root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-28 12:24:35 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

cuda : minor indentation

Browse Source
This commit is contained in:
Georgi Gerganov 2023-10-23 20:36:50 +03:00
parent 8d8d54f834
commit 84d4ca0e47
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 7 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
ggml-cuda.cu
View File

@ -4326,13 +4326,13 @@ static __global__ void mul_mat_vec_nc_f16_f32( // nc == non-contiguous
const half * x = (const half *) vx; const half * x = (const half *) vx;
const int row_x = blockDim.y*blockIdx.y + threadIdx.y; const int row_x = blockDim.y*blockIdx.y + threadIdx.y;
const int channel = blockDim.z*blockIdx.z + threadIdx.z; const int channel = blockDim.z*blockIdx.z + threadIdx.z;
const int channel_x = channel / channel_x_divisor; const int channel_x = channel / channel_x_divisor;
const int nrows_y = ncols_x; const int nrows_y = ncols_x;
const int nrows_dst = nrows_x; const int nrows_dst = nrows_x;
const int row_dst = row_x; const int row_dst = row_x;
const int idst = channel*nrows_dst + row_dst; const int idst = channel*nrows_dst + row_dst;
@ -4345,13 +4345,13 @@ static __global__ void mul_mat_vec_nc_f16_f32( // nc == non-contiguous
break; break;
} }
const int ix = channel_x*channel_stride_x + row_x*row_stride_x + col_x;
const float xi = __half2float(x[ix]);
const int row_y = col_x; const int row_y = col_x;
const int ix = channel_x*channel_stride_x + row_x*row_stride_x + col_x;
const int iy = channel*nrows_y + row_y; const int iy = channel*nrows_y + row_y;
const float xi = __half2float(x[ix]);
tmp += xi * y[iy]; tmp += xi * y[iy];
} }