aboutsummaryrefslogtreecommitdiffstats
path: root/host/lib/convert/sse2_sc16_to_fc64.cpp
diff options
context:
space:
mode:
authorMartin Braun <martin.braun@ettus.com>2019-07-18 15:36:11 -0700
committerMartin Braun <martin.braun@ettus.com>2019-11-26 11:49:10 -0800
commit9df26a9d89ef8fb50a667428066f3ef1732245c9 (patch)
treeaa8aa5adf1c40e0aecb3e45a527511af96e05ca1 /host/lib/convert/sse2_sc16_to_fc64.cpp
parentfed32af0806a730e0f4202003dc49cb736c832fb (diff)
downloaduhd-9df26a9d89ef8fb50a667428066f3ef1732245c9.tar.gz
uhd-9df26a9d89ef8fb50a667428066f3ef1732245c9.tar.bz2
uhd-9df26a9d89ef8fb50a667428066f3ef1732245c9.zip
convert: sse2: Apply clang-format
Diffstat (limited to 'host/lib/convert/sse2_sc16_to_fc64.cpp')
-rw-r--r--host/lib/convert/sse2_sc16_to_fc64.cpp140
1 files changed, 70 insertions, 70 deletions
diff --git a/host/lib/convert/sse2_sc16_to_fc64.cpp b/host/lib/convert/sse2_sc16_to_fc64.cpp
index 7f22fd07f..45821ac9f 100644
--- a/host/lib/convert/sse2_sc16_to_fc64.cpp
+++ b/host/lib/convert/sse2_sc16_to_fc64.cpp
@@ -11,95 +11,95 @@
using namespace uhd::convert;
-DECLARE_CONVERTER(sc16_item32_le, 1, fc64, 1, PRIORITY_SIMD){
- const item32_t *input = reinterpret_cast<const item32_t *>(inputs[0]);
- fc64_t *output = reinterpret_cast<fc64_t *>(outputs[0]);
+DECLARE_CONVERTER(sc16_item32_le, 1, fc64, 1, PRIORITY_SIMD)
+{
+ const item32_t* input = reinterpret_cast<const item32_t*>(inputs[0]);
+ fc64_t* output = reinterpret_cast<fc64_t*>(outputs[0]);
- const __m128d scalar = _mm_set1_pd(scale_factor/(1 << 16));
- const __m128i zeroi = _mm_setzero_si128();
+ const __m128d scalar = _mm_set1_pd(scale_factor / (1 << 16));
+ const __m128i zeroi = _mm_setzero_si128();
- #define convert_item32_1_to_fc64_1_nswap_guts(_al_) \
- for (; i+3 < nsamps; i+=4){ \
- /* load from input */ \
- __m128i tmpi = _mm_loadu_si128(reinterpret_cast<const __m128i *>(input+i)); \
- \
- /* unpack + swap 16-bit pairs */ \
- tmpi = _mm_shufflelo_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1)); \
- tmpi = _mm_shufflehi_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1)); \
+#define convert_item32_1_to_fc64_1_nswap_guts(_al_) \
+ for (; i + 3 < nsamps; i += 4) { \
+ /* load from input */ \
+ __m128i tmpi = _mm_loadu_si128(reinterpret_cast<const __m128i*>(input + i)); \
+ \
+ /* unpack + swap 16-bit pairs */ \
+ tmpi = _mm_shufflelo_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1)); \
+ tmpi = _mm_shufflehi_epi16(tmpi, _MM_SHUFFLE(2, 3, 0, 1)); \
__m128i tmpilo = _mm_unpacklo_epi16(zeroi, tmpi); /* value in upper 16 bits */ \
- __m128i tmpihi = _mm_unpackhi_epi16(zeroi, tmpi); \
- \
- /* convert and scale */ \
- __m128d tmp0 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
- tmpilo = _mm_unpackhi_epi64(tmpilo, zeroi); \
- __m128d tmp1 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
- __m128d tmp2 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
- tmpihi = _mm_unpackhi_epi64(tmpihi, zeroi); \
- __m128d tmp3 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
- \
- /* store to output */ \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+0), tmp0); \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+1), tmp1); \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+2), tmp2); \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+3), tmp3); \
- } \
+ __m128i tmpihi = _mm_unpackhi_epi16(zeroi, tmpi); \
+ \
+ /* convert and scale */ \
+ __m128d tmp0 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
+ tmpilo = _mm_unpackhi_epi64(tmpilo, zeroi); \
+ __m128d tmp1 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
+ __m128d tmp2 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
+ tmpihi = _mm_unpackhi_epi64(tmpihi, zeroi); \
+ __m128d tmp3 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
+ \
+ /* store to output */ \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 0), tmp0); \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 1), tmp1); \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 2), tmp2); \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 3), tmp3); \
+ }
size_t i = 0;
- //dispatch according to alignment
- if ((size_t(output) & 0xf) == 0){
+ // dispatch according to alignment
+ if ((size_t(output) & 0xf) == 0) {
convert_item32_1_to_fc64_1_nswap_guts(_)
- }
- else{
+ } else {
convert_item32_1_to_fc64_1_nswap_guts(u_)
}
- //convert remainder
- item32_sc16_to_xx<uhd::htowx>(input+i, output+i, nsamps-i, scale_factor);
+ // convert remainder
+ item32_sc16_to_xx<uhd::htowx>(input + i, output + i, nsamps - i, scale_factor);
}
-DECLARE_CONVERTER(sc16_item32_be, 1, fc64, 1, PRIORITY_SIMD){
- const item32_t *input = reinterpret_cast<const item32_t *>(inputs[0]);
- fc64_t *output = reinterpret_cast<fc64_t *>(outputs[0]);
+DECLARE_CONVERTER(sc16_item32_be, 1, fc64, 1, PRIORITY_SIMD)
+{
+ const item32_t* input = reinterpret_cast<const item32_t*>(inputs[0]);
+ fc64_t* output = reinterpret_cast<fc64_t*>(outputs[0]);
- const __m128d scalar = _mm_set1_pd(scale_factor/(1 << 16));
- const __m128i zeroi = _mm_setzero_si128();
+ const __m128d scalar = _mm_set1_pd(scale_factor / (1 << 16));
+ const __m128i zeroi = _mm_setzero_si128();
- #define convert_item32_1_to_fc64_1_bswap_guts(_al_) \
- for (; i+3 < nsamps; i+=4){ \
- /* load from input */ \
- __m128i tmpi = _mm_loadu_si128(reinterpret_cast<const __m128i *>(input+i)); \
- \
- /* byteswap + unpack -> byteswap 16 bit words */ \
- tmpi = _mm_or_si128(_mm_srli_epi16(tmpi, 8), _mm_slli_epi16(tmpi, 8)); \
- __m128i tmpilo = _mm_unpacklo_epi16(zeroi, tmpi); /* value in upper 16 bits */ \
- __m128i tmpihi = _mm_unpackhi_epi16(zeroi, tmpi); \
- \
- /* convert and scale */ \
- __m128d tmp0 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
- tmpilo = _mm_unpackhi_epi64(tmpilo, zeroi); \
- __m128d tmp1 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
- __m128d tmp2 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
- tmpihi = _mm_unpackhi_epi64(tmpihi, zeroi); \
- __m128d tmp3 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
- \
- /* store to output */ \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+0), tmp0); \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+1), tmp1); \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+2), tmp2); \
- _mm_store ## _al_ ## pd(reinterpret_cast<double *>(output+i+3), tmp3); \
- } \
+#define convert_item32_1_to_fc64_1_bswap_guts(_al_) \
+ for (; i + 3 < nsamps; i += 4) { \
+ /* load from input */ \
+ __m128i tmpi = _mm_loadu_si128(reinterpret_cast<const __m128i*>(input + i)); \
+ \
+ /* byteswap + unpack -> byteswap 16 bit words */ \
+ tmpi = _mm_or_si128(_mm_srli_epi16(tmpi, 8), _mm_slli_epi16(tmpi, 8)); \
+ __m128i tmpilo = _mm_unpacklo_epi16(zeroi, tmpi); /* value in upper 16 bits */ \
+ __m128i tmpihi = _mm_unpackhi_epi16(zeroi, tmpi); \
+ \
+ /* convert and scale */ \
+ __m128d tmp0 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
+ tmpilo = _mm_unpackhi_epi64(tmpilo, zeroi); \
+ __m128d tmp1 = _mm_mul_pd(_mm_cvtepi32_pd(tmpilo), scalar); \
+ __m128d tmp2 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
+ tmpihi = _mm_unpackhi_epi64(tmpihi, zeroi); \
+ __m128d tmp3 = _mm_mul_pd(_mm_cvtepi32_pd(tmpihi), scalar); \
+ \
+ /* store to output */ \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 0), tmp0); \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 1), tmp1); \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 2), tmp2); \
+ _mm_store##_al_##pd(reinterpret_cast<double*>(output + i + 3), tmp3); \
+ }
size_t i = 0;
- //dispatch according to alignment
- if ((size_t(output) & 0xf) == 0){
+ // dispatch according to alignment
+ if ((size_t(output) & 0xf) == 0) {
convert_item32_1_to_fc64_1_bswap_guts(_)
- }
- else{
+ } else {
convert_item32_1_to_fc64_1_bswap_guts(u_)
}
- //convert remainder
- item32_sc16_to_xx<uhd::htonx>(input+i, output+i, nsamps-i, scale_factor);
+ // convert remainder
+ item32_sc16_to_xx<uhd::htonx>(input + i, output + i, nsamps - i, scale_factor);
}