#include <stdlib.h>

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

// Test for vmerge.vim and vmerge.vvm instructions.
namespace vmerge_test {
namespace {

using namespace test_v_helpers;

uint8_t src_vector_1[MAXVL_BYTES];
uint8_t src_vector_2[MAXVL_BYTES];
uint8_t src_mask_vector[MAXVL_BYTES];
uint8_t dest_vector[MAXVL_BYTES];

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

TEST_F(VmergeTest, vmerge_vim) {
  for (int i = 0; i < AVL_COUNT; i++) {
    int32_t avl = AVLS[i];
    int vlmax;
    int vl;
    /* For non narrowing instructions all vectors have same type*/
    std::tie(vlmax, vl) = vector_test_setup<uint8_t>(
        VLMUL::LMUL_M1, avl, {dest_vector, src_vector_2, src_mask_vector});
    if (avl > vlmax) {
      continue;
    }

    fill_random_vector<uint8_t>(src_vector_2, vl);
    fill_random_vector<uint8_t>(src_mask_vector, vl);
    const int8_t test_val = 12;

    // Load vector registers
    __asm__ volatile("vle8.v v16, (%0)" : : "r"(src_vector_2));
    __asm__ volatile("vle8.v v0, (%0)" : : "r"(src_mask_vector));

    // Run target instruction
    __asm__ volatile("vmerge.vim v24, v16, %[IMM], v0" ::[IMM] "n"(test_val));

    // Store result vector register
    __asm__ volatile("vse8.v v24, (%0)" : : "r"(dest_vector));

    // Check vector elements
    constexpr uint32_t kShift = 3;  // shift for uint8_t
    for (int idx = 0; idx < vl; idx++) {
      uint32_t mask_idx = idx >> kShift;
      uint32_t mask_pos = idx & ~(mask_idx << kShift);
      if (src_mask_vector[mask_idx] & (1 << mask_pos)) {
        ASSERT_EQ(dest_vector[idx], test_val);
      } else {
        ASSERT_EQ(dest_vector[idx], src_vector_2[idx]);
      }
    }
  }
}

TEST_F(VmergeTest, vmerge_vvm) {
  for (int i = 0; i < AVL_COUNT; i++) {
    int32_t avl = AVLS[i];
    int vlmax;
    int vl;
    /* For non narrowing instructions all vectors have same type*/
    std::tie(vlmax, vl) = vector_test_setup<uint8_t>(
        VLMUL::LMUL_M1, avl,
        {dest_vector, src_vector_1, src_vector_2, src_mask_vector});
    if (avl > vlmax) {
      continue;
    }

    fill_random_vector<uint8_t>(src_vector_1, vl);
    fill_random_vector<uint8_t>(src_vector_2, vl);
    fill_random_vector<uint8_t>(src_mask_vector, vl);

    // Load vector registers
    __asm__ volatile("vle8.v v8, (%0)" : : "r"(src_vector_1));
    __asm__ volatile("vle8.v v16, (%0)" : : "r"(src_vector_2));
    __asm__ volatile("vle8.v v0, (%0)" : : "r"(src_mask_vector));

    // Run target instruction
    __asm__ volatile("vmerge.vvm v24, v16, v8, v0");

    // Store result vector register
    __asm__ volatile("vse8.v v24, (%0)" : : "r"(dest_vector));

    // Check vector elements
    constexpr uint32_t kShift = 3;  // shift for uint8_t
    for (int idx = 0; idx < vl; idx++) {
      uint32_t mask_idx = idx >> kShift;
      uint32_t mask_pos = idx & ~(mask_idx << kShift);
      if (src_mask_vector[mask_idx] & (1 << mask_pos)) {
        ASSERT_EQ(dest_vector[idx], src_vector_1[idx]);
      } else {
        ASSERT_EQ(dest_vector[idx], src_vector_2[idx]);
      }
    }
  }
}

}  // namespace
}  // namespace vmerge_test