whispercpp 1.3.0 → 1.3.1

data/ext/ggml/src/ggml-sycl/rope.cpp

@@ -0,0 +1,276 @@
+#include "rope.hpp"
+
+struct rope_corr_dims {
+    float v[2];
+};
+
+static float rope_yarn_ramp(const float low, const float high, const int i0) {
+    const float y = (i0 / 2 - low) / sycl::max(0.001f, high - low);
+    return 1.0f - sycl::min(1.0f, sycl::max(0.0f, y));
+}
+
+// YaRN algorithm based on LlamaYaRNScaledRotaryEmbedding.py from https://github.com/jquesnelle/yarn
+// MIT licensed. Copyright (c) 2023 Jeffrey Quesnelle and Bowen Peng.
+static void rope_yarn(
+    float theta_extrap, float freq_scale, rope_corr_dims corr_dims, int64_t i0, float ext_factor, float mscale,
+    float * cos_theta, float * sin_theta) {
+    // Get n-d rotational scaling corrected for extrapolation
+    float theta_interp = freq_scale * theta_extrap;
+    float theta = theta_interp;
+    if (ext_factor != 0.0f) {
+        float ramp_mix = rope_yarn_ramp(corr_dims.v[0], corr_dims.v[1], i0) * ext_factor;
+        theta = theta_interp * (1 - ramp_mix) + theta_extrap * ramp_mix;
+
+        // Get n-d magnitude scaling corrected for interpolation
+        mscale *= 1.0f + 0.1f * sycl::log(1.0f / freq_scale);
+    }
+    *cos_theta = sycl::cos(theta) * mscale;
+    *sin_theta = sycl::sin(theta) * mscale;
+}
+
+template<typename T, bool has_ff>
+static void rope_norm(
+    const T * x, T * dst, int ne0, int n_dims, const int32_t * pos, float freq_scale, int p_delta_rows,
+    float ext_factor, float attn_factor, rope_corr_dims corr_dims, float theta_scale, const float * freq_factors,
+    const sycl::nd_item<3> &item_ct1) {
+    const int i0 = 2 * (item_ct1.get_local_range(1) * item_ct1.get_group(1) +
+                         item_ct1.get_local_id(1));
+
+    if (i0 >= ne0) {
+        return;
+    }
+
+    const int row = item_ct1.get_local_range(2) * item_ct1.get_group(2) +
+                    item_ct1.get_local_id(2);
+
+    if (i0 >= n_dims) {
+        const int i = row*ne0 + i0;
+
+        dst[i + 0] = x[i + 0];
+        dst[i + 1] = x[i + 1];
+
+        return;
+    }
+
+    const int i = row*ne0 + i0;
+    const int i2 = row/p_delta_rows;
+
+    const float theta_base = pos[i2] * sycl::pow(theta_scale, i0 / 2.0f);
+
+    const float freq_factor = has_ff ? freq_factors[i0/2] : 1.0f;
+
+    float cos_theta;
+    float sin_theta;
+
+    rope_yarn(theta_base/freq_factor, freq_scale, corr_dims, i0, ext_factor, attn_factor, &cos_theta, &sin_theta);
+
+    const float x0 = x[i + 0];
+    const float x1 = x[i + 1];
+
+    dst[i + 0] = x0*cos_theta - x1*sin_theta;
+    dst[i + 1] = x0*sin_theta + x1*cos_theta;
+}
+
+template<typename T, bool has_ff>
+static void rope_neox(
+    const T * x, T * dst, int ne0, int n_dims, const int32_t * pos, float freq_scale, int p_delta_rows,
+    float ext_factor, float attn_factor, rope_corr_dims corr_dims, float theta_scale, const float * freq_factors,
+    const sycl::nd_item<3> &item_ct1) {
+    const int i0 = 2 * (item_ct1.get_local_range(1) * item_ct1.get_group(1) +
+                         item_ct1.get_local_id(1));
+
+    if (i0 >= ne0) {
+        return;
+    }
+
+    const int row = item_ct1.get_local_range(2) * item_ct1.get_group(2) +
+                    item_ct1.get_local_id(2);
+
+    if (i0 >= n_dims) {
+        const int i = row*ne0 + i0;
+
+        dst[i + 0] = x[i + 0];
+        dst[i + 1] = x[i + 1];
+
+        return;
+    }
+
+    const int i  = row*ne0 + i0/2;
+    const int i2 = row/p_delta_rows;
+
+    const float theta_base = pos[i2] * sycl::pow(theta_scale, i0 / 2.0f);
+
+    const float freq_factor = has_ff ? freq_factors[i0/2] : 1.0f;
+
+    float cos_theta;
+    float sin_theta;
+
+    rope_yarn(theta_base/freq_factor, freq_scale, corr_dims, i0, ext_factor, attn_factor, &cos_theta, &sin_theta);
+
+    const float x0 = x[i + 0];
+    const float x1 = x[i + n_dims/2];
+
+    dst[i + 0]        = x0*cos_theta - x1*sin_theta;
+    dst[i + n_dims/2] = x0*sin_theta + x1*cos_theta;
+}
+
+template <typename T>
+static void rope_norm_sycl(
+    const T *x, T *dst, int ne0, int n_dims, int nr, const int32_t *pos, float freq_scale, int p_delta_rows,
+    float freq_base, float ext_factor, float attn_factor, rope_corr_dims corr_dims, const float * freq_factors, queue_ptr stream) {
+    GGML_ASSERT(ne0 % 2 == 0);
+    const sycl::range<3> block_dims(1, SYCL_ROPE_BLOCK_SIZE, 1);
+    const int num_blocks_x = (ne0 + 2*SYCL_ROPE_BLOCK_SIZE - 1) / (2*SYCL_ROPE_BLOCK_SIZE);
+    const sycl::range<3> block_nums(1, num_blocks_x, nr);
+
+    const float theta_scale = powf(freq_base, -2.0f/n_dims);
+
+    dpct::has_capability_or_fail(stream->get_device(),
+                                     {sycl::aspect::fp16});
+
+    if (freq_factors == nullptr) {
+        /*
+        DPCT1049:40: The work-group size passed to the SYCL kernel may exceed
+        the limit. To get the device limit, query
+        info::device::max_work_group_size. Adjust the work-group size if needed.
+        */
+        stream->parallel_for(
+            sycl::nd_range<3>(block_nums * block_dims, block_dims),
+            [=](sycl::nd_item<3> item_ct1) {
+                rope_norm<T, false>(x, dst, ne0, n_dims, pos, freq_scale, p_delta_rows,
+                               ext_factor, attn_factor, corr_dims, theta_scale, freq_factors,
+                               item_ct1);
+            });
+    } else {
+        /*
+        DPCT1049:41: The work-group size passed to the SYCL kernel may exceed
+        the limit. To get the device limit, query
+        info::device::max_work_group_size. Adjust the work-group size if needed.
+        */
+        stream->parallel_for(
+            sycl::nd_range<3>(block_nums * block_dims, block_dims),
+            [=](sycl::nd_item<3> item_ct1) {
+                rope_norm<T, true>(x, dst, ne0, n_dims, pos, freq_scale, p_delta_rows,
+                              ext_factor, attn_factor, corr_dims, theta_scale, freq_factors,
+                              item_ct1);
+            });
+    }
+}
+
+template <typename T>
+static void rope_neox_sycl(
+    const T *x, T *dst, int ne0, int n_dims, int nr, const int32_t *pos, float freq_scale, int p_delta_rows,
+    float freq_base, float ext_factor, float attn_factor, rope_corr_dims corr_dims, const float * freq_factors, queue_ptr stream) {
+    GGML_ASSERT(ne0 % 2 == 0);
+    const sycl::range<3> block_dims(1, SYCL_ROPE_BLOCK_SIZE, 1);
+    const int num_blocks_x = (ne0 + 2*SYCL_ROPE_BLOCK_SIZE - 1) / (2*SYCL_ROPE_BLOCK_SIZE);
+    const sycl::range<3> block_nums(1, num_blocks_x, nr);
+
+    const float theta_scale = powf(freq_base, -2.0f/n_dims);
+
+    dpct::has_capability_or_fail(stream->get_device(),
+                                    {sycl::aspect::fp16});
+
+    if (freq_factors == nullptr) {
+        stream->parallel_for(
+            sycl::nd_range<3>(block_nums * block_dims, block_dims),
+            [=](sycl::nd_item<3> item_ct1) {
+                rope_neox<T, false>(x, dst, ne0, n_dims, pos, freq_scale,
+                                    p_delta_rows, ext_factor, attn_factor,
+                                    corr_dims, theta_scale, freq_factors,
+                                    item_ct1);
+            });
+    } else {
+        stream->parallel_for(
+            sycl::nd_range<3>(block_nums * block_dims, block_dims),
+            [=](sycl::nd_item<3> item_ct1) {
+                rope_neox<T, true>(x, dst, ne0, n_dims, pos, freq_scale,
+                                    p_delta_rows, ext_factor, attn_factor,
+                                    corr_dims, theta_scale, freq_factors,
+                                    item_ct1);
+            });
+    }
+}
+
+void ggml_sycl_op_rope(
+    ggml_backend_sycl_context & ctx, const ggml_tensor *src0, const ggml_tensor *src1, ggml_tensor *dst,
+    const float *src0_dd, const float *src1_dd, float *dst_dd, const queue_ptr &main_stream) {
+    const ggml_tensor * src2 = dst->src[2];
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16);
+    GGML_ASSERT( dst->type == GGML_TYPE_F32 ||  dst->type == GGML_TYPE_F16);
+    GGML_ASSERT(src0->type == dst->type);
+
+    const int64_t ne00 = src0->ne[0];
+    const int64_t ne01 = src0->ne[1];
+    const int64_t nr = ggml_nrows(src0);
+
+    //const int n_past      = ((int32_t *) dst->op_params)[0];
+    const int n_dims      = ((int32_t *) dst->op_params)[1];
+    const int mode        = ((int32_t *) dst->op_params)[2];
+    //const int n_ctx       = ((int32_t *) dst->op_params)[3];
+    const int n_ctx_orig  = ((int32_t *) dst->op_params)[4];
+
+    // RoPE alteration for extended context
+    float freq_base;
+    float freq_scale;
+    float ext_factor;
+    float attn_factor;
+    float beta_fast;
+    float beta_slow;
+
+    memcpy(&freq_base,   (int32_t *) dst->op_params +  5, sizeof(float));
+    memcpy(&freq_scale,  (int32_t *) dst->op_params +  6, sizeof(float));
+    memcpy(&ext_factor,  (int32_t *) dst->op_params +  7, sizeof(float));
+    memcpy(&attn_factor, (int32_t *) dst->op_params +  8, sizeof(float));
+    memcpy(&beta_fast,   (int32_t *) dst->op_params +  9, sizeof(float));
+    memcpy(&beta_slow,   (int32_t *) dst->op_params + 10, sizeof(float));
+
+    const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
+
+    const int32_t * pos = (const int32_t *) src1_dd;
+
+    const float * freq_factors = nullptr;
+    if (src2 != nullptr) {
+        freq_factors = (const float *) src2->data;
+    }
+
+    rope_corr_dims corr_dims;
+    ggml_rope_yarn_corr_dims(n_dims, n_ctx_orig, freq_base, beta_fast, beta_slow, corr_dims.v);
+
+    // compute
+    if (is_neox) {
+        if (src0->type == GGML_TYPE_F32) {
+            rope_neox_sycl(
+                (const float *)src0_dd, (float *)dst_dd, ne00, n_dims, nr, pos, freq_scale, ne01, freq_base, ext_factor,
+                attn_factor, corr_dims, freq_factors, main_stream
+            );
+        } else if (src0->type == GGML_TYPE_F16) {
+            rope_neox_sycl(
+                (const sycl::half *)src0_dd, (sycl::half *)dst_dd, ne00, n_dims, nr, pos, freq_scale, ne01, freq_base, ext_factor,
+                attn_factor, corr_dims, freq_factors, main_stream
+            );
+        } else {
+            GGML_ABORT("fatal error");
+        }
+    } else {
+        if (src0->type == GGML_TYPE_F32) {
+            rope_norm_sycl(
+                (const float *)src0_dd, (float *)dst_dd, ne00, n_dims, nr, pos, freq_scale, ne01, freq_base, ext_factor,
+                attn_factor, corr_dims, freq_factors, main_stream
+            );
+        } else if (src0->type == GGML_TYPE_F16) {
+            rope_norm_sycl(
+                (const sycl::half *)src0_dd, (sycl::half *)dst_dd, ne00, n_dims, nr, pos, freq_scale, ne01, freq_base, ext_factor,
+                attn_factor, corr_dims, freq_factors, main_stream
+            );
+        } else {
+            GGML_ABORT("fatal error");
+        }
+    }
+
+    GGML_UNUSED(src1);
+    GGML_UNUSED(dst);
+    GGML_UNUSED(src1_dd);
+    GGML_UNUSED(ctx);
+}

data/ext/ggml/src/ggml-sycl/softmax.cpp

@@ -0,0 +1,251 @@
+#include "norm.hpp"
+
+template <bool vals_smem, int ncols_template, int block_size_template>
+static void soft_max_f32(const float * x, const float * mask, float * dst, const int ncols_par,
+                         const int nrows_y, const float scale, const float max_bias, const float m0,
+                         const float m1, uint32_t n_head_log2, const sycl::nd_item<3> &item_ct1, float *buf) {
+    const int ncols = ncols_template == 0 ? ncols_par : ncols_template;
+
+    const int tid = item_ct1.get_local_id(2);
+    const int rowx = item_ct1.get_group(2);
+    const int rowy = rowx % nrows_y; // broadcast the mask (y) in the row dimension
+
+    const int block_size = block_size_template == 0 ? item_ct1.get_local_range(2) : block_size_template;
+
+    const int warp_id = item_ct1.get_local_id(2) / WARP_SIZE;
+    const int lane_id = item_ct1.get_local_id(2) % WARP_SIZE;
+    const int nthreads = block_size;
+    const int nwarps = nthreads / WARP_SIZE;
+    size_t nreduce = nwarps / WARP_SIZE;
+    float slope = 1.0f;
+
+    // ALiBi
+    if (max_bias > 0.0f) {
+        const uint32_t h = rowx/nrows_y; // head index
+
+        const float base = h < n_head_log2 ? m0 : m1;
+        const int   exp  = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1;
+
+        slope = sycl::pow(base, float(exp));
+    }
+
+    float *vals = vals_smem ? buf + std::max(nwarps, WARP_SIZE) : dst + rowx * ncols;
+    float max_val = -INFINITY;
+
+    for (int col0 = 0; col0 < ncols; col0 += block_size) {
+        const int col = col0 + tid;
+
+        if (ncols_template == 0 && col >= ncols) {
+            break;
+        }
+
+        const int ix = rowx*ncols + col;
+        const int iy = rowy*ncols + col;
+
+        const float val = x[ix]*scale + (mask ? slope*mask[iy] : 0.0f);
+
+        vals[col] = val;
+        max_val = sycl::max(max_val, val);
+    }
+
+    // find the max value in the block
+    max_val = warp_reduce_max(max_val, item_ct1);
+    if (block_size > WARP_SIZE) {
+        if (warp_id == 0) {
+            buf[lane_id] = -INFINITY;
+            for (size_t i = 1; i < nreduce; i += 1) {
+                buf[lane_id + i * WARP_SIZE] = -INFINITY;
+            }
+        }
+        item_ct1.barrier(sycl::access::fence_space::local_space);
+
+        if (lane_id == 0) {
+            buf[warp_id] = max_val;
+        }
+        item_ct1.barrier(sycl::access::fence_space::local_space);
+        max_val = buf[lane_id];
+        for (size_t i = 1; i < nreduce; i += 1) {
+            max_val = std::max(max_val, buf[lane_id + i * WARP_SIZE]);
+        }
+        max_val = warp_reduce_max(max_val, item_ct1);
+    }
+
+    float tmp = 0.f;
+#pragma unroll
+    for (int col0 = 0; col0 < ncols; col0 += block_size) {
+        const int col = col0 + tid;
+                if (ncols_template == 0 && col >= ncols) {
+            break;
+        }
+
+        const float val = sycl::native::exp(vals[col] - max_val);
+        tmp += val;
+        vals[col] = val;
+    }
+
+    // find the sum of exps in the block
+    tmp = warp_reduce_sum(tmp, item_ct1);
+    if (block_size > WARP_SIZE) {
+        item_ct1.barrier(sycl::access::fence_space::local_space);
+        if (warp_id == 0) {
+            buf[lane_id] = 0.f;
+            for (size_t i = 1; i < nreduce; i += 1) {
+                buf[lane_id + i * WARP_SIZE] = 0.f;
+            }
+        }
+        item_ct1.barrier(sycl::access::fence_space::local_space);
+
+        if (lane_id == 0) {
+            buf[warp_id] = tmp;
+        }
+        item_ct1.barrier(sycl::access::fence_space::local_space);
+
+        tmp = buf[lane_id];
+        for (size_t i = 1; i < nreduce; i += 1) {
+            tmp += buf[lane_id + i * WARP_SIZE];
+        }
+        tmp = warp_reduce_sum(tmp, item_ct1);
+    }
+
+    const float inv_sum = 1.f / tmp;
+
+#pragma unroll
+    for (int col0 = 0; col0 < ncols; col0 += block_size) {
+        const int col = col0 + tid;
+
+        if (ncols_template == 0 && col >= ncols) {
+            return;
+        }
+
+        const int idst = rowx*ncols + col;
+        dst[idst] = vals[col] * inv_sum;
+    }
+}
+
+template <bool vals_smem, int ncols_template, int block_size_template>
+static void soft_max_f32_submitter(const float * x, const float * mask, float * dst, const int ncols_par,
+                                   const int nrows_y, const float scale, const float max_bias, const float m0,
+                                   const float m1, uint32_t n_head_log2, sycl::range<3> block_nums, sycl::range<3> block_dims,
+                                   const size_t n_local_scratch, queue_ptr stream) {
+    stream->submit([&](sycl::handler &cgh) {
+        sycl::local_accessor<float, 1> local_buf_acc(n_local_scratch, cgh);
+
+        cgh.parallel_for(
+            sycl::nd_range<3>(block_nums * block_dims, block_dims),
+            [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] {
+                soft_max_f32<vals_smem, ncols_template, block_size_template>(x, mask, dst, ncols_par,
+                                                                             nrows_y, scale, max_bias, m0,
+                                                                             m1, n_head_log2, item_ct1,
+                                                                             get_pointer(local_buf_acc));
+            });
+    });
+}
+
+static void soft_max_f32_sycl(const float * x, const float * mask,
+                              float * dst, const int ncols_x, const int nrows_x,
+                              const int nrows_y, const float scale, const float max_bias,
+                              queue_ptr stream, int device) {
+    int nth = WARP_SIZE;
+    int max_block_size = ggml_sycl_info().max_work_group_sizes[device];
+    while (nth < ncols_x && nth < max_block_size) nth *= 2;
+    if (nth>max_block_size) nth = max_block_size;
+
+    const sycl::range<3> block_dims(1, 1, nth);
+    const sycl::range<3> block_nums(1, 1, nrows_x);
+    const size_t n_val_tmp = nth / WARP_SIZE;
+    const size_t n_local_scratch = (GGML_PAD(ncols_x, WARP_SIZE) + n_val_tmp);
+
+    const uint32_t n_head_kv   = nrows_x/nrows_y;
+    const uint32_t n_head_log2 = 1u << (uint32_t) floorf(log2f((float) n_head_kv));
+
+    const float m0 = powf(2.0f, -(max_bias       ) / n_head_log2);
+    const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2);
+
+    const size_t local_mem_size = stream->get_device().get_info<sycl::info::device::local_mem_size>();
+    if (n_local_scratch*sizeof(float) < local_mem_size) {
+        if (ncols_x > max_block_size) {
+            soft_max_f32_submitter<true, 0, 0>(x, mask, dst, ncols_x, nrows_y, scale,
+                                               max_bias, m0, m1, n_head_log2, block_nums,
+                                               block_dims, n_local_scratch, stream);
+            return;
+        }
+        switch (ncols_x) {
+            case 32:
+                soft_max_f32_submitter<true, 32, 32>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                     max_bias, m0, m1, n_head_log2, block_nums,
+                                                     block_dims, n_local_scratch, stream);
+                break;
+            case 64:
+                soft_max_f32_submitter<true, 64, 64>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                     max_bias, m0, m1, n_head_log2, block_nums,
+                                                     block_dims, n_local_scratch, stream);
+                break;
+            case 128:
+                soft_max_f32_submitter<true, 128, 128>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                       max_bias, m0, m1, n_head_log2, block_nums,
+                                                       block_dims, n_local_scratch, stream);
+                break;
+            case 256:
+                soft_max_f32_submitter<true, 256, 256>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                       max_bias, m0, m1, n_head_log2, block_nums,
+                                                       block_dims, n_local_scratch, stream);
+                break;
+            case 512:
+                soft_max_f32_submitter<true, 512, 512>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                       max_bias, m0, m1, n_head_log2, block_nums,
+                                                       block_dims, n_local_scratch, stream);
+                break;
+            case 1024:
+                soft_max_f32_submitter<true, 1024, 1024>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                         max_bias, m0, m1, n_head_log2, block_nums,
+                                                         block_dims, n_local_scratch, stream);
+                break;
+            case 2048:
+                soft_max_f32_submitter<true, 2048, 1024>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                         max_bias, m0, m1, n_head_log2, block_nums,
+                                                         block_dims, n_local_scratch, stream);
+                break;
+            case 4096:
+                soft_max_f32_submitter<true, 4096, 1024>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                         max_bias, m0, m1, n_head_log2, block_nums,
+                                                         block_dims, n_local_scratch, stream);
+                break;
+            default:
+                soft_max_f32_submitter<true, 0, 0>(x, mask, dst, ncols_x, nrows_y, scale,
+                                                   max_bias, m0, m1, n_head_log2, block_nums,
+                                                   block_dims, n_local_scratch, stream);
+                break;
+        }
+    } else {
+        soft_max_f32_submitter<false, 0, 0>(x, mask, dst, ncols_x, nrows_y, scale,
+                                            max_bias, m0, m1, n_head_log2, block_nums,
+                                            block_dims, WARP_SIZE, stream);
+    }
+}
+
+void ggml_sycl_op_soft_max(ggml_backend_sycl_context & ctx, const ggml_tensor *src0,
+                                  const ggml_tensor *src1, ggml_tensor *dst,
+                                  const float *src0_dd, const float *src1_dd,
+                                  float *dst_dd,
+                                  const queue_ptr &main_stream) {
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT( dst->type == GGML_TYPE_F32);
+
+#pragma message("TODO: add ggml_sycl_op_soft_max() F16 src1 support")
+#pragma message("ref:  https://github.com/ggerganov/llama.cpp/pull/5021")
+    GGML_ASSERT(!src1 || src1->type == GGML_TYPE_F32); // src1 contains mask and it is optional
+
+    const int64_t ne00 = src0->ne[0];
+    const int64_t nrows_x = ggml_nrows(src0);
+    const int64_t nrows_y = src0->ne[1];
+
+    float scale = 1.0f;
+    float max_bias = 0.0f;
+
+    memcpy(&scale, dst->op_params + 0, sizeof(float));
+    memcpy(&max_bias, dst->op_params + 1, sizeof(float));
+
+    soft_max_f32_sycl(src0_dd, src1 ? src1_dd : nullptr, dst_dd, ne00,
+        nrows_x, nrows_y, scale, max_bias, main_stream, ctx.device);
+}

data/ext/ggml/src/ggml-sycl/tsembd.cpp

@@ -0,0 +1,72 @@
+//
+// MIT license
+// Copyright (C) 2024 Intel Corporation
+// SPDX-License-Identifier: MIT
+//
+
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+
+#include "tsembd.hpp"
+
+static void timestep_embedding_f32(
+        const float * timesteps, float * dst, const int nb1,
+        const int dim, const int max_period, const sycl::nd_item<3> &item_ct1) {
+    // item_ct1.get_group(1)(blockIDx.y): idx of timesteps->ne[0]
+    // item_ct1.get_group(2) (blockIDx.x): idx of ((dim + 1) / 2) / BLOCK_SIZE
+    int i = item_ct1.get_group(1);
+    int j = item_ct1.get_local_id(2) + item_ct1.get_group(2) * item_ct1.get_local_range(2);
+    float * embed_data = (float *)((char *)dst +  i*nb1);
+
+    if (dim % 2 != 0 && j == ((dim + 1) / 2)) {
+        embed_data[dim] = 0.f;
+    }
+
+    int half = dim / 2;
+    if (j >= half) {
+        return;
+    }
+
+    float timestep = timesteps[i];
+    float freq = (float)sycl::native::exp(-(sycl::log((float)max_period)) * j / half);
+    float arg = timestep * freq;
+    embed_data[j] = sycl::cos(arg);
+    embed_data[j + half] = sycl::sin(arg);
+}
+
+static void timestep_embedding_f32_sycl(
+        const float * x, float * dst, const int ne00, const int nb1,
+        const int dim, const int max_period, const queue_ptr& stream) {
+    // As the kernel returns when thread.idx is larger than dim/2, the half_ceil does not need to pad
+    int half_ceil = dim / 2;
+    int num_blocks = (half_ceil + SYCL_TIMESTEP_EMBEDDING_BLOCK_SIZE - 1) / SYCL_TIMESTEP_EMBEDDING_BLOCK_SIZE;
+    sycl::range<3> block_dims(1, 1, SYCL_TIMESTEP_EMBEDDING_BLOCK_SIZE);
+    sycl::range<3> gridDim(1, ne00, num_blocks);
+    stream->parallel_for(
+        sycl::nd_range<3>(
+            gridDim * block_dims, block_dims),
+        [=](sycl::nd_item<3> item_ct1) {
+            timestep_embedding_f32(
+                x, dst, nb1, dim, max_period, item_ct1
+            );
+        });
+}
+
+void ggml_sycl_op_timestep_embedding(ggml_backend_sycl_context & ctx, const ggml_tensor *src0,
+    const ggml_tensor *src1, ggml_tensor * dst) {
+    const float * src0_d = (const float *)src0->data;
+    float * dst_d = (float *)dst->data;
+    dpct::queue_ptr stream = ctx.stream();
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT(dst->type == GGML_TYPE_F32);
+
+    const int dim = dst->op_params[0];
+    const int max_period = dst->op_params[1];
+
+    timestep_embedding_f32_sycl(src0_d, dst_d, src0->ne[0], dst->nb[1], dim, max_period, stream);
+    GGML_UNUSED(src1);
+}