mirror of https://github.com/PentHertz/srsLTE.git
a44671fc77
clang does not unroll those loops even though it supports the
-funroll-loops command line option, adding various #pragma unroll
options also does not help.
The unroll is needed to make the second argument a constant integer.
Enable the SSE/AVX turbo rate matching when compiling in debug mode.
srsLTE/lib/src/phy/fec/rm_turbo.c:590:33: error: argument to '__builtin_ia32_vec_ext_v16qi' must be a constant integer
int8_t x = (int8_t) _mm_extract_epi8(xVal, j+8);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/lib64/clang/7.0.1/include/smmintrin.h:1048:23: note: expanded from macro '_mm_extract_epi8'
(int)(unsigned char)__builtin_ia32_vec_ext_v16qi((__v16qi)(__m128i)(X), \
^
srsLTE/lib/src/phy/fec/rm_turbo.c:591:35: error: argument to '__builtin_ia32_vec_ext_v8hi' must be a constant integer
uint16_t l = (uint16_t) _mm_extract_epi16(lutVal2, j);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/usr/lib64/clang/7.0.1/include/emmintrin.h:4273:24: note: expanded from macro '_mm_extract_epi16'
(int)(unsigned short)__builtin_ia32_vec_ext_v8hi((__v8hi)(__m128i)(a), \
|
||
|---|---|---|
| .. | ||
| examples | ||
| include | ||
| src | ||
| test | ||
| CMakeLists.txt | ||