Merge pull request #482 from howjmay/vmla_lane

test: Add tests for vmla_lane_*
howjmay · Jul 30, 2024 · b82fe37 · b82fe37
2 parents d1f1aed + e47c422
commit b82fe37
Showing 1 changed file with 107 additions and 7 deletions.
diff --git a/tests/impl.cpp b/tests/impl.cpp
@@ -32341,9 +32341,9 @@ result_t test_vmulq_lane_u32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 
 result_t test_vmla_lane_s16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 #ifdef ENABLE_TEST_ALL
-  const int16_t *_a = (int16_t *)impl.test_cases_int_pointer1;
-  const int16_t *_b = (int16_t *)impl.test_cases_int_pointer2;
-  const int16_t *_c = (int16_t *)impl.test_cases_int_pointer3;
+  const int16_t *_a = (const int16_t *)impl.test_cases_int_pointer1;
+  const int16_t *_b = (const int16_t *)impl.test_cases_int_pointer2;
+  const int16_t *_c = (const int16_t *)impl.test_cases_int_pointer3;
   int16_t _d[4];
   int16x4_t a = vld1_s16(_a);
   int16x4_t b = vld1_s16(_b);
@@ -32366,13 +32366,113 @@ result_t test_vmla_lane_s16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 #endif  // ENABLE_TEST_ALL
 }
 
-result_t test_vmla_lane_s32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+result_t test_vmla_lane_s32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const int32_t *_a = (const int32_t *)impl.test_cases_int_pointer1;
+  const int32_t *_b = (const int32_t *)impl.test_cases_int_pointer2;
+  const int32_t *_c = (const int32_t *)impl.test_cases_int_pointer3;
+  int32_t _d[4];
+  int32x2_t a = vld1_s32(_a);
+  int32x2_t b = vld1_s32(_b);
+  int32x2_t c = vld1_s32(_c);
+  int32x2_t d;
+
+#define TEST_IMPL(IDX)               \
+  for (int i = 0; i < 2; i++) {      \
+    _d[i] = _a[i] + _b[i] * _c[IDX]; \
+  }                                  \
+  d = vmla_lane_s32(a, b, c, IDX);   \
+  CHECK_RESULT(validate_int32(d, _d[0], _d[1]))
+
+  IMM_2_ITER
+#undef TEST_IMPL
+
+  return TEST_SUCCESS;
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
+
+result_t test_vmla_lane_f32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const float *_a = (const float *)impl.test_cases_float_pointer1;
+  const float *_b = (const float *)impl.test_cases_float_pointer2;
+  const float *_c = (const float *)impl.test_cases_float_pointer3;
+  float _d[4];
+  float32x2_t a = vld1_f32(_a);
+  float32x2_t b = vld1_f32(_b);
+  float32x2_t c = vld1_f32(_c);
+  float32x2_t d;
+
+#define TEST_IMPL(IDX)               \
+  for (int i = 0; i < 2; i++) {      \
+    _d[i] = _a[i] + _b[i] * _c[IDX]; \
+  }                                  \
+  d = vmla_lane_f32(a, b, c, IDX);   \
+  CHECK_RESULT(validate_float_error(d, _d[0], _d[1], 0.0001f))
+
+  IMM_2_ITER
+#undef TEST_IMPL
+
+  return TEST_SUCCESS;
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
+
+result_t test_vmla_lane_u16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const uint16_t *_a = (const uint16_t *)impl.test_cases_int_pointer1;
+  const uint16_t *_b = (const uint16_t *)impl.test_cases_int_pointer2;
+  const uint16_t *_c = (const uint16_t *)impl.test_cases_int_pointer3;
+  uint16_t _d[4];
+  uint16x4_t a = vld1_u16(_a);
+  uint16x4_t b = vld1_u16(_b);
+  uint16x4_t c = vld1_u16(_c);
+  uint16x4_t d;
+
+#define TEST_IMPL(IDX)               \
+  for (int i = 0; i < 4; i++) {      \
+    _d[i] = _a[i] + _b[i] * _c[IDX]; \
+  }                                  \
+  d = vmla_lane_u16(a, b, c, IDX);   \
+  CHECK_RESULT(validate_uint16(d, _d[0], _d[1], _d[2], _d[3]))
+
+  IMM_4_ITER
+#undef TEST_IMPL
+
+  return TEST_SUCCESS;
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
+
+result_t test_vmla_lane_u32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const uint32_t *_a = (const uint32_t *)impl.test_cases_int_pointer1;
+  const uint32_t *_b = (const uint32_t *)impl.test_cases_int_pointer2;
+  const uint32_t *_c = (const uint32_t *)impl.test_cases_int_pointer3;
+  uint32_t _d[4];
+  uint32x2_t a = vld1_u32(_a);
+  uint32x2_t b = vld1_u32(_b);
+  uint32x2_t c = vld1_u32(_c);
+  uint32x2_t d;
 
-result_t test_vmla_lane_f32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+#define TEST_IMPL(IDX)               \
+  for (int i = 0; i < 2; i++) {      \
+    _d[i] = _a[i] + _b[i] * _c[IDX]; \
+  }                                  \
+  d = vmla_lane_u32(a, b, c, IDX);   \
+  CHECK_RESULT(validate_uint32(d, _d[0], _d[1]))
 
-result_t test_vmla_lane_u16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+  IMM_2_ITER
+#undef TEST_IMPL
 
-result_t test_vmla_lane_u32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+  return TEST_SUCCESS;
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
 
 result_t test_vmlaq_lane_s16(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 #ifdef ENABLE_TEST_ALL