptr instead of deprecated get_pointer in wkv6
Make the mul_mat_vec shaders support N>1 (as a spec constant, NUM_COLS) where
the batch_strides are overloaded to hold the row strides. Put the loads from the
B matrix in the innermost loop because it should cache better.
Share some code for reducing the result values to memory in mul_mat_vec_base.
* tests: Add im2col perf tests
* vulkan: optimize im2col, more elements per thread
* vulkan: increase small tile size for NV_coopmat2
* vulkan: change im2col to 512 elements per workgroup
Warning types fixed (observed under MSYS2 GCC 14.2.0):
* format '%ld' expects argument of type 'long int', but argument has type 'size_t'
* llama.cpp/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp:81:46: warning: missing initializer for member '_STARTUPINFOA::lpDesktop' [-Wmissing-field-initializers] (emitted for all struct field except first)
* more perfo with llamafile tinyblas on x86_64.
- add bf16 suport
- change dispache strategie (thanks:
https://github.com/ikawrakow/ik_llama.cpp/pull/71 )
- reduce memory bandwidth
simple tinyblas dispache and more cache freindly
* tinyblas dynamic dispaching
* sgemm: add M blocs.
* - git 2.47 use short id of len 9.
- show-progress is not part of GNU Wget2
* remove not stable test
Change the code to do 16b loads when possible and extract the appropriate
component late, so the code is effectively decoding a pair of elements and
then selecting one. This can allow more commoning to happen in the compiler
when neighboring elements are loaded.
* Migrate to tensor->buffer for checking backend buffer type: 1
* SYCL: common.cpp try to migrate away from tensor->backend
* SYCL: fix assertions and add proper comments
* SYCL: remove extra space
* SYCL: Add back static to ggml_backend_buffer_is_sycl_split function
* SYCL: Add pragma directive to suppress warning spam
* SYCL: Integrate debug logs with GGML_LOG and other fixes
* Revert "SYCL: Integrate debug logs with GGML_LOG and other fixes"
This reverts commit 2607b7de0f0d2f4f1f690226f86fa861aa39cb97.
Let's keep the current SYCL specific logging mechanism for now
* SYCL: Use GGML_SYCL_DEBUG after reverting
* SYCL: reg_get_proc_address func, update to the current func signature
* SYCL: Refactor SYCL buffer checks in ggml_sycl_cpy_tensor_2d
This commit attempts to improve the log message for the inputs of the
splits in the sched_print_assignments function.
The motivation for this change is that currently even if there are no
inputs a colon is displayed at the end of the line, which can make it a
little confusing when reading the output as it could be interpreted as
the line below are inputs when they are in fact nodes. With this change
the colon will only be printed if there actually are inputs.
* ensure mul mat shaders work on systems with subgroup size less than 32
more fixes
add test
* only s_warptile_mmq needs to be run with 32 threads or more
* [cl][adreno] Add Adreno GPU support
Add new OpenCL backend to support Adreno GPUs
---------
Co-authored-by: Skyler Szot <quic_sszot@quicinc.com>
Co-authored-by: Shangqing Gu <quic_shawngu@quicinc.com>
Co-authored-by: Alexander Angus <quic_aangus@quicinc.com>
Co-authored-by: Hongqiang Wang <quic_wangh@quicinc.com>
Co-authored-by: Max Krasnyansky <quic_maxk@quicinc.com>
* [cl][ci] Add workflow for CL
* [cl][adreno] Fix memory leak for non SMALL_ALLOC path
* opencl: integrate backend dyn.load interface and fix compiler and format warnings
* opencl: remove small-alloc support and fix build errors for non-opencl platforms
* opencl: fixed merge conflict (MUSA added twice in cmake)
* opencl-ci: use RUNNER_TEMP instead of github.workspace
* opencl: fix embed tool invocation with python3
* opencl: CI workflow fixes
* opencl: Clean up small-alloc in CMake files
* opencl: cleanup ggml-opencl2 header file
* opencl: use ulong for offsets and strides in ADD kernel
* opencl: use cl_ulong for all offsets
* opencl: use cl_ulong for sizes and strides
* opencl: use `GGML_LOG_xxx` instead of `fprintf(stderr, ...)`
* opencl: rename backend `opencl2` -> `opencl`
* opencl: rename kernel files `ggml-opencl2` -> `ggml-opencl`
* opencl: make OpenCL required, remove redundant lib and inc directories
* `ggml-base`, `..` and `.` are added by `ggml_add_backend_library`
* opencl: rename backend - funcs, structs, etc `opencl2` -> `opencl`
* opencl: remove copyright marker since main license already covers
* opencl: replace some more OPENCL2 leftovers
* opencl: remove limits on `tensor_extra`
* opencl: use pools for `tensor_extra`
* opencl: fix compiler warnings with GCC and Clang
Still getting the warning about clCreateCmdQueue being obsolete.
Will fix that separately.
* opencl: fail gracefully if opencl devices are not available
Also for unsupported GPUs.
* opencl: fix MSVC builds (string length error)
* opencl: check for various requirements, allow deprecated API
* opencl: update log message for unsupported GPUs
---------
Co-authored-by: Skyler Szot <quic_sszot@quicinc.com>
Co-authored-by: Shangqing Gu <quic_shawngu@quicinc.com>
Co-authored-by: Alexander Angus <quic_aangus@quicinc.com>
Co-authored-by: Hongqiang Wang <quic_wangh@quicinc.com>
Co-authored-by: Max Krasnyansky <quic_maxk@quicinc.com>
* Fix crash caused by ggml_backend_load_all when launching on AndroidActivity.
Details:
Calling ggml_backend_load_all during initialization in the AndroidActivity project leads to a crash with the error:
terminating with uncaught exception of type std::__ndk1::__fs::filesystem::filesystem_error: filesystem error: in directory_iterator::directory_iterator(...): Permission denied [./].
This issue occurs because AndroidActivity restricts file access due to sandboxing.
Reproduction:
In the example folder, the LlamaAndroid project can reproduce the crash by calling ggml_backend_load_all first in Java_android_llama_cpp_LLamaAndroid_backend_1init.
* Update ggml/src/ggml-backend-reg.cpp
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* double the number of rows per workgroup
* Update ggml-vulkan.cpp
* Vulkan: Add VK_EXT_subgroup_size_control support to ensure full subgroups for coopmats
* only increase the number of rows for amd and subgroup size 64
* fix missing NUM_ROWS for mul_mat_vec_iq4_nl_f16_f32, untested
* use subgroup min and max to check for gcn (requires https://github.com/ggerganov/llama.cpp/pull/10721)
* manual merge ggml-vulkan.cpp
* set min and max subgroup size in any case
* Also double the number of rows for Intel GPUs
* Try to reduce some unused and typecast warnings
* Reduce compiler warnings step 2
* add a newline at the end of the file
* Initialize nreduce as size_t
* [SYCL] Remove pragma directives from mmq.cpp
* SYCL: mmq add condition to prevent blocks_per_tile_x_row variable from becoming 0
* SYCL softmax: Initialize nreduce as size_t
* ggml-sycl.cpp: fix some trailing whitespaces
* SYCL: remove the unused variables instead of commenting it out
* SYCL poo2d kernel: set NAN for invalid pooling op
* SYCL gemm.hpp: remove pragma directives
* SYCL gemm.hpp: use const cast to properly support dnnl::memory
* SYCL: wkv6 remove a comment
* SYCL: clean comments step 2
* SYCL: clean comments and variables step 3
* SYCL: Use GGML_UNUSED for unused variables
* SYCL: remove extra empty lines and a comment
* Remove TODO
* cleanup spaces
* add a stdout for unsupported op
* use sycl printf over fprintf
* remove prints for CI
* SYCL ggml-sycl: pool2D use sycl::nan and remove if-else block
---------
Co-authored-by: Abhilash Majumder <30946547+abhilash1910@users.noreply.github.com>
* faster uncontiguous concat
* Use a lambda to avoid code duplication
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* Update ggml/src/ggml-cuda/concat.cu
* add constexpr and static assert
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* Vulkan: Add VK_EXT_subgroup_size_control support to ensure full subgroups for coopmats
* Fix subgroup size control extension support check
Add accf32 and accf16 checks for coopmats
* Also disable coopmats on amdvlk
* feat: load all backends from a user-provided search path
* fix: Windows search path
* refactor: rename `ggml_backend_load_all_in_search_path` to `ggml_backend_load_all_from_path`
* refactor: rename `search_path` to `dir_path`
* fix: change `NULL` to `nullptr`
Co-authored-by: Diego Devesa <slarengh@gmail.com>
* fix: change `NULL` to `nullptr`
---------
Co-authored-by: Diego Devesa <slarengh@gmail.com>
Vulkan doesn't mandate a specific rounding mode, but the shader_float_controls
feature allows rounding mode to be requested if the implementation supports it.
* Renames NVIDIA GPU-architecture flags to avoid name clashes with WinAPI. (e.g. CC_PASCAL, GPU architecture or WinAPI pascal compiler flag?)
* Reverts erroneous rename in SYCL-code.
* Renames GGML_CUDA_MIN_CC_DP4A to GGML_CUDA_CC_DP4A.
* Renames the rest of the compute capability macros for consistency.
There are some bugs in the 1.3.296 SDK, so disable this. It isn't strictly
necessary anyway.
Add missing dependency on vulkan-shaders-gen, so shaders get recompiled when it
changes.
Fix coopmat support reporting when glslc doesn't support NV_coopmat2.
This commit removes the return statement from ggml_gallocr_allocate_node
function.
The motivation behind this change is to make the code more readable and
consistent.
* ggml : add check for grad_accs
This commit adds a check for grad_accs in ggml_graph_get_grad and
ggml_graph_get_grad_acc functions. This is necessary to avoid segfaults
when grad_accs is not initialized.
The motivation for this change is that I find it nice to be able to
print out a computation graph using ggml_graph_print but this function
segfaults when grad_accs is not initialized:
```console
(gdb) p g1
$2 = (ggml_cgraph *) 0x7ffff66004b0
(gdb) p *g1
$3 = {size = 2048, n_nodes = 1, n_leafs = 2, nodes = 0x7ffff6600500,
grads = 0x0, grad_accs = 0x0, leafs = 0x7ffff6604500,
visited_hash_set = {size = 4099, used = 0x7ffff6610518,
keys = 0x7ffff6608500}, order = GGML_CGRAPH_EVAL_ORDER_LEFT_TO_RIGHT}
(gdb) p ggml_graph_print(g1)
=== GRAPH ===
n_nodes = 1
Program received signal SIGSEGV, Segmentation fault.
0x0000555555579775 in ggml_graph_get_grad
(cgraph=0x7ffff66004b0,node=0x7ffff6600340)
at /ggml/ggml/src/ggml.c:5990
5990 return igrad != GGML_HASHSET_FULL &&
ggml_bitset_get(cgraph->visited_hash_set.used, igrad) ?
cgraph->grads[igrad] : NULL;
```
* squash! ggml : add check for grad_accs
Fix the check in ggml_graph_get_grad. The check was incorrectly using
cgraph->grad_accs instead of cgraph->grads.
* rename ggml-cpu-aarch64.c to .cpp
* reformat extra cpu backend.
- clean Q4_0_N_M and IQ4_0_N_M
- remove from "file" tensor type
- allow only with dynamic repack
- extract cpu extra bufts and convert to C++
- hbm
- "aarch64"
- more generic use of extra buffer
- generalise extra_supports_op
- new API for "cpu-accel":
- amx
- aarch64
* clang-format
* Clean Q4_0_N_M ref
Enable restrict on C++
* add op GGML_OP_MUL_MAT_ID for Q4_0_N_M with runtime repack
* added/corrected control on tensor size for Q4 repacking.
* Update ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* add debug logs on repacks.
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Vulkan: Implement VK_KHR_cooperative_matrix support in the matrix matrix multiplication shader
* Improve performance with better q4_k and q5_k dequant and store unrolling
* Add Vulkan MUL_MAT and MUL_MAT_ID accumulator precision selection
* Rework mulmat shader selection and compilation logic, avoid compiling shaders that won't get used by device
* Vulkan: Implement accumulator switch for specific mul mat mat shaders
* Vulkan: Unroll more loops for more mul mat mat performance
* Vulkan: Add VK_AMD_shader_core_properties2 support to read Compute Unit count for split_k logic
* Disable coopmat support on AMD proprietary driver
* Remove redundant checks
* Add environment variable GGML_VK_DISABLE_COOPMAT to disable VK_KHR_cooperative_matrix support
* Fix rebase typo
* Fix coopmat2 MUL_MAT_ID pipeline selection
* metal : Extend how Llama.cpp locates metal resources (llama/10675)
* It searches the resource file in the directory where the current
binary is located as well.
* Resolves symbolic links.
Rationale:
When we plug this dependency into a Bazel build and run it in the
context of Bazel (e.g. testing):
* the execution directory is often very different from where the files
are located and no direct control over this (Bazel sandboxing),
* the Bazel sandbox often use symbolic links to make files available.
With this patch, we can have the resource file added to the target,
can build and run tests in the context of Bazel.
* Update ggml/src/ggml-metal/ggml-metal.m
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Update ggml/src/ggml-metal/ggml-metal.m
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* [SYCL] Move to Compile Time backend selection on oneMKL Interface for NVIDIA backend
Move to compile time selection to backend to avoid latency at run time.
Add it to all mkl gemm calls and only for NVIDIA backend.
Signed-off-by: nscipione <nicolo.scipione@codeplay.com>
* Formatting
* Address PR comments to increase readibility
---------
Signed-off-by: nscipione <nicolo.scipione@codeplay.com>
* ggml_pad_reflect_1d defined in header
* implemented on CPU
* called the forward pass
* impl Metal kernel
* added Metal kernel
* added OP_PAD_REFLECT_1D in test-backend-ops.cpp
* add test-pad-reflect-1d test case
* test case support multiple backend
* subgroup 64 version with subgroup add. 15% faster
scalable version
tested for subgroup sizes 16-128
* check for subgroup multiple of 16 and greater than 16
* subgroup sizes are always a power of 2 (https://github.com/KhronosGroup/GLSL/issues/45)
* force 16 sequential threads per block
* make 16 subgroup size a constant
This is an incremental improvement over #9118 to get work to the GPU a bit
sooner. The first part is to start with a smaller number of nodes before
the first submit, and ramp it up to the current 100 nodes/submit. The
second part is to reduce the dryrun overhead for all the nodes that just
need to request descriptor space.
With these changes I get around 1-2% speedup on RTX 4070 combined with my
old Haswell-era CPU.
* CANN: Fix the bug build fail on Ascend310P under two cases:
1) Manual specify SOC_TYPE
2) Under some unusual compile environment
* Update the cann backend News content: Support F16 and F32 data type model for Ascend 310P NPU.
* fix CANN compile fail bug: the assert in ascend kernel function doesn't supportted on some CANN version
There have been reports of failure to compile on systems with <= 32KB
of shared memory (e.g. #10037). This change makes the large tile size
fall back to a smaller size if necessary, and makes mul_mat_id fall
back to CPU if there's only 16KB of shared memory.
* improve inferencing performance for ascend npu.
Co-authored-by: Frank Mai <thxCode@thxcode0824@gmail.com>
* some modification after review
* some modifications after review
* restore some modifications
* restore some modifications
---------
Co-authored-by: shanshan shen <shanshanshen333@gmail.com>
Co-authored-by: Frank Mai <thxCode@thxcode0824@gmail.com>
The vulkan-shaders-gen was not parsing the --no-clean argument correctly.
Because the previous code was parsing the arguments which have a value only
and the --no-clean argument does not have a value, it was not being parsed
correctly. This commit can now correctly parse arguments that don't have values.
* llama : accept a list of devices to use to offload a model
* accept `--dev none` to completely disable offloading
* fix dev list with dl backends
* rename env parameter to LLAMA_ARG_DEVICE for consistency
* CANN Support Ascend310P to accelerate F32 and F16 Model
* Add compile option soc type macro ASCEND_310P to ggml-cann lib
* Remove unused code
* Remove the ascend soc_type hard code compile option in CMakelist.txt
* vulkan: Use pipeline_robustness to disable robustness in mul_mat_vec.
Add some early returns for nonexistent rows in mul_mat_vec shaders. These
can only be hit when dispatching a 2D grid of workgroups. Fix the logic
for the 2D grid of workgroups to round up.
Enable the pipeline robustness extension if it's available, and use it to
disable robustness for these pipelines. The instructions to do the bounds
checking contend for the same ALU resources as the bit twiddling dequant
instructions.
* vulkan: Add GLSL structure aliases for quant types to allow larger loads
In Vulkan it's not possible to cast pointer types, so instead you have to
declare an aliased binding for the memory with a different type. This
commit adds aliases for the quant formats using 16b ints, and in a few
places where the struct size is a multiple of 4 also using 32b ints.
Currently only q4_k's aliases are used, but others will be used in
subsequent commits.
* vulkan: use larger loads in q5_k and q6_k shaders.
Similar to the optimization I did in q4_k recently, this vectorizes some loads
and reduces the number of bit twiddling instructions.
* vulkan: use larger K step per iteration in mul_mat_vec.
Add vec4 dequantization functions, and use them to do K=8 per iteration in
mul_mat_vec. This uses 16b loads for the quant values and 128b loads for B
which helps reduce the load on the memory system.
The K_PER_ITER==2 logic is still there, just for F16/F32, and really only
because they support unaligned sizes.
Tweak the num_iters/unrolling logic to be simpler and catch a couple missed
unrolling opportunities.
* kqmax_new_j in every thread within warp is same after operate at line 199,this reduce can be omit
* same problem in vec32
---------
Co-authored-by: ZhaoXiaoYu <zhao.xiaoyu@zte.com.cn>
* Add option to set the SYCL architecture for all targets
* Convert GGML_SYCL_HIP_TARGET to the more generic GGML_SYCL_ARCH option
* Document that setting GGML_SYCL_ARCH can improve the performance
* vulkan: Optimize soft_max
Large soft_max could already saturate memory, but small/medium sizes were
pretty slow. The bulk of the gains for them comes from using a smaller
workgroup size, and making the workgroup size match the subgroup size also
makes the barriers much cheaper.
Cache some values in locals to avoid refetching/recomputing. And stamp
out a few "template instantiations" so smaller cases will fully unroll.
Add a missing early return for OOB rows. This happens when there are more
than 512 rows and the dispatch is 512 x H.
* vulkan: Further soft_max optimizations
Restore the workgroup size of 512 case, use it for >1024.
Use unrollable loops for more iteration counts.
* metal : add kernel arg structs (wip)
* metal : fattn args
ggml-ci
* metal : cont + avoid potential int overflow [no ci]
* metal : mul mat struct (wip)
* cont : mul mat vec
* cont : pass by reference
* cont : args is first argument
* cont : use char ptr
* cont : shmem style
* cont : thread counters style
* cont : mul mm id
ggml-ci
* cont : int safety + register optimizations
ggml-ci
* metal : GGML_OP_CONCAT
ggml-ci
* metal : GGML_OP_ADD, GGML_OP_SUB, GGML_OP_MUL, GGML_OP_DIV
* metal : GGML_OP_REPEAT
* metal : GGML_OP_CPY
* metal : GGML_OP_RMS_NORM
* metal : GGML_OP_NORM
* metal : add TODOs for rest of ops
* ggml : add ggml-metal-impl.h
ggml-ci
Johannes Gäßler
Gilad S
Jeff Bolz
Ruben Ortlam
s-goto-11
Prashant Vithule
Daniel Bevenius
Akarshan Biswas
hipudding
Georgi Gerganov
Diego Devesa
Charles Xu
Chenguang Li
Aman Gupta
mnehete32
Aaron Teo
compilade
matiaslin
uvos
xctan
rmatif
shalinib-ibm
Sigbjørn Skjæret
Yoshi_likes_e4
Qeeweew
lhez
Ihar Hrachyshka
Acly
Reese Levine
Yavor Ivanov
Dong Won Kim
Xuan-Son Nguyen
R0CKSTAR
SHUAI YANG
Marvin Gießing
Carlos Zoido
rmatif
Jason Ni
Jonathan Graehl
Judd
Oliver Simons
Tak-RS
Romain Biessy
David Zhao
AN Long
Christian Kastner
leejet
lhez
Srihari-mcw
diannao
Alberto Cabrera Pérez
Kai Pastor
Erik Scholz
deepsek
Gabe Goodhart
Chris Rohlf
Donghyeon Jeong
lixing-star
chen fan
Ervin Áron Tasnádi
Peter0x44
Oliver Simons
Neo Zhang Jianyu
Anton Mitkov
Tarek Dakhran
Slobodan Josic
Miaoqian Lin
Eve
luyhcsu
Nicolò Scipione
zhouwg
Eric Zhang
Björn Ganster
xiaobing318
Xinpeng Dou
Radoslav Gerganov
bandoti
uvos
Markus Tavenrath
fanyang
Anton Mitkov
Ewan Crawford
Isaac McFadyen
xctan
Masato Nakasaka
rmatif
Atharva Dubey
Max Krasnyansky
Shawn yang
Yibo Cai
Vineel Abhinav
leo-pony
Daniel Tang
Simon Booth
Bizhao Shi
Henry Linjamäki
Svetlozar Georgiev
Łukasz Ślusarczyk
Yibo Cai
Dan Johansson
David Huang
R0CKSTAR
Justin Santa Barbara
Jesse Gross
Jared Tweed
Aaron Teo
Ville Vesilehto
SXX
Jeffrey Morgan
Alan Gray
kimminsu
Ewan Crawford
cmdr2
Prajwal B Mehendarkar
Piotr Kubaj
Ronny Brendel
Gaurav Garg
a3sh
Junil Kim
Wagner Bruna
Jay
Sandro Hanea
Icenowy Zheng
amritahs-ibm
stduhpf
蕭澧邦
Guus Waals
fj-y-saito
Molly Sophia
Daniele
aubreyli
fairydreaming
jklincn
BB-fat
Jason C.H
Rémy O
xiaofei
Henry Linjamäki
simon886212
Plamen Minev
vmobilis
mgroeber9110
ag2s20150909
William Tambellini
Daniele
Vladimir Vuksanovic
petterreinholdtsen
cmdr2
Gian-Carlo Pascutto
Akarshan Biswas
PureJourney
Bodhi
Hale Chan
Adrian Kretz
Jinyang He
Xuan-Son Nguyen
Richard
Weizhao Ouyang
Sheldon Robinson
Maxim Evtush
Danny Milosavljevic
Karol Kontny
junchao-zhao
Charles Duffy
fxzjshm
Jhen-Jie Hong
Patrick Peng
midnight
Olivier Chafik
uvos
Rémy Oudompheng
Nikita Sarychev
someone13574
Haus1
Ihar Hrachyshka
Bernhard M. Wiedemann
amd-dwang
issixx
Andreas Kieslinger
hydai
Mathieu Baudier
matt23654
Peter
Djip007
yuri@FreeBSD
Adrien Gallouët
Zhiyuan Li
HimariO
谢乃闻
Robert Ormandi
gn64
Frankie Robertson
PAB
Adrien Gallouët
Shupei Fan
Chenguang Li
Sergio López
Shanshan Shen
haopeng
mahorozte