* Vectorize load instructions in dmmv f16 CUDA kernel
Replaces scalar with vector load instructions, which substantially
improves performance on NVIDIA HBM GPUs, e.g. gives a 1.27X overall
speedup for Meta-Llama-3-8B-Instruct-F16 BS1 inference evaluation on
H100 SXM 80GB HBM3. On GDDR GPUs, there is a slight (1.01X) speedup.
* addressed comment
* Update ggml/src/ggml-cuda/dmmv.cu
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
---------
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* ggml : do not use BLAS with types without to_float
* ggml : return pointer from ggml_internal_get_type_traits to avoid unnecessary copies
* ggml : rename ggml_internal_get_type_traits -> ggml_get_type_traits
it's not really internal if everybody uses it
* docs : clarify building Android on Termux
* docs : update building Android on Termux
* docs : add cross-compiling for Android
* cmake : link dl explicitly for Android
* ggml : add metal backend registry / device
ggml-ci
* metal : fix names [no ci]
* metal : global registry and device instances
ggml-ci
* cont : alternative initialization of global objects
ggml-ci
* llama : adapt to backend changes
ggml-ci
* fixes
* metal : fix indent
* metal : fix build when MTLGPUFamilyApple3 is not available
ggml-ci
* fix merge
* metal : avoid unnecessary singleton accesses
ggml-ci
* metal : minor fix [no ci]
* metal : g_state -> g_ggml_ctx_dev_main [no ci]
* metal : avoid reference of device context in the backend context
ggml-ci
* metal : minor [no ci]
* metal : fix maxTransferRate check
* metal : remove transfer rate stuff
---------
Co-authored-by: slaren <slarengh@gmail.com>
* Single allocation of encode_async block with non-ARC capture in ggml-metal.m
* Moving Block_release to the deallocation code
* Release encode block when re-setting encoding buffer count if needed
* Update ggml/src/ggml-metal.m
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Add scaffolding for ggml logging macros
* Metal backend now uses GGML logging
* Cuda backend now uses GGML logging
* Cann backend now uses GGML logging
* Add enum tag to parameters
* Use C memory allocation funcs
* Fix compile error
* Use GGML_LOG instead of GGML_PRINT
* Rename llama_state to llama_logger_state
* Prevent null format string
* Fix whitespace
* Remove log callbacks from ggml backends
* Remove cuda log statement
* vulkan : do not use tensor->extra
This patch allows using the Vulkan backend with the RPC backend as
tensor->extra is no longer used.
Ref: #8536
* Adapt GGML_VULKAN_CHECK_RESULTS to extra removal (#2)
---------
Co-authored-by: 0cc4m <picard12@live.de>
When the device's warp size is less than 16,
it is possible for loadstride_a (mul_mm.comp:114)
and loadstride_b (mul_mm.comp:115) to be set to 0.
Because they are calculated as: the workgroup size,
multiplied by LOAD_VEC_* (which can be 1) and divided by 16.
And the workgroup size is set to be the same as the
warp/subgroup size.
The loadstride_* variables are used as increments in the
loops that populate the buffers used for the multiplication.
When they are 0 they cause an infinite loop.
But infinite loops without side-effects are UB and the
values of loadstride_* are known at compile time.
So, the compiler quietly optimizes all the loops away.
As a consequence, the buffers are not populated and
the multiplication result is just a matrix with all elements
set to 0.
We prevent the UB by making sure that the workgroup size
will never be less than 16, even if our device has a
smaller warp size (e.g. 8).
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
a return before a barrier (that happens only in some threads in
a workgroup) leads to UB.
While the old code actually works on some devices,
it fails on some others (i.e. "smaller" GPUs).
BTW, I think it would be better to set specialization constants
when the graph is built, in that way the local workgroup
could be sized appropriately.
But it would take a lot of work.
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* ggml: Added run-time detection of neon, i8mm and sve
Adds run-time detection of the Arm instructions set features
neon, i8mm and sve for Linux and Apple build targets.
* ggml: Extend feature detection to include non aarch64 Arm arch
* ggml: Move definition of ggml_arm_arch_features to the global data section
* ggml : remove assert for AArch64 GEMV and GEMM Q4 kernels
* added fallback mechanism when the offline re-quantized model is not
optimized for the underlying target.
* fix for build errors
* remove prints from the low-level code
* Rebase to the latest upstream
Make sure n_barrier and n_barrier_passed do not share the cache line to avoid cache line bouncing.
This optimization shows performance improvements even for n_threads <= 8 cases.
Resurect TSAN (Thread Sanitizer) check so that we can avoid doing expensive read-modify-write
in the normal case and just use thread-fence as originally intended.
---
Here is the original description and suggestions from Willy Tarreau :
There's currently some false sharing between n_barrier and
n_barrier_passed that is amplified in ggml_barrier() by the fact that
all threads need to increment n_barrier when entering, while all
previous threads continue to read n_barrier_passed, waiting for the last
one to release them all. The side effect is that all these readers are
slowing down all new threads by making the cache line bounce back and
forth between readers and writers.
Just placing them in two distinct cache lines is sufficient to boost
the performance by 21% on a 80-core ARM server compared to the
no-openmp version, and by 3% compared to the openmp version.
Note that the variables could have been spread apart in the structure
as well, but it doesn't seem that the size of this threadpool struct is
critical so here we're simply aligning them.
Finally, the same issue was present when leaving the barrier since all
threads had to update the n_barrier_passed counter, though only one
would add a non-zero value. This alone is responsible for half of the
cost due to undesired serialization.
It might be possible that using a small array of n_barrier counters
could make things even faster on many-core systems, but it would likely
complicate the logic needed to detect the last thread.
Co-authored-by: Willy Tarreau <w@1wt.eu>
* AVX512 version of ggml_gemm_q4_0_8x8_q8_0
* Remove zero vector parameter passing
* Rename functions and rearrange order of macros
* Edit commments
* style : minor adjustments
* Update x to start from 0
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* threadpool: skip polling for unused threads
Currently all threads do N polling rounds even if only 1 thread is active (n_threads_cur == 1).
This commit adds a check to skip the polling for unused threads (ith >= n_threads_cur).
n_threads_cur is now an atomic_int to explicitly tell thread sanitizer that it is written
from one thread and read from other threads (not a race conditions).
* threadpool: further simplify and improve ggml_barrier
Avoid using strict memory order while polling, yet make sure that all threads go through
full memory barrier (memory fence) on ggml_barrier entrace and exit.
* threads: add simple barrier test
This test does lots of small, parallel matmul ops where the barriers in between dominate the overhead.
* threadpool: improve thread sync for new-graphs
Using the same tricks as ggml_barrier. All the polling is done with relaxed memory order
to keep it efficient, once the new graph is detected we do full fence using read-modify-write
with strict memory order.
* threadpool: improve abort handling
Do not use threadpool->ec (exit code) to decide whether to exit the compute loop.
threadpool->ec is not atomic which makes thread-sanitizer rightfully unhappy about it.
Instead introduce atomic threadpool->abort flag used for this. This is consistent with
how we handle threadpool->stop or pause.
While at it add an explicit atomic_load for n_threads_cur for consistency.
* test-barrier: release threadpool before releasing the context
fixes use-after-free detected by gcc thread-sanitizer on x86-64
for some reason llvm sanitizer is not detecting this issue.
* sycl : update support condition to im2col
Signed-off-by: Alberto Cabrera <alberto.cabrera@codeplay.com>
* Added TODO to remind supporting FP32 im2col
---------
Signed-off-by: Alberto Cabrera <alberto.cabrera@codeplay.com>
* Overlap cmdbuffer creation and cmdbuffer execution in Vulkan backend by submitting smaller cmdbuffers early.
* fix compile issues
* Fix issues where the last submit wasn't executed or handled properly.
* remove trailing whitespace
* Repair GGML_VULKAN_CHECK_RESULTS
* Increase submit counter only if actual work has been submitted and increase submit count to 100.
* Fix some nodes are not checked with GGML_VULKAN_CHECK_RESULTS enabled.
* add check malloc result on device
* update for review comments, check all malloc_device() result
---------
Co-authored-by: arthw <14088817+arthw@users.noreply.github.com>
sin and cos failed test-backend-ops because they
tried to dereference a context pointer that is null
on dry runs.
This commit prevents that segfault.
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
test-backend-ops fails because ggml_cont aborts
when invoked passing an unsupported type.
This commit makes ggml_cont tests pass
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* tests: add gradient checking to test-backend-ops
* remove old comment
* reorder includes
* adjust SIN/COS parameters
* add documentation, use supports_op if possible
* ggml_cont: fix issue with transposed tensors when one dimension is 1
when using multiple threads, it is not enough
to check for the tensors to be contiguous for
ggml_compute_forward_dup_same_cont to work correctly.
The tensors strides also need to match.
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* Add ggml_cont tests
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* Remove dead code
it isn't possible to reach this code because
all these functions are invoked by ggml_compute_forward_dup
if and only if src0->type != dst->type
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
* Make ggml_compute_forward_dup_same_cont work with contiguous tensors
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
---------
Signed-off-by: Salvatore Mesoraca <s.mesoraca16@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* Improve Vulkan shader builds system
- Add dependency to vulkan-shaders-gen to rebuild shaders when changing the shader compilation utility.
- Add option to generate debug info for Vulkan shaders to provide shader source to Vulkan shader profiling tools
* remove not required self dependency
* ggml-quants : 1.625 bpw ternary packing for BitNet 1.58b
* ggml-quants : faster 1.625 bpw AVX2 vec_dot
Not using a lookup table anymore makes it match q4_0 speed.
* gguf-py : fix formatting
* llama : remove spaces on empty line
* ggml-quants : subtract 1 when back in epi8
This makes the 1.625 bpw type go faster than q4_0. Still not the fastest.
* ggml-quants : Q2_2 now faster than Q4_K on with AVX2
* ggml-quants : cleanup Q1_3 code formatting
* ggml-quants : ARM NEON vec_dot for q2_2 and q1_3
* ggml-quants : use ceiling division when quantizing q1_3
* convert-hf : simplify BitNet pre-quantization
This still results in the exact same tensor weights and scales,
but it reveals some weirdness in the current algorithm.
* convert-hf : allow converting the weird BitNet 1.3B
Its FFN size is 5460 which is not convenient.
The offending tensors are kept in F16,
which makes the final model 5.01 bpw.
* bitnet : replace 1.58b with b1.58, as in the paper
* ggml-quants : fix build failure on Windows
* ggml-quants : attempt to fix Arm 32-bit support
* ggml : add some informative comments in q1_3 vec_dot
* ggml : add TQ1_0 and TQ2_0 ternary quantization types
* ggml : even faster TQ2_0
* ggml : also faster TQ1_0
Same optimization as for TQ2_0 by offsetting the sum instead of the weights.
This makes TQ1_0 almost as fast as Q8_0 on AVX2.
* ggml : fix build issues in certain environments
* ggml : add NEON vec_dot implementation for TQ1_0 and TQ2_0
* ggml : avoid directly using vmlal_high_s8, for 32-bit ARM compat
The compiler seems smart enough to use the same instruction
even when using vget_high_s8 instead.
* ggml : remove q1_3 and q2_2
No more 1.625 bpw and 2.000 bpw,
now instead using 1.6875 bpw and 2.0625 bpw
with TQ1_0 and TQ2_0, respectively.
* llama : remove the separate scale tensors of BitNet b1.58
They won't be needed, since the remaining ternary quant types have
built-in scales.
* ggml-quants : rename fields of TQ1_0 and TQ2_0 structs for consistency
* ggml-quants : allow using vdotq_s32 in TQ2_0 vec_dot
Not yet tested on hardware which supports it,
might not work or might not even compile. But also it might.
It should make the performance better on recent ARM CPUs.
* ggml-quants : remove comment about possible format change of TQ2_0
Making it slightly more convenient for AVX512
but less convenient for everything else is not worth the trouble.
* gguf-py : Numpy (de)quantization for TQ1_0 and TQ2_0
* ggml-quants : use roundf instead of nearest_int for TQ1_0 and TQ2_0
This does not change anything for ternary models,
since their values should never end up being in halfway cases anyway.
* convert : allow direct conversion to TQ1_0 and TQ2_0
The token embeddings and output tensors are kept in F16
to allow quantizing them to Q4_K and Q6_K with llama-quantize.
* llama : handle fallback for TQ1_0 and TQ2_0 with Q4_0
Q4_0 is not completely symmetric (so not lossless for ternary models),
but it should be good enough.
* ggml-quants : allow using ARM dot product instructions for TQ1_0
* ggml-quants : deduplicate TQ1_0 and TQ2_0 __ARM_FEATURE_DOTPROD support
* ggml : remove unused ggml_mul special case
It would otherwise conflict with the more general
optimization coming with Mamba-2.
* ggml : handle TQ1_0 and TQ2_0 in dequantization-based operators
* test-backend-ops : add TQ1_0 and TQ2_0 comments for later
Not yet adding uncommented, because some backends like SYCL and Metal
do not properly handle unknown types in supports_op for GGML_OP_MUL_MAT.
(and Metal also doesn't handle it with GGML_OP_GET_ROWS)
Support for TQ1_0 and TQ2_0 for other backends than CPU
will be added in follow-up pull requests.
* Add AVX2 based implementations for quantize_q8_0_4x8, ggml_gemv_q4_0_8x8_q8_0 and ggml_gemm_q4_0_8x8_q8_0 functions
* Update code to fix issues occuring due to non alignment of elements to be processed as multiple of 16 in MSVC
* Update comments and indentation
* Make updates to reduce number of load instructions
* Introduce ggml_compute_threadpool
- OpenMP functional: check
- Vanilla ggml functional: Check
- ggml w/threadpool functional: Check
- OpenMP no regression: No glaring problems
- Vanilla ggml no regression: No glaring problems
- ggml w/threadpool no regression: No glaring problems
* Minor fixes
* fixed use after release bug
* fixed a harmless race condition
* Fix Android bulid issue
* fix more race conditions
* fix deadlock for cases where cgraph.n_nodes == 1
and fix --poll case
* threadpool: use cpu_get_num_math to set the default number of threadpool threads
This way we avoid using E-Cores and Hyperthreaded siblings.
* bench: create fresh threadpool for each test
For benchmarking it's better to start a fresh pool for each test with the exact number of threads
needed for that test. Having larger pools is suboptimal (causes more load, etc).
* atomics: always use stdatomics with clang and use relaxed memory order when polling in ggml_barrier
This also removes sched_yield() calls from ggml_barrier() to match OpenMP behavior.
* threadpool: make polling the default to match openmp behavior
All command line args now allow for setting poll to 0 (false).
* threadpool: do not wakeup threads in already paused threadpool
* fix potential race condition in check_for_work
* threadpool: do not create two threadpools if their params are identical
* threadpool: reduce pause/resume/wakeup overhead in common cases
We now start threadpool in paused state only if we have two.
The resume is now implicit (ie new work) which allows for reduced locking and context-switch overhead.
* threadpool: add support for hybrid polling
poll params (--poll, ...) now specify "polling level", i.e. how aggresively we poll before waiting on cond.var.
poll=0 means no polling, 1 means poll for 128K rounds then wait, 2 for 256K rounds, ...
The default value of 50 (ie 50x128K rounds) seems like a decent default across modern platforms.
We can tune this further as things evolve.
* threadpool: reduce the number of barrier required
New work is now indicated with an atomic counter that is incremented for
each new graph that needs to be computed.
This removes the need for extra barrier for clearing the "new_work" and
removes the special case for trivial graphs.
* threadpool: remove special-casing for disposable threadpools
With the efficient hybrid polling there is no need to make disposable pools any different.
This simplifies the overall logic and reduces branching.
Include n_threads in debug print for disposable threadpool.
Declare pause and stop flags as atomic_bool
This doesn't actually generate any memory barriers and simply informs
the thread sanitizer that these flags can be written & read by different
threads without locking.
* threadpool: do not clear barrier counters between graphs computes (fixes race with small graphs)
This fixes the race condition with very small graphs where the main thread happens to
start a new graph while the workers are just about to exit from barriers.
* threadpool: use relaxed order for chunk sync
Full memory barrier is an overkill for this since each thread works on different chunk
* threadpool: remove abort_callback from threadpool state
* threadpool: better naming for thread/cpumask releated functions
* threadpool: consistent use of int type for n_threads params
* threadpool: add support for ggml_threadpool_params_default/init
Also removes the need for explicit mask_specified param.
all-zero cpumask means use default (usually inherited) cpu affinity mask.
* threadpool: move typedef into ggml.h
* threadpool: fix apply_priority() function name
* threadpool: fix swift wrapper errors due to n_threads int type cleanup
* threadpool: enable --cpu-mask and other threadpool related options only if threadpool is enabled
* threadpool: replace checks for compute_thread ret code with proper status check
* threadpool: simplify threadpool init logic and fix main thread affinity application
Most of the init code is now exactly the same between threadpool and openmp.
* threadpool: update threadpool resume/pause function names
* threadpool: enable openmp by default for now
* threadpool: don't forget to free workers state when omp is enabled
* threadpool: avoid updating process priority on the platforms that do not require it
On Windows we need to change overall process priority class in order to set thread priorities,
but on Linux, Mac, etc we do not need to touch the overall process settings.
* threadpool: update calling thread prio and affinity only at start/resume
This avoids extra syscalls for each graph_compute()
* llama-bench: turn threadpool params into vectors, add output headers, etc
* llama-bench: add support for cool off between tests --delay
This helps for long running tests on platforms that are thermally limited (phones, laptops, etc).
--delay (disabled by default) introduces the sleep for N seconds before starting each test.
* threadpool: move process priority setting into the apps (bench and cli)
This avoids changing the overall process priority on Windows for the apps
that use ggml/llama.cpp directy.
* threadpool: move all pause/resume logic into ggml
* threadpool: futher api cleanup and prep for future refactoring
All threadpool related functions and structs use ggml_threadpool prefix.
* threadpool: minor indent fixes
* threadpool: improve setprioty error message
* Update examples/llama-bench/llama-bench.cpp
Co-authored-by: slaren <slarengh@gmail.com>
* threadpool: fix indent in set_threadpool call
* use int32_t for n_thread type in public llama.cpp API
* threadpool: use _new and _free instead of _create and _release
* fix two more public APIs to use int32_t for n_threads
* build: set _GNU_SOURCE for Adroid
---------
Co-authored-by: Max Krasnyansky <quic_maxk@quicinc.com>
Co-authored-by: fmz <quic_fzaghlou@quic.com>
Co-authored-by: Max Krasnyansky <max.krasnyansky@gmail.com>
Co-authored-by: slaren <slarengh@gmail.com>
* llama : advanced batch splits
This includes equal-sequence-length batch splits which are useful
to simplify recurrent model operators.
* llama : always make recurrent state slots contiguous
* ggml : simplify mamba operators
* llama : fix integer signedness mixing
* llama : logits_all has priority over batch->logits
Otherwise, the server embeddings tests failed.
This was likely an existing problem but was only detected here
because of an additional assertion.
* llama : apply suggestions
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* llama : fix t5 segfault
* llama : fix Mamba session save and restore
* llama : minor cosmetic changes
* llama : rename llama_reorder_outputs to llama_output_reorder
Also move it closer to llama_output_reserve.
* llama : fix pooled embeddings when using batches with equal_seqs
* minor : add struct members for clarity
ggml-ci
* llama : fix T5 segfault again
* llama : fix Mamba pooled embeddings with multiple sequences
Until the pooled embeddings are refactored to allow splitting
across ubatches for causal embeddings,
recurrent models can only process a single sequence per ubatch
when calculating pooled embeddings.
* llama : add llama_model_is_recurrent to simplify figuring that out
This will make it easier to more cleanly support RWKV-v6 and Mamba-2.
* llama : fix simple splits when the batch contains embeddings
---------
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
* llava: Add ACC OP for GPU acceleration to the Vulkan backend in the LLAVA CLIP model.
- The CLIP model now prioritizes the Vulkan backend over the CPU when vulkan available.
- A GGML_OP_ACC shader has been added.
- The encoding performance of the CLIP model improved from 4.2s on the CPU to 0.9s on the GPU.
Signed-off-by: Changyeon Kim <cyzero.kim@samsung.com>
* fix-up coding style.
Signed-off-by: Changyeon Kim <cyzero.kim@samsung.com>
* Fix-up the missing initial parameter to resolve the compilation warning.
Signed-off-by: Changyeon Kim <cyzero.kim@samsung.com>
* [fix] Add missing parameters.
Signed-off-by: Changyeon Kim <cyzero.kim@samsung.com>
* [fix] Use nb1 and nb2 for dst.
Signed-off-by: Changyeon Kim <cyzero.kim@samsung.com>
* Fix check results ggml_acc call
---------
Signed-off-by: Changyeon Kim <cyzero.kim@samsung.com>
Co-authored-by: 0cc4m <picard12@live.de>
* fallback mmvq to mul_mat
* mmvq in cuda path
* Update ggml/src/ggml-sycl.cpp
Co-authored-by: Alberto Cabrera Pérez <alberto.cabrera@codeplay.com>
---------
Co-authored-by: Alberto Cabrera Pérez <alberto.cabrera@codeplay.com>
* ggml : move rope type enum to ggml.h
This commit moves the `llama_rope_type` enum from `llama.h` to
`ggml.h` and changes its name to `ggml_rope_type`.
The motivation for this change is to address the TODO in `llama.h` and
use the enum in ggml.
Note: This commit does not change the `mode` parameter to be of type
`enum ggml_rope_type`. The name `mode` and its usage suggest that it
might be more generic and possibly used as a bit field for multiple
flags. Further investigation/discussion may be needed to determine
if `mode` should be restricted to RoPE types.
* squash! ggml : move rope type enum to ggml.h
This commit removes GGML_ROPE_TYPE_NONE and GGML_ROPE_TYPE_GLM from
ggml.h, and back the llama_rope_type enum.
I've kept the assert for GGML_ROPE_TYPE_GLM as I'm not sure if it is
safe to remove it yet.
* squash! ggml : move rope type enum to ggml.h
This commit removes the enum ggml_rope_type from ggml.h and replaces it
with a define (GGML_ROPE_TYPE_NEOX). This define is used in the code to
check if the mode is set to GPT-NeoX. Also the enum llama_rope_type has
been updated to reflect this change.
* squash! ggml : move rope type enum to ggml.h
This commit contains a suggestion enable the GGML_ROPE_TYPE_NEOX
macro/define to be passed to the shader compiler.
* squash! ggml : move rope type enum to ggml.h
This commit fixes the editorconfig-checker warnings.
* squash! ggml : move rope type enum to ggml.h
Update comment for ggml_rope function.
* Revert "squash! ggml : move rope type enum to ggml.h"
This reverts commit 6261222bd0.
* squash! ggml : move rope type enum to ggml.h
Add GGML_ROPE_TYPE_NEOX to rope_common.comp.
* remove extra line
---------
Co-authored-by: slaren <slarengh@gmail.com>
* Optimize Vulkan backend for better CPU performance and less GPU synchronization overhead.
- Allocation overhead for the temporary std::vectors was easily detectable with a sampling profiler and simple to remove.
- ggml_vk_sync_buffer introduce a full pipeline sync which has a significant cost on the GPU side, sometimes larger than the actual kernel execution. Adding only barriers for shader read/writes and transfers seems to be sufficient looking at the code which either launches compute kernels or copies tensors.
* Fix small typo
---------
Co-authored-by: 0cc4m <picard12@live.de>
* ggml: use vulkan as gpu backend when available
Signed-off-by: Matt Stephenson <mstephenson6@users.noreply.github.com>
* whisper: enable using vk as default buffer type
Signed-off-by: Matt Stephenson <mstephenson6@users.noreply.github.com>
---------
Signed-off-by: Matt Stephenson <mstephenson6@users.noreply.github.com>
When using CMake to build with Vulkan support, compiling
vulkan-shaders-gen fails due to missing a CMakeLists.txt specification
to link vulkan-shaders-gen with the threading library, resulting in the
following error.
[5/172] Linking CXX executable bin/vulkan-shaders-gen
FAILED: bin/vulkan-shaders-gen
: && /usr/bin/c++ ggml/src/vulkan-shaders/CMakeFiles/vulkan-shaders-gen.dir/vulkan-shaders-gen.cpp.o -o bin/vulkan-shaders-gen && :
ld: error: undefined symbol: pthread_create
>>> referenced by vulkan-shaders-gen.cpp
>>> ggml/src/vulkan-shaders/CMakeFiles/vulkan-shaders-gen.dir/vulkan-shaders-gen.cpp.o:(std::__1::__libcpp_thread_create[abi:se180100](pthread**,
>>> void* (*)(void*), void*))
c++: error: linker command failed with exit code 1 (use -v to see invocation)
[6/172] Generating build details from Git
-- Found Git: /usr/local/bin/git (found version "2.45.2")
ninja: build stopped: subcommand failed.
Add the CMakeLists.txt specification to link vulkan-shaders-gen with the
threading library and fix the above error.
Fixes#8834
* Fix compilation issue in `vulkan-shaders-gen`
e31a4f6797 broke compilation on w64devkit. Including `algorithm` seems to fix that.
* Guard it under `#ifdef _WIN32`
* Fix Vulkan repeat op
* Implement Vulkan concat op
* Delete old Vulkan shader generator
* Implement Vulkan im2col op
* Implement Vulkan unary gelu_quick op
* Implement Vulkan group_norm op
* Implement Vulkan timestep_embedding op
* Implement Vulkan upscale op
* Fix Vulkan vk_context tensor extra index issue
* Fix Vulkan matmul shader parameter bug
* Properly fix Vulkan matmul shader parameter bug
* Add Vulkan ADD f16 + f32 -> f16 operator support
* Implement Vulkan tanh op
* Fix Vulkan group count too large Validation error on non-Nvidia GPUs
* Throw error when too much memory is requested
* Fix another Vulkan group count too large Validation error on non-Nvidia GPUs
* Fix matmul MMQ condition
* Implement Vulkan pad op
* Fix Vulkan crash when tensor is used multiple times in a compute graph
* Add Vulkan CONCAT f16 + f16 -> f16 op
* Add Vulkan LEAKY_RELU op
This commit moves the comment for the c parameter from ggml_rope to
ggml_rope_ext. The comment is currently incorrect as ggml_rope does not
have a c parameter (freq_factors tensor).
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* ggml : reading the runtime sve config of the cpu
* change to one time init to prevent performance drop
* prefix variable to avoid possible conflicts
* revert xxhash fix and add brackets
---------
Co-authored-by: domke <673751-domke@users.noreply.gitlab.com>
* add truncate_bf16
* truncate intermediate fp32 if converting bf16 to bf16
* fix masking in __compute_fp32_to_bf16
* np.int16 no longer used
* missing cast and additional numpy 2.x fix
* ggml-impl : do not flush bf16 subnormals to zero
* ggml : add reference fp32 to bf16 conversion
The fast version is no longer equivalent for all platforms
because of the handling of subnormal values.
* gguf-py : remove flush to zero for bf16 subnormals
* gguf-py : remove float32 truncation to bf16
Rounding achieves the same thing in the cases where this was used.
* missed prototype update in merge
* merge cleanup
---------
Co-authored-by: Francis Couture-Harpin <git@compilade.net>
* Adding support for unified memory
* adding again the documentation about unified memory
* refactoring: Moved the unified memory code in the correct location.
* Fixed compilation error when using hipblas
* cleaning up the documentation
* Updating the documentation
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
* adding one more case where the PR should not be enabled
---------
Co-authored-by: matteo serva <matteo.serva@gmail.com>
Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
In these codes, we want to retain the value that they previously held
when mask[i] is false. So we should use undisturbed. With the default
agnostic policy of rvv intrinsic, these values can be held or be
written with 1s.
Co-authored-by: carter.li <carter.li@starfivetech.com>
* chore: Fix compiler warnings, add help text, improve CLI options
* Add prototypes for function definitions
* Invert logic of --no-clean option to be more intuitive
* Provide a new help prompt with clear instructions
* chore : Add ignore rule for vulkan shader generator
Signed-off-by: teleprint-me <77757836+teleprint-me@users.noreply.github.com>
* Update ggml/src/vulkan-shaders/vulkan-shaders-gen.cpp
Co-authored-by: 0cc4m <picard12@live.de>
* chore : Remove void and apply C++ style empty parameters
* chore : Remove void and apply C++ style empty parameters
---------
Signed-off-by: teleprint-me <77757836+teleprint-me@users.noreply.github.com>
Co-authored-by: 0cc4m <picard12@live.de>
* Update doc for MUSA
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* Add GGML_MUSA in Makefile
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* Add GGML_MUSA in CMake
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* CUDA => MUSA
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* MUSA adds support for __vsubss4
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
* Fix CI build failure
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
---------
Signed-off-by: Xiaodong Ye <xiaodong.ye@mthreads.com>
Apply a loop tiling technique to the generic path, which provides
performance upside for ISAs with enough registers to take advantage
of it. Also helps the compiler optimize this path.
* Add support for float16 tensors in 1d pooling operations
* Add support for float16 input tensors in 2d pooling operations
* code cleanup
remove unnecessary casting during srow ptr initialization
---------
Co-authored-by: vanaka11 <vanaka1189@gmail.com>
This prevents invalid frees when destroying a partially initialized
vk_buffer_struct. For example, this could happen in ggml_vk_create_buffer
when running out of device memory.
Co-authored-by: Tony Wasserka <neobrain@users.noreply.github.com>
This commit removes an UNUSED macro call that is not needed as the
variable n0 is used in the code and will not produce a warning.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
`ggml_init` can fail if no unused context is found. In that case, a NULL-pointer deref will happen later in the code during a call to `ggml_set_on_alloc`.
This fixes it by bailing out if no context is found.
* Improvements for Windows with Snapdragon X
* Revert "Improvements for Windows with Snapdragon X"
This reverts commit bf21397ae5.
* Improvements for Windows with Snapdragon X
* WOA build clarifications
* WIndows on ARM build clarifications
* cmake build for Windows clarifications
* Update docs/build.md
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
---------
Co-authored-by: AndreasKunar <andreaskmsn.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
The check gating the use of `__builtin_amdgc_sdot4` specifically checks for gfx1030. This causes a severe perf regression for anything gfx103? that's not gfx1030 and not using `HSA_OVERRIDE_GFX_VERSION` (if you've built ROCm to support it). We already have a generic RDNA2 define, let's use it.
* [CANN] Add Ascend NPU backend
Ascend is a full-stack AI computing infrastructure for industry
applications and services based on Huawei Ascend processors and
software.
CANN (Compute Architecture of Neural Networks), developped by
Huawei, is a heterogeneous computing architecture for AI.
Co-authored-by: wangshuai09 <391746016@qq.com>
* delete trailing whitespaces
* Modify the code based on review comment
* Rename LLAMA_CANN to GGML_CANN
* Make ggml-common.h private
* add ggml_cann prefix for acl funcs
* Add logging for CANN backend
* Delete Trailing whitespace
---------
Co-authored-by: wangshuai09 <391746016@qq.com>
* lora: load to devide buft
* add patch tensor function
* correct tensor patch
* llama_lora_adapter_apply
* correct ggml_backend_tensor_copy
* add llm_build_mm
* fix auto merge
* update based on review comments
* add convert script
* no more transpose A
* add f16 convert
* add metadata check
* add sanity check
* fix ftype
* add requirements
* fix requirements
* fix outfile
* conversion: only allow selected models
* fix types
* cuda : do not use dmmv if the tensor does not have enough cols
* llama : lora fixes
* do not disable mmap with lora
Co-authored-by: slaren <slarengh@gmail.com>
* llm_build_lora_mm_id
* convert_lora : MoE LoRA conversion support
* convert_lora : prefer safetensors, similarly to convert_hf
* convert_hf : simplify modify_tensors for InternLM2
* convert_lora : lazy conversion
* llama : load and use alpha from LoRA adapters
* llama : use llm_build_lora_mm in most model graphs
* auto scale
* Revert "auto scale"
This reverts commit 42415a4874.
* remove redundant params
* Apply suggestions from code review
Co-authored-by: slaren <slarengh@gmail.com>
* change kv metadata
* move add_type to __init__
* convert_hf : move add_type to main()
* convert_lora : use the GGUFWriter from Model instead of overwriting it
---------
Co-authored-by: slaren <slarengh@gmail.com>
Co-authored-by: Francis Couture-Harpin <git@compilade.net>
This commit adds a macro guard to pragma GCC to avoid the following
warning on windows:
```console
C:\llama.cpp\ggml\src\ggml-aarch64.c(17,9): warning C4068:
unknown pragma 'GCC' [C:\lama.cpp\build\ggml\src\ggml.vcxproj]
```
* fix part of mul_mat_id
* skip the bfloat 16 sycl ut
Signed-off-by: Chen Xi <xi2chen@intel.com>
---------
Signed-off-by: Chen Xi <xi2chen@intel.com>
Co-authored-by: Meng, Hengyu <hengyu.meng@intel.com>
Co-authored-by: Chen Xi <xi2chen@intel.com>
* cuda : suppress 'noreturn' warn in no_device_code
This commit adds a while(true) loop to the no_device_code function in
common.cuh. This is done to suppress the warning:
```console
/ggml/src/ggml-cuda/template-instances/../common.cuh:346:1: warning:
function declared 'noreturn' should not return [-Winvalid-noreturn]
346 | }
| ^
```
The motivation for this is to reduce the number of warnings when
compilng with GGML_HIPBLAS=ON.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* squash! cuda : suppress 'noreturn' warn in no_device_code
Update __trap macro instead of using a while loop to suppress the
warning.
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
---------
Signed-off-by: Daniel Bevenius <daniel.bevenius@gmail.com>
* Arm AArch64: optimized GEMV and GEMM kernels for q4_0_q8_0, and q8_0_q8_0 quantization
* Arm AArch64: add optimized GEMV and GEMM asm kernels for q4_0_q8_0 quantization and refactor code to address llama.cpp pr#5780 suggestions
* Arm AArch64: add optimized GEMV and GEMM asm kernels for q4_0_q8_0 quantization and refactor code to address llama.cpp pr#5780 suggestions
* Arm AArch64: add optimized GEMV and GEMM asm kernels for q4_0_q8_0 quantization and refactor code to address llama.cpp pr#5780 suggestions
* Arm AArch64: add optimized GEMV and GEMM asm kernels for q4_0_q8_0 quantization and refactor code to address llama.cpp pr#5780 suggestions
* Arm AArch64: add copyright claim only to ggml-aarch64.cpp and ggml-aarch64.h files
* Arm AArch64: minor code refactoring for rebase
* Arm AArch64: minor code refactoring for resolving a build issue with cmake
* Arm AArch64: minor code refactoring to split the Q4_0_AARC64 type into three separate types: Q4_0_4_4, Q4_0_4_8, and Q4_0_8_8
* Arm AArch64: minor code change for resolving a build issue with server-windows
* retrigger checks
* Arm AArch64: minor code changes for rebase
* Arm AArch64: minor changes to skip the pr#7433 vec_dot code for arm cpus with SVE VL not equal to 256 bits
* Arm AArch64: remove stale LLAMA_QKK_64 from CMakeLists.txt and delete build.zig
* Arm AArch64: add reference scalar gemm and gemv, and avoid dynamic memory allocations during quantization for Q4_0_4_4, Q4_0_4_8, and Q4_0_8_8
* Arm AArch64: add multithreaded quantization support for the new types: Q4_0_4_4, Q4_0_4_8, and Q4_0_8_8
* Arm AArch64: minor code refactoring
* Arm AArch64: simplify logic for calling gemm and gemv functions in ggml_compute_forward_mul_mat
* Arm AArch64: minimize changes in ggml_compute_forward_mul_mat
* Arm AArch64: minor code refactoring, and add reference scalar code to quantize routines for new quant types
* Arm AArch64: minor code refactoring
* Arm AArch64: minor code refactoring
* Arm AArch64: minor code refactoring
* rebase on the latest master commit 3fd62a6 and adapt to the new directory structure
* Arm AArch64: remove a redundant comment
* Arm AArch64: add pragma in ggml-aarch64.c to turn -Woverlength-strings warning off
* Arm AArch64: use __aarch64__ check to guard 64-bit neon kernels
* Arm AArch64: update docs/build.md README to include compile time flags for buiilding the Q4_0_4_4 quant type
* conv transpose 1d passing test for 1d input and kernel
* working for different input and output channel counts, added test for variable stride
* initial draft appears to work with stride other than 1
* working with all old and new conv1d tests
* added a test for large tensors
* removed use cuda hardcoding
* restored test-conv-transpose.c
* removed unused arugments, and fixed bug where test failure would cause subsequent tests to fail
* fixed accumulator bug
* added test to test-backend-ops
* fixed mistake
* addressed review
* fixed includes
* removed blank lines
* style and warning fixes
* return failure when test fails
* fix supports_op
---------
Co-authored-by: slaren <slarengh@gmail.com>
* py : type-check all Python scripts with Pyright
* server-tests : use trailing slash in openai base_url
* server-tests : add more type annotations
* server-tests : strip "chat" from base_url in oai_chat_completions
* server-tests : model metadata is a dict
* ci : disable pip cache in type-check workflow
The cache is not shared between branches, and it's 250MB in size,
so it would become quite a big part of the 10GB cache limit of the repo.
* py : fix new type errors from master branch
* tests : fix test-tokenizer-random.py
Apparently, gcc applies optimisations even when pre-processing,
which confuses pycparser.
* ci : only show warnings and errors in python type-check
The "information" level otherwise has entries
from 'examples/pydantic_models_to_grammar.py',
which could be confusing for someone trying to figure out what failed,
considering that these messages can safely be ignored
even though they look like errors.
