root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-11-14 06:49:54 +00:00
Code Issues Actions 14 Packages Projects Releases Wiki Activity

CUDA: fixed mmvq kernel for bs 2,3,4 and -sm row (#5386)

Browse Source
This commit is contained in:
Johannes Gäßler 2024-02-07 12:40:26 +01:00 committed by GitHub
parent 10afa6f1d1
commit aa7ab99be2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 39 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

66
ggml-cuda.cu
View File

@ -5313,7 +5313,7 @@ template <bool need_check> static __global__ void
template <int ncols_y_template, int qk, int qi, typename block_q_t, int vdr, vec_dot_q_cuda_t vec_dot_q_cuda> template <int ncols_y_template, int qk, int qi, typename block_q_t, int vdr, vec_dot_q_cuda_t vec_dot_q_cuda>
static __global__ void mul_mat_vec_q( static __global__ void mul_mat_vec_q(
const void * __restrict__ vx, const void * __restrict__ vy, float * __restrict__ dst, const void * __restrict__ vx, const void * __restrict__ vy, float * __restrict__ dst,
const int ncols_x, const int nrows_x, const int nrows_y, const int ncols_y_par) { const int ncols_x, const int nrows_x, const int nrows_y, const int ncols_y_par, const int nrows_dst) {
const int ncols_y = ncols_y_template != 0 ? ncols_y_template : ncols_y_par; const int ncols_y = ncols_y_template != 0 ? ncols_y_template : ncols_y_par;
@ -5352,7 +5352,7 @@ static __global__ void mul_mat_vec_q(
tmp[j] = warp_reduce_sum(tmp[j]); tmp[j] = warp_reduce_sum(tmp[j]);
if (threadIdx.x == 0) { if (threadIdx.x == 0) {
dst[j*nrows_x + row] = tmp[j]; dst[j*nrows_dst + row] = tmp[j];
} }
} }
} }
@ -6828,7 +6828,7 @@ static void convert_mul_mat_vec_f16_cuda(const void * vx, const dfloat * y, floa
template <int qk, int qi, typename block_q_t, int vdr, vec_dot_q_cuda_t vec_dot> template <int qk, int qi, typename block_q_t, int vdr, vec_dot_q_cuda_t vec_dot>
static void mul_mat_vec_q_cuda( static void mul_mat_vec_q_cuda(
const void * vx, const void * vy, float * dst, const void * vx, const void * vy, float * dst,
const int ncols_x, const int nrows_x, const int nrows_y, const int ncols_y, cudaStream_t stream) { const int ncols_x, const int nrows_x, const int nrows_y, const int ncols_y, const int nrows_dst, cudaStream_t stream) {
GGML_ASSERT(ncols_x % qk == 0); GGML_ASSERT(ncols_x % qk == 0);
GGML_ASSERT(ncols_y <= 4); GGML_ASSERT(ncols_y <= 4);
@ -6839,40 +6839,40 @@ static void mul_mat_vec_q_cuda(
switch (ncols_y) { switch (ncols_y) {
case 1: case 1:
mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot> mul_mat_vec_q<1, qk, qi, block_q_t, vdr, vec_dot>
<<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
break; break;
case 2: case 2:
mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot> mul_mat_vec_q<2, qk, qi, block_q_t, vdr, vec_dot>
<<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
break; break;
case 3: case 3:
mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot> mul_mat_vec_q<3, qk, qi, block_q_t, vdr, vec_dot>
<<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
break; break;
case 4: case 4:
mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot> mul_mat_vec_q<4, qk, qi, block_q_t, vdr, vec_dot>
<<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
break; break;
// case 5: // case 5:
// mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot> // mul_mat_vec_q<5, qk, qi, block_q_t, vdr, vec_dot>
// <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); // <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
// break; // break;
// case 6: // case 6:
// mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot> // mul_mat_vec_q<6, qk, qi, block_q_t, vdr, vec_dot>
// <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); // <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
// break; // break;
// case 7: // case 7:
// mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot> // mul_mat_vec_q<7, qk, qi, block_q_t, vdr, vec_dot>
// <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); // <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
// break; // break;
// case 8: // case 8:
// mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot> // mul_mat_vec_q<8, qk, qi, block_q_t, vdr, vec_dot>
// <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); // <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
// break; // break;
default: default:
GGML_ASSERT(false); GGML_ASSERT(false);
// mul_mat_vec_q<0, qk, qi, block_q_t, vdr, vec_dot> // mul_mat_vec_q<0, qk, qi, block_q_t, vdr, vec_dot>
// <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y); // <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols_x, nrows_x, nrows_y, ncols_y, nrows_dst);
break; break;
} }
} }
@ -8391,7 +8391,7 @@ static void ggml_cuda_op_mul_mat_q(
CUDA_CHECK(cudaGetDevice(&id)); CUDA_CHECK(cudaGetDevice(&id));
// the main device has a larger memory buffer to hold the results from all GPUs // the main device has a larger memory buffer to hold the results from all GPUs
// nrows_dst == nrows of the matrix that the dequantize_mul_mat kernel writes into // nrows_dst == nrows of the matrix that the kernel writes into
const int64_t nrows_dst = dst->backend == GGML_BACKEND_GPU && id == g_main_device ? ne0 : row_diff; const int64_t nrows_dst = dst->backend == GGML_BACKEND_GPU && id == g_main_device ? ne0 : row_diff;
switch (src0->type) { switch (src0->type) {
@ -8525,58 +8525,70 @@ static void ggml_cuda_op_mul_mat_vec_q(
const int64_t ne00 = src0->ne[0]; const int64_t ne00 = src0->ne[0];
const int64_t row_diff = row_high - row_low; const int64_t row_diff = row_high - row_low;
const int64_t ne10 = src1->ne[0];
GGML_ASSERT(ne10 % QK8_1 == 0);
const int64_t ne0 = dst->ne[0];
int id;
CUDA_CHECK(cudaGetDevice(&id));
// the main device has a larger memory buffer to hold the results from all GPUs
// nrows_dst == nrows of the matrix that the kernel writes into
const int64_t nrows_dst = dst->backend == GGML_BACKEND_GPU && id == g_main_device ? ne0 : row_diff;
switch (src0->type) { switch (src0->type) {
case GGML_TYPE_Q4_0: case GGML_TYPE_Q4_0:
mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1> mul_mat_vec_q_cuda<QK4_0, QI4_0, block_q4_0, VDR_Q4_0_Q8_1_MMVQ, vec_dot_q4_0_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q4_1: case GGML_TYPE_Q4_1:
mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1> mul_mat_vec_q_cuda<QK4_1, QI4_1, block_q4_1, VDR_Q4_1_Q8_1_MMVQ, vec_dot_q4_1_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q5_0: case GGML_TYPE_Q5_0:
mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1> mul_mat_vec_q_cuda<QK5_0, QI5_0, block_q5_0, VDR_Q5_0_Q8_1_MMVQ, vec_dot_q5_0_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q5_1: case GGML_TYPE_Q5_1:
mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1> mul_mat_vec_q_cuda<QK5_1, QI5_1, block_q5_1, VDR_Q5_1_Q8_1_MMVQ, vec_dot_q5_1_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q8_0: case GGML_TYPE_Q8_0:
mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1> mul_mat_vec_q_cuda<QK8_0, QI8_0, block_q8_0, VDR_Q8_0_Q8_1_MMVQ, vec_dot_q8_0_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q2_K: case GGML_TYPE_Q2_K:
mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1> mul_mat_vec_q_cuda<QK_K, QI2_K, block_q2_K, VDR_Q2_K_Q8_1_MMVQ, vec_dot_q2_K_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q3_K: case GGML_TYPE_Q3_K:
mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1> mul_mat_vec_q_cuda<QK_K, QI3_K, block_q3_K, VDR_Q3_K_Q8_1_MMVQ, vec_dot_q3_K_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q4_K: case GGML_TYPE_Q4_K:
mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1> mul_mat_vec_q_cuda<QK_K, QI4_K, block_q4_K, VDR_Q4_K_Q8_1_MMVQ, vec_dot_q4_K_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q5_K: case GGML_TYPE_Q5_K:
mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1> mul_mat_vec_q_cuda<QK_K, QI5_K, block_q5_K, VDR_Q5_K_Q8_1_MMVQ, vec_dot_q5_K_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_Q6_K: case GGML_TYPE_Q6_K:
mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1> mul_mat_vec_q_cuda<QK_K, QI6_K, block_q6_K, VDR_Q6_K_Q8_1_MMVQ, vec_dot_q6_K_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_IQ2_XXS: case GGML_TYPE_IQ2_XXS:
mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1> mul_mat_vec_q_cuda<QK_K, QI2_XXS, block_iq2_xxs, 1, vec_dot_iq2_xxs_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_IQ2_XS: case GGML_TYPE_IQ2_XS:
mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1> mul_mat_vec_q_cuda<QK_K, QI2_XS, block_iq2_xs, 1, vec_dot_iq2_xs_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
case GGML_TYPE_IQ3_XXS: case GGML_TYPE_IQ3_XXS:
mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1> mul_mat_vec_q_cuda<QK_K, QI3_XXS, block_iq3_xxs, 1, vec_dot_iq3_xxs_q8_1>
(src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, stream); (src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, src1_padded_row_size, src1_ncols, nrows_dst, stream);
break; break;
default: default:
GGML_ASSERT(false); GGML_ASSERT(false);
@ -9909,7 +9921,7 @@ static void ggml_cuda_mul_mat(const ggml_tensor * src0, const ggml_tensor * src1
ggml_cuda_op_mul_mat(src0, src1, dst, ggml_cuda_op_dequantize_mul_mat_vec, false); ggml_cuda_op_mul_mat(src0, src1, dst, ggml_cuda_op_dequantize_mul_mat_vec, false);
} }
} else { } else {
if (src1->ne[1] <= 4 && min_compute_capability >= MIN_CC_DP4A && ggml_is_quantized(src0->type)) { if (src1->ne[1] <= 4 && min_compute_capability >= MIN_CC_DP4A && ggml_is_quantized(src0->type) && src1->type == GGML_TYPE_F32) {
ggml_cuda_op_mul_mat(src0, src1, dst, ggml_cuda_op_mul_mat_vec_q, true); ggml_cuda_op_mul_mat(src0, src1, dst, ggml_cuda_op_mul_mat_vec_q, true);
} else if (use_mul_mat_q) { } else if (use_mul_mat_q) {
ggml_cuda_op_mul_mat(src0, src1, dst, ggml_cuda_op_mul_mat_q, true); ggml_cuda_op_mul_mat(src0, src1, dst, ggml_cuda_op_mul_mat_q, true);