ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

ggml : portability fixes for VS 2017 (#12150)

Browse source 
* Add include files for std::min/max and std::toupper/tolower

* win32: move _USE_MATH_DEFINES before includes to ensure M_PI is defined

* Use GGML_RESTRICT instead of "restrict" keyword everywhere, and use "__restrict" in MSVC plain C mode

* win32: only use __restrict in MSVC if C11/C17 support is not enabled

---------

Co-authored-by: Marcus Groeber <Marcus.Groeber@cerence.com>
This commit is contained in:
mgroeber9110 2025-03-04 17:53:26 +01:00 • committed by GitHub
parent 20a9b8f5e1
commit 5bbe6a9fe9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
20 changed files with 547 additions and 529 deletions
Download patch file Download diff file Expand all files Collapse all files

26
ggml/src/ggml-cpu/ggml-cpu.c
View file

@ -247,9 +247,9 @@ 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] = {
@ -1451,7 +1451,7 @@ 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);
@ -1494,7 +1494,7 @@ 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);
@ -1562,7 +1562,7 @@ 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);
@ -1606,10 +1606,10 @@ 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);
@ -1659,7 +1659,7 @@ 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));
@ -1690,7 +1690,7 @@ 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));
@ -1722,10 +1722,10 @@ 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);