vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-11 17:44:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 3

sync - fix cmake failing to build with c++11, updated glslc.exe to handle coopmat, sync sdtype count, aarch repack flags

Browse source
This commit is contained in:
Concedo 2024-12-13 17:08:10 +08:00
parent de64b9198c
commit 7e1abf3aaf
5 changed files with 20 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

1
CMakeLists.txt
View file

@ -68,6 +68,7 @@ find_package(Threads REQUIRED)
add_compile_definitions(LOG_DISABLE_LOGS) add_compile_definitions(LOG_DISABLE_LOGS)
add_compile_definitions(GGML_USE_CPU) add_compile_definitions(GGML_USE_CPU)
add_compile_definitions(GGML_USE_CPU_AARCH64)
file(GLOB GGML_SOURCES_CUDA "ggml/src/ggml-cuda/*.cu") file(GLOB GGML_SOURCES_CUDA "ggml/src/ggml-cuda/*.cu")
list(APPEND GGML_SOURCES_CUDA "ggml/src/ggml-cuda/ggml-cuda.cu") list(APPEND GGML_SOURCES_CUDA "ggml/src/ggml-cuda/ggml-cuda.cu")

4
Makefile
View file

@ -54,8 +54,8 @@ ifdef KCPP_DEBUG
CFLAGS = -g -O0 CFLAGS = -g -O0
CXXFLAGS = -g -O0 CXXFLAGS = -g -O0
endif endif
CFLAGS += -I. -Iggml/include -Iggml/src -Iggml/src/ggml-cpu -Iinclude -Isrc -I./include -I./include/CL -I./otherarch -I./otherarch/tools -I./otherarch/sdcpp -I./otherarch/sdcpp/thirdparty -I./include/vulkan -O3 -fno-finite-math-only -std=c11 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE -DGGML_USE_CPU CFLAGS += -I. -Iggml/include -Iggml/src -Iggml/src/ggml-cpu -Iinclude -Isrc -I./include -I./include/CL -I./otherarch -I./otherarch/tools -I./otherarch/sdcpp -I./otherarch/sdcpp/thirdparty -I./include/vulkan -O3 -fno-finite-math-only -std=c11 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE -DGGML_USE_CPU -DGGML_USE_CPU_AARCH64
CXXFLAGS += -I. -Iggml/include -Iggml/src -Iggml/src/ggml-cpu -Iinclude -Isrc -I./common -I./include -I./include/CL -I./otherarch -I./otherarch/tools -I./otherarch/sdcpp -I./otherarch/sdcpp/thirdparty -I./include/vulkan -O3 -fno-finite-math-only -std=c++11 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE -DGGML_USE_CPU CXXFLAGS += -I. -Iggml/include -Iggml/src -Iggml/src/ggml-cpu -Iinclude -Isrc -I./common -I./include -I./include/CL -I./otherarch -I./otherarch/tools -I./otherarch/sdcpp -I./otherarch/sdcpp/thirdparty -I./include/vulkan -O3 -fno-finite-math-only -std=c++11 -fPIC -DLOG_DISABLE_LOGS -D_GNU_SOURCE -DGGML_USE_CPU -DGGML_USE_CPU_AARCH64
ifndef KCPP_DEBUG ifndef KCPP_DEBUG
CFLAGS += -DNDEBUG -s CFLAGS += -DNDEBUG -s
CXXFLAGS += -DNDEBUG -s CXXFLAGS += -DNDEBUG -s

13
ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp
View file

@ -3821,7 +3821,9 @@ static int repack_iq4_nl_to_iq4_nl_4_bl(struct ggml_tensor * t, int interleave_b
GGML_UNUSED(data_size); GGML_UNUSED(data_size);
} }
namespace ggml::cpu::aarch64 { namespace ggml {
namespace cpu {
namespace aarch64 { //ggml::cpu::aarch64
// repack // repack
template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS> template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS>
int repack(struct ggml_tensor *, const void *, size_t); int repack(struct ggml_tensor *, const void *, size_t);
@ -4137,6 +4139,8 @@ static const tensor_traits<block_q4_0, 8, 8> q4_0_8x8_q8_0;
// instance for IQ4 // instance for IQ4
static const tensor_traits<block_iq4_nl, 4, 4> iq4_nl_4x4_q8_0; static const tensor_traits<block_iq4_nl, 4, 4> iq4_nl_4x4_q8_0;
}
}
} // namespace ggml::cpu::aarch64 } // namespace ggml::cpu::aarch64
static const ggml::cpu::tensor_traits * ggml_aarch64_get_optimal_repack_type(const struct ggml_tensor * cur) { static const ggml::cpu::tensor_traits * ggml_aarch64_get_optimal_repack_type(const struct ggml_tensor * cur) {
@ -4210,7 +4214,9 @@ static size_t ggml_backend_cpu_aarch64_buffer_type_get_alignment(ggml_backend_bu
GGML_UNUSED(buft); GGML_UNUSED(buft);
} }
namespace ggml::cpu::aarch64 { namespace ggml {
namespace cpu {
namespace aarch64 { //ggml::cpu::aarch64
class extra_buffer_type : ggml::cpu::extra_buffer_type { class extra_buffer_type : ggml::cpu::extra_buffer_type {
bool supports_op(ggml_backend_dev_t, const struct ggml_tensor * op) override { bool supports_op(ggml_backend_dev_t, const struct ggml_tensor * op) override {
if ( op->op == GGML_OP_MUL_MAT && if ( op->op == GGML_OP_MUL_MAT &&
@ -4257,6 +4263,9 @@ class extra_buffer_type : ggml::cpu::extra_buffer_type {
return nullptr; return nullptr;
} }
}; };
}
}
} // namespace ggml::cpu::aarch64 } // namespace ggml::cpu::aarch64
ggml_backend_buffer_type_t ggml_backend_cpu_aarch64_buffer_type(void) { ggml_backend_buffer_type_t ggml_backend_cpu_aarch64_buffer_type(void) {

4
ggml/src/ggml-cpu/ggml-cpu-traits.cpp
View file

@ -3,10 +3,12 @@
#include "ggml-backend-impl.h" #include "ggml-backend-impl.h"
#include "ggml-backend.h" #include "ggml-backend.h"
namespace ggml::cpu { namespace ggml {
namespace cpu {
tensor_traits::~tensor_traits() {} tensor_traits::~tensor_traits() {}
extra_buffer_type::~extra_buffer_type() {} extra_buffer_type::~extra_buffer_type() {}
}
} // namespace ggml::cpu } // namespace ggml::cpu
bool ggml_cpu_extra_compute_forward(struct ggml_compute_params * params, struct ggml_tensor * op) { bool ggml_cpu_extra_compute_forward(struct ggml_compute_params * params, struct ggml_tensor * op) {

4
ggml/src/ggml-cpu/ggml-cpu-traits.h
View file

@ -15,7 +15,8 @@ bool ggml_cpu_extra_work_size(int n_threads, const struct ggml_tensor * op, size
#ifdef __cplusplus #ifdef __cplusplus
} }
namespace ggml::cpu { namespace ggml {
namespace cpu {
// register in tensor->extra // register in tensor->extra
class tensor_traits { class tensor_traits {
public: public:
@ -30,6 +31,7 @@ class extra_buffer_type {
virtual bool supports_op(ggml_backend_dev_t dev, const struct ggml_tensor * op) = 0; virtual bool supports_op(ggml_backend_dev_t dev, const struct ggml_tensor * op) = 0;
virtual tensor_traits * get_tensor_traits(const struct ggml_tensor * op) = 0; virtual tensor_traits * get_tensor_traits(const struct ggml_tensor * op) = 0;
}; };
}
} // namespace ggml::cpu } // namespace ggml::cpu
// implemented in ggml-cpu.cpp. // implemented in ggml-cpu.cpp.