whispercpp 1.3.4 → 1.3.5

data/ext/sources/ggml/src/ggml-sycl/roll.hpp

@@ -0,0 +1,20 @@
+//
+// 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
+//
+
+#ifndef GGML_SYCL_ROLL_HPP
+#define GGML_SYCL_ROLL_HPP
+
+#include "common.hpp"
+
+void ggml_sycl_roll(ggml_backend_sycl_context & ctx, ggml_tensor *dst);
+
+#endif // GGML_SYCL_ROLL_HPP

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

@@ -119,7 +119,7 @@ static void rope_multi(const T * x, T * dst, const int ne0, const int ne1, const
                         const size_t s2, const int n_dims, const int32_t * pos, const float freq_scale,
                         const float ext_factor, const float attn_factor, const rope_corr_dims corr_dims,
                         const float theta_scale, const float * freq_factors, const mrope_sections sections,
-                        const sycl::nd_item<3> & item_ct1) {
+                        const bool is_imrope, const sycl::nd_item<3> & item_ct1) {
     // get index pos
     const int i0 = 2 * (item_ct1.get_group(1) * item_ct1.get_local_range(1) + item_ct1.get_local_id(1));
     if (i0 >= ne0) {
@@ -143,17 +143,29 @@ static void rope_multi(const T * x, T * dst, const int ne0, const int ne1, const
 
 
     float theta_base = 0.0;
-    if (sector < sections.v[0]) {
-        theta_base = pos[channel_x]*sycl::pow(theta_scale, i0/2.0f);
-    }
-    else if (sector >= sections.v[0] && sector < sec_w) {
-        theta_base = pos[channel_x + ne2 * 1]*sycl::pow(theta_scale, i0/2.0f);
-    }
-    else if (sector >= sec_w && sector < sec_w + sections.v[2]) {
-        theta_base = pos[channel_x + ne2 * 2]*sycl::pow(theta_scale, i0/2.0f);
-    }
-    else if (sector >= sec_w + sections.v[2]) {
-        theta_base = pos[channel_x + ne2 * 3]*sycl::pow(theta_scale, i0/2.0f);
+    if (is_imrope) {
+        if (sector % 3 == 1 && sector < 3 * sections.v[1]) {
+            theta_base = pos[channel_x + ne2 * 1]*sycl::pow(theta_scale, i0/2.0f);
+        } else if (sector % 3 == 2 && sector < 3 * sections.v[2]) {
+            theta_base = pos[channel_x + ne2 * 2]*sycl::pow(theta_scale, i0/2.0f);
+        } else if (sector % 3 == 0 && sector < 3 * sections.v[0]) {
+            theta_base = pos[channel_x]*sycl::pow(theta_scale, i0/2.0f);
+        } else {
+            theta_base = pos[channel_x + ne2 * 3]*sycl::pow(theta_scale, i0/2.0f);
+        }
+    } else {
+        if (sector < sections.v[0]) {
+            theta_base = pos[channel_x]*sycl::pow(theta_scale, i0/2.0f);
+        }
+        else if (sector >= sections.v[0] && sector < sec_w) {
+            theta_base = pos[channel_x + ne2 * 1]*sycl::pow(theta_scale, i0/2.0f);
+        }
+        else if (sector >= sec_w && sector < sec_w + sections.v[2]) {
+            theta_base = pos[channel_x + ne2 * 2]*sycl::pow(theta_scale, i0/2.0f);
+        }
+        else if (sector >= sec_w + sections.v[2]) {
+            theta_base = pos[channel_x + ne2 * 3]*sycl::pow(theta_scale, i0/2.0f);
+        }
     }
 
     const float freq_factor = has_ff ? freq_factors[i0 / 2] : 1.0f;
@@ -281,7 +293,7 @@ static void rope_multi_sycl(const T * x, T * dst, const int ne0, const int ne1,
                              const size_t s2, const int n_dims, const int nr, const int32_t * pos,
                              const float freq_scale, const float freq_base, const float ext_factor,
                              const float attn_factor, const rope_corr_dims corr_dims, const float * freq_factors,
-                             const mrope_sections sections, queue_ptr stream) {
+                             const mrope_sections sections, const bool is_imrope, queue_ptr stream) {
     GGML_ASSERT(ne0 % 2 == 0);
     const sycl::range<3>    block_dims(1, SYCL_ROPE_BLOCK_SIZE, 1);
     const int               n_blocks_y = ceil_div(ne0, (2 * SYCL_ROPE_BLOCK_SIZE));
@@ -297,12 +309,12 @@ static void rope_multi_sycl(const T * x, T * dst, const int ne0, const int ne1,
     if (freq_factors == nullptr) {
         stream->parallel_for(nd_range, [=](sycl::nd_item<3> item_ct1) {
             rope_multi<T, false>(x, dst, ne0, ne1, ne2, s1, s2, n_dims, pos, freq_scale, ext_factor, attn_factor,
-                                  corr_dims, theta_scale, freq_factors, sections, item_ct1);
+                                  corr_dims, theta_scale, freq_factors, sections, is_imrope, item_ct1);
         });
     } else {
         stream->parallel_for(nd_range, [=](sycl::nd_item<3> item_ct1) {
             rope_multi<T, true>(x, dst, ne0, ne1, ne2, s1, s2, n_dims, pos, freq_scale, ext_factor, attn_factor,
-                                 corr_dims, theta_scale, freq_factors, sections, item_ct1);
+                                 corr_dims, theta_scale, freq_factors, sections, is_imrope, item_ct1);
         });
     }
 }
@@ -381,6 +393,7 @@ inline void ggml_sycl_op_rope(ggml_backend_sycl_context & ctx, ggml_tensor *dst)
 
     const bool is_neox = mode & GGML_ROPE_TYPE_NEOX;
     const bool is_mrope = mode & GGML_ROPE_TYPE_MROPE;
+    const bool is_imrope = mode == GGML_ROPE_TYPE_IMROPE;
     const bool is_vision = mode == GGML_ROPE_TYPE_VISION;
 
     if (is_mrope) {
@@ -422,11 +435,11 @@ inline void ggml_sycl_op_rope(ggml_backend_sycl_context & ctx, ggml_tensor *dst)
         if (dst->src[0]->type == GGML_TYPE_F16) {
             rope_multi_sycl((const sycl::half *)dst->src[0]->data, (sycl::half *)dst->data, ne00, ne01, ne02, s01,
                 s02, n_dims, nr, pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims,
-                freq_factors, sections, main_stream);
+                freq_factors, sections, is_imrope, main_stream);
         } else if (dst->src[0]->type == GGML_TYPE_F32) {
             rope_multi_sycl((const float *) dst->src[0]->data, (float *) dst->data, ne00, ne01, ne02, s01, s02, n_dims,
                              nr, pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims, freq_factors, sections,
-                             main_stream);
+                             is_imrope, main_stream);
         } else {
             GGML_ABORT("Fatal error: Tensor type unsupported!");
         }

data/ext/sources/ggml/src/ggml-sycl/set.cpp

@@ -0,0 +1,73 @@
+#include "presets.hpp"
+#include "common.hpp"
+#include "ggml.h"
+#include "set.hpp"
+#include <cstdint>
+#include <sycl/sycl.hpp>
+using namespace sycl;
+
+// Internal function: perform element-wise set operation for each thread
+inline void set_f32(const float* src, float* dst,
+                    const int64_t ne0, const int64_t ne1,
+                    const int64_t ne2, const int64_t ne3,
+                    const int64_t nb[3], const int64_t src_nb[3],
+                    const int64_t offset_elem,
+                    const nd_item<1>& item)
+{
+    const size_t idx = item.get_global_id(0);
+    const size_t total = ne0 * ne1 * ne2 * ne3;
+    if (idx >= total) return;
+
+    // Convert linear index to 4D indices
+    const size_t i3 = idx / (ne2 * ne1 * ne0);
+    const size_t rem = idx % (ne2 * ne1 * ne0);
+    const size_t i2 = rem / (ne1 * ne0);
+    const size_t rem2 = rem % (ne1 * ne0);
+    const size_t i1 = rem2 / ne0;
+    const size_t i0 = rem2 % ne0;
+
+    // Compute source and destination indices and copy
+    dst[i0 + i1*nb[0] + i2*nb[1] + i3*nb[2] + offset_elem] =
+        src[i0 + i1*src_nb[0] + i2*src_nb[1] + i3*src_nb[2]];
+}
+
+// Main function: prepare GPU queue and launch parallel_for
+void ggml_sycl_op_set(ggml_backend_sycl_context& ctx, ggml_tensor* dst) {
+    const ggml_tensor* src0 = dst->src[0];
+    const ggml_tensor* src1 = dst->src[1];
+
+    // Ensure shapes and types are compatible
+    GGML_ASSERT(ggml_are_same_shape(src0, dst));
+    GGML_ASSERT(ggml_is_contiguous(dst) && ggml_is_contiguous(src0));
+    GGML_ASSERT(dst->type == src0->type && src0->type == src1->type && dst->type == GGML_TYPE_F32);
+
+    const int32_t* opts = (const int32_t*) dst->op_params;
+    const int64_t nb[3]     = {opts[0]/sizeof(float), opts[1]/sizeof(float), opts[2]/sizeof(float)};
+    const int64_t offset_elem = opts[3] / sizeof(float);
+    const bool inplace = opts[4];
+
+    float* dst_ptr = (float*) dst->data;
+    const float* src0_ptr = (const float*) src0->data;
+    const float* src1_ptr = (const float*) src1->data;
+
+    queue_ptr stream = ctx.stream();
+
+    // Copy src0 to dst if not inplace
+    if (!inplace)
+        stream->memcpy(dst_ptr, src0_ptr, ggml_nbytes(dst));
+
+    const int64_t ne[4] = {src1->ne[0], src1->ne[1], src1->ne[2], src1->ne[3]};
+    const int64_t src_nb[3] = {src1->nb[1]/sizeof(float), src1->nb[2]/sizeof(float), src1->nb[3]/sizeof(float)};
+
+    const size_t total_threads = ne[0]*ne[1]*ne[2]*ne[3];
+    const size_t grid_size = ((total_threads + SYCL_SET_BLOCK_SIZE - 1) / SYCL_SET_BLOCK_SIZE) * SYCL_SET_BLOCK_SIZE;
+
+    // Copy src0 to dst if not inplace
+    stream->parallel_for(
+        nd_range<1>(range<1>(grid_size), range<1>(SYCL_SET_BLOCK_SIZE)),
+        [=](nd_item<1> item) {
+            set_f32(src1_ptr, dst_ptr,
+                ne[0], ne[1], ne[2], ne[3],
+                nb, src_nb, offset_elem, item); }
+    );
+}

data/ext/sources/ggml/src/ggml-sycl/set.hpp

@@ -0,0 +1,5 @@
+#pragma once
+#include "backend.hpp"
+#include "ggml.h"
+
+void ggml_sycl_op_set(ggml_backend_sycl_context & ctx, ggml_tensor * dst);