Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[SYCL] Support OP EXPM1, support all UT cases of FLOOR, TRUNC, ROUND (#24363)

Browse Source 
* support OP EXPM1, support all UT cases of FLOOR, TRUNC, ROUND

* fix conflict

* rebase, support new UT case of repeat, concat
This commit is contained in:
Neo Zhang
2026-06-16 08:34:29 +03:00
committed by GitHub
parent 4196b477da
commit fdd109883d
7 changed files with 101 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-sycl/element_wise.cpp
+25 -51
View File
@@ -124,6 +124,11 @@ static __dpct_inline__ T op_exp(T x) {
return sycl::exp(x);
}
template<typename T>
static __dpct_inline__ T op_expm1(T x) {
return sycl::expm1(x);
}
template<typename T>
static __dpct_inline__ T op_log(T x) {
if (x <= static_cast<T>(0)) {
@@ -266,13 +271,6 @@ static void unary_op_clamp_kernel(const T * x, T * dst, const int k, const sycl:
}
}
template<typename T>
static void unary_op_floor_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
dst[i] = op_floor(x[i]);
}
}
template<typename T>
static void unary_op_ceil_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
@@ -280,20 +278,6 @@ static void unary_op_ceil_kernel(const T * x, T * dst, const int k, const sycl::
}
}
template<typename T>
static void unary_op_round_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
dst[i] = op_round(x[i]);
}
}
template<typename T>
static void unary_op_trunc_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
dst[i] = op_trunc(x[i]);
}
}
template<typename T>
static void clamp(const T * x, T * dst, const float min, const float max, const int k,
const sycl::nd_item<1> &item_ct1) {
@@ -605,6 +589,12 @@ static inline void ggml_sycl_op_exp(ggml_backend_sycl_context & ctx, ggml_tensor
});
}
static inline void ggml_sycl_op_expm1(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
return op_expm1(x);
});
}
static inline void ggml_sycl_op_log(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
[](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
@@ -728,16 +718,9 @@ static inline void ggml_sycl_op_clamp(ggml_backend_sycl_context & ctx, ggml_tens
}
static inline void ggml_sycl_op_floor(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
[](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
const int num_blocks = ceil_div(k_elements, 256);
stream->parallel_for(
sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(256),
sycl::range<1>(256)),
[=](sycl::nd_item<1> item_ct1) {
unary_op_floor_kernel(src, dst_ptr, k_elements, item_ct1);
});
});
ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
return op_floor(x);
});
}
static inline void ggml_sycl_op_ceil(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
@@ -747,29 +730,15 @@ static inline void ggml_sycl_op_ceil(ggml_backend_sycl_context & ctx, ggml_tenso
}
static inline void ggml_sycl_op_round(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
[](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
const int num_blocks = ceil_div(k_elements, 256);
stream->parallel_for(
sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(256),
sycl::range<1>(256)),
[=](sycl::nd_item<1> item_ct1) {
unary_op_round_kernel(src, dst_ptr, k_elements, item_ct1);
});
});
ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
return op_round(x);
});
}
static inline void ggml_sycl_op_trunc(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
[](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
const int num_blocks = ceil_div(k_elements, 256);
stream->parallel_for(
sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(256),
sycl::range<1>(256)),
[=](sycl::nd_item<1> item_ct1) {
unary_op_trunc_kernel(src, dst_ptr, k_elements, item_ct1);
});
});
ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
return op_trunc(x);
});
}
static inline void ggml_sycl_op_acc(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
@@ -1018,6 +987,11 @@ void ggml_sycl_exp(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
ggml_sycl_op_exp(ctx, dst);
}
void ggml_sycl_expm1(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/1);
ggml_sycl_op_expm1(ctx, dst);
}
void ggml_sycl_log(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
scope_op_debug_print scope_dbg_print(__func__, dst, /*num_src=*/1);
ggml_sycl_op_log(ctx, dst);