Commit 3f5bf8db authored by Rohit Gupta's avatar Rohit Gupta

Merge branch 'enhancement-10-harmony' of...

Merge branch 'enhancement-10-harmony' of gitlab.eurecom.fr:oai/openairinterface5g into enhancement-10-harmony
parents c33ca8d9 f77abdd6
......@@ -493,6 +493,11 @@ void compute_beta8(llr_t* alpha,llr_t* beta,llr_t *m_11,llr_t* m_10,unsigned sho
#endif
if (frame_length > 6143) {
LOG_E(PHY,"compute_beta: frame_length %d\n",frame_length);
return;
}
// we are supposed to run compute_alpha just before compute_beta
// so the initial states of backward computation can be set from last value of alpha states (forward computation)
......
......@@ -818,11 +818,14 @@ int32_t generate_prach( PHY_VARS_UE *ue, uint8_t eNB_id, uint8_t subframe, uint1
break;
case 75:
memset((void*)prachF,0,4*19432);
memset((void*)prachF,0,4*18432);
break;
case 100:
memset((void*)prachF,0,4*24576);
if (ue->frame_parms.threequarter_fs == 0)
memset((void*)prachF,0,4*24576);
else
memset((void*)prachF,0,4*18432);
break;
}
......@@ -886,6 +889,9 @@ int32_t generate_prach( PHY_VARS_UE *ue, uint8_t eNB_id, uint8_t subframe, uint1
break;
}
if (ue->frame_parms.threequarter_fs == 1)
Ncp=(Ncp*3)>>2;
prach2 = prach+(Ncp<<1);
// do IDFT
......@@ -986,20 +992,39 @@ int32_t generate_prach( PHY_VARS_UE *ue, uint8_t eNB_id, uint8_t subframe, uint1
break;
case 100:
if (prach_fmt == 4) {
idft4096(prachF,prach2,1);
memmove( prach, prach+8192, Ncp<<2 );
prach_len = 4096+Ncp;
} else {
idft24576(prachF,prach2);
memmove( prach, prach+49152, Ncp<<2 );
prach_len = 24576+Ncp;
if (prach_fmt>1) {
memmove( prach2+49152, prach2, 98304 );
prach_len = 2* 24576+Ncp;
if (ue->frame_parms.threequarter_fs == 0) {
if (prach_fmt == 4) {
idft4096(prachF,prach2,1);
memmove( prach, prach+8192, Ncp<<2 );
prach_len = 4096+Ncp;
} else {
idft24576(prachF,prach2);
memmove( prach, prach+49152, Ncp<<2 );
prach_len = 24576+Ncp;
if (prach_fmt>1) {
memmove( prach2+49152, prach2, 98304 );
prach_len = 2* 24576+Ncp;
}
}
}
else {
if (prach_fmt == 4) {
idft3072(prachF,prach2);
//TODO: account for repeated format in dft output
memmove( prach, prach+6144, Ncp<<2 );
prach_len = 3072+Ncp;
} else {
idft18432(prachF,prach2);
memmove( prach, prach+36864, Ncp<<2 );
prach_len = 18432+Ncp;
printf("Generated prach for 100 PRB, 3/4 sampling\n");
if (prach_fmt>1) {
memmove( prach2+36834, prach2, 73728 );
prach_len = 2*18432+Ncp;
}
}
}
break;
}
......@@ -1026,18 +1051,18 @@ int32_t generate_prach( PHY_VARS_UE *ue, uint8_t eNB_id, uint8_t subframe, uint1
((int16_t*)ue->common_vars.txdata[0])[2*i+1] = prach[2*j+1]<<4;
}
#if defined(EXMIMO)
// handle switch before 1st TX subframe, guarantee that the slot prior to transmission is switch on
for (k=prach_start - (ue->frame_parms.samples_per_tti>>1) ; k<prach_start ; k++) {
if (k<0)
ue->common_vars.txdata[0][k+ue->frame_parms.samples_per_tti*LTE_NUMBER_OF_SUBFRAMES_PER_FRAME] &= 0xFFFEFFFE;
else if (k>(ue->frame_parms.samples_per_tti*LTE_NUMBER_OF_SUBFRAMES_PER_FRAME))
ue->common_vars.txdata[0][k-ue->frame_parms.samples_per_tti*LTE_NUMBER_OF_SUBFRAMES_PER_FRAME] &= 0xFFFEFFFE;
else
ue->common_vars.txdata[0][k] &= 0xFFFEFFFE;
}
// handle switch before 1st TX subframe, guarantee that the slot prior to transmission is switch on
for (k=prach_start - (ue->frame_parms.samples_per_tti>>1) ; k<prach_start ; k++) {
if (k<0)
ue->common_vars.txdata[0][k+ue->frame_parms.samples_per_tti*LTE_NUMBER_OF_SUBFRAMES_PER_FRAME] &= 0xFFFEFFFE;
else if (k>(ue->frame_parms.samples_per_tti*LTE_NUMBER_OF_SUBFRAMES_PER_FRAME))
ue->common_vars.txdata[0][k-ue->frame_parms.samples_per_tti*LTE_NUMBER_OF_SUBFRAMES_PER_FRAME] &= 0xFFFEFFFE;
else
ue->common_vars.txdata[0][k] &= 0xFFFEFFFE;
}
#endif
#else
for (i=0; i<prach_len; i++) {
((int16_t*)(&ue->common_vars.txdata[0][prach_start]))[2*i] = prach[2*i];
((int16_t*)(&ue->common_vars.txdata[0][prach_start]))[2*i+1] = prach[2*i+1];
......@@ -1496,7 +1521,7 @@ void rx_prach(PHY_VARS_eNB *eNB,
#ifdef PRACH_DEBUG
if (en>40) {
// if (en>40) {
k = (12*n_ra_prb) - 6*eNB->frame_parms.N_RB_UL;
if (k<0)
......@@ -1510,7 +1535,7 @@ void rx_prach(PHY_VARS_eNB *eNB,
write_output("prach_rxF_comp0.m","prach_rxF_comp0",prachF,1024,1,1);
write_output("prach_ifft0.m","prach_t0",prach_ifft[0],1024,1,1);
exit(-1);
}
// }
#endif
} // new dft
......
......@@ -190,6 +190,9 @@ void generate_pucch1x(int32_t **txdataF,
else {
d = (frame_parms->Ncp==0) ? 2 : 0;
h= (nprime0+d)%(c*Nprime_div_deltaPUCCH_Shift);
#ifdef DEBUG_PUCCH_TX
printf("[PHY] PUCCH: h %d, d %d\n",h,d);
#endif
nprime1 = (h/c) + (h%c)*Nprime_div_deltaPUCCH_Shift;
}
......@@ -208,7 +211,7 @@ void generate_pucch1x(int32_t **txdataF,
n_oc1<<=1;
#ifdef DEBUG_PUCCH_TX
printf("[PHY] PUCCH: noc0 %d noc11 %d\n",n_oc0,n_oc1);
printf("[PHY] PUCCH: noc0 %d noc1 %d\n",n_oc0,n_oc1);
#endif
nprime=nprime0;
......
......@@ -1200,6 +1200,7 @@ unsigned int ulsch_decoding(PHY_VARS_eNB *eNB,eNB_rxtx_proc_t *proc,
y[i2] = c*ulsch_llr[i++];
i2=(i2+(Cmux<<2)-3);
*/
// slightly more optimized version (equivalent to above) for 16QAM to improve computational performance
*(__m64 *)&y[i2] = _mm_sign_pi16(*(__m64*)&ulsch_llr[i],*(__m64*)&cseq[i]);i+=4;i2+=(Cmux<<2);
......@@ -1531,6 +1532,7 @@ unsigned int ulsch_decoding(PHY_VARS_eNB *eNB,eNB_rxtx_proc_t *proc,
// RI
// rank 1
if ((ulsch_harq->O_RI == 1) && (Qprime_RI > 0)) {
ulsch_harq->o_RI[0] = ((ulsch_harq->q_RI[0] + ulsch_harq->q_RI[Q_m/2]) > 0) ? 0 : 1;
}
......
......@@ -5229,7 +5229,7 @@ void dft8192(int16_t *x,int16_t *y,int scale)
xtmpp = xtmp;
for (i=0; i<32; i++) {
for (i=0; i<16; i++) {
transpose4_ooff_simd256(x256 ,xtmpp,512);
transpose4_ooff_simd256(x256+2,xtmpp+1,512);
transpose4_ooff_simd256(x256+4,xtmpp+2,512);
......@@ -5267,7 +5267,7 @@ void dft8192(int16_t *x,int16_t *y,int scale)
}
dft4096((int16_t*)(xtmp),(int16_t*)ytmp,1);
dft4096((int16_t*)(xtmp+1024),(int16_t*)(ytmp+512),1);
dft4096((int16_t*)(xtmp+512),(int16_t*)(ytmp+512),1);
for (i=0; i<512; i++) {
......@@ -5319,7 +5319,7 @@ void idft8192(int16_t *x,int16_t *y,int scale)
xtmpp = xtmp;
for (i=0; i<32; i++) {
for (i=0; i<16; i++) {
transpose4_ooff_simd256(x256 ,xtmpp,512);
transpose4_ooff_simd256(x256+2,xtmpp+1,512);
transpose4_ooff_simd256(x256+4,xtmpp+2,512);
......@@ -5357,7 +5357,7 @@ void idft8192(int16_t *x,int16_t *y,int scale)
}
idft4096((int16_t*)(xtmp),(int16_t*)ytmp,1);
idft4096((int16_t*)(xtmp+1024),(int16_t*)(ytmp+512),1);
idft4096((int16_t*)(xtmp+512),(int16_t*)(ytmp+512),1);
for (i=0; i<512; i++) {
......@@ -5680,15 +5680,58 @@ void idft12288(int16_t *input, int16_t *output)
// write_output("out.m","out",output,6144,1,1);
}
#include "twiddle18432.h"
// 6144 x 3
void dft18432(int16_t *input, int16_t *output)
{
void dft18432(int16_t *input, int16_t *output) {
int i,i2,j;
uint32_t tmp[3][6144] __attribute__((aligned(32)));
uint32_t tmpo[3][6144] __attribute__((aligned(32)));
for (i=0,j=0; i<6144; i++) {
tmp[0][i] = ((uint32_t *)input)[j++];
tmp[1][i] = ((uint32_t *)input)[j++];
tmp[2][i] = ((uint32_t *)input)[j++];
}
dft6144((int16_t*)(tmp[0]),(int16_t*)(tmpo[0]));
dft6144((int16_t*)(tmp[1]),(int16_t*)(tmpo[1]));
dft6144((int16_t*)(tmp[2]),(int16_t*)(tmpo[2]));
for (i=0,i2=0; i<12288; i+=8,i2+=4) {
bfly3((simd_q15_t*)(&tmpo[0][i2]),(simd_q15_t*)(&tmpo[1][i2]),(simd_q15_t*)(&tmpo[2][i2]),
(simd_q15_t*)(output+i),(simd_q15_t*)(output+12288+i),(simd_q15_t*)(output+24576+i),
(simd_q15_t*)(twa18432+i),(simd_q15_t*)(twb18432+i));
}
_mm_empty();
_m_empty();
}
void idft18432(int16_t *input, int16_t *output)
{
void idft18432(int16_t *input, int16_t *output) {
int i,i2,j;
uint32_t tmp[3][6144] __attribute__((aligned(32)));
uint32_t tmpo[3][6144] __attribute__((aligned(32)));
for (i=0,j=0; i<6144; i++) {
tmp[0][i] = ((uint32_t *)input)[j++];
tmp[1][i] = ((uint32_t *)input)[j++];
tmp[2][i] = ((uint32_t *)input)[j++];
}
idft6144((int16_t*)(tmp[0]),(int16_t*)(tmpo[0]));
idft6144((int16_t*)(tmp[1]),(int16_t*)(tmpo[1]));
idft6144((int16_t*)(tmp[2]),(int16_t*)(tmpo[2]));
for (i=0,i2=0; i<12288; i+=8,i2+=4) {
ibfly3((simd_q15_t*)(&tmpo[0][i2]),(simd_q15_t*)(&tmpo[1][i2]),(simd_q15_t*)(&tmpo[2][i2]),
(simd_q15_t*)(output+i),(simd_q15_t*)(output+12288+i),(simd_q15_t*)(output+24576+i),
(simd_q15_t*)(twa18432+i),(simd_q15_t*)(twb18432+i));
}
_mm_empty();
_m_empty();
}
#include "twiddle24576.h"
......@@ -5733,8 +5776,8 @@ void dft24576(int16_t *input, int16_t *output)
void idft24576(int16_t *input, int16_t *output)
{
int i,i2,j;
uint32_t tmp[3][16384] __attribute__((aligned(32)));
uint32_t tmpo[3][16384] __attribute__((aligned(32)));
uint32_t tmp[3][8192] __attribute__((aligned(32)));
uint32_t tmpo[3][8192] __attribute__((aligned(32)));
for (i=0,j=0; i<8192; i++) {
tmp[0][i] = ((uint32_t *)input)[j++];
......@@ -5746,13 +5789,6 @@ void idft24576(int16_t *input, int16_t *output)
idft8192((int16_t*)(tmp[1]),(int16_t*)(tmpo[1]),1);
idft8192((int16_t*)(tmp[2]),(int16_t*)(tmpo[2]),1);
/*
for (i=1; i<8192; i++) {
tmpo[0][i] = tmpo[0][i<<1];
tmpo[1][i] = tmpo[1][i<<1];
tmpo[2][i] = tmpo[2][i<<1];
}*/
/*
write_output("in.m","in",input,24576,1,1);
write_output("out0.m","o0",tmpo[0],8192,1,1);
......@@ -19129,6 +19165,55 @@ int main(int argc, char**argv)
write_output("y2048.m","y2048",y,2048,1,1);
write_output("x2048.m","x2048",x,2048,1,1);
memset((void*)x,0,2048*sizeof(int32_t));
for (i=2;i<2402;i++) {
if ((taus() & 1)==0)
((int16_t*)x)[i] = 364;
else
((int16_t*)x)[i] = -364;
}
for (i=2*(4096-1200);i<8192;i++) {
if ((taus() & 1)==0)
((int16_t*)x)[i] = 364;
else
((int16_t*)x)[i] = -364;
}
reset_meas(&ts);
for (i=0; i<10000; i++) {
start_meas(&ts);
idft4096((int16_t *)x,(int16_t *)y,1);
stop_meas(&ts);
}
printf("\n\n4096-point(%f cycles)\n",(double)ts.diff/(double)ts.trials);
write_output("y4096.m","y4096",y,4096,1,1);
write_output("x4096.m","x4096",x,4096,1,1);
memset((void*)x,0,8192*sizeof(int32_t));
for (i=2;i<4802;i++) {
if ((taus() & 1)==0)
((int16_t*)x)[i] = 364;
else
((int16_t*)x)[i] = -364;
}
for (i=2*(4096-1200);i<8192;i++) {
if ((taus() & 1)==0)
((int16_t*)x)[i] = 364;
else
((int16_t*)x)[i] = -364;
}
reset_meas(&ts);
for (i=0; i<10000; i++) {
start_meas(&ts);
idft8192((int16_t *)x,(int16_t *)y,1);
stop_meas(&ts);
}
printf("\n\n8192-point(%f cycles)\n",(double)ts.diff/(double)ts.trials);
write_output("y8192.m","y8192",y,8192,1,1);
write_output("x8192.m","x8192",x,8192,1,1);
return(0);
}
This source diff could not be displayed because it is too large. You can view the blob instead.
This source diff could not be displayed because it is too large. You can view the blob instead.
This source diff could not be displayed because it is too large. You can view the blob instead.
This source diff could not be displayed because it is too large. You can view the blob instead.
......@@ -104,13 +104,15 @@ int main(int argc, char **argv)
double delay_avg=0;
double ue_speed = 0;
int NCS_config = 1,rootSequenceIndex=0;
int threequarter_fs = 0;
logInit();
number_of_cards = 1;
while ((c = getopt (argc, argv, "hHaA:Cr:p:g:n:s:S:t:x:y:v:V:z:N:F:d:Z:L:R:")) != -1) {
while ((c = getopt (argc, argv, "hHaA:Cr:p:g:n:s:S:t:x:y:v:V:z:N:F:d:Z:L:R:E")) != -1) {
switch (c) {
case 'a':
printf("Running AWGN simulation\n");
......@@ -182,6 +184,10 @@ int main(int argc, char **argv)
break;
case 'E':
threequarter_fs=1;
break;
case 'n':
n_frames = atoi(optarg);
break;
......@@ -316,7 +322,7 @@ int main(int argc, char **argv)
Nid_cell,
3,
N_RB_DL,
0,
threequarter_fs,
osf,
0);
......@@ -395,8 +401,9 @@ int main(int argc, char **argv)
eNB->frame_parms.prach_config_common.prach_ConfigInfo.highSpeedFlag=hs_flag;
eNB->frame_parms.prach_config_common.prach_ConfigInfo.prach_FreqOffset=0;
eNB->node_function = eNodeB_3GPP;
eNB->proc.subframe_rx = subframe;
eNB->node_function = eNodeB_3GPP;
eNB->proc.subframe_rx = subframe;
eNB->proc.subframe_prach = subframe;
/* N_ZC not used later, so prach_fmt is also useless, don't set */
//prach_fmt = get_prach_fmt(eNB->frame_parms.prach_config_common.prach_ConfigInfo.prach_ConfigIndex,
......@@ -530,7 +537,7 @@ int main(int argc, char **argv)
write_output("rxsig0.m","rxs0",
&eNB->common_vars.rxdata[0][0][subframe*frame_parms->samples_per_tti],
frame_parms->samples_per_tti,1,1);
write_output("rxsigF0.m","rxsF0", &eNB->common_vars.rxdataF[0][0][0],512*nsymb*2,2,1);
write_output("rxsigF0.m","rxsF0", eNB->prach_vars.rxsigF[0],6144,1,1);
write_output("prach_preamble.m","prachp",&eNB->X_u[0],839,1,1);
}
}
......
......@@ -66,7 +66,11 @@ int main(int argc, char **argv)
uint8_t snr1set=0;
//mod_sym_t **txdataF;
int **txdata;
double s_re[2][30720],s_im[2][30720],r_re[2][30720],r_im[2][30720];
double s_re0[30720],s_re1[30720],s_im0[30720],s_im1[30720],r_re0[30720],r_im0[30720],r_re1[30720],r_im1[30720];
double *s_re[2]={s_re0,s_re1};
double *s_im[2]={s_im0,s_im1};
double *r_re[2]={r_re0,r_re1};
double *r_im[2]={r_im0,r_im1};
double ricean_factor=0.0000005,iqim=0.0;
int trial, n_trials, ntrials=1, n_errors;
......@@ -352,12 +356,15 @@ int main(int argc, char **argv)
init_ncs_cell(&UE->frame_parms,UE->ncs_cell);
eNB->frame_parms.pucch_config_common.deltaPUCCH_Shift = 1;
eNB->frame_parms.pucch_config_common.nRB_CQI = 0;
eNB->frame_parms.pucch_config_common.nCS_AN = 0;
UE->frame_parms.pucch_config_common.deltaPUCCH_Shift = 1;
UE->frame_parms.pucch_config_common.nRB_CQI = 0;
UE->frame_parms.pucch_config_common.nCS_AN = 0;
init_ul_hopping(&eNB->frame_parms);
init_ul_hopping(&UE->frame_parms);
eNB->frame_parms.pucch_config_common.deltaPUCCH_Shift = 2;
eNB->frame_parms.pucch_config_common.nRB_CQI = 4;
eNB->frame_parms.pucch_config_common.nCS_AN = 6;
UE->frame_parms.pucch_config_common.deltaPUCCH_Shift = 2;
UE->frame_parms.pucch_config_common.nRB_CQI = 4;
UE->frame_parms.pucch_config_common.nCS_AN = 6;
pucch_payload = 0;
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment