* ggml_qnn_op_config now manager the construction of ggml_qnn_tensor
* wip
* add interface ggml_qnn_op_config
* add ggml_qnn_list_op_config
* add create_tensor and move tensor bind to execute
* wip
* rename: ggml_qnn_list_op_config -> ggml_qnn_matmul_op_config
* add tensortype to allow native tensor
* remove ggml_tensor param at ggml_qnn_tensor::create_tensor
* postpone the tensor id allocation to add_node
* add ggml_qnn_op_config_base
* trival change to reduct the param of function
* split bind_tensors into bind_input_tensors and bind_output_tensors
* implement ggml_qnn_single_op_config::create_tensors
next will set the prameter of transpose
* tensor: add bind buffer
* add parameter tensor type
* implement add_tensor_param
* set qnn_instance only at constructor
* set transpose tensor param
* move create_op_constructor into op-config module
* create QNN_OP_MAT_MUL from ggml_qnn_matmul_op_config
* try fix crash
* fix compiling error at older ndk (r23c)
* fix crash
* fix parameter tensor name
* update tensor dimension assignment and add TODO
* fix mat_mul graph creating
* fix MUL_MAT_256x16x10x1_256x1x10x1_16x1x10x1
* append type to graph cache key
* wip
* fix supported op
* update comment
* disable op other than add and mat_mul
* add convert op to adapt multi input/output format
* disable f16 for cpu backend according to official doc
https://docs.qualcomm.com/bundle/publicresource/topics/80-63442-50/cpu_backend.html#supported-operations
* add supported data types flags in each backend
* remove unused functions
* append output type to graph key
* fix gpu backend by disable the different data type op
* fix cpu backend support ops
* fix duplicated tensor name
* append op name
* suppress warning
* remove unused code
* 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
nullname
Diego Devesa
R0CKSTAR
Andrew Minh Nguyen
Georgi Gerganov
Paul Tsochantaris
SRHMorris
Daniel Bevenius
Johannes Gäßler
Jack Mousseau
bandoti
Ouadie EL FAROUKI
Alberto Cabrera Pérez
Radoslav Gerganov
Salvatore Mesoraca
Borislav Stanimirov
Jeff Bolz
Dan Johansson
Markus Tavenrath
Charles Xu
Dou Xinpeng
Eric Zhang
Max Krasnyansky
Ivan