Skip to content

Commit

Permalink
wip
Browse files Browse the repository at this point in the history
  • Loading branch information
@howjmay
howjmay committed Mar 4, 2024
1 parent 132c50c commit 7ed5da0
Show file tree
Hide file tree
Showing 2 changed files with 36 additions and 32 deletions.
12 changes: 6 additions & 6 deletions sse2neon.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1774,7 +1774,7 @@ FORCE_INLINE void _sse2neon_set_fpcr(uint64_t value)
#if defined(_MSC_VER)
_WriteStatusReg(ARM64_FPCR, value);
#else
__asm__ __volatile__("msr FPCR, %0" ::"r"(value)); /* write */
__asm__ __volatile__("msr FPCR, %0" ::"r"(value)); /* write */
#endif
}

Expand Down Expand Up @@ -2385,7 +2385,7 @@ FORCE_INLINE void _sse2neon_mm_set_flush_zero_mode(unsigned int flag)
#if defined(__aarch64__) || defined(_M_ARM64)
_sse2neon_set_fpcr(r.value);
#else
__asm__ __volatile__("vmsr FPSCR, %0" ::"r"(r)); /* write */
__asm__ __volatile__("vmsr FPSCR, %0" ::"r"(r)); /* write */
#endif
}

Expand Down Expand Up @@ -2449,7 +2449,7 @@ FORCE_INLINE void _MM_SET_ROUNDING_MODE(int rounding)
#if defined(__aarch64__) || defined(_M_ARM64)
_sse2neon_set_fpcr(r.value);
#else
__asm__ __volatile__("vmsr FPSCR, %0" ::"r"(r)); /* write */
__asm__ __volatile__("vmsr FPSCR, %0" ::"r"(r)); /* write */
#endif
}

Expand Down Expand Up @@ -8644,9 +8644,9 @@ FORCE_INLINE __m128i _mm_aesenc_si128(__m128i a, __m128i RoundKey)
#define SSE2NEON_AES_B2W(b0, b1, b2, b3) \
(((uint32_t) (b3) << 24) | ((uint32_t) (b2) << 16) | \
((uint32_t) (b1) << 8) | (uint32_t) (b0))
// muliplying 'x' by 2 in GF(2^8)
// muliplying 'x' by 2 in GF(2^8)
#define SSE2NEON_AES_F2(x) ((x << 1) ^ (((x >> 7) & 1) * 0x011b /* WPOLY */))
// muliplying 'x' by 3 in GF(2^8)
// muliplying 'x' by 3 in GF(2^8)
#define SSE2NEON_AES_F3(x) (SSE2NEON_AES_F2(x) ^ x)
#define SSE2NEON_AES_U0(p) \
SSE2NEON_AES_B2W(SSE2NEON_AES_F2(p), p, p, SSE2NEON_AES_F3(p))
Expand Down Expand Up @@ -9175,7 +9175,7 @@ FORCE_INLINE void _sse2neon_mm_set_denormals_zero_mode(unsigned int flag)
#if defined(__aarch64__) || defined(_M_ARM64)
_sse2neon_set_fpcr(r.value);
#else
__asm__ __volatile__("vmsr FPSCR, %0" ::"r"(r)); /* write */
__asm__ __volatile__("vmsr FPSCR, %0" ::"r"(r)); /* write */
#endif
}

Expand Down
56 changes: 30 additions & 26 deletions tests/impl.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2816,17 +2816,17 @@ result_t test_mm_shuffle_pi16(const SSE2NEONTestImpl &impl, uint32_t iter)
const int16_t *_a = (const int16_t *) impl.mTestIntPointer1;
__m64 a;
__m64 d;
int16_t _d[4];
#define TEST_IMPL(IDX) \
a = load_m64(_a); \
d = _mm_shuffle_pi16(a, IDX); \
\
int16_t _d[4];
#define TEST_IMPL(IDX) \
a = load_m64(_a); \
d = _mm_shuffle_pi16(a, IDX); \
\
_d[0] = _a[IDX & 0x3]; \
_d[1] = _a[(IDX >> 2) & 0x3]; \
_d[2] = _a[(IDX >> 4) & 0x3]; \
_d[3] = _a[(IDX >> 6) & 0x3]; \
if (VALIDATE_INT16_M64(d, _d) != TEST_SUCCESS) { \
return TEST_FAIL; \
return TEST_FAIL; \
}

IMM_256_ITER
Expand Down Expand Up @@ -5846,7 +5846,7 @@ result_t test_mm_shuffle_epi32(const SSE2NEONTestImpl &impl, uint32_t iter)

#define TEST_IMPL(IDX) \
int32_t d##IDX[4]; \
d##IDX[0] = _a[((IDX) &0x3)]; \
d##IDX[0] = _a[((IDX) & 0x3)]; \
d##IDX[1] = _a[((IDX >> 2) & 0x3)]; \
d##IDX[2] = _a[((IDX >> 4) & 0x3)]; \
d##IDX[3] = _a[((IDX >> 6) & 0x3)]; \
Expand Down Expand Up @@ -5911,31 +5911,35 @@ result_t test_mm_shufflehi_epi16(const SSE2NEONTestImpl &impl, uint32_t iter)

result_t test_mm_shufflelo_epi16(const SSE2NEONTestImpl &impl, uint32_t iter)
{
#if (__GNUC__ == 8 && __GNUC_MINOR__ == 3)
return TEST_UNIMPL;
#else
const int16_t *_a = (const int16_t *) impl.mTestIntPointer1;
__m128i a, c;

#define TEST_IMPL(IDX) \
int16_t d##IDX[8]; \
d##IDX[0] = (int16_t) (((const int64_t *) _a)[0] >> ((IDX & 0x3) * 16)); \
d##IDX[1] = \
(int16_t) (((const int64_t *) _a)[0] >> (((IDX >> 2) & 0x3) * 16)); \
d##IDX[2] = \
(int16_t) (((const int64_t *) _a)[0] >> (((IDX >> 4) & 0x3) * 16)); \
d##IDX[3] = \
(int16_t) (((const int64_t *) _a)[0] >> (((IDX >> 6) & 0x3) * 16)); \
d##IDX[4] = _a[4]; \
d##IDX[5] = _a[5]; \
d##IDX[6] = _a[6]; \
d##IDX[7] = _a[7]; \
\
a = load_m128i(_a); \
c = _mm_shufflelo_epi16(a, IDX); \
\
CHECK_RESULT(VALIDATE_INT16_M128(c, d##IDX))
int16_t _d[8];

#define TEST_IMPL(IDX) \
_d[0] = (int16_t) (((const int64_t *) _a)[0] >> ((IDX & 0x3) * 16)); \
_d[1] = \
(int16_t) (((const int64_t *) _a)[0] >> (((IDX >> 2) & 0x3) * 16)); \
_d[2] = \
(int16_t) (((const int64_t *) _a)[0] >> (((IDX >> 4) & 0x3) * 16)); \
_d[3] = \
(int16_t) (((const int64_t *) _a)[0] >> (((IDX >> 6) & 0x3) * 16)); \
_d[4] = _a[4]; \
_d[5] = _a[5]; \
_d[6] = _a[6]; \
_d[7] = _a[7]; \
\
a = load_m128i(_a); \
c = _mm_shufflelo_epi16(a, IDX); \
\
CHECK_RESULT(VALIDATE_INT16_M128(c, _d))

IMM_256_ITER
#undef TEST_IMPL
return TEST_SUCCESS;
#endif
}

result_t test_mm_sll_epi16(const SSE2NEONTestImpl &impl, uint32_t iter)
Expand Down

0 comments on commit 7ed5da0

Please sign in to comment.