vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Add missing barrier

Browse source
This commit is contained in:
0cc4m 2023-08-02 22:04:43 +02:00
parent c638955bfa
commit 44bbc85aaf
1 changed files with 1 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ggml-vulkan.cpp
View file

@ -2181,6 +2181,7 @@ static void ggml_vk_mul_mat_vec_q_f16(const ggml_tensor * src0, const ggml_tenso
// copy dst to host // copy dst to host
float * d = (float *) ((char *) dst->data + i02*nb2 + i03*nb3); float * d = (float *) ((char *) dst->data + i02*nb2 + i03*nb3);
ggml_vk_sync_buffers(s.buffer, { { d_D, d_offset, d_sz } }, compq, vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eTransferRead, true);
ggml_vk_buffer_read_async(&d_D, d_offset, d, sizeof(float) * d_ne, compq, {}, {}, &s); ggml_vk_buffer_read_async(&d_D, d_offset, d, sizeof(float) * d_ne, compq, {}, {}, &s);
ggml_vk_end_submission(s, {}, {}); ggml_vk_end_submission(s, {}, {});