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

Also fix UMA handling for prealloc buffers

Browse source
This commit is contained in:
0cc4m 2024-01-29 21:22:16 +01:00
parent 54fb5c6b6c
commit f185d860e9
1 changed files with 24 additions and 18 deletions
Download patch file Download diff file Expand all files Collapse all files

42
ggml-vulkan.cpp
View file

@ -722,6 +722,24 @@ static vk_buffer ggml_vk_create_buffer_check(size_t size, vk::MemoryPropertyFlag
}
}
static vk_buffer ggml_vk_create_buffer_device(size_t size) {
vk_buffer buf;
try {
buf = ggml_vk_create_buffer(size, vk::MemoryPropertyFlagBits::eDeviceLocal);
} catch (const vk::SystemError& e) {
if (vk_device.uma) {
// Fall back to host memory type
buf = ggml_vk_create_buffer_check(size, vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
} else {
std::cerr << "ggml_vulkan: Device memory allocation of size " << size << " failed." << std::endl;
std::cerr << "ggml_vulkan: " << e.what() << std::endl;
throw e;
}
}
return buf;
}
static void ggml_vk_destroy_buffer(vk_buffer& buf) {
if (buf.size == 0) {
return;
@ -3772,35 +3790,35 @@ void ggml_vk_preallocate_buffers() {
if (vk_prealloc_qx.size > 0) {
ggml_vk_destroy_buffer(vk_prealloc_qx);
}
vk_prealloc_qx = ggml_vk_create_buffer_check(vk_prealloc_size_qx, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_prealloc_qx = ggml_vk_create_buffer_device(vk_prealloc_size_qx);
}
if (vk_prealloc_size_qy > 0 && vk_prealloc_qy.size < vk_prealloc_size_qy) {
// Resize buffer
if (vk_prealloc_qy.size > 0) {
ggml_vk_destroy_buffer(vk_prealloc_qy);
}
vk_prealloc_qy = ggml_vk_create_buffer_check(vk_prealloc_size_qy, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_prealloc_qy = ggml_vk_create_buffer_device(vk_prealloc_size_qy);
}
if (vk_prealloc_size_x > 0 && vk_prealloc_x.size < vk_prealloc_size_x) {
// Resize buffer
if (vk_prealloc_x.size > 0) {
ggml_vk_destroy_buffer(vk_prealloc_x);
}
vk_prealloc_x = ggml_vk_create_buffer_check(vk_prealloc_size_x, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_prealloc_x = ggml_vk_create_buffer_device(vk_prealloc_size_x);
}
if (vk_prealloc_size_y > 0 && vk_prealloc_y.size < vk_prealloc_size_y) {
// Resize buffer
if (vk_prealloc_y.size > 0) {
ggml_vk_destroy_buffer(vk_prealloc_y);
}
vk_prealloc_y = ggml_vk_create_buffer_check(vk_prealloc_size_y, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_prealloc_y = ggml_vk_create_buffer_device(vk_prealloc_size_y);
}
if (vk_prealloc_size_split_k > 0 && vk_prealloc_split_k.size < vk_prealloc_size_split_k) {
// Resize buffer
if (vk_prealloc_split_k.size > 0) {
ggml_vk_destroy_buffer(vk_prealloc_split_k);
}
vk_prealloc_split_k = ggml_vk_create_buffer_check(vk_prealloc_size_split_k, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_prealloc_split_k = ggml_vk_create_buffer_device(vk_prealloc_size_split_k);
}
if (vk_staging_size > 0 && vk_staging.size < vk_staging_size) {
// Resize buffer
@ -4301,19 +4319,7 @@ GGML_CALL static ggml_backend_buffer_t ggml_backend_vk_buffer_type_alloc_buffer(
#ifdef VK_DEBUG
std::cerr << "ggml_backend_vk_buffer_type_alloc_buffer(" << size << ")" << std::endl;
#endif
vk_buffer dev_buffer;
try {
dev_buffer = ggml_vk_create_buffer(size, vk::MemoryPropertyFlagBits::eDeviceLocal);
} catch (const vk::SystemError& e) {
if (vk_device.uma) {
// Fall back to host memory type
dev_buffer = ggml_vk_create_buffer_check(size, vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
} else {
std::cerr << "ggml_vulkan: Device memory allocation of size " << size << " failed." << std::endl;
std::cerr << "ggml_vulkan: " << e.what() << std::endl;
throw e;
}
}
vk_buffer dev_buffer = ggml_vk_create_buffer_device(size);
ggml_backend_vk_buffer_context * ctx = new ggml_backend_vk_buffer_context(dev_buffer);