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

try fix

Browse source
This commit is contained in:
JohannesGaessler 2023-12-31 15:47:35 +01:00
parent f0a176bd43
commit 452749e792
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-cuda.cu
View file

@ -8216,7 +8216,7 @@ static void ggml_cuda_op_mul_mat(
cudaStream_t stream = g_cudaStreams[id][is];
// wait for main GPU data if necessary
if (split && (id != g_main_device || is != 0)) { // TODO is this correct?
if (split && (id != g_main_device || is != is0)) { // TODO is this correct?
CUDA_CHECK(cudaStreamWaitEvent(stream, src0_extra->events[g_main_device][is0], 0));
}
@ -8322,7 +8322,7 @@ static void ggml_cuda_op_mul_mat(
}
// add event for the main device to wait on until other device is done
if (split && (id != g_main_device || is != 0)) { // TODO is this correct?
if (split && (id != g_main_device || is != is0)) { // TODO is this correct?
CUDA_CHECK(cudaEventRecord(src0_extra->events[id][is], stream));
}
}