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

cann: fix multi-npu exec error

Browse source
This commit is contained in:
wangshuai09 2024-07-26 09:28:17 +00:00
parent 01245f5b16
commit 937a12c1bc
1 changed files with 15 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

30
ggml/src/ggml-cann.cpp
View file

@ -1559,23 +1559,18 @@ GGML_CALL static bool ggml_backend_cann_cpy_tensor_async(
return false; return false;
} }
// need open both directions for memcpyasync between devices.
ggml_cann_set_device(cann_ctx_dst->device);
ACL_CHECK(aclrtDeviceEnablePeerAccess(cann_ctx_src->device, 0));
ggml_cann_set_device(cann_ctx_src->device); ggml_cann_set_device(cann_ctx_src->device);
ACL_CHECK(aclrtDeviceEnablePeerAccess(cann_ctx_dst->device, 0)); ACL_CHECK(aclrtDeviceEnablePeerAccess(cann_ctx_dst->device, 0));
ACL_CHECK(aclrtMemcpyAsync(dst->data, copy_size, src->data, copy_size, ACL_CHECK(aclrtMemcpyAsync(dst->data, copy_size, src->data, copy_size,
ACL_MEMCPY_DEVICE_TO_DEVICE, ACL_MEMCPY_DEVICE_TO_DEVICE,
cann_ctx_dst->stream())); cann_ctx_src->stream()));
// record event on src stream //TODO: workaround for Event didn`t work here.
if (!cann_ctx_src->copy_event) { aclrtSynchronizeStream(cann_ctx_src->stream());
ACL_CHECK(aclrtCreateEvent(&cann_ctx_src->copy_event));
}
ACL_CHECK(
aclrtRecordEvent(cann_ctx_src->copy_event, cann_ctx_src->stream()));
// wait on dst stream for the copy to complete
ACL_CHECK(aclrtStreamWaitEvent(cann_ctx_dst->stream(),
cann_ctx_src->copy_event));
} else { } else {
// src and dst are on the same backend // src and dst are on the same backend
ACL_CHECK(aclrtMemcpyAsync(dst->data, copy_size, src->data, copy_size, ACL_CHECK(aclrtMemcpyAsync(dst->data, copy_size, src->data, copy_size,
@ -1773,9 +1768,14 @@ static bool ggml_backend_buft_is_cann(ggml_backend_buffer_type_t buft) {
*/ */
GGML_CALL static bool ggml_backend_cann_supports_buft( GGML_CALL static bool ggml_backend_cann_supports_buft(
ggml_backend_t backend, ggml_backend_buffer_type_t buft) { ggml_backend_t backend, ggml_backend_buffer_type_t buft) {
return buft->iface.get_name == ggml_backend_cann_buffer_type_name; if (ggml_backend_buft_is_cann(buft)) {
ggml_backend_cann_context * cann_ctx =
GGML_UNUSED(backend); (ggml_backend_cann_context *)backend->context;
ggml_backend_cann_buffer_type_context * buft_ctx =
(ggml_backend_cann_buffer_type_context *)buft->context;
return buft_ctx->device == cann_ctx->device;
}
return false;
} }
/** /**