sync : leftovers (ggml/0)

ggml-ci
This commit is contained in:
Georgi Gerganov 2024-11-15 21:43:41 +02:00
parent 44d1cbdfe9
commit 401fbea326
3 changed files with 12 additions and 9 deletions

View File

@ -211,10 +211,13 @@ elseif (CMAKE_OSX_ARCHITECTURES STREQUAL "x86_64" OR CMAKE_GENERATOR_PLATFORM_LW
endif() endif()
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64") elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64")
message(STATUS "PowerPC detected") message(STATUS "PowerPC detected")
execute_process(COMMAND bash -c "grep POWER10 /proc/cpuinfo | head -n 1" execute_process(COMMAND bash -c "grep POWER10 /proc/cpuinfo | head -n 1" OUTPUT_VARIABLE POWER10_M)
OUTPUT_VARIABLE POWER10_M) string(FIND "${POWER10_M}" "POWER10" substring_index)
string(FIND ${POWER10_M} "POWER10" substring_index) if (NOT DEFINED substring_index OR "${substring_index}" STREQUAL "")
if(${substring_index} GREATER_EQUAL 0) set(substring_index -1)
endif()
if (${substring_index} GREATER_EQUAL 0)
list(APPEND ARCH_FLAGS -mcpu=power10) list(APPEND ARCH_FLAGS -mcpu=power10)
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64le") elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64le")
list(APPEND ARCH_FLAGS -mcpu=powerpc64le) list(APPEND ARCH_FLAGS -mcpu=powerpc64le)

View File

@ -3553,7 +3553,7 @@ static enum ggml_status ggml_metal_graph_compute(
dispatch_apply(n_cb, ctx->d_queue, ctx->encode_async); dispatch_apply(n_cb, ctx->d_queue, ctx->encode_async);
// wait for completion and check status of each command buffer // wait for completion and check status of each command buffer
// needed to detect if the device ran out-of-memory for example (llama/1881) // needed to detect if the device ran out-of-memory for example (ggml/1881)
{ {
id<MTLCommandBuffer> command_buffer = ctx->command_buffers[n_cb]; id<MTLCommandBuffer> command_buffer = ctx->command_buffers[n_cb];
[command_buffer waitUntilCompleted]; [command_buffer waitUntilCompleted];

View File

@ -1763,7 +1763,8 @@ static void ggml_vk_print_gpu_info(size_t idx) {
fp16 = fp16 && vk12_features.shaderFloat16; fp16 = fp16 && vk12_features.shaderFloat16;
std::string device_name = props2.properties.deviceName.data(); std::string device_name = props2.properties.deviceName.data();
std::cerr << GGML_VK_NAME << idx << ": " << device_name << " (" << driver_props.driverName << ") | uma: " << uma << " | fp16: " << fp16 << " | warp size: " << subgroup_size << std::endl; GGML_LOG_DEBUG("ggml_vulkan: %d = %s (%s) | uma: %d | fp16: %d | warp size: %d\n",
idx, device_name.c_str(), driver_props.driverName, uma, fp16, subgroup_size);
if (props2.properties.deviceType == vk::PhysicalDeviceType::eCpu) { if (props2.properties.deviceType == vk::PhysicalDeviceType::eCpu) {
std::cerr << "ggml_vulkan: Warning: Device type is CPU. This is probably not the device you want." << std::endl; std::cerr << "ggml_vulkan: Warning: Device type is CPU. This is probably not the device you want." << std::endl;
@ -1821,8 +1822,7 @@ void ggml_vk_instance_init() {
}; };
validation_features.setPNext(nullptr); validation_features.setPNext(nullptr);
instance_create_info.setPNext(&validation_features); instance_create_info.setPNext(&validation_features);
GGML_LOG_DEBUG("ggml_vulkan: Validation layers enabled\n");
std::cerr << "ggml_vulkan: Validation layers enabled" << std::endl;
} }
vk_instance.instance = vk::createInstance(instance_create_info); vk_instance.instance = vk::createInstance(instance_create_info);
@ -1936,8 +1936,8 @@ void ggml_vk_instance_init() {
vk_instance.device_indices.push_back(0); vk_instance.device_indices.push_back(0);
} }
} }
GGML_LOG_DEBUG("ggml_vulkan: Found %d Vulkan devices:\n", vk_instance.device_indices.size());
std::cerr << "ggml_vulkan: Found " << vk_instance.device_indices.size() << " Vulkan devices:" << std::endl;
for (size_t i = 0; i < vk_instance.device_indices.size(); i++) { for (size_t i = 0; i < vk_instance.device_indices.size(); i++) {
ggml_vk_print_gpu_info(i); ggml_vk_print_gpu_info(i);