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

Merge branch 'upstream' into concedo_experimental

Browse source 
# Conflicts:
#	.github/workflows/build.yml
#	README.md
#	common/common.cpp
#	examples/embedding/embedding.cpp
#	examples/json_schema_to_grammar.py
#	examples/llama.android/llama/src/main/cpp/llama-android.cpp
#	examples/llama.swiftui/README.md
#	examples/llama.swiftui/llama.swiftui.xcodeproj/project.pbxproj
#	examples/lookahead/lookahead.cpp
#	examples/parallel/parallel.cpp
#	examples/passkey/passkey.cpp
#	ggml/CMakeLists.txt
#	ggml/src/CMakeLists.txt
#	ggml/src/ggml-cpu/CMakeLists.txt
#	requirements.txt
#	requirements/requirements-all.txt
#	scripts/fetch_server_test_models.py
#	tests/test-chat.cpp
#	tests/test-json-schema-to-grammar.cpp
This commit is contained in:
Concedo 2025-03-06 18:54:58 +08:00
commit ec43d2b147
52 changed files with 5600 additions and 946 deletions
Download patch file Download diff file Expand all files Collapse all files

34
ggml/src/ggml-cpu/ggml-cpu.c
View file

@ -252,9 +252,9 @@ typedef pthread_t ggml_thread_t;
static const size_t CACHE_LINE_SIZE_F32 = CACHE_LINE_SIZE/sizeof(float);
static void ggml_vec_dot_f32(int n, float * restrict s, size_t bs, const float * restrict x, size_t bx, const float * restrict y, size_t by, int nrc);
static void ggml_vec_dot_f16(int n, float * restrict s, size_t bs, ggml_fp16_t * restrict x, size_t bx, ggml_fp16_t * restrict y, size_t by, int nrc);
static void ggml_vec_dot_bf16(int n, float * restrict s, size_t bs, ggml_bf16_t * restrict x, size_t bx, ggml_bf16_t * restrict y, size_t by, int nrc);
static void ggml_vec_dot_f32(int n, float * GGML_RESTRICT s, size_t bs, const float * GGML_RESTRICT x, size_t bx, const float * GGML_RESTRICT y, size_t by, int nrc);
static void ggml_vec_dot_f16(int n, float * GGML_RESTRICT s, size_t bs, ggml_fp16_t * GGML_RESTRICT x, size_t bx, ggml_fp16_t * GGML_RESTRICT y, size_t by, int nrc);
static void ggml_vec_dot_bf16(int n, float * GGML_RESTRICT s, size_t bs, ggml_bf16_t * GGML_RESTRICT x, size_t bx, ggml_bf16_t * GGML_RESTRICT y, size_t by, int nrc);
static const struct ggml_type_traits_cpu type_traits_cpu[GGML_TYPE_COUNT] = {
[GGML_TYPE_F32] = {
@ -1456,7 +1456,7 @@ inline static void ggml_vec_div_f16 (const int n, ggml_fp16_t * z, const ggml_fp
}
}
static void ggml_vec_dot_f32(int n, float * restrict s, size_t bs, const float * restrict x, size_t bx, const float * restrict y, size_t by, int nrc) {
static void ggml_vec_dot_f32(int n, float * GGML_RESTRICT s, size_t bs, const float * GGML_RESTRICT x, size_t bx, const float * GGML_RESTRICT y, size_t by, int nrc) {
assert(nrc == 1);
UNUSED(nrc);
UNUSED(bx);
@ -1499,7 +1499,7 @@ static void ggml_vec_dot_f32(int n, float * restrict s, size_t bs, const float *
*s = sumf;
}
static void ggml_vec_dot_bf16(int n, float * restrict s, size_t bs, ggml_bf16_t * restrict x, size_t bx, ggml_bf16_t * restrict y, size_t by, int nrc) {
static void ggml_vec_dot_bf16(int n, float * GGML_RESTRICT s, size_t bs, ggml_bf16_t * GGML_RESTRICT x, size_t bx, ggml_bf16_t * GGML_RESTRICT y, size_t by, int nrc) {
assert(nrc == 1);
UNUSED(nrc);
UNUSED(bx);
@ -1567,7 +1567,7 @@ static void ggml_vec_dot_bf16(int n, float * restrict s, size_t bs, ggml_bf16_t
*s = sumf;
}
static void ggml_vec_dot_f16(int n, float * restrict s, size_t bs, ggml_fp16_t * restrict x, size_t bx, ggml_fp16_t * restrict y, size_t by, int nrc) {
static void ggml_vec_dot_f16(int n, float * GGML_RESTRICT s, size_t bs, ggml_fp16_t * GGML_RESTRICT x, size_t bx, ggml_fp16_t * GGML_RESTRICT y, size_t by, int nrc) {
assert(nrc == 1);
UNUSED(nrc);
UNUSED(bx);
@ -1611,10 +1611,10 @@ static void ggml_vec_dot_f16(int n, float * restrict s, size_t bs, ggml_fp16_t *
// compute GGML_VEC_DOT_UNROLL dot products at once
// xs - x row stride in bytes
inline static void ggml_vec_dot_f16_unroll(const int n, const int xs, float * restrict s, void * restrict xv, ggml_fp16_t * restrict y) {
inline static void ggml_vec_dot_f16_unroll(const int n, const int xs, float * GGML_RESTRICT s, void * GGML_RESTRICT xv, ggml_fp16_t * GGML_RESTRICT y) {
ggml_float sumf[GGML_VEC_DOT_UNROLL] = { 0.0 };
ggml_fp16_t * restrict x[GGML_VEC_DOT_UNROLL];
ggml_fp16_t * GGML_RESTRICT x[GGML_VEC_DOT_UNROLL];
for (int i = 0; i < GGML_VEC_DOT_UNROLL; ++i) {
x[i] = (ggml_fp16_t *) ((char *) xv + i*xs);
@ -1664,7 +1664,7 @@ inline static void ggml_vec_dot_f16_unroll(const int n, const int xs, float * re
}
}
inline static void ggml_vec_mad_f32(const int n, float * restrict y, const float * restrict x, const float v) {
inline static void ggml_vec_mad_f32(const int n, float * GGML_RESTRICT y, const float * GGML_RESTRICT x, const float v) {
#if defined(GGML_SIMD)
const int np = (n & ~(GGML_F32_STEP - 1));
@ -1695,7 +1695,7 @@ inline static void ggml_vec_mad_f32(const int n, float * restrict y, const float
#endif
}
inline static void ggml_vec_mad_f16(const int n, ggml_fp16_t * restrict y, const ggml_fp16_t * restrict x, const float v) {
inline static void ggml_vec_mad_f16(const int n, ggml_fp16_t * GGML_RESTRICT y, const ggml_fp16_t * GGML_RESTRICT x, const float v) {
#if defined(GGML_SIMD)
const int np = (n & ~(GGML_F16_STEP - 1));
@ -1727,10 +1727,10 @@ inline static void ggml_vec_mad_f16(const int n, ggml_fp16_t * restrict y, const
}
// xs and vs are byte strides of x and v
inline static void ggml_vec_mad_f32_unroll(const int n, const int xs, const int vs, float * restrict y, const float * restrict xv, const float * restrict vv) {
inline static void ggml_vec_mad_f32_unroll(const int n, const int xs, const int vs, float * GGML_RESTRICT y, const float * GGML_RESTRICT xv, const float * GGML_RESTRICT vv) {
const float * restrict x[GGML_VEC_MAD_UNROLL];
const float * restrict v[GGML_VEC_MAD_UNROLL];
const float * GGML_RESTRICT x[GGML_VEC_MAD_UNROLL];
const float * GGML_RESTRICT v[GGML_VEC_MAD_UNROLL];
for (int i = 0; i < GGML_VEC_MAD_UNROLL; ++i) {
x[i] = (const float *) ((const char *) xv + i*xs);
@ -15485,6 +15485,14 @@ int ggml_cpu_has_amx_int8(void) {
#endif
}
int ggml_cpu_has_bmi2(void) {
#if defined(__BMI2__)
return 1;
#else
return 0;
#endif
}
int ggml_cpu_has_fma(void) {
#if defined(__FMA__)
return 1;