/*
 * Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
 *
 * SPDX-License-Identifier: MIT
 */

#pragma clang diagnostic ignored "-Wunused-parameter"
#include <hip_test_common.hh>

#define N 512

#if defined(__HIP_PLATFORM_AMD__) || CUDA_VERSION < CUDA_12000

texture<float, 1, hipReadModeElementType> tex_ref;

static __global__ void kernel(float* out) {
#if !__HIP_NO_IMAGE_SUPPORT
  int x = blockIdx.x * blockDim.x + threadIdx.x;
  if (x < N) {
    out[x] = tex1Dfetch(tex_ref, x);
  }
#endif
}

HIP_TEST_CASE(Unit_hipBindTexture_Positive) {
  CHECK_IMAGE_SUPPORT
  size_t offset = 0;
  float* tex_buf;

  hipChannelFormatDesc channel_desc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);

  HIP_CHECK(hipMalloc(&tex_buf, N * sizeof(float)));

  SECTION("With Channel Descriptor") {
    HIP_CHECK(hipBindTexture(&offset, tex_ref, reinterpret_cast<void*>(tex_buf), channel_desc,
                             N * sizeof(float)));
  }

  SECTION("Without Channel Descriptor") {
    HIP_CHECK(
        hipBindTexture(&offset, tex_ref, reinterpret_cast<void*>(tex_buf), N * sizeof(float)));
  }

  HIP_CHECK(hipUnbindTexture(&tex_ref));
  HIP_CHECK(hipFree(tex_buf));
}

HIP_TEST_CASE(Unit_hipBindTexture_1DfetchVerification) {
  CHECK_IMAGE_SUPPORT

  float* tex_buf;
  float val[N], output[N];
  size_t offset = 0;
  float* dev_buf;
  for (int i = 0; i < N; i++) {
    val[i] = i;
    output[i] = 0.0;
  }
  hipChannelFormatDesc channel_desc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);

  HIP_CHECK(hipMalloc(&tex_buf, N * sizeof(float)));
  HIP_CHECK(hipMalloc(&dev_buf, N * sizeof(float)));
  HIP_CHECK(hipMemcpy(tex_buf, val, N * sizeof(float), hipMemcpyHostToDevice));

  tex_ref.addressMode[0] = hipAddressModeClamp;
  tex_ref.addressMode[1] = hipAddressModeClamp;
  tex_ref.filterMode = hipFilterModePoint;
  tex_ref.normalized = 0;

  HIP_CHECK(hipBindTexture(&offset, tex_ref, reinterpret_cast<void*>(tex_buf), channel_desc,
                           N * sizeof(float)));
  HIP_CHECK(hipGetTextureAlignmentOffset(&offset, &tex_ref));

  dim3 dimBlock(64, 1, 1);
  dim3 dimGrid(N / dimBlock.x, 1, 1);

  hipLaunchKernelGGL(kernel, dim3(dimGrid), dim3(dimBlock), 0, 0, dev_buf);
  HIP_CHECK(hipGetLastError());
  HIP_CHECK(hipDeviceSynchronize());
  HIP_CHECK(hipMemcpy(output, dev_buf, N * sizeof(float), hipMemcpyDeviceToHost));
  for (int i = 0; i < N; i++) {
    if (output[i] != val[i]) {
      INFO("Mismatch at index : " << i << ", output[i] " << output[i] << ", val[i] " << val[i]);
      REQUIRE(false);
    }
  }

  HIP_CHECK(hipUnbindTexture(&tex_ref));
  HIP_CHECK(hipFree(tex_buf));
  HIP_CHECK(hipFree(dev_buf));
}

HIP_TEST_CASE(Unit_hipBindTexture_Negative) {
  CHECK_IMAGE_SUPPORT
  size_t offset = 0;
  float* tex_buf;

  hipChannelFormatDesc channel_desc = hipCreateChannelDesc(32, 0, 0, 0, hipChannelFormatKindFloat);

  HIP_CHECK(hipMalloc(&tex_buf, N * sizeof(float)));

  SECTION("Invalid texture reference") {
#if HT_AMD
    HIP_CHECK_ERROR(hipBindTexture(&offset, nullptr, reinterpret_cast<void*>(tex_buf),
                                   &channel_desc, N * sizeof(float)),
                    hipErrorInvalidSymbol);
#else
    HIP_CHECK_ERROR(hipBindTexture(&offset, nullptr, reinterpret_cast<void*>(tex_buf),
                                   &channel_desc, N * sizeof(float)),
                    hipErrorInvalidTexture);
#endif
  }

  SECTION("Device memory is nullptr") {
    HIP_CHECK_ERROR(hipBindTexture(&offset, tex_ref, nullptr, channel_desc, N * sizeof(float)),
                    hipErrorNotFound);
  }

  SECTION("Invalid hipChannelFormatDesc") {
    hipChannelFormatDesc invalid_channel_desc{-1, -1, -1, -1, hipChannelFormatKindSigned};
    HIP_CHECK_ERROR(hipBindTexture(&offset, tex_ref, reinterpret_cast<void*>(tex_buf),
                                   invalid_channel_desc, N * sizeof(float)),
                    hipErrorInvalidChannelDescriptor);
  }

  if (tex_buf) {
    HIP_CHECK(hipFree(tex_buf));
  }
}

#endif