Skip to content

Commit

Permalink
Merge branch 'c98' into next
Browse files Browse the repository at this point in the history 
Now VOLK_GNSSSDR is built with the default c98 standard instead of c11.
This allows the usage of more compilers
  • Loading branch information
@carlesfernandez
carlesfernandez committed May 31, 2016
2 parents 0aa980a + 2f339d2 commit d490191
Show file tree
Hide file tree
Showing 32 changed files with 844 additions and 674 deletions.
4 changes: 2 additions & 2 deletions src/algorithms/libs/volk_gnsssdr_module/volk_gnsssdr/CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,8 +27,8 @@ enable_language(CXX)
enable_language(C)
enable_testing()

set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -Wall")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -std=c11 -Wall")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall")

option(ENABLE_STRIP "Create a stripped volk_gnsssdr_profile binary (without shared libraries)" OFF)

Expand Down
43 changes: 22 additions & 21 deletions ...bs/volk_gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_convert_32fc.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,8 @@

static inline void volk_gnsssdr_16ic_convert_32fc_generic(lv_32fc_t* outputVector, const lv_16sc_t* inputVector, unsigned int num_points)
{
for(unsigned int i = 0; i < num_points; i++)
unsigned int i;
for(i = 0; i < num_points; i++)
{
outputVector[i] = lv_cmake((float)lv_creal(inputVector[i]), (float)lv_cimag(inputVector[i]));
}
Expand All @@ -76,22 +77,19 @@ static inline void volk_gnsssdr_16ic_convert_32fc_generic(lv_32fc_t* outputVecto
static inline void volk_gnsssdr_16ic_convert_32fc_a_sse2(lv_32fc_t* outputVector, const lv_16sc_t* inputVector, unsigned int num_points)
{
const unsigned int sse_iters = num_points / 2;

unsigned int i;
const lv_16sc_t* _in = inputVector;
lv_32fc_t* _out = outputVector;
__m128 a;
for(unsigned int number = 0; number < sse_iters; number++)

for(i = 0; i < sse_iters; i++)
{
a = _mm_set_ps((float)(lv_cimag(_in[1])), (float)(lv_creal(_in[1])), (float)(lv_cimag(_in[0])), (float)(lv_creal(_in[0]))); // //load (2 byte imag, 2 byte real) x 2 into 128 bits reg
a = _mm_set_ps((float)(lv_cimag(_in[1])), (float)(lv_creal(_in[1])), (float)(lv_cimag(_in[0])), (float)(lv_creal(_in[0]))); // load (2 byte imag, 2 byte real) x 2 into 128 bits reg
_mm_store_ps((float*)_out, a);
_in += 2;
_out += 2;
//*_out++ = lv_cmake((float)lv_creal(*_in),(float)lv_cimag(*_in));
//_in++;
//*_out++ = lv_cmake((float)lv_creal(*_in),(float)lv_cimag(*_in));
//_in++;
}
for (unsigned int i = 0; i < (num_points % 2); ++i)
for (i = 0; i < (num_points % 2); ++i)
{
*_out++ = lv_cmake((float)lv_creal(*_in), (float)lv_cimag(*_in));
_in++;
Expand All @@ -106,18 +104,19 @@ static inline void volk_gnsssdr_16ic_convert_32fc_a_sse2(lv_32fc_t* outputVector
static inline void volk_gnsssdr_16ic_convert_32fc_u_sse2(lv_32fc_t* outputVector, const lv_16sc_t* inputVector, unsigned int num_points)
{
const unsigned int sse_iters = num_points / 2;

unsigned int i;
const lv_16sc_t* _in = inputVector;
lv_32fc_t* _out = outputVector;
__m128 a;
for(unsigned int number = 0; number < sse_iters; number++)

for(i = 0; i < sse_iters; i++)
{
a = _mm_set_ps((float)(lv_cimag(_in[1])), (float)(lv_creal(_in[1])), (float)(lv_cimag(_in[0])), (float)(lv_creal(_in[0]))); // //load (2 byte imag, 2 byte real) x 2 into 128 bits reg
_mm_storeu_ps((float*)_out, a);
_in += 2;
_out += 2;
}
for (unsigned int i = 0; i < (num_points % 2); ++i)
for (i = 0; i < (num_points % 2); ++i)
{
*_out++ = lv_cmake((float)lv_creal(*_in), (float)lv_cimag(*_in));
_in++;
Expand All @@ -132,19 +131,20 @@ static inline void volk_gnsssdr_16ic_convert_32fc_u_sse2(lv_32fc_t* outputVector
static inline void volk_gnsssdr_16ic_convert_32fc_u_axv(lv_32fc_t* outputVector, const lv_16sc_t* inputVector, unsigned int num_points)
{
const unsigned int sse_iters = num_points / 4;

unsigned int i;
const lv_16sc_t* _in = inputVector;
lv_32fc_t* _out = outputVector;
__m256 a;
for(unsigned int number = 0; number < sse_iters; number++)

for(i = 0; i < sse_iters; i++)
{
a = _mm256_set_ps((float)(lv_cimag(_in[3])), (float)(lv_creal(_in[3])), (float)(lv_cimag(_in[2])), (float)(lv_creal(_in[2])), (float)(lv_cimag(_in[1])), (float)(lv_creal(_in[1])), (float)(lv_cimag(_in[0])), (float)(lv_creal(_in[0]))); // //load (2 byte imag, 2 byte real) x 2 into 128 bits reg
_mm256_storeu_ps((float*)_out, a);
_in += 4;
_out += 4;
}
_mm256_zeroupper();
for (unsigned int i = 0; i < (num_points % 4); ++i)
for(i = 0; i < (num_points % 4); ++i)
{
*_out++ = lv_cmake((float)lv_creal(*_in), (float)lv_cimag(*_in));
_in++;
Expand All @@ -158,19 +158,20 @@ static inline void volk_gnsssdr_16ic_convert_32fc_u_axv(lv_32fc_t* outputVector,
static inline void volk_gnsssdr_16ic_convert_32fc_a_axv(lv_32fc_t* outputVector, const lv_16sc_t* inputVector, unsigned int num_points)
{
const unsigned int sse_iters = num_points / 4;

unsigned int i;
const lv_16sc_t* _in = inputVector;
lv_32fc_t* _out = outputVector;
__m256 a;
for(unsigned int number = 0; number < sse_iters; number++)

for(i = 0; i < sse_iters; i++)
{
a = _mm256_set_ps((float)(lv_cimag(_in[3])), (float)(lv_creal(_in[3])), (float)(lv_cimag(_in[2])), (float)(lv_creal(_in[2])), (float)(lv_cimag(_in[1])), (float)(lv_creal(_in[1])), (float)(lv_cimag(_in[0])), (float)(lv_creal(_in[0]))); // //load (2 byte imag, 2 byte real) x 2 into 128 bits reg
_mm256_store_ps((float*)_out, a);
_in += 4;
_out += 4;
}
_mm256_zeroupper();
for (unsigned int i = 0; i < (num_points % 4); ++i)
for(i = 0; i < (num_points % 4); ++i)
{
*_out++ = lv_cmake((float)lv_creal(*_in), (float)lv_cimag(*_in));
_in++;
Expand All @@ -185,15 +186,15 @@ static inline void volk_gnsssdr_16ic_convert_32fc_a_axv(lv_32fc_t* outputVector,
static inline void volk_gnsssdr_16ic_convert_32fc_neon(lv_32fc_t* outputVector, const lv_16sc_t* inputVector, unsigned int num_points)
{
const unsigned int sse_iters = num_points / 2;

unsigned int i;
const lv_16sc_t* _in = inputVector;
lv_32fc_t* _out = outputVector;

int16x4_t a16x4;
int32x4_t a32x4;
float32x4_t f32x4;

for(unsigned int number = 0; number < sse_iters; number++)
for(i = 0; i < sse_iters; i++)
{
a16x4 = vld1_s16((const int16_t*)_in);
__builtin_prefetch(_in + 4);
Expand All @@ -203,7 +204,7 @@ static inline void volk_gnsssdr_16ic_convert_32fc_neon(lv_32fc_t* outputVector,
_in += 2;
_out += 2;
}
for (unsigned int i = 0; i < (num_points % 2); ++i)
for (i = 0; i < (num_points % 2); ++i)
{
*_out++ = lv_cmake((float)lv_creal(*_in), (float)lv_cimag(*_in));
_in++;
Expand Down
3 changes: 2 additions & 1 deletion ..._gnsssdr_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_resampler_fast_16ic.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -71,8 +71,9 @@
static inline void volk_gnsssdr_16ic_resampler_fast_16ic_generic(lv_16sc_t* result, const lv_16sc_t* local_code, float rem_code_phase_chips, float code_phase_step_chips, int code_length_chips, unsigned int num_output_samples)
{
int local_code_chip_index;
unsigned int n;
//fesetround(FE_TONEAREST);
for (unsigned int n = 0; n < num_output_samples; n++)
for (n = 0; n < num_output_samples; n++)
{
// resample code for current tap
local_code_chip_index = round(code_phase_step_chips * (float)n + rem_code_phase_chips - 0.5f);
Expand Down
25 changes: 16 additions & 9 deletions ...r_module/volk_gnsssdr/kernels/volk_gnsssdr/volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -47,10 +47,11 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_generic(lv_16sc_
float code_phase_step_chips = 0.1;
int code_length_chips = 2046;
int num_out_vectors = 3;
unsigned int n;
float* rem_code_phase_chips = (float*)volk_gnsssdr_malloc(sizeof(float) * num_out_vectors, volk_gnsssdr_get_alignment());

lv_16sc_t** result_aux = (lv_16sc_t**)volk_gnsssdr_malloc(sizeof(lv_16sc_t*) * num_out_vectors, volk_gnsssdr_get_alignment());
for(unsigned int n = 0; n < num_out_vectors; n++)

for(n = 0; n < num_out_vectors; n++)
{
rem_code_phase_chips[n] = -0.234;
result_aux[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t) * num_points, volk_gnsssdr_get_alignment());
Expand All @@ -59,7 +60,7 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_generic(lv_16sc_

memcpy((lv_16sc_t*)result, (lv_16sc_t*)result_aux[0], sizeof(lv_16sc_t) * num_points);
volk_gnsssdr_free(rem_code_phase_chips);
for(unsigned int n = 0; n < num_out_vectors; n++)
for(n = 0; n < num_out_vectors; n++)
{
volk_gnsssdr_free(result_aux[n]);
}
Expand All @@ -75,9 +76,11 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_a_sse2(lv_16sc_t
float code_phase_step_chips = 0.1;
int code_length_chips = 2046;
int num_out_vectors = 3;
unsigned int n;
float * rem_code_phase_chips = (float*)volk_gnsssdr_malloc(sizeof(float) * num_out_vectors, volk_gnsssdr_get_alignment());
lv_16sc_t** result_aux = (lv_16sc_t**)volk_gnsssdr_malloc(sizeof(lv_16sc_t*) * num_out_vectors, volk_gnsssdr_get_alignment());
for(unsigned int n = 0; n < num_out_vectors; n++)

for(n = 0; n < num_out_vectors; n++)
{
rem_code_phase_chips[n] = -0.234;
result_aux[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t) * num_points, volk_gnsssdr_get_alignment());
Expand All @@ -86,7 +89,7 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_a_sse2(lv_16sc_t

memcpy(result, result_aux[0], sizeof(lv_16sc_t) * num_points);
volk_gnsssdr_free(rem_code_phase_chips);
for(unsigned int n = 0; n < num_out_vectors; n++)
for(n = 0; n < num_out_vectors; n++)
{
volk_gnsssdr_free(result_aux[n]);
}
Expand All @@ -102,9 +105,11 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_u_sse2(lv_16sc_t
float code_phase_step_chips = 0.1;
int code_length_chips = 2046;
int num_out_vectors = 3;
unsigned int n;
float * rem_code_phase_chips = (float*)volk_gnsssdr_malloc(sizeof(float) * num_out_vectors, volk_gnsssdr_get_alignment());
lv_16sc_t** result_aux = (lv_16sc_t**)volk_gnsssdr_malloc(sizeof(lv_16sc_t*) * num_out_vectors, volk_gnsssdr_get_alignment());
for(unsigned int n = 0; n < num_out_vectors; n++)

for(n = 0; n < num_out_vectors; n++)
{
rem_code_phase_chips[n] = -0.234;
result_aux[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t) * num_points, volk_gnsssdr_get_alignment());
Expand All @@ -113,7 +118,7 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_u_sse2(lv_16sc_t

memcpy(result, result_aux[0], sizeof(lv_16sc_t) * num_points);
volk_gnsssdr_free(rem_code_phase_chips);
for(unsigned int n = 0; n < num_out_vectors; n++)
for(n = 0; n < num_out_vectors; n++)
{
volk_gnsssdr_free(result_aux[n]);
}
Expand All @@ -129,9 +134,11 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_neon(lv_16sc_t*
float code_phase_step_chips = 0.1;
int code_length_chips = 2046;
int num_out_vectors = 3;
unsigned int n;
float * rem_code_phase_chips = (float*)volk_gnsssdr_malloc(sizeof(float) * num_out_vectors, volk_gnsssdr_get_alignment());
lv_16sc_t** result_aux = (lv_16sc_t**)volk_gnsssdr_malloc(sizeof(lv_16sc_t*) * num_out_vectors, volk_gnsssdr_get_alignment());
for(unsigned int n = 0; n < num_out_vectors; n++)

for(n = 0; n < num_out_vectors; n++)
{
rem_code_phase_chips[n] = -0.234;
result_aux[n] = (lv_16sc_t*)volk_gnsssdr_malloc(sizeof(lv_16sc_t) * num_points, volk_gnsssdr_get_alignment());
Expand All @@ -140,7 +147,7 @@ static inline void volk_gnsssdr_16ic_resamplerfastxnpuppet_16ic_neon(lv_16sc_t*

memcpy(result, result_aux[0], sizeof(lv_16sc_t) * num_points);
volk_gnsssdr_free(rem_code_phase_chips);
for(unsigned int n = 0; n < num_out_vectors; n++)
for(n = 0; n < num_out_vectors; n++)
{
volk_gnsssdr_free(result_aux[n]);
}
Expand Down
Loading

0 comments on commit d490191

Please sign in to comment.