948ef47106
Some checks failed
Bindings Tests (Go) / ubuntu-22 (push) Failing after 31s
CI (sycl) / ubuntu-22-cmake-sycl (linux/arm/v7, icx, icpx, ON) (push) Failing after 3s
CI (sycl) / ubuntu-22-cmake-sycl (linux/ppc64le, icx, icpx, ON) (push) Failing after 2s
CI (sycl) / ubuntu-22-cmake-sycl-fp16 (linux/amd64, icx, icpx, ON) (push) Failing after 3s
CI (sycl) / ubuntu-22-cmake-sycl-fp16 (linux/arm/v7, icx, icpx, ON) (push) Failing after 3s
CI (sycl) / ubuntu-22-cmake-sycl-fp16 (linux/arm64, icx, icpx, ON) (push) Failing after 3s
CI (sycl) / ubuntu-22-cmake-sycl-fp16 (linux/ppc64le, icx, icpx, ON) (push) Failing after 3s
CI (vad) / vad (push) Failing after 3s
CI (wasm) / emscripten (Release) (push) Failing after 4s
Bindings Tests (Ruby) / ubuntu-22 (push) Failing after 37s
CI (android) / android (push) Failing after 16s
CI (android) / android_java (push) Failing after 3s
CI (clang) / ubuntu-22-clang (linux/amd64, Debug) (push) Failing after 3s
CI (clang) / ubuntu-22-clang (linux/amd64, Release) (push) Failing after 2s
CI (clang) / ubuntu-22-clang (linux/ppc64le, Debug) (push) Failing after 3s
CI (clang) / ubuntu-22-clang (linux/ppc64le, Release) (push) Failing after 3s
CI (cpu) / ggml-ci-x64-cpu-low-perf (push) Failing after 3s
CI (cpu) / ggml-ci-x64-cpu-high-perf (push) Failing after 2s
CI (gcc) / ubuntu-22-gcc (linux/amd64, Debug) (push) Failing after 3s
CI (gcc) / ubuntu-22-gcc (linux/amd64, Release) (push) Failing after 3s
CI (gcc) / ubuntu-22-gcc (linux/ppc64le, Debug) (push) Failing after 3s
CI (gcc) / ubuntu-22-gcc (linux/ppc64le, Release) (push) Failing after 4s
CI (gcc) / ubuntu-22-gcc-arm-v7 (linux/arm/v7, Debug) (push) Failing after 3s
CI (gcc) / ubuntu-22-gcc-arm-v7 (linux/arm/v7, Release) (push) Failing after 3s
CI (quantize) / quantize (push) Failing after 3s
CI (sanitize) / ubuntu-22-gcc-sanitized (ADDRESS) (push) Failing after 3s
CI (sanitize) / ubuntu-22-gcc-sanitized (THREAD) (push) Failing after 3s
CI (sanitize) / ubuntu-22-gcc-sanitized (UNDEFINED) (push) Failing after 2s
CI (sycl) / ubuntu-22-cmake-sycl (linux/amd64, icx, icpx, ON) (push) Failing after 3s
CI (sycl) / ubuntu-22-cmake-sycl (linux/arm64, icx, icpx, ON) (push) Failing after 3s
Examples WASM / deploy-wasm-github-pages (push) Failing after 3s
Examples Tests / addon_node-ubuntu-22 (20.x) (push) Failing after 3s
Examples Tests / addon_node-ubuntu-22 (22.x) (push) Failing after 2s
CI (coreml) / coreml-base-en (push) Has been cancelled
CI (self-hosted) / gpu-vulkan (push) Has been cancelled
CI (self-hosted) / npu-amd-windows (push) Has been cancelled
CI (self-hosted) / gpu-vulkan-nvidia-cm2 (push) Has been cancelled
CI (clang) / ubuntu-22-clang-arm64 (Debug) (push) Has been cancelled
CI (clang) / ubuntu-22-clang-arm64 (Release) (push) Has been cancelled
CI (self-hosted) / gpu-cuda (push) Has been cancelled
CI (self-hosted) / gpu-vulkan-nvidia-cm (push) Has been cancelled
CI (cpu) / ggml-ci-arm64-cpu-low-perf (push) Has been cancelled
CI (cpu) / ggml-ci-arm64-cpu-high-perf (push) Has been cancelled
CI (cpu) / ggml-ci-arm64-cpu-high-perf-sve (push) Has been cancelled
CI (freebsd) / freeBSD-latest (push) Has been cancelled
CI (self-hosted) / gpu-metal (push) Has been cancelled
CI (gcc) / ubuntu-22-gcc-arm64 (Debug) (push) Has been cancelled
CI (gcc) / ubuntu-22-gcc-arm64 (Release) (push) Has been cancelled
CI (macOS) / macOS-latest (generic/platform=iOS) (push) Has been cancelled
CI (macOS) / macOS-latest (generic/platform=macOS) (push) Has been cancelled
CI (macOS) / macOS-latest (generic/platform=tvOS) (push) Has been cancelled
CI (self-hosted) / npu-amd-linux (push) Has been cancelled
CI (windows) / windows-msys2 (Release, clang, clang-x86_64, CLANG64) (push) Has been cancelled
CI (windows) / windows-msys2 (Release, gcc, ucrt-x86_64, UCRT64) (push) Has been cancelled
Publish Docker image / Push Docker image to Docker Hub (map[dockerfile:.devops/main-cuda.Dockerfile platform:linux/amd64 runs_on:ubuntu-24.04 tag:main-cuda]) (push) Has been cancelled
Publish Docker image / Push Docker image to Docker Hub (map[dockerfile:.devops/main-intel.Dockerfile platform:linux/amd64 runs_on:ubuntu-24.04 tag:main-intel]) (push) Has been cancelled
Publish Docker image / Push Docker image to Docker Hub (map[dockerfile:.devops/main-musa.Dockerfile platform:linux/amd64 runs_on:ubuntu-24.04 tag:main-musa]) (push) Has been cancelled
Publish Docker image / Push Docker image to Docker Hub (map[dockerfile:.devops/main-vulkan.Dockerfile platform:linux/amd64 runs_on:ubuntu-24.04 tag:main-vulkan]) (push) Has been cancelled
Publish Docker image / Push Docker image to Docker Hub (map[dockerfile:.devops/main-vulkan.Dockerfile platform:linux/arm64 runs_on:ubuntu-24.04-arm tag:main-vulkan-arm64]) (push) Has been cancelled
Publish Docker image / Push Docker image to Docker Hub (map[dockerfile:.devops/main.Dockerfile platform:linux/amd64 runs_on:ubuntu-24.04 tag:main]) (push) Has been cancelled
Publish Docker image / Push Docker image to Docker Hub (map[dockerfile:.devops/main.Dockerfile platform:linux/arm64 runs_on:ubuntu-24.04-arm tag:main-arm64]) (push) Has been cancelled
Mark inactive issues stale / mark-issues (push) Failing after 1s
107 lines
11 KiB
C
107 lines
11 KiB
C
#pragma once
|
|
|
|
#define GGML_COMMON_DECL_C
|
|
#include "ggml-common.h"
|
|
|
|
#include "ggml.h"
|
|
|
|
// GGML CPU internal header
|
|
|
|
#ifdef __cplusplus
|
|
extern "C" {
|
|
#endif
|
|
|
|
// Quantization
|
|
void quantize_row_q1_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q2_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q4_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q4_1(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q5_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q5_1(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q8_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q8_1(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
|
|
void quantize_row_mxfp4(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_nvfp4(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
|
|
void quantize_row_q2_K(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q3_K(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q4_K(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q5_K(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q6_K(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_q8_K(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
|
|
void quantize_row_tq1_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_tq2_0(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
|
|
void quantize_row_iq4_nl (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void quantize_row_iq4_xs (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
|
|
// Dot product
|
|
void ggml_vec_dot_q1_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q2_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q4_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q4_1_q8_1(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q5_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q5_1_q8_1(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q8_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
void ggml_vec_dot_mxfp4_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_nvfp4_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
void ggml_vec_dot_q2_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q3_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q4_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q5_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q6_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
void ggml_vec_dot_tq1_0_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_tq2_0_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
void ggml_vec_dot_iq2_xxs_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq2_xs_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq2_s_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq3_xxs_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq1_s_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq1_m_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq4_nl_q8_0 (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq4_xs_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq3_s_q8_K (int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
// Generic implementation
|
|
void quantize_row_q8_0_generic(const float * GGML_RESTRICT x, void * GGML_RESTRICT vy, int64_t k);
|
|
void quantize_row_q8_1_generic(const float * GGML_RESTRICT x, void * GGML_RESTRICT vy, int64_t k);
|
|
void quantize_row_q8_K_generic(const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int64_t k);
|
|
void ggml_vec_dot_q1_0_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q2_0_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q4_0_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q4_1_q8_1_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q5_0_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q5_1_q8_1_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q8_0_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
void ggml_vec_dot_mxfp4_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_nvfp4_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
void ggml_vec_dot_tq1_0_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_tq2_0_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
void ggml_vec_dot_q2_K_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q3_K_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q4_K_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q5_K_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_q6_K_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq2_xxs_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq2_xs_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq2_s_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq3_xxs_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq3_s_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq1_s_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq1_m_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq4_nl_q8_0_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
void ggml_vec_dot_iq4_xs_q8_K_generic(int n, float * GGML_RESTRICT s, size_t bs, const void * GGML_RESTRICT vx, size_t bx, const void * GGML_RESTRICT vy, size_t by, int nrc);
|
|
|
|
#ifdef __cplusplus
|
|
}
|
|
#endif
|