#include <limits.h>
#include <riscv_vector.h>
#include <springbok.h>
#include <stdio.h>
#include <stdlib.h>

#include <bit>
#include <tuple>
#include <type_traits>

#include "pw_unit_test/framework.h"
#include "test_v_helpers.h"

namespace vmv_s_x_test {
namespace {

using namespace test_v_helpers;

uint8_t test_vector_1[MAXVL_BYTES];
uint8_t reference_vector_1[MAXVL_BYTES];

class VmvsxTest : public ::testing::Test {
 protected:
  void SetUp() override { zero_vector_registers(); }
  void TearDown() override { zero_vector_registers(); }
};

template <typename T>
void vmv_s_x_test(const int32_t test_value) {
  for (int lmul = LMUL_M1; lmul <= LMUL_M8; lmul++) {
    // TODO(gkielian): increase upper bound to LMUL_MF2 after support is added
    // TODO(gkielian): skip lmul==4 (unused) on adding fractional lmul to range

    for (int32_t i = 0; i < AVL_COUNT; i++) {
      int32_t avl = AVLS[i];
      int32_t vlmax;
      int32_t vl;

      // reset vector elements to zero and perform appropriate vsetvl
      std::tie(vlmax, vl) = vector_test_setup<T>(
          static_cast<VLMUL>(lmul), avl, {test_vector_1, reference_vector_1});
      if (avl > vlmax) {
        continue;
      }

      // recast test and ref vector to appropriate sew
      T *test_ptr_vec = reinterpret_cast<T *>(test_vector_1);
      T *ref_ptr_vec = reinterpret_cast<T *>(reference_vector_1);

      // populate test and ref vectors
      fill_random_vector<T>(ref_ptr_vec, avl);

      // load test_value to scalar register
      __asm__ volatile("lw x5, (%0)" : : "r"(&test_value));

      switch (sizeof(T)) {
        case sizeof(int8_t):
          __asm__ volatile("vle8.v v8, (%0)" : : "r"(ref_ptr_vec));
          break;
        case sizeof(int16_t):
          __asm__ volatile("vle16.v v8, (%0)" : : "r"(ref_ptr_vec));
          break;
        case sizeof(int32_t):
          __asm__ volatile("vle32.v v8, (%0)" : : "r"(ref_ptr_vec));
          break;
        default:
          EXPECT_TRUE(false);
          break;
      }

      // load value of x5 into v8[0], truncating x5 to fit into the current SEW
      __asm__ volatile("vmv.s.x v8, x5");

      switch (sizeof(T)) {
        case sizeof(int8_t):
          __asm__ volatile("vse8.v v8, (%0)" : : "r"(test_ptr_vec));
          break;
        case sizeof(int16_t):
          __asm__ volatile("vse16.v v8, (%0)" : : "r"(test_ptr_vec));
          break;
        case sizeof(int32_t):
          __asm__ volatile("vse32.v v8, (%0)" : : "r"(test_ptr_vec));
          break;
        default:
          EXPECT_TRUE(false);
          break;
      }

      // perform operations on reference vector
      ref_ptr_vec[0] = static_cast<T>(test_value);

      assert_vec_elem_eq<T>(vlmax, test_ptr_vec, ref_ptr_vec);
    }
  }
}

TEST_F(VmvsxTest, vmv_s_x_test_e8) {
  vmv_s_x_test<int8_t>(INT32_MIN);
  vmv_s_x_test<int8_t>(INT32_MAX);
  vmv_s_x_test<int8_t>(INT8_MIN);
  vmv_s_x_test<int8_t>(INT8_MAX);
  vmv_s_x_test<int8_t>(0);
}

TEST_F(VmvsxTest, vmv_s_x_test_e16) {
  vmv_s_x_test<int16_t>(INT32_MIN);
  vmv_s_x_test<int16_t>(INT32_MAX);
  vmv_s_x_test<int16_t>(INT8_MIN);
  vmv_s_x_test<int16_t>(INT8_MAX);
  vmv_s_x_test<int16_t>(0);
}

TEST_F(VmvsxTest, vmv_s_x_test_e32) {
  vmv_s_x_test<int32_t>(INT32_MIN);
  vmv_s_x_test<int32_t>(INT32_MAX);
  vmv_s_x_test<int32_t>(INT8_MIN);
  vmv_s_x_test<int32_t>(INT8_MAX);
  vmv_s_x_test<int32_t>(0);
}

}  // namespace
}  // namespace vmv_s_x_test