CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

[SYCL] Fix WARP_SIZE=16 bug of Intel GPU (#8266)

Browse Source 
* fix group_norm ut

* split softmax

* fix softmax

* add concat support condition

* revert debug code

* move QK_WARP_SIZE to presets.hpp
This commit is contained in:
luoyu-intel
2024-07-05 05:06:13 +00:00
committed by GitHub
parent e235b267a2
commit a9554e20b6
8 changed files with 301 additions and 257 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/CMakeLists.txt
View File

@@ -490,7 +490,7 @@ if (GGML_SYCL)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=nvptx64-nvidia-cuda") set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=nvptx64-nvidia-cuda")
add_compile_definitions(GGML_SYCL_WARP_SIZE=32) add_compile_definitions(GGML_SYCL_WARP_SIZE=32)
else() else()
add_compile_definitions(GGML_SYCL_WARP_SIZE=32) add_compile_definitions(GGML_SYCL_WARP_SIZE=16)
endif() endif()
file(GLOB GGML_HEADERS_SYCL "ggml-sycl/*.hpp") file(GLOB GGML_HEADERS_SYCL "ggml-sycl/*.hpp")

241
ggml/src/ggml-sycl.cpp
View File

@@ -892,117 +892,6 @@ static void diag_mask_inf_f32(const float * x, float * dst, const int ncols, con
dst[i] = x[i] - (col > n_past + row % rows_per_channel) * FLT_MAX; dst[i] = x[i] - (col > n_past + row % rows_per_channel) * FLT_MAX;
} }
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;
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 + 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;
}
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];
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;
}
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];
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;
}
}
static void scale_f32(const float * x, float * dst, const float scale, const int k, static void scale_f32(const float * x, float * dst, const float scale, const int k,
const sycl::nd_item<3> &item_ct1) { const sycl::nd_item<3> &item_ct1) {
const int i = item_ct1.get_local_range(2) * item_ct1.get_group(2) + const int i = item_ct1.get_local_range(2) * item_ct1.get_group(2) +
@@ -1890,106 +1779,6 @@ static void diag_mask_inf_f32_sycl(const float *x, float *dst,
}); });
} }
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,
local_buf_acc.get_pointer());
});
});
}
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_local_scratch = (GGML_PAD(ncols_x, WARP_SIZE) + WARP_SIZE);
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);
}
}
template <typename T> template <typename T>
static void im2col_sycl(const float *x, T *dst, int IW, int IH, static void im2col_sycl(const float *x, T *dst, int IW, int IH,
int OW, int OH, int KW, int KH, int IC, int OW, int OH, int KW, int KH, int IC,
@@ -3009,33 +2798,6 @@ inline void ggml_sycl_op_diag_mask_inf(ggml_backend_sycl_context & ctx, const gg
(void) src1_dd; (void) src1_dd;
} }
inline 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);
}
inline void ggml_sycl_op_scale(ggml_backend_sycl_context & ctx, const ggml_tensor *src0, const ggml_tensor *src1, inline void ggml_sycl_op_scale(ggml_backend_sycl_context & ctx, const ggml_tensor *src0, const ggml_tensor *src1,
ggml_tensor *dst, const float *src0_dd, ggml_tensor *dst, const float *src0_dd,
const float *src1_dd, float *dst_dd, const float *src1_dd, float *dst_dd,
@@ -5532,7 +5294,8 @@ GGML_CALL static bool ggml_backend_sycl_supports_op(ggml_backend_t backend, cons
case GGML_OP_CONCAT: case GGML_OP_CONCAT:
{ {
ggml_type src0_type = op->src[0]->type; ggml_type src0_type = op->src[0]->type;
return src0_type != GGML_TYPE_I32 && src0_type != GGML_TYPE_I16; int dim = op->op_params[0];
return ggml_is_contiguous(op->src[0]) && ggml_is_contiguous(op->src[1]) && src0_type != GGML_TYPE_I32 && src0_type != GGML_TYPE_I16 && dim == 2;
} break; } break;
case GGML_OP_DUP: case GGML_OP_DUP:
case GGML_OP_NONE: case GGML_OP_NONE:

1
ggml/src/ggml-sycl/backend.hpp
View File

@@ -21,5 +21,6 @@
#include "mmvq.hpp" #include "mmvq.hpp"
#include "rope.hpp" #include "rope.hpp"
#include "norm.hpp" #include "norm.hpp"
#include "softmax.hpp"
#endif // GGML_SYCL_BACKEND_HPP #endif // GGML_SYCL_BACKEND_HPP

31
ggml/src/ggml-sycl/dmmv.cpp
View File

@@ -3,6 +3,7 @@
#include "dequantize.hpp" #include "dequantize.hpp"
#include "presets.hpp" #include "presets.hpp"
static void convert_f16(const void * vx, const int ib, const int iqs, dfloat2 & v){ static void convert_f16(const void * vx, const int ib, const int iqs, dfloat2 & v){
const sycl::half *x = (const sycl::half *)vx; const sycl::half *x = (const sycl::half *)vx;
@@ -227,7 +228,7 @@ static void dequantize_mul_mat_vec_q2_k(const void *__restrict__ vx,
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll #pragma unroll
for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) {
tmp += tmp +=
dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask);
} }
@@ -346,7 +347,7 @@ static void dequantize_mul_mat_vec_q3_k(const void *__restrict__ vx,
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll #pragma unroll
for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) {
tmp += tmp +=
dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask);
} }
@@ -499,7 +500,7 @@ static void dequantize_mul_mat_vec_q4_k(const void *__restrict__ vx,
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll #pragma unroll
for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) {
tmp += tmp +=
dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask);
} }
@@ -633,7 +634,7 @@ static void dequantize_mul_mat_vec_q5_k(const void *__restrict__ vx,
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll #pragma unroll
for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) {
tmp += tmp +=
dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask);
} }
@@ -748,7 +749,7 @@ static void dequantize_mul_mat_vec_q6_k(const void * __restrict__ vx, const floa
// sum up partial sums and write back result // sum up partial sums and write back result
#pragma unroll #pragma unroll
for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) {
tmp += tmp +=
dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask);
} }
@@ -873,10 +874,10 @@ static void dequantize_mul_mat_vec_q2_K_sycl(const void *vx, const float *y,
const int ny = 2; // very slightly faster than 1 even when K_QUANTS_PER_ITERATION = 2 const int ny = 2; // very slightly faster than 1 even when K_QUANTS_PER_ITERATION = 2
const int block_num_y = (nrows + ny - 1) / ny; const int block_num_y = (nrows + ny - 1) / ny;
const sycl::range<3> block_nums(1, 1, block_num_y); const sycl::range<3> block_nums(1, 1, block_num_y);
const sycl::range<3> block_dims(1, ny, WARP_SIZE); const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE);
stream->parallel_for( stream->parallel_for(
sycl::nd_range<3>(block_nums * block_dims, block_dims), sycl::nd_range<3>(block_nums * block_dims, block_dims),
[=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] {
dequantize_mul_mat_vec_q2_k(vx, y, dst, ncols, nrows, item_ct1); dequantize_mul_mat_vec_q2_k(vx, y, dst, ncols, nrows, item_ct1);
}); });
} }
@@ -889,10 +890,10 @@ static void dequantize_mul_mat_vec_q3_K_sycl(const void *vx, const float *y,
const int ny = 2 / K_QUANTS_PER_ITERATION; const int ny = 2 / K_QUANTS_PER_ITERATION;
const int block_num_y = (nrows + ny - 1) / ny; const int block_num_y = (nrows + ny - 1) / ny;
const sycl::range<3> block_nums(1, 1, block_num_y); const sycl::range<3> block_nums(1, 1, block_num_y);
const sycl::range<3> block_dims(1, ny, WARP_SIZE); const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE);
stream->parallel_for( stream->parallel_for(
sycl::nd_range<3>(block_nums * block_dims, block_dims), sycl::nd_range<3>(block_nums * block_dims, block_dims),
[=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] {
dequantize_mul_mat_vec_q3_k(vx, y, dst, ncols, nrows, item_ct1); dequantize_mul_mat_vec_q3_k(vx, y, dst, ncols, nrows, item_ct1);
}); });
} }
@@ -905,10 +906,10 @@ static void dequantize_mul_mat_vec_q4_K_sycl(const void *vx, const float *y,
const int ny = 2 / K_QUANTS_PER_ITERATION; const int ny = 2 / K_QUANTS_PER_ITERATION;
const int block_num_y = (nrows + ny - 1) / ny; const int block_num_y = (nrows + ny - 1) / ny;
const sycl::range<3> block_nums(1, 1, block_num_y); const sycl::range<3> block_nums(1, 1, block_num_y);
const sycl::range<3> block_dims(1, ny, WARP_SIZE); const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE);
stream->parallel_for( stream->parallel_for(
sycl::nd_range<3>(block_nums * block_dims, block_dims), sycl::nd_range<3>(block_nums * block_dims, block_dims),
[=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] {
dequantize_mul_mat_vec_q4_k(vx, y, dst, ncols, nrows, item_ct1); dequantize_mul_mat_vec_q4_k(vx, y, dst, ncols, nrows, item_ct1);
}); });
} }
@@ -918,10 +919,10 @@ static void dequantize_mul_mat_vec_q5_K_sycl(const void *vx, const float *y,
const int nrows, const int nrows,
dpct::queue_ptr stream) { dpct::queue_ptr stream) {
GGML_ASSERT(ncols % QK_K == 0); GGML_ASSERT(ncols % QK_K == 0);
const sycl::range<3> block_dims(1, 1, WARP_SIZE); const sycl::range<3> block_dims(1, 1, QK_WARP_SIZE);
stream->parallel_for( stream->parallel_for(
sycl::nd_range<3>(sycl::range<3>(1, 1, nrows) * block_dims, block_dims), sycl::nd_range<3>(sycl::range<3>(1, 1, nrows) * block_dims, block_dims),
[=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] {
dequantize_mul_mat_vec_q5_k(vx, y, dst, ncols, item_ct1); dequantize_mul_mat_vec_q5_k(vx, y, dst, ncols, item_ct1);
}); });
} }
@@ -934,10 +935,10 @@ static void dequantize_mul_mat_vec_q6_K_sycl(const void *vx, const float *y,
const int ny = 2 / K_QUANTS_PER_ITERATION; const int ny = 2 / K_QUANTS_PER_ITERATION;
const int block_num_y = (nrows + ny - 1) / ny; const int block_num_y = (nrows + ny - 1) / ny;
const sycl::range<3> block_nums(1, 1, block_num_y); const sycl::range<3> block_nums(1, 1, block_num_y);
const sycl::range<3> block_dims(1, ny, WARP_SIZE); const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE);
stream->parallel_for( stream->parallel_for(
sycl::nd_range<3>(block_nums * block_dims, block_dims), sycl::nd_range<3>(block_nums * block_dims, block_dims),
[=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] {
dequantize_mul_mat_vec_q6_k(vx, y, dst, ncols, nrows, item_ct1); dequantize_mul_mat_vec_q6_k(vx, y, dst, ncols, nrows, item_ct1);
}); });
} }

8
ggml/src/ggml-sycl/norm.cpp
View File

@@ -57,6 +57,7 @@ static void group_norm_f32(const float* x, float* dst, const int group_size, con
const int nwarps = nthreads / WARP_SIZE; const int nwarps = nthreads / WARP_SIZE;
assert(nwarps % WARP_SIZE == 0); assert(nwarps % WARP_SIZE == 0);
start += item_ct1.get_local_id(2); start += item_ct1.get_local_id(2);
int nreduce = nwarps / WARP_SIZE;
if (end >= ne_elements) { if (end >= ne_elements) {
end = ne_elements; end = ne_elements;
@@ -87,7 +88,6 @@ static void group_norm_f32(const float* x, float* dst, const int group_size, con
*/ */
item_ct1.barrier(); item_ct1.barrier();
tmp = 0.f; tmp = 0.f;
int nreduce = nwarps / WARP_SIZE;
for (size_t i = 0; i < nreduce; i += 1) for (size_t i = 0; i < nreduce; i += 1)
{ {
tmp += s_sum[lane_id + i * WARP_SIZE]; tmp += s_sum[lane_id + i * WARP_SIZE];
@@ -122,7 +122,11 @@ static void group_norm_f32(const float* x, float* dst, const int group_size, con
better performance if there is no access to global memory. better performance if there is no access to global memory.
*/ */
item_ct1.barrier(); item_ct1.barrier();
tmp = s_sum[lane_id]; tmp = 0.f;
for (size_t i = 0; i < nreduce; i += 1)
{
tmp += s_sum[lane_id + i * WARP_SIZE];
}
tmp = warp_reduce_sum(tmp, item_ct1); tmp = warp_reduce_sum(tmp, item_ct1);
} }

1
ggml/src/ggml-sycl/presets.hpp
View File

@@ -62,4 +62,5 @@ static_assert(K_QUANTS_PER_ITERATION == 1 || K_QUANTS_PER_ITERATION == 2, "K_QUA
#define MUL_MAT_SRC1_COL_STRIDE 128 #define MUL_MAT_SRC1_COL_STRIDE 128
#define QK_WARP_SIZE 32
#endif // GGML_SYCL_PRESETS_HPP #endif // GGML_SYCL_PRESETS_HPP

250
ggml/src/ggml-sycl/softmax.cpp Normal file
View File

@@ -0,0 +1,250 @@
#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;
int 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,
local_buf_acc.get_pointer());
});
});
}
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_local_scratch = (GGML_PAD(ncols_x, WARP_SIZE) + WARP_SIZE);
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);
}

24
ggml/src/ggml-sycl/softmax.hpp Normal file
View File

@@ -0,0 +1,24 @@
//
// 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_SOFTMAX_HPP
#define GGML_SYCL_SOFTMAX_HPP
#include "common.hpp"
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);
#endif // GGML_SYCL_SOFTMAX_HPP