Fix test_mm_dp_pd failure when compiling with -O2 (#643)

The test_mm_dp_pd failed when running "make check" with -O2 passed to
ARCH_CFLAGS. Debug investigation revealed that a buffer initialized with
floats was being  cast to a buffer of doubles. The test failure occurred
due to differences  between the dot product obtained with intrinsics and
the one calculated with C++ doubles. This discrepancy arose because the
numbers were too large, causing rounding errors in the dot product
calculation.

Co-authored-by: Alexander Orlov <aorlov@aligntech.com>

tests/impl.cpp

@@ -8384,22 +8384,24 @@ result_t test_mm_cvtepu8_epi64(const SSE2NEONTestImpl &impl, uint32_t iter)
  return validateInt64(ret, i0, i1);
 }
 
-#define MM_DP_PD_TEST_CASE_WITH(imm8) \
- do { \
- const double *_a = (const double *) impl.mTestFloatPointer1; \
- const double *_b = (const double *) impl.mTestFloatPointer2; \
- const int imm = imm8; \
- double d[2]; \
- double sum = 0; \
- for (size_t i = 0; i < 2; i++) \
- sum += ((imm) & (1 << (i + 4))) ? _a[i] * _b[i] : 0; \
- for (size_t i = 0; i < 2; i++) \
- d[i] = (imm & (1 << i)) ? sum : 0; \
- __m128d a = load_m128d(_a); \
- __m128d b = load_m128d(_b); \
- __m128d ret = _mm_dp_pd(a, b, imm); \
- if (validateDouble(ret, d[0], d[1]) != TEST_SUCCESS) \
- return TEST_FAIL; \
+#define MM_DP_PD_TEST_CASE_WITH(imm8) \
+ do { \
+ const double _a[] = {impl.mTestFloatPointer1[0], \
+ impl.mTestFloatPointer1[1]}; \
+ const double _b[] = {impl.mTestFloatPointer2[0], \
+ impl.mTestFloatPointer2[1]}; \
+ const int imm = imm8; \
+ double d[2] = {0}; \
+ double sum = 0; \
+ for (size_t i = 0; i < 2; i++) \
+ sum += ((imm) & (1 << (i + 4))) ? _a[i] * _b[i] : 0; \
+ for (size_t i = 0; i < 2; i++) \
+ d[i] = (imm & (1 << i)) ? sum : 0; \
+ __m128d a = load_m128d(_a); \
+ __m128d b = load_m128d(_b); \
+ __m128d ret = _mm_dp_pd(a, b, imm); \
+ if (validateDouble(ret, d[0], d[1]) != TEST_SUCCESS) \
+ return TEST_FAIL; \
  } while (0)
 
 #define GENERATE_MM_DP_PD_TEST_CASES \