CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-31 08:51:55 +00:00
Code Issues Packages Projects Releases Wiki Activity

sycl: quantize and reorder the input to q8_1 when reorder is enabled (#13826)

Browse Source 
* [WIP]: fuse q8 quantization and reorder

* wip2: fuse q8 quantization and reorder

* working q8 reorder commit

* restored common.hpp

* remove debug prints

* remove unnecessary headers and remove trailing whitespace

* Update ggml/src/ggml-sycl/ggml-sycl.cpp

Co-authored-by: Alberto Cabrera Pérez <alberto.cabrera@intel.com>

---------

Co-authored-by: Alberto Cabrera Pérez <alberto.cabrera@intel.com>
This commit is contained in:
Atharva Dubey
2025-06-02 10:12:20 +01:00
committed by GitHub
parent 7675c555a1
commit 663445b0de
3 changed files with 120 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

79
ggml/src/ggml-sycl/ggml-sycl.cpp
View File

@@ -1434,6 +1434,59 @@ static void quantize_q8_1(const float * __restrict__ x, void * __restrict__ vy,
reinterpret_cast<sycl::half &>(y[ib].ds.y()) = sum; reinterpret_cast<sycl::half &>(y[ib].ds.y()) = sum;
} }
template <int ElementsPerWI>
static __dpct_inline__ void quantize_and_reorder_q8_1(const float * __restrict__ x, void * reordered_q8_tensor,
const int kx, const int kx_padded, const sycl::nd_item<1> & it) {
/*
Quantizes and reorders the resultant q8 tensor in a per row fashion
Each sub-group calculates one quant block. i.e. QK8_1 quant values and the d and sum values
*/
auto subgroup_id = it.get_group(0);
auto wi_id = it.get_local_id(0);
const int num_blocks_per_row = kx / QK8_1;
auto row = subgroup_id / num_blocks_per_row;
auto col = subgroup_id % num_blocks_per_row;
auto row_offset = row * (kx_padded / QK8_1) * sizeof(block_q8_1);
auto col_offset = QK8_1 * col + wi_id * ElementsPerWI;
auto quant_ptr = (int8_t *) ((char *) reordered_q8_tensor + row_offset + col_offset);
auto ds_ptr = (sycl::half2 *) ((char *) reordered_q8_tensor + row_offset + kx + col * sizeof(sycl::half2));
sycl::vec<float, ElementsPerWI> wi_f32_vals;
sycl::vec<int8_t, ElementsPerWI> quantized_values;
auto float_ptr_offset = subgroup_id * QK8_1 + ElementsPerWI * wi_id;
wi_f32_vals = *reinterpret_cast<const sycl::vec<float, ElementsPerWI> *>(x + float_ptr_offset);
float sum = 0.0f;
float amax = 0.0f;
#pragma unroll(ElementsPerWI)
for (int i = 0; i < ElementsPerWI; i++) {
sum += wi_f32_vals[i];
amax = sycl::fmax(amax, sycl::fabs(wi_f32_vals[i]));
quantized_values[i] = 0;
}
sum = sycl::reduce_over_group(it.get_group(), sum, sycl::plus<float>());
amax = sycl::reduce_over_group(it.get_group(), amax, sycl::maximum<float>());
float d = amax == 0 ? 1 : amax / 127;
#pragma unroll(ElementsPerWI)
for (int i = 0; i < ElementsPerWI; i++) {
quantized_values[i] = sycl::round(wi_f32_vals[i] / d);
}
d = amax == 0 ? 0 : d;
*reinterpret_cast<sycl::vec<int8_t, ElementsPerWI> *>(quant_ptr) = quantized_values;
if (wi_id == 0) {
*ds_ptr = sycl::half2(sycl::half(d), sycl::half(sum));
}
}
static void mul_mat_p021_f16_f32( static void mul_mat_p021_f16_f32(
const void * __restrict__ vx, const float * __restrict__ y, float * __restrict__ dst, const void * __restrict__ vx, const float * __restrict__ y, float * __restrict__ dst,
const int ncols_x, const int nrows_x, const int nchannels_x, const int nchannels_y, const int ncols_x, const int nrows_x, const int nchannels_x, const int nchannels_y,
@@ -1718,24 +1771,31 @@ static void pool2d_nchw_kernel(
o_ptr[cur_oh * ow + cur_ow] = res; o_ptr[cur_oh * ow + cur_ow] = res;
} }
static void quantize_row_q8_1_sycl(const float *x, void *vy, const int kx, static void quantize_row_q8_1_sycl(const float * x, void * vy, const int kx, const int ky, const int kx_padded,
const int ky, const int kx_padded, bool reorder_q8_tensor, queue_ptr stream) {
queue_ptr stream) { if (reorder_q8_tensor) {
auto local_range = std::size_t(WARP_SIZE);
auto num_quant_blocks = ky * (kx / QK8_1);
auto global_range = num_quant_blocks * local_range;
stream->parallel_for(sycl::nd_range<1>({ global_range }, { local_range }),
[=](sycl::nd_item<1> it) [[sycl::reqd_sub_group_size(WARP_SIZE)]] {
quantize_and_reorder_q8_1<QK8_1 / WARP_SIZE>(x, vy, kx, kx_padded, it);
});
} else {
const int block_num_x = (kx_padded + SYCL_QUANTIZE_BLOCK_SIZE - 1) / SYCL_QUANTIZE_BLOCK_SIZE; const int block_num_x = (kx_padded + SYCL_QUANTIZE_BLOCK_SIZE - 1) / SYCL_QUANTIZE_BLOCK_SIZE;
const sycl::range<3> num_blocks(1, ky, block_num_x); const sycl::range<3> num_blocks(1, ky, block_num_x);
int constexpr QUANT_BLOCK_TILE = QK8_1 / WARP_SIZE; int constexpr QUANT_BLOCK_TILE = QK8_1 / WARP_SIZE;
static_assert(QK8_1 % WARP_SIZE == 0); static_assert(QK8_1 % WARP_SIZE == 0);
const sycl::range<3> block_size(1, 1, SYCL_QUANTIZE_BLOCK_SIZE / QUANT_BLOCK_TILE); const sycl::range<3> block_size(1, 1, SYCL_QUANTIZE_BLOCK_SIZE / QUANT_BLOCK_TILE);
{ {
dpct::has_capability_or_fail(stream->get_device(), dpct::has_capability_or_fail(stream->get_device(), { sycl::aspect::fp16 });
{sycl::aspect::fp16});
stream->parallel_for( stream->parallel_for(sycl::nd_range<3>(num_blocks * block_size, block_size),
sycl::nd_range<3>(num_blocks * block_size, block_size),
[=](sycl::nd_item<3> item_ct1) [[sycl::reqd_sub_group_size(WARP_SIZE)]] { [=](sycl::nd_item<3> item_ct1) [[sycl::reqd_sub_group_size(WARP_SIZE)]] {
quantize_q8_1<QUANT_BLOCK_TILE>(x, vy, kx, kx_padded, item_ct1); quantize_q8_1<QUANT_BLOCK_TILE>(x, vy, kx, kx_padded, item_ct1);
}); });
} }
}
} }
static void ggml_mul_mat_p021_f16_f32_sycl(const void *vx, const float *y, static void ggml_mul_mat_p021_f16_f32_sycl(const void *vx, const float *y,
@@ -2446,9 +2506,10 @@ static void ggml_sycl_op_mul_mat(ggml_backend_sycl_context & ctx, const ggml_ten
dev[i].src1_ddq = dev[i].src1_ddq_alloc.alloc(ctx.pool(i), nrows1*src1_padded_col_size*q8_1_ts/q8_1_bs); dev[i].src1_ddq = dev[i].src1_ddq_alloc.alloc(ctx.pool(i), nrows1*src1_padded_col_size*q8_1_ts/q8_1_bs);
if (src1_on_device && src1_is_contiguous) { if (src1_on_device && src1_is_contiguous) {
bool reorder_q8_tensor = src0->extra && ((ggml_tensor_extra_gpu *)src0->extra)->optimized_feature.reorder;
scope_op_debug_print scope_dbg_print(__func__, "/quantize_row_q8_1_sycl", dst, scope_op_debug_print scope_dbg_print(__func__, "/quantize_row_q8_1_sycl", dst,
/*num_src=*/2, " : converting src1 to Q8_1"); /*num_src=*/2, " : converting src1 to Q8_1");
quantize_row_q8_1_sycl(dev[i].src1_ddf, dev[i].src1_ddq, ne10, nrows1, src1_padded_col_size, stream); quantize_row_q8_1_sycl(dev[i].src1_ddf, dev[i].src1_ddq, ne10, nrows1, src1_padded_col_size, reorder_q8_tensor, stream);
/* /*
DPCT1010:90: SYCL uses exceptions to report errors and does not DPCT1010:90: SYCL uses exceptions to report errors and does not
use the error codes. The call was replaced with 0. You need to use the error codes. The call was replaced with 0. You need to
@@ -2554,7 +2615,7 @@ static void ggml_sycl_op_mul_mat(ggml_backend_sycl_context & ctx, const ggml_ten
if (convert_src1_to_q8_1 && !src1_is_contiguous) { if (convert_src1_to_q8_1 && !src1_is_contiguous) {
scope_op_debug_print scope_dbg_print(__func__, "/quantize_row_q8_1_sycl", dst, scope_op_debug_print scope_dbg_print(__func__, "/quantize_row_q8_1_sycl", dst,
/*num_src=*/2, " : converting src1 to Q8_1"); /*num_src=*/2, " : converting src1 to Q8_1");
quantize_row_q8_1_sycl(src1_ddf_i, src1_ddq_i, ne10, src1_ncols, src1_padded_col_size, stream); quantize_row_q8_1_sycl(src1_ddf_i, src1_ddq_i, ne10, src1_ncols, src1_padded_col_size, false, stream);
/* /*
DPCT1010:92: SYCL uses exceptions to report errors and does DPCT1010:92: SYCL uses exceptions to report errors and does
not use the error codes. The call was replaced with 0. You not use the error codes. The call was replaced with 0. You

6
ggml/src/ggml-sycl/mmvq.cpp
View File

@@ -29,8 +29,6 @@ static void mul_mat_vec_q_reorder(const void * __restrict__ vx, const void * __r
static_assert(blocks_per_subgroup > 0); static_assert(blocks_per_subgroup > 0);
static_assert(block_elements_per_subgroup > 0); static_assert(block_elements_per_subgroup > 0);
const block_q8_1 * y = (const block_q8_1 *) vy;
float partial_sum = 0.0f; float partial_sum = 0.0f;
for (int i = sg.get_local_linear_id() / block_elements_per_subgroup; i < blocks_per_row; i += blocks_per_subgroup) { for (int i = sg.get_local_linear_id() / block_elements_per_subgroup; i < blocks_per_row; i += blocks_per_subgroup) {
const int ibx = row * blocks_per_row + i; // x block index const int ibx = row * blocks_per_row + i; // x block index
@@ -40,13 +38,15 @@ static void mul_mat_vec_q_reorder(const void * __restrict__ vx, const void * __r
// Y block index that aligns with ibx // Y block index that aligns with ibx
const int iby = i * block_type::block_to_q8_1_ratio(); const int iby = i * block_type::block_to_q8_1_ratio();
const int8_t* q8_1_quant_ptr = (const int8_t*)vy + iby * QK8_1;
const sycl::half2* q8_1_ds_ptr = (const sycl::half2*)((const char*)vy + ncols + iby * sizeof(sycl::half2));
#pragma unroll #pragma unroll
for (int elem = 0; elem < block_elements_per_subgroup; elem += WARP_SIZE) { for (int elem = 0; elem < block_elements_per_subgroup; elem += WARP_SIZE) {
// x block quant index when casting the quants to int // x block quant index when casting the quants to int
const int iqs = elem + block_traits::vdr_mmvq * (sg.get_local_linear_id() % block_elements_per_subgroup); const int iqs = elem + block_traits::vdr_mmvq * (sg.get_local_linear_id() % block_elements_per_subgroup);
partial_sum += reorder_vec_dot_q_sycl()(vx, bx_offset, d_offset, &y[iby], iqs, nblocks); partial_sum += reorder_vec_dot_q_sycl()(vx, bx_offset, d_offset, q8_1_quant_ptr, q8_1_ds_ptr, iqs, nblocks);
} }
} }

45
ggml/src/ggml-sycl/vecdotq.hpp
View File

@@ -285,21 +285,21 @@ template <> struct reorder_vec_dot_q_sycl<GGML_TYPE_Q4_0> {
} }
__dpct_inline__ float operator()(const void * __restrict__ vbq, const int ibx_offset, const int d_offset, __dpct_inline__ float operator()(const void * __restrict__ vbq, const int ibx_offset, const int d_offset,
const block_q8_1 * __restrict__ bq8_1, const int & iqs, int /* nblocks */) { const int8_t* q8_1_quant_ptr, const sycl::half2* q8_1_ds, const int & iqs, int /* nblocks */) {
const uint8_t * bq4_0 = static_cast<const uint8_t *>(vbq) + ibx_offset; const uint8_t * bq4_0 = static_cast<const uint8_t *>(vbq) + ibx_offset;
const ggml_half d = *(reinterpret_cast<const ggml_half *>(static_cast<const uint8_t *>(vbq) + d_offset)); const ggml_half d = *(reinterpret_cast<const ggml_half *>(static_cast<const uint8_t *>(vbq) + d_offset));
int v[q4_0_traits::vdr_mmvq]; int v[q4_0_traits::vdr_mmvq];
int u[2 * q4_0_traits::vdr_mmvq]; int u[2 * q4_0_traits::vdr_mmvq];
#pragma unroll
#pragma unroll
for (size_t i = 0; i < q4_0_traits::vdr_mmvq; ++i) { for (size_t i = 0; i < q4_0_traits::vdr_mmvq; ++i) {
v[i] = get_int_from_uint8(bq4_0, iqs + i); v[i] = get_int_from_uint8(bq4_0, iqs + i);
u[2 * i + 0] = get_int_from_int8_aligned(bq8_1->qs, iqs + i); u[2 * i + 0] = get_int_from_int8_aligned(q8_1_quant_ptr, iqs + i);
u[2 * i + 1] = get_int_from_int8_aligned(bq8_1->qs, iqs + i + q4_0_traits::qi); u[2 * i + 1] = get_int_from_int8_aligned(q8_1_quant_ptr, iqs + i + q4_0_traits::qi);
} }
return vec_dot_q4_0_q8_1_impl(v, u, d, bq8_1->ds); return vec_dot_q4_0_q8_1_impl(v, u, d, *q8_1_ds);
}; };
}; };
@@ -347,7 +347,7 @@ template <> struct reorder_vec_dot_q_sycl<GGML_TYPE_Q4_K> {
using q4_k_traits = typename q4_k_block::traits; using q4_k_traits = typename q4_k_block::traits;
float operator()(const void * __restrict__ vbq, const int ibx_offset, const int d_offset, float operator()(const void * __restrict__ vbq, const int ibx_offset, const int d_offset,
const block_q8_1 * __restrict__ bq8_1, const int & iqs, int nblocks) { const int8_t* q8_1_quant_ptr, const sycl::half2* q8_1_ds, const int & iqs, int nblocks) {
const int ib = ibx_offset / (QK_K / 2); const int ib = ibx_offset / (QK_K / 2);
const uint8_t * base = static_cast<const uint8_t *>(vbq); const uint8_t * base = static_cast<const uint8_t *>(vbq);
@@ -360,7 +360,38 @@ template <> struct reorder_vec_dot_q_sycl<GGML_TYPE_Q4_K> {
const int * q4 = (const int *) (qs + 16 * bq8_offset + 4 * ((iqs / 2) % 4)); const int * q4 = (const int *) (qs + 16 * bq8_offset + 4 * ((iqs / 2) % 4));
const uint16_t * scales = (const uint16_t *) scs; const uint16_t * scales = (const uint16_t *) scs;
return vec_dot_q4_K_q8_1_common(q4, scales, *dms, bq8_1, iqs); int v[2];
int u[2 * QR4_K];
float d8[QR4_K];
v[0] = q4[0];
v[1] = q4[4];
uint16_t aux[2];
const int j = (QR4_K * ((iqs / 2) / (QI8_1 / 2))) / 2;
if (j < 2) {
aux[0] = scales[j + 0] & 0x3f3f;
aux[1] = scales[j + 2] & 0x3f3f;
} else {
aux[0] = ((scales[j + 2] >> 0) & 0x0f0f) | ((scales[j - 2] & 0xc0c0) >> 2);
aux[1] = ((scales[j + 2] >> 4) & 0x0f0f) | ((scales[j - 0] & 0xc0c0) >> 2);
}
const uint8_t * sc = (const uint8_t *) aux;
const uint8_t * m = sc + 2;
for (int i = 0; i < QR4_K; ++i) {
const int8_t* quant_base_ptr = q8_1_quant_ptr + (bq8_offset + i) * QK8_1;
sycl::half2 ds_values = *(q8_1_ds + bq8_offset + i);
d8[i] = ds_values[0];
const int * q8 = (const int *) quant_base_ptr + ((iqs / 2) % 4);
u[2 * i + 0] = q8[0];
u[2 * i + 1] = q8[4];
}
return vec_dot_q4_K_q8_1_impl_vmmq(v, u, sc, m, *dms, d8);
} }
}; };