samples/util/util.c

// An example based on iree/samples/simple_embedding.

#include <springbok.h>
#include <stdio.h>

#include "iree/base/api.h"
#include "iree/hal/api.h"
#include "iree/modules/hal/module.h"
#include "iree/vm/api.h"
#include "iree/vm/bytecode_module.h"
#include "samples/util/util.h"

// A function to create the HAL device from the different backend targets.
// The HAL device is returned based on the implementation, and it must be
// released by the caller
extern iree_status_t create_sample_device(iree_hal_device_t **device);

extern const iree_const_byte_span_t load_bytecode_module_data();

extern iree_status_t load_input_data(const MlModel *model, void **buffer);

extern iree_status_t check_output_data(const MlModel *model,
                                       iree_hal_buffer_mapping_t *mapped_memory,
                                       int index_output);

extern const MlModel kModel;

// Prepare the input buffers and buffer_views based on the data type. They must
// be released by the caller.
static iree_status_t prepare_input_hal_buffer_views(
    const MlModel *model, iree_hal_device_t *device, void **arg0_buffer,
    iree_hal_buffer_view_t **arg0_buffer_view) {
  iree_status_t result = iree_ok_status();
  *arg0_buffer = iree_aligned_alloc(
      sizeof(uint32_t), model->input_size_bytes * model->input_length);
  if (*arg0_buffer == NULL) {
    result = iree_make_status(IREE_STATUS_RESOURCE_EXHAUSTED);
  }

  // Populate initial value
  load_input_data(model, arg0_buffer);

  // Wrap buffers in shaped buffer views.
  // The buffers can be mapped on the CPU and that can also be used
  // on the device. Not all devices support this, but the ones we have now do.

  iree_hal_memory_type_t input_memory_type =
      IREE_HAL_MEMORY_TYPE_HOST_LOCAL | IREE_HAL_MEMORY_TYPE_DEVICE_VISIBLE;

  result = iree_hal_buffer_view_wrap_or_clone_heap_buffer(
      iree_hal_device_allocator(device), model->input_shape,
      model->num_input_dim, model->hal_element_type, input_memory_type,
      IREE_HAL_ENCODING_TYPE_DENSE_ROW_MAJOR, IREE_HAL_MEMORY_ACCESS_READ,
      IREE_HAL_BUFFER_USAGE_ALL,
      iree_make_byte_span(*arg0_buffer,
                          model->input_size_bytes * model->input_length),
      iree_allocator_null(), arg0_buffer_view);
  return result;
}

iree_status_t run(const MlModel *model) {
  IREE_RETURN_IF_ERROR(iree_hal_module_register_types());

  iree_vm_instance_t *instance = NULL;
  iree_status_t result =
      iree_vm_instance_create(iree_allocator_system(), &instance);

  iree_hal_device_t *device = NULL;
  if (iree_status_is_ok(result)) {
    result = create_sample_device(&device);
  }
  iree_vm_module_t *hal_module = NULL;
  if (iree_status_is_ok(result)) {
    result =
        iree_hal_module_create(device, iree_allocator_system(), &hal_module);
  }
  // Load bytecode module from the embedded data.
  const iree_const_byte_span_t module_data = load_bytecode_module_data();

  iree_vm_module_t *bytecode_module = NULL;
  if (iree_status_is_ok(result)) {
    result = iree_vm_bytecode_module_create(module_data, iree_allocator_null(),
                                            iree_allocator_system(),
                                            &bytecode_module);
  }

  // Allocate a context that will hold the module state across invocations.
  iree_vm_context_t *context = NULL;
  iree_vm_module_t *modules[] = {hal_module, bytecode_module};
  if (iree_status_is_ok(result)) {
    result = iree_vm_context_create_with_modules(
        instance, &modules[0], IREE_ARRAYSIZE(modules), iree_allocator_system(),
        &context);
  }
  iree_vm_module_release(hal_module);
  iree_vm_module_release(bytecode_module);

  // Lookup the entry point function.
  // Note that we use the synchronous variant which operates on pure type/shape
  // erased buffers.
  iree_vm_function_t main_function;
  if (iree_status_is_ok(result)) {
    result = (iree_vm_context_resolve_function(
        context, iree_make_cstring_view(model->entry_func), &main_function));
  }

  // Prepare the input buffers.
  void *arg0_buffer = NULL;
  iree_hal_buffer_view_t *arg0_buffer_view = NULL;
  if (iree_status_is_ok(result)) {
    result = prepare_input_hal_buffer_views(model, device, &arg0_buffer,
                                            &arg0_buffer_view);
  }

  // Setup call inputs with our buffers.
  iree_vm_list_t *inputs = NULL;
  if (iree_status_is_ok(result)) {
    result = iree_vm_list_create(
        /*element_type=*/NULL,
        /*capacity=*/1, iree_allocator_system(), &inputs);
  }
  iree_vm_ref_t arg0_buffer_view_ref =
      iree_hal_buffer_view_move_ref(arg0_buffer_view);
  if (iree_status_is_ok(result)) {
    result = iree_vm_list_push_ref_move(inputs, &arg0_buffer_view_ref);
  }

  // Prepare outputs list to accept the results from the invocation.
  // The output vm list is allocated statically.
  iree_vm_list_t *outputs = NULL;
  if (iree_status_is_ok(result)) {
    result = iree_vm_list_create(
        /*element_type=*/NULL,
        /*capacity=*/1, iree_allocator_system(), &outputs);
  }

  // Invoke the function.
  if (iree_status_is_ok(result)) {
    result = iree_vm_invoke(context, main_function,
                            /*policy=*/NULL, inputs, outputs,
                            iree_allocator_system());
  }

  for (int index_output = 0; index_output < model->num_output; index_output++) {
    iree_hal_buffer_view_t *ret_buffer_view = NULL;
    if (iree_status_is_ok(result)) {
      // Get the result buffers from the invocation.
      ret_buffer_view = (iree_hal_buffer_view_t *)iree_vm_list_get_ref_deref(
          outputs, index_output, iree_hal_buffer_view_get_descriptor());
      if (ret_buffer_view == NULL) {
        result = iree_make_status(IREE_STATUS_NOT_FOUND,
                                  "can't find return buffer view");
      }
    }
    // Read back the results and ensure we got the right values.
    iree_hal_buffer_mapping_t mapped_memory;
    if (iree_status_is_ok(result)) {
      result = iree_hal_buffer_map_range(
          iree_hal_buffer_view_buffer(ret_buffer_view),
          IREE_HAL_MEMORY_ACCESS_READ, 0, IREE_WHOLE_BUFFER, &mapped_memory);
    }
    if (iree_status_is_ok(result)) {
      result = check_output_data(model, &mapped_memory, index_output);
      iree_hal_buffer_unmap_range(&mapped_memory);
    }
  }

  iree_vm_list_release(inputs);
  iree_vm_list_release(outputs);
  iree_aligned_free(arg0_buffer);
  iree_vm_context_release(context);
  IREE_IGNORE_ERROR(iree_hal_allocator_statistics_fprint(
        stdout, iree_hal_device_allocator(device)));
  iree_hal_device_release(device);
  iree_vm_instance_release(instance);
  return result;
}

int main() {
  const MlModel *model_ptr = &kModel;
  const iree_status_t result = run(model_ptr);
  int ret = (int)iree_status_code(result);
  if (!iree_status_is_ok(result)) {
    iree_status_fprint(stderr, result);
    iree_status_free(result);
  } else {
    LOG_INFO("%s finished successfully", model_ptr->model_name);
  }

  return ret;
}