static inline __m256i mulhi_epu64(__m256i x, __m256i y) {
__m256i x_hi = _mm256_srli_epi64(x, 32);
__m256i y_hi = _mm256_srli_epi64(y, 32);
// __m256i mask = _mm256_set1_epi64x(0xFFFFFFFFL);
// __m256i x_lo = _mm256_and_si256(x, mask);
// __m256i y_lo = _mm256_and_si256(y, mask);
/// masking is unnecessary because _mm256_mul_epu32 does it for us (for free):
__m256i x_lo = x;
__m256i y_lo = y;
/////////////
__m256i result = _mm256_mul_epu32(x_lo,y_lo);
result = _mm256_srli_epi64(result, 32);
__m256i result1 = _mm256_mul_epu32(x_hi,y_lo);
__m256i result2 = _mm256_mul_epu32(x_lo,y_hi);
result = _mm256_add_epi64(result, result1);
result = _mm256_add_epi64(result, result2);
result = _mm256_srli_epi64(result, 32);
__m256i result3 = _mm256_mul_epu32(x_hi,y_hi);
result = _mm256_add_epi64(result, result3);
return result;
}
// assume N is divisible by 4
uint32_t vectorsum(uint32_t * z, uint32_t N, uint32_t * accesses, uint32_t nmbr) {
__m256i Nvec = _mm256_set1_epi32(N);
__m128i sum = _mm_setzero_si128();
for(uint32_t j = 0; j < nmbr ; j+=4) {
__m256i fourints = _mm256_loadu_si256((const __m256i *)(accesses + j));
__m256i four64bitsproducts = _mm256_mul_epu32(fourints, Nvec);
__m256i fourtop32ints = _mm256_srli_epi64(four64bitsproducts,32);
__m128i four32ints = _mm256_i64gather_epi32 (z,fourtop32ints , 4);
sum = _mm_add_epi32(sum, four32ints);
}
uint32_t buffer[4];
_mm_storeu_si128((__m128i *)buffer,sum);
return buffer[0] + buffer[1] + buffer[2] + buffer[3];
}
__m256i test_mm256_mul_epu32(__m256i a, __m256i b) {
// CHECK: @llvm.x86.avx2.pmulu.dq
return _mm256_mul_epu32(a, b);
}
__m256i test_mm256_mul_epu32(__m256i a, __m256i b) {
// CHECK-LABEL: test_mm256_mul_epu32
// CHECK: call <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32> %{{.*}}, <8 x i32> %{{.*}})
return _mm256_mul_epu32(a, b);
}
static FORCE_INLINE void FlowInter_8px_AVX2(
int w, PixelType *pdst,
const PixelType *prefB, const PixelType *prefF,
const int16_t *VXFullB, const int16_t *VXFullF,
const int16_t *VYFullB, const int16_t *VYFullF,
const uint8_t *MaskB, const uint8_t *MaskF,
int nPelLog,
const __m256i &dwords_time256, const __m256i &dwords_256_time256,
const __m256i &dwords_ref_pitch, const __m256i &dwords_hoffsets) {
__m256i dwords_w = _mm256_add_epi32(_mm256_set1_epi32(w << nPelLog), dwords_hoffsets);
__m256i dstF = lookup_AVX2(VXFullF, VYFullF, prefF, w, dwords_time256, dwords_ref_pitch, dwords_w);
__m256i dstB = lookup_AVX2(VXFullB, VYFullB, prefB, w, dwords_256_time256, dwords_ref_pitch, dwords_w);
__m256i dstF0 = _mm256_i32gather_epi32((const int *)prefF, dwords_w, sizeof(PixelType));
__m256i dstB0 = _mm256_i32gather_epi32((const int *)prefB, dwords_w, sizeof(PixelType));
dstF0 = _mm256_and_si256(dstF0, _mm256_set1_epi32((1 << (sizeof(PixelType) * 8)) - 1));
dstB0 = _mm256_and_si256(dstB0, _mm256_set1_epi32((1 << (sizeof(PixelType) * 8)) - 1));
__m256i maskf = _mm256_cvtepu8_epi32(_mm_loadl_epi64((const __m128i *)&MaskF[w]));
__m256i maskb = _mm256_cvtepu8_epi32(_mm_loadl_epi64((const __m128i *)&MaskB[w]));
const __m256i dwords_255 = _mm256_set1_epi32(255);
__m256i maskf_inv = _mm256_sub_epi32(dwords_255, maskf);
__m256i maskb_inv = _mm256_sub_epi32(dwords_255, maskb);
__m256i dstF_maskf_inv, dstB_maskb_inv, dstF0_maskb, dstB0_maskf;
if (sizeof(PixelType) == 1) {
dstF_maskf_inv = _mm256_mullo_epi16(dstF, maskf_inv);
dstB_maskb_inv = _mm256_mullo_epi16(dstB, maskb_inv);
dstF0_maskb = _mm256_mullo_epi16(dstF0, maskb);
dstB0_maskf = _mm256_mullo_epi16(dstB0, maskf);
} else {
dstF_maskf_inv = _mm256_mullo_epi32(dstF, maskf_inv);
dstB_maskb_inv = _mm256_mullo_epi32(dstB, maskb_inv);
dstF0_maskb = _mm256_mullo_epi32(dstF0, maskb);
dstB0_maskf = _mm256_mullo_epi32(dstB0, maskf);
}
__m256i f = _mm256_add_epi32(dstF0_maskb, dstB_maskb_inv);
__m256i b = _mm256_add_epi32(dstB0_maskf, dstF_maskf_inv);
if (sizeof(PixelType) == 1) {
f = _mm256_mullo_epi32(f, maskf);
b = _mm256_mullo_epi32(b, maskb);
f = _mm256_add_epi32(f, dwords_255);
b = _mm256_add_epi32(b, dwords_255);
f = _mm256_srai_epi32(f, 8);
b = _mm256_srai_epi32(b, 8);
} else {
const __m256i qwords_255 = _mm256_set1_epi64x(255);
__m256i tempf = _mm256_mul_epu32(f, maskf);
__m256i tempb = _mm256_mul_epu32(b, maskb);
tempf = _mm256_add_epi64(tempf, qwords_255);
tempb = _mm256_add_epi64(tempb, qwords_255);
tempf = _mm256_srli_epi64(tempf, 8);
tempb = _mm256_srli_epi64(tempb, 8);
f = _mm256_srli_epi64(f, 32);
b = _mm256_srli_epi64(b, 32);
f = _mm256_mul_epu32(f, _mm256_srli_epi64(maskf, 32));
b = _mm256_mul_epu32(b, _mm256_srli_epi64(maskb, 32));
f = _mm256_add_epi64(f, qwords_255);
b = _mm256_add_epi64(b, qwords_255);
f = _mm256_srli_epi64(f, 8);
b = _mm256_srli_epi64(b, 8);
f = _mm256_or_si256(tempf, _mm256_slli_epi64(f, 32));
b = _mm256_or_si256(tempb, _mm256_slli_epi64(b, 32));
}
f = _mm256_add_epi32(f, dstF_maskf_inv);
b = _mm256_add_epi32(b, dstB_maskb_inv);
f = _mm256_add_epi32(f, dwords_255);
b = _mm256_add_epi32(b, dwords_255);
f = _mm256_srai_epi32(f, 8);
b = _mm256_srai_epi32(b, 8);
if (sizeof(PixelType) == 1) {
f = _mm256_madd_epi16(f, dwords_256_time256);
b = _mm256_madd_epi16(b, dwords_time256);
} else {
f = _mm256_mullo_epi32(f, dwords_256_time256);
b = _mm256_mullo_epi32(b, dwords_time256);
}
__m256i dst = _mm256_add_epi32(f, b);
dst = _mm256_srai_epi32(dst, 8);
dst = _mm256_packus_epi32(dst, dst);
dst = _mm256_permute4x64_epi64(dst, 0xe8); // 0b11101000 - copy third qword to second qword
__m128i dst128 = _mm256_castsi256_si128(dst);
if (sizeof(PixelType) == 1) {
dst128 = _mm_packus_epi16(dst128, dst128);
_mm_storel_epi64((__m128i *)&pdst[w], dst128);
} else {
_mm_storeu_si128((__m128i *)&pdst[w], dst128);
}
}
