added avx2 optimized turbo decoder for 16-bit LLR. This decoder parallelizes by decoding 2 code segments concurrently. requires updates dlsch_decoding.c to identify when new parallel version can be used. other minor changes related to memory allocations for future avx2 optimizations (32-byte alignment).
This commit is contained in:
Raymond Knopp
@@ -134,7 +134,7 @@ else (CMAKE_SYSTEM_PROCESSOR STREQUAL "armv7l")
|
||||
set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -mavx2")
|
||||
endif()
|
||||
if (CPUINFO MATCHES "sse4_2")
|
||||
set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -msse4.2")
|
||||
set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -mavx2 -msse4.2")
|
||||
endif()
|
||||
if (CPUINFO MATCHES "sse4_1")
|
||||
set(C_FLAGS_PROCESSOR "${C_FLAGS_PROCESSOR} -msse4.1")
|
||||
@@ -168,7 +168,7 @@ set(CMAKE_CXX_FLAGS
|
||||
# these changes are related to hardcoded path to include .h files
|
||||
add_definitions(-DCMAKER)
|
||||
set(CMAKE_C_FLAGS_DEBUG "${CMAKE_C_FLAGS} -g -DMALLOC_CHECK_=3")
|
||||
set(CMAKE_C_FLAGS_RELWITHDEBINFO "${CMAKE_C_FLAGS} -g -DMALLOC_CHECK_=3 -O2")
|
||||
set(CMAKE_C_FLAGS_RELWITHDEBINFO "${CMAKE_C_FLAGS} -g -DMALLOC_CHECK_=3 -O3")
|
||||
|
||||
# Below has been put in comment because does not work with
|
||||
# SVN authentication.
|
||||
@@ -840,6 +840,7 @@ set(PHY_SRC
|
||||
${OPENAIR1_DIR}/PHY/CODING/crc_byte.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/3gpplte_turbo_decoder_sse_8bit.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/3gpplte_turbo_decoder_sse_16bit.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/3gpplte_turbo_decoder_avx2_16bit.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/lte_rate_matching.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/rate_matching.c
|
||||
${OPENAIR1_DIR}/PHY/CODING/viterbi.c
|
||||
|
||||
File diff suppressed because it is too large
Load Diff
@@ -63,11 +63,17 @@
|
||||
#include "mex.h"
|
||||
#endif
|
||||
|
||||
|
||||
#define print_shorts(s,x) printf("%s %d,%d,%d,%d,%d,%d,%d,%d\n",s,(x)[0],(x)[1],(x)[2],(x)[3],(x)[4],(x)[5],(x)[6],(x)[7])
|
||||
|
||||
//#define DEBUG_LOGMAP
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
#define print_shorts(s,x) fprintf(fdsse4,"%s %d,%d,%d,%d,%d,%d,%d,%d\n",s,(x)[0],(x)[1],(x)[2],(x)[3],(x)[4],(x)[5],(x)[6],(x)[7])
|
||||
#endif
|
||||
|
||||
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
FILE *fdsse4;
|
||||
#endif
|
||||
|
||||
typedef int16_t llr_t; // internal decoder LLR data is 16-bit fixed
|
||||
typedef int16_t channel_t;
|
||||
@@ -99,7 +105,7 @@ void log_map16(llr_t* systematic,
|
||||
{
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
msg("log_map, frame_length %d\n",frame_length);
|
||||
fprintf(fdsse4,"log_map, frame_length %d\n",frame_length);
|
||||
#endif
|
||||
|
||||
start_meas(gamma_stats) ;
|
||||
@@ -135,7 +141,7 @@ void compute_gamma16(llr_t* m11,llr_t* m10,llr_t* systematic,channel_t* y_parity
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
msg("compute_gamma, %p,%p,%p,%p,framelength %d\n",m11,m10,systematic,y_parity,frame_length);
|
||||
fprintf(fdsse4,"compute_gamma (sse_16bit), %p,%p,%p,%p,framelength %d\n",m11,m10,systematic,y_parity,frame_length);
|
||||
#endif
|
||||
|
||||
K1=frame_length>>3;
|
||||
@@ -150,7 +156,7 @@ void compute_gamma16(llr_t* m11,llr_t* m10,llr_t* systematic,channel_t* y_parity
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
printf("Loop index k, m11,m10\n");
|
||||
fprintf(fdsse4,"Loop index k %d\n", k);
|
||||
print_shorts("sys",(int16_t*)&systematic128[k]);
|
||||
print_shorts("yp",(int16_t*)&y_parity128[k]);
|
||||
print_shorts("m11",(int16_t*)&m11_128[k]);
|
||||
@@ -164,7 +170,15 @@ void compute_gamma16(llr_t* m11,llr_t* m10,llr_t* systematic,channel_t* y_parity
|
||||
m10_128[k] = _mm_srai_epi16(_mm_subs_epi16(systematic128[k+term_flag],y_parity128[k]),1);
|
||||
#elif defined(__arm__)
|
||||
m11_128[k] = vhaddq_s16(systematic128[k+term_flag],y_parity128[k]);
|
||||
m10_128[k] = vhsubq_s16(systematic128[k+term_flag],y_parity128[k]);
|
||||
m10_128[k] = vhsubq_s16(systematic128[k+term_flag],y_parity128[k]);
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"Loop index k %d (term flag %d)\n", k,term_flag);
|
||||
print_shorts("sys",(int16_t*)&systematic128[k]);
|
||||
print_shorts("yp",(int16_t*)&y_parity128[k]);
|
||||
print_shorts("m11",(int16_t*)&m11_128[k]);
|
||||
print_shorts("m10",(int16_t*)&m10_128[k]);
|
||||
#endif
|
||||
}
|
||||
|
||||
@@ -188,7 +202,9 @@ void compute_alpha16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,unsigned s
|
||||
#endif
|
||||
l2 = L>>3;
|
||||
K1 = (frame_length>>3);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"compute_alpha (sse_16bit)\n");
|
||||
#endif
|
||||
for (l=K1;; l=l2,rerun_flag=1) {
|
||||
#if defined(__x86_64__) || defined(__i386__)
|
||||
alpha128 = (__m128i *)alpha;
|
||||
@@ -218,7 +234,7 @@ void compute_alpha16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,unsigned s
|
||||
alpha128[7] = vdupq_n_s16(-MAX/2);
|
||||
#endif
|
||||
#ifdef DEBUG_LOGMAP
|
||||
printf("Initial alpha\n");
|
||||
fprintf(fdsse4,"Initial alpha\n");
|
||||
print_shorts("a0",(int16_t*)&alpha128[0]);
|
||||
print_shorts("a1",(int16_t*)&alpha128[1]);
|
||||
print_shorts("a2",(int16_t*)&alpha128[2]);
|
||||
@@ -258,7 +274,7 @@ void compute_alpha16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,unsigned s
|
||||
alpha[48] = -MAX/2;
|
||||
alpha[56] = -MAX/2;
|
||||
#ifdef DEBUG_LOGMAP
|
||||
printf("Second run\n");
|
||||
fprintf(fdsse4,"Second run\n");
|
||||
print_shorts("a0",(int16_t*)&alpha128[0]);
|
||||
print_shorts("a1",(int16_t*)&alpha128[1]);
|
||||
print_shorts("a2",(int16_t*)&alpha128[2]);
|
||||
@@ -390,7 +406,7 @@ void compute_alpha16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,unsigned s
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
printf("Loop index %d, mb\n",k);
|
||||
fprintf(fdsse4,"Loop index %d\n",k);
|
||||
print_shorts("mb0",(int16_t*)&m_b0);
|
||||
print_shorts("mb1",(int16_t*)&m_b1);
|
||||
print_shorts("mb2",(int16_t*)&m_b2);
|
||||
@@ -400,7 +416,7 @@ void compute_alpha16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,unsigned s
|
||||
print_shorts("mb6",(int16_t*)&m_b6);
|
||||
print_shorts("mb7",(int16_t*)&m_b7);
|
||||
|
||||
printf("Loop index %d, new\n",k);
|
||||
fprintf(fdsse4,"Loop index %d, new\n",k);
|
||||
print_shorts("new0",(int16_t*)&new0);
|
||||
print_shorts("new1",(int16_t*)&new1);
|
||||
print_shorts("new2",(int16_t*)&new2);
|
||||
@@ -410,7 +426,7 @@ void compute_alpha16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,unsigned s
|
||||
print_shorts("new6",(int16_t*)&new6);
|
||||
print_shorts("new7",(int16_t*)&new7);
|
||||
|
||||
printf("Loop index %d, after max\n",k);
|
||||
fprintf(fdsse4,"Loop index %d, after max\n",k);
|
||||
print_shorts("a0",(int16_t*)&a0);
|
||||
print_shorts("a1",(int16_t*)&a1);
|
||||
print_shorts("a2",(int16_t*)&a2);
|
||||
@@ -420,7 +436,7 @@ void compute_alpha16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,unsigned s
|
||||
print_shorts("a6",(int16_t*)&a6);
|
||||
print_shorts("a7",(int16_t*)&a7);
|
||||
|
||||
printf("Loop index %d\n",k);
|
||||
fprintf(fdsse4,"Loop index %d\n",k);
|
||||
print_shorts("a0",(int16_t*)&alpha_ptr[0]);
|
||||
print_shorts("a1",(int16_t*)&alpha_ptr[1]);
|
||||
print_shorts("a2",(int16_t*)&alpha_ptr[2]);
|
||||
@@ -463,25 +479,29 @@ void compute_beta16(llr_t* alpha,llr_t* beta,llr_t *m_11,llr_t* m_10,unsigned sh
|
||||
llr_t beta0,beta1;
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
msg("compute_beta, %p,%p,%p,%p,framelength %d,F %d\n",
|
||||
beta,m_11,m_10,alpha,frame_length,F);
|
||||
fprintf(fdsse4,"compute_beta, %p,%p,%p,%p,framelength %d,F %d\n",
|
||||
beta,m_11,m_10,alpha,frame_length,F);
|
||||
#endif
|
||||
|
||||
|
||||
// termination for beta initialization
|
||||
|
||||
// printf("beta init: offset8 %d\n",offset8_flag);
|
||||
// fprintf(fdsse4,"beta init: offset8 %d\n",offset8_flag);
|
||||
m11=(int16_t)m_11[2+frame_length];
|
||||
m10=(int16_t)m_10[2+frame_length];
|
||||
|
||||
// printf("m11,m10 %d,%d\n",m11,m10);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"m11,m10 %d,%d\n",m11,m10);
|
||||
#endif
|
||||
|
||||
beta0 = -m11;//M0T_TERM;
|
||||
beta1 = m11;//M1T_TERM;
|
||||
m11=(int16_t)m_11[1+frame_length];
|
||||
m10=(int16_t)m_10[1+frame_length];
|
||||
|
||||
// printf("m11,m10 %d,%d\n",m11,m10);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"m11,m10 %d,%d\n",m11,m10);
|
||||
#endif
|
||||
|
||||
beta0_2 = beta0-m11;//+M0T_TERM;
|
||||
beta1_2 = beta0+m11;//+M1T_TERM;
|
||||
@@ -489,8 +509,9 @@ void compute_beta16(llr_t* alpha,llr_t* beta,llr_t *m_11,llr_t* m_10,unsigned sh
|
||||
beta3_2 = beta1-m10;//+M3T_TERM;
|
||||
m11=(int16_t)m_11[frame_length];
|
||||
m10=(int16_t)m_10[frame_length];
|
||||
// printf("m11,m10 %d,%d (%p)\n",m11,m10,m_11+frame_length);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"m11,m10 %d,%d\n",m11,m10);
|
||||
#endif
|
||||
beta0_16 = beta0_2-m11;//+M0T_TERM;
|
||||
beta1_16 = beta0_2+m11;//+M1T_TERM;
|
||||
beta2_16 = beta1_2+m10;//+M2T_TERM;
|
||||
@@ -536,6 +557,17 @@ void compute_beta16(llr_t* alpha,llr_t* beta,llr_t *m_11,llr_t* m_10,unsigned sh
|
||||
beta_ptr[5] = alpha128[5+(frame_length)];
|
||||
beta_ptr[6] = alpha128[6+(frame_length)];
|
||||
beta_ptr[7] = alpha128[7+(frame_length)];
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"beta init \n");
|
||||
print_shorts("b0",(int16_t*)&beta_ptr[0]);
|
||||
print_shorts("b1",(int16_t*)&beta_ptr[1]);
|
||||
print_shorts("b2",(int16_t*)&beta_ptr[2]);
|
||||
print_shorts("b3",(int16_t*)&beta_ptr[3]);
|
||||
print_shorts("b4",(int16_t*)&beta_ptr[4]);
|
||||
print_shorts("b5",(int16_t*)&beta_ptr[5]);
|
||||
print_shorts("b6",(int16_t*)&beta_ptr[6]);
|
||||
print_shorts("b7",(int16_t*)&beta_ptr[7]);
|
||||
#endif
|
||||
} else {
|
||||
#if defined(__x86_64__) || defined(__i386__)
|
||||
beta128 = (__m128i*)&beta[0];
|
||||
@@ -558,6 +590,17 @@ void compute_beta16(llr_t* alpha,llr_t* beta,llr_t *m_11,llr_t* m_10,unsigned sh
|
||||
beta_ptr[5] = (int16x8_t)vshrq_n_s64((int64x2_t)beta128[5],16); beta_ptr[5] = vsetq_lane_s16(beta[43],beta_ptr[5],4);
|
||||
beta_ptr[6] = (int16x8_t)vshrq_n_s64((int64x2_t)beta128[6],16); beta_ptr[6] = vsetq_lane_s16(beta[51],beta_ptr[6],4);
|
||||
beta_ptr[7] = (int16x8_t)vshrq_n_s64((int64x2_t)beta128[7],16); beta_ptr[7] = vsetq_lane_s16(beta[59],beta_ptr[7],4);
|
||||
#endif
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"beta init (second run) \n");
|
||||
print_shorts("b0",(int16_t*)&beta_ptr[0]);
|
||||
print_shorts("b1",(int16_t*)&beta_ptr[1]);
|
||||
print_shorts("b2",(int16_t*)&beta_ptr[2]);
|
||||
print_shorts("b3",(int16_t*)&beta_ptr[3]);
|
||||
print_shorts("b4",(int16_t*)&beta_ptr[4]);
|
||||
print_shorts("b5",(int16_t*)&beta_ptr[5]);
|
||||
print_shorts("b6",(int16_t*)&beta_ptr[6]);
|
||||
print_shorts("b7",(int16_t*)&beta_ptr[7]);
|
||||
#endif
|
||||
}
|
||||
|
||||
@@ -582,6 +625,17 @@ void compute_beta16(llr_t* alpha,llr_t* beta,llr_t *m_11,llr_t* m_10,unsigned sh
|
||||
beta_ptr[7] = vsetq_lane_s16(beta7_16,beta_ptr[7],7);
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"beta init (after insert) \n");
|
||||
print_shorts("b0",(int16_t*)&beta_ptr[0]);
|
||||
print_shorts("b1",(int16_t*)&beta_ptr[1]);
|
||||
print_shorts("b2",(int16_t*)&beta_ptr[2]);
|
||||
print_shorts("b3",(int16_t*)&beta_ptr[3]);
|
||||
print_shorts("b4",(int16_t*)&beta_ptr[4]);
|
||||
print_shorts("b5",(int16_t*)&beta_ptr[5]);
|
||||
print_shorts("b6",(int16_t*)&beta_ptr[6]);
|
||||
print_shorts("b7",(int16_t*)&beta_ptr[7]);
|
||||
#endif
|
||||
int loopval=((rerun_flag==0)?0:((frame_length-L)>>3));
|
||||
|
||||
for (k=(frame_length>>3)-1; k>=loopval; k--) {
|
||||
@@ -684,6 +738,18 @@ void compute_beta16(llr_t* alpha,llr_t* beta,llr_t *m_11,llr_t* m_10,unsigned sh
|
||||
beta_ptr[7] = vqsubq_s16(beta_ptr[7],beta_max);
|
||||
#endif
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"Loop index %d, mb\n",k);
|
||||
fprintf(fdsse4,"beta init (after max)\n");
|
||||
print_shorts("b0",(int16_t*)&beta_ptr[0]);
|
||||
print_shorts("b1",(int16_t*)&beta_ptr[1]);
|
||||
print_shorts("b2",(int16_t*)&beta_ptr[2]);
|
||||
print_shorts("b3",(int16_t*)&beta_ptr[3]);
|
||||
print_shorts("b4",(int16_t*)&beta_ptr[4]);
|
||||
print_shorts("b5",(int16_t*)&beta_ptr[5]);
|
||||
print_shorts("b6",(int16_t*)&beta_ptr[6]);
|
||||
print_shorts("b7",(int16_t*)&beta_ptr[7]);
|
||||
#endif
|
||||
|
||||
}
|
||||
|
||||
@@ -721,7 +787,7 @@ void compute_ext16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,llr_t* ext,
|
||||
//
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
msg("compute_ext, %p, %p, %p, %p, %p, %p ,framelength %d\n",alpha,beta,m_11,m_10,ext,systematic,frame_length);
|
||||
fprintf(fdsse4,"compute_ext (sse_16bit), %p, %p, %p, %p, %p, %p ,framelength %d\n",alpha,beta,m_11,m_10,ext,systematic,frame_length);
|
||||
#endif
|
||||
|
||||
alpha_ptr = alpha128;
|
||||
@@ -736,7 +802,7 @@ void compute_ext16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,llr_t* ext,
|
||||
ext_128 = (__m128i*)&ext[k<<3];
|
||||
|
||||
/*
|
||||
printf("EXT %03d\n",k);
|
||||
fprintf(fdsse4,"EXT %03d\n",k);
|
||||
print_shorts("a0:",&alpha_ptr[0]);
|
||||
print_shorts("a1:",&alpha_ptr[1]);
|
||||
print_shorts("a2:",&alpha_ptr[2]);
|
||||
@@ -816,15 +882,15 @@ void compute_ext16(llr_t* alpha,llr_t* beta,llr_t* m_11,llr_t* m_10,llr_t* ext,
|
||||
// print_shorts("m10_1:",&m10_1);
|
||||
|
||||
*ext_128 = _mm_subs_epi16(m10_1,m01_1);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"ext %p\n",ext_128);
|
||||
print_shorts("ext:",(int16_t*)ext_128);
|
||||
print_shorts("m11:",(int16_t*)m11_128);
|
||||
print_shorts("m10:",(int16_t*)m10_128);
|
||||
print_shorts("m10_1:",(int16_t*)&m10_1);
|
||||
print_shorts("m01_1:",(int16_t*)&m01_1);
|
||||
#endif
|
||||
|
||||
/*
|
||||
print_shorts("ext:",ext_128);
|
||||
print_shorts("m11:",m11_128);
|
||||
print_shorts("m10:",m10_128);
|
||||
print_shorts("m10_1:",&m10_1);
|
||||
print_shorts("m01_1:",&m01_1);
|
||||
print_shorts("syst:",systematic_128);
|
||||
*/
|
||||
#elif defined(__arm__)
|
||||
m11_128 = (int16x8_t*)&m_11[k<<3];
|
||||
m10_128 = (int16x8_t*)&m_10[k<<3];
|
||||
@@ -927,7 +993,7 @@ void init_td16()
|
||||
// j-=(n-1);
|
||||
|
||||
pi2tab16[ind][i] = j;
|
||||
// printf("pi2[%d] = %d\n",i,j);
|
||||
// fprintf(fdsse4,"pi2[%d] = %d\n",i,j);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -989,7 +1055,7 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
#if defined(__x86_64__) || defined(__i386__)
|
||||
__m128i *yp128;
|
||||
__m128i tmp, zeros=_mm_setzero_si128();
|
||||
register __m128i tmpe;
|
||||
__m128i tmpe;
|
||||
#elif defined(__arm__)
|
||||
int16x8_t *yp128;
|
||||
// int16x8_t tmp128[(n+8)>>3];
|
||||
@@ -1000,12 +1066,20 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
#endif
|
||||
int offset8_flag=0;
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fdsse4 = fopen("dump_sse4.txt","w");
|
||||
|
||||
|
||||
printf("tc sse4_16 (y) %p\n",y);
|
||||
#endif
|
||||
|
||||
if (crc_type > 3) {
|
||||
msg("Illegal crc length!\n");
|
||||
printf("Illegal crc length!\n");
|
||||
return 255;
|
||||
}
|
||||
|
||||
|
||||
|
||||
start_meas(init_stats);
|
||||
|
||||
|
||||
@@ -1013,7 +1087,7 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
for (iind=0; iind < 188 && f1f2mat[iind].nb_bits != n; iind++);
|
||||
|
||||
if ( iind == 188 ) {
|
||||
msg("Illegal frame length!\n");
|
||||
printf("Illegal frame length!\n");
|
||||
return 255;
|
||||
}
|
||||
|
||||
@@ -1059,62 +1133,74 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
|
||||
#if defined(__x86_64__) || defined(__i386__)
|
||||
tmpe = _mm_load_si128(yp128);
|
||||
// fprintf(fdsse4,"yp128 %p\n",yp128);
|
||||
// print_shorts("tmpe",(int16_t *)&tmpe);
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,0);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,1);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,2);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init0: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
j=pi2_p[1];
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,3);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,4);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,5);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init1: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
j=pi2_p[2];
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,6);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,7);
|
||||
tmpe = _mm_load_si128(&yp128[1]);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,0);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init2: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
j=pi2_p[3];
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,1);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,2);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,3);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init3: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
j=pi2_p[4];
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,4);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,5);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,6);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init4: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
j=pi2_p[5];
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,7);
|
||||
tmpe = _mm_load_si128(&yp128[2]);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,0);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,1);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init5: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
|
||||
j=pi2_p[6];
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,2);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,3);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,4);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init6: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
j=pi2_p[7];
|
||||
|
||||
s[j] = _mm_extract_epi16(tmpe,5);
|
||||
yp1[j] = _mm_extract_epi16(tmpe,6);
|
||||
yp2[j] = _mm_extract_epi16(tmpe,7);
|
||||
// printf("init: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"init7: j %d, s[j] %d yp1[j] %d yp2[j] %d\n",j,s[j],yp1[j],yp2[j]);
|
||||
#endif
|
||||
|
||||
#elif defined(__arm__)
|
||||
s[j] = vgetq_lane_s16(yp128[0],0);
|
||||
@@ -1172,7 +1258,7 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
yp1[i] = *yp;
|
||||
yp++;
|
||||
#ifdef DEBUG_LOGMAP
|
||||
msg("Term 1 (%d): %d %d\n",i,s[i],yp1[i]);
|
||||
fprintf(fdsse4,"Term 1 (%d): %d %d\n",i,s[i],yp1[i]);
|
||||
#endif //DEBUG_LOGMAP
|
||||
}
|
||||
|
||||
@@ -1184,12 +1270,12 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
yp2[i-8] = *yp;
|
||||
yp++;
|
||||
#ifdef DEBUG_LOGMAP
|
||||
msg("Term 2 (%d): %d %d\n",i-3,s[i],yp2[i-8]);
|
||||
fprintf(fdsse4,"Term 2 (%d): %d %d\n",i-3,s[i],yp2[i-8]);
|
||||
#endif //DEBUG_LOGMAP
|
||||
}
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
msg("\n");
|
||||
fprintf(fdsse4,"\n");
|
||||
#endif //DEBUG_LOGMAP
|
||||
|
||||
stop_meas(init_stats);
|
||||
@@ -1201,7 +1287,7 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
while (iteration_cnt++ < max_iterations) {
|
||||
|
||||
#ifdef DEBUG_LOGMAP
|
||||
printf("\n*******************ITERATION %d (n %d), ext %p\n\n",iteration_cnt,n,ext);
|
||||
fprintf(fdsse4,"\n*******************ITERATION %d (n %d), ext %p\n\n",iteration_cnt,n,ext);
|
||||
#endif //DEBUG_LOGMAP
|
||||
|
||||
start_meas(intl1_stats);
|
||||
@@ -1209,24 +1295,29 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
pi4_p=pi4tab16[iind];
|
||||
|
||||
for (i=0; i<(n>>3); i++) { // steady-state portion
|
||||
|
||||
#if defined(__x86_64__) || defined(__i386__)
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],0);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],1);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],2);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],3);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],4);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],5);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],6);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],((llr_t*)ext)[*pi4_p++],7);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],0);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],1);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],2);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],3);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],4);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],5);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],6);
|
||||
((__m128i *)systematic2)[i]=_mm_insert_epi16(((__m128i *)systematic2)[i],ext[*pi4_p++],7);
|
||||
|
||||
#elif defined(__arm__)
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],0);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],1);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],2);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],3);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],4);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],5);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],6);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(((llr_t*)ext)[*pi4_p++],((int16x8_t*)systematic2)[i],7);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],0);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],1);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],2);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],3);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],4);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],5);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],6);
|
||||
((int16x8_t*)systematic2)[i]=vsetq_lane_s16(ext[*pi4_p++],((int16x8_t*)systematic2)[i],7);
|
||||
#endif
|
||||
#ifdef DEBUG_LOGMAP
|
||||
print_shorts("syst2",(int16_t*)&((__m128i *)systematic2)[i]);
|
||||
#endif
|
||||
}
|
||||
|
||||
@@ -1261,6 +1352,9 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
tmp=vsetq_lane_s16(ext2[*pi5_p++],tmp,6);
|
||||
tmp=vsetq_lane_s16(ext2[*pi5_p++],tmp,7);
|
||||
((int16x8_t *)systematic1)[i] = vqaddq_s16(vqsubq_s16(tmp,((int16x8_t*)ext)[i]),((int16x8_t *)systematic0)[i]);
|
||||
#endif
|
||||
#ifdef DEBUG_LOGMAP
|
||||
print_shorts("syst1",(int16_t*)&((__m128i *)systematic1)[i]);
|
||||
#endif
|
||||
}
|
||||
|
||||
@@ -1278,6 +1372,9 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
tmp=_mm_insert_epi16(tmp, ((llr_t*)ext2)[*pi6_p++],2);
|
||||
tmp=_mm_insert_epi16(tmp, ((llr_t*)ext2)[*pi6_p++],1);
|
||||
tmp=_mm_insert_epi16(tmp, ((llr_t*)ext2)[*pi6_p++],0);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
print_shorts("tmp",(int16_t*)&tmp);
|
||||
#endif
|
||||
tmp=_mm_cmpgt_epi8(_mm_packs_epi16(tmp,zeros),zeros);
|
||||
decoded_bytes[i]=(unsigned char)_mm_movemask_epi8(tmp);
|
||||
#elif defined(__arm__)
|
||||
@@ -1297,6 +1394,10 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
uint64x2_t Mask = vpaddlq_u32(vpaddlq_u16(vandq_u16(vcgtq_s16(tmp,zeros), Powers)));
|
||||
uint64x1_t Mask64 = vget_high_u64(Mask)+vget_low_u64(Mask);
|
||||
decoded_bytes[i] = (uint8_t)Mask64;
|
||||
#endif
|
||||
#ifdef DEBUG_LOGMAP
|
||||
print_shorts("tmp",(int16_t*)&tmp);
|
||||
fprintf(fdsse4,"decoded_bytes[%d] %x\n",i,decoded_bytes[i]);
|
||||
#endif
|
||||
}
|
||||
}
|
||||
@@ -1344,6 +1445,9 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
}
|
||||
|
||||
stop_meas(intl2_stats);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fprintf(fdsse4,"oldcrc %x, crc %x\n",oldcrc,crc);
|
||||
#endif
|
||||
|
||||
if ((crc == oldcrc) && (crc!=0)) {
|
||||
return(iteration_cnt);
|
||||
@@ -1374,8 +1478,12 @@ unsigned char phy_threegpplte_turbo_decoder16(short *y,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// fprintf(fdsse4,"crc %x, oldcrc %x\n",crc,oldcrc);
|
||||
|
||||
// printf("crc %x, oldcrc %x\n",crc,oldcrc);
|
||||
#ifdef DEBUG_LOGMAP
|
||||
fclose(fdsse4);
|
||||
#endif
|
||||
|
||||
#if defined(__x86_64__) || defined(__i386__)
|
||||
_mm_empty();
|
||||
|
||||
@@ -483,6 +483,24 @@ uint8_t phy_threegpplte_turbo_decoder16(int16_t *y,
|
||||
time_stats_t *intl1_stats,
|
||||
time_stats_t *intl2_stats);
|
||||
|
||||
uint8_t phy_threegpplte_turbo_decoder16avx2(int16_t *y,
|
||||
int16_t *y2,
|
||||
uint8_t *decoded_bytes,
|
||||
uint8_t *decoded_bytes2,
|
||||
uint16_t n,
|
||||
uint16_t interleaver_f1,
|
||||
uint16_t interleaver_f2,
|
||||
uint8_t max_iterations,
|
||||
uint8_t crc_type,
|
||||
uint8_t F,
|
||||
time_stats_t *init_stats,
|
||||
time_stats_t *alpha_stats,
|
||||
time_stats_t *beta_stats,
|
||||
time_stats_t *gamma_stats,
|
||||
time_stats_t *ext_stats,
|
||||
time_stats_t *intl1_stats,
|
||||
time_stats_t *intl2_stats);
|
||||
|
||||
/*!
|
||||
\brief This routine performs max-logmap detection for the 3GPP turbo code (with termination). It is optimized for SIMD processing and 8-bit
|
||||
LLR arithmetic, and requires SSE2,SSSE3 and SSE4.1 (gcc >=4.3 and appropriate CPU)
|
||||
|
||||
@@ -895,7 +895,9 @@ void phy_init_lte_top(LTE_DL_FRAME_PARMS *lte_frame_parms)
|
||||
|
||||
init_td8();
|
||||
init_td16();
|
||||
|
||||
#ifdef __AVX2__
|
||||
init_td16avx2();
|
||||
#endif
|
||||
|
||||
lte_sync_time_init(lte_frame_parms);
|
||||
|
||||
|
||||
@@ -26,187 +26,187 @@
|
||||
Address : Eurecom, Campus SophiaTech, 450 Route des Chappes, CS 50193 - 06904 Biot Sophia Antipolis cedex, FRANCE
|
||||
|
||||
*******************************************************************************/
|
||||
short filt24_0[24] __attribute__((aligned(16))) ={
|
||||
short filt24_0[24] __attribute__((aligned(32))) ={
|
||||
2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_0_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_0_dcl[24] __attribute__((aligned(32))) ={
|
||||
2341,4681,7022,9362,11703,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_0_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_0_dcr[24] __attribute__((aligned(32))) ={
|
||||
2730,5461,8192,10922,13653,16384,14043,11703,9362,7022,4681,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
short filt24_1[24] __attribute__((aligned(16))) ={
|
||||
short filt24_1[24] __attribute__((aligned(32))) ={
|
||||
0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_1_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_1_dcl[24] __attribute__((aligned(32))) ={
|
||||
0,4681,7022,9362,11703,14043,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_1_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_1_dcr[24] __attribute__((aligned(32))) ={
|
||||
0,2730,5461,8192,10922,13653,16384,14043,11703,9362,7022,4681,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
|
||||
short filt24_2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_2[24] __attribute__((aligned(32))) ={
|
||||
0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_2_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_2_dcl[24] __attribute__((aligned(32))) ={
|
||||
0,0,2341,4681,7022,9362, 11703,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_2_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_2_dcr[24] __attribute__((aligned(32))) ={
|
||||
0,0,2730,5461,8192,10922,13653,16384,14043,11703,9362,4681,2341,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
// X X X Y | X X X X | X Y X X
|
||||
short filt24_3[24] __attribute__((aligned(16))) ={
|
||||
short filt24_3[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_3_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_3_dcl[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,2341,4681,7022,9362,14043,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
// X X X Y | X X DC X X | X Y X X
|
||||
short filt24_3_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_3_dcr[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,2730,5461,8192,10922,13653,16384,14043,11703,7022,4681,2341,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
|
||||
short filt24_4[24] __attribute__((aligned(16))) ={
|
||||
short filt24_4[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_4_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_4_dcl[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,2341,7022,9362,11703,14043,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_4_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_4_dcr[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,2730,5461,8192,10922,13653,16384,14043,11703,7022,4681,2341,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
short filt24_5[24] __attribute__((aligned(16))) ={
|
||||
short filt24_5[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0
|
||||
};
|
||||
// X X X Y | X X DC X X | X Y X X
|
||||
short filt24_5_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_5_dcl[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,2341,4681,9362,11703,14043,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_5_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_5_dcr[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,2730,5461,8192,10922,13653,16384,11703,9362,7022,4681,2730,0,0,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
|
||||
short filt24_6[24] __attribute__((aligned(16))) ={
|
||||
short filt24_6[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_6_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_6_dcl[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,4681,7022,9362,11703,14043,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_6_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_6_dcr[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,14043,11703,9362,7022,4681,0,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
|
||||
short filt24_7[24] __attribute__((aligned(16))) ={
|
||||
short filt24_7[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_7_dcl[24] __attribute__((aligned(16))) ={
|
||||
short filt24_7_dcl[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,0,4681,7022,9362,11703,14043,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_7_dcr[24] __attribute__((aligned(16))) ={
|
||||
short filt24_7_dcr[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,14043,11703,9362,7022,4681,0,0,0,0,0,0
|
||||
};
|
||||
|
||||
short filt24_0l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_0l[24] __attribute__((aligned(32))) ={
|
||||
30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_1l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_1l[24] __attribute__((aligned(32))) ={
|
||||
0,30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_2l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_2l[24] __attribute__((aligned(32))) ={
|
||||
0,0,30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_3l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_3l[24] __attribute__((aligned(32))) ={
|
||||
//0,0,0,30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0};
|
||||
0,0,0,0,0,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_4l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_4l[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_5l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_5l[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_6l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_6l[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_7l[24] __attribute__((aligned(16))) ={
|
||||
short filt24_7l[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,0,30037,27306,24576,21845,19114,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_0l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_0l2[24] __attribute__((aligned(32))) ={
|
||||
2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_1l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_1l2[24] __attribute__((aligned(32))) ={
|
||||
0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_2l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_2l2[24] __attribute__((aligned(32))) ={
|
||||
-2730,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_3l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_3l2[24] __attribute__((aligned(32))) ={
|
||||
-5461,-2730,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_4l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_4l2[24] __attribute__((aligned(32))) ={
|
||||
-8192,-5461,-2730,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_5l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_5l2[24] __attribute__((aligned(32))) ={
|
||||
0,-8192,-5461,-2730,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_6l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_6l2[24] __attribute__((aligned(32))) ={
|
||||
-13653,-10922,-8192,-5461,-2730,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_7l2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_7l2[24] __attribute__((aligned(32))) ={
|
||||
0,-13653,-10922,-8192,-5461,-2730,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_0r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_0r[24] __attribute__((aligned(32))) ={
|
||||
2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_1r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_1r[24] __attribute__((aligned(32))) ={
|
||||
0,2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_2r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_2r[24] __attribute__((aligned(32))) ={
|
||||
0,0,2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_3r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_3r[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_4r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_4r[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_5r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_5r[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_6r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_6r[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_7r[24] __attribute__((aligned(16))) ={
|
||||
short filt24_7r[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,19114,21845,24576,27306,30037,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_0r2[24] __attribute__((aligned(16))) ={ /****/
|
||||
short filt24_0r2[24] __attribute__((aligned(32))) ={ /****/
|
||||
2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653,0,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_1r2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_1r2[24] __attribute__((aligned(32))) ={
|
||||
0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653,0,0,0,0,0,0
|
||||
};
|
||||
short filt24_2r2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_2r2[24] __attribute__((aligned(32))) ={
|
||||
0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653,0,0,0,0,0
|
||||
};
|
||||
short filt24_3r2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_3r2[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653,0,0,0,0
|
||||
};
|
||||
short filt24_4r2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_4r2[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653,0,0,0
|
||||
};
|
||||
short filt24_5r2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_5r2[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653,0,0
|
||||
};
|
||||
short filt24_6r2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_6r2[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653,0
|
||||
};
|
||||
short filt24_7r2[24] __attribute__((aligned(16))) ={
|
||||
short filt24_7r2[24] __attribute__((aligned(32))) ={
|
||||
0,0,0,0,0,0,0,2730,5461,8192,10922,13653,16384,13653,10922,8192,5461,2730,0,-2730,-5461,-8192,-10922,-13653
|
||||
};
|
||||
|
||||
@@ -52,8 +52,8 @@
|
||||
int* sync_corr_ue0 = NULL;
|
||||
int* sync_corr_ue1 = NULL;
|
||||
int* sync_corr_ue2 = NULL;
|
||||
int sync_tmp[2048*4] __attribute__((aligned(16)));
|
||||
short syncF_tmp[2048*2] __attribute__((aligned(16)));
|
||||
int sync_tmp[2048*4] __attribute__((aligned(32)));
|
||||
short syncF_tmp[2048*2] __attribute__((aligned(32)));
|
||||
|
||||
|
||||
|
||||
|
||||
@@ -56,8 +56,8 @@ void lte_sync_timefreq(PHY_VARS_UE *ue,int band,unsigned int DL_freq)
|
||||
{
|
||||
#if defined(__x86_64__) || defined(__i386__)
|
||||
UE_SCAN_INFO_t *scan_info = &ue->scan_info[band];
|
||||
int16_t spectrum[12288] __attribute__((aligned(16)));
|
||||
int16_t spectrum_p5ms[12288] __attribute__((aligned(16)));
|
||||
int16_t spectrum[12288] __attribute__((aligned(32)));
|
||||
int16_t spectrum_p5ms[12288] __attribute__((aligned(32)));
|
||||
int i,f,band_idx;
|
||||
__m128i autocorr0[256/4],autocorr1[256/4],autocorr2[256/4];
|
||||
__m128i autocorr0_t[256/4],autocorr1_t[256/4],autocorr2_t[256/4];
|
||||
|
||||
@@ -186,6 +186,27 @@ uint32_t dlsch_decoding(PHY_VARS_UE *phy_vars_ue,
|
||||
uint8_t crc_type;
|
||||
#ifdef DEBUG_DLSCH_DECODING
|
||||
uint16_t i;
|
||||
#endif
|
||||
#ifdef __AVX2__
|
||||
int Kr_last,skipped_last=0;
|
||||
uint8_t (*tc_2cw)(int16_t *y,
|
||||
int16_t *y2,
|
||||
uint8_t *,
|
||||
uint8_t *,
|
||||
uint16_t,
|
||||
uint16_t,
|
||||
uint16_t,
|
||||
uint8_t,
|
||||
uint8_t,
|
||||
uint8_t,
|
||||
time_stats_t *,
|
||||
time_stats_t *,
|
||||
time_stats_t *,
|
||||
time_stats_t *,
|
||||
time_stats_t *,
|
||||
time_stats_t *,
|
||||
time_stats_t *);
|
||||
|
||||
#endif
|
||||
uint8_t (*tc)(int16_t *y,
|
||||
uint8_t *,
|
||||
@@ -203,6 +224,9 @@ uint32_t dlsch_decoding(PHY_VARS_UE *phy_vars_ue,
|
||||
time_stats_t *,
|
||||
time_stats_t *);
|
||||
|
||||
|
||||
|
||||
|
||||
if (!dlsch_llr) {
|
||||
printf("dlsch_decoding.c: NULL dlsch_llr pointer\n");
|
||||
return(dlsch->max_turbo_iterations);
|
||||
@@ -223,8 +247,12 @@ uint32_t dlsch_decoding(PHY_VARS_UE *phy_vars_ue,
|
||||
return(dlsch->max_turbo_iterations);
|
||||
}
|
||||
|
||||
if (llr8_flag == 0)
|
||||
if (llr8_flag == 0) {
|
||||
#ifdef __AVX2__
|
||||
tc_2cw = phy_threegpplte_turbo_decoder16avx2;
|
||||
#endif
|
||||
tc = phy_threegpplte_turbo_decoder16;
|
||||
}
|
||||
else
|
||||
tc = phy_threegpplte_turbo_decoder8;
|
||||
|
||||
@@ -300,6 +328,10 @@ uint32_t dlsch_decoding(PHY_VARS_UE *phy_vars_ue,
|
||||
LOG_E(PHY,"Illegal harq_process->C %d > %d\n",harq_process->C,MAX_NUM_DLSCH_SEGMENTS/bw_scaling);
|
||||
return((1+dlsch->max_turbo_iterations));
|
||||
}
|
||||
#ifdef DEBUG_DLSCH_DECODING
|
||||
printf("Segmentation: C %d, Cminus %d, Kminus %d, Kplus %d\n",harq_process->C,harq_process->Cminus,harq_process->Kminus,harq_process->Kplus);
|
||||
#endif
|
||||
|
||||
for (r=0; r<harq_process->C; r++) {
|
||||
|
||||
|
||||
@@ -414,15 +446,11 @@ uint32_t dlsch_decoding(PHY_VARS_UE *phy_vars_ue,
|
||||
printf("\n");
|
||||
*/
|
||||
|
||||
#ifndef __AVX2__
|
||||
if (err_flag == 0) {
|
||||
|
||||
start_meas(dlsch_turbo_decoding_stats);
|
||||
#ifdef TURBO_S
|
||||
ret = phy_threegpplte_turbo_decoder_scalar
|
||||
#else
|
||||
|
||||
ret = tc
|
||||
#endif
|
||||
(&harq_process->d[r][96],
|
||||
harq_process->c[r],
|
||||
Kr,
|
||||
@@ -442,7 +470,130 @@ uint32_t dlsch_decoding(PHY_VARS_UE *phy_vars_ue,
|
||||
|
||||
stop_meas(dlsch_turbo_decoding_stats);
|
||||
}
|
||||
#else
|
||||
if ((harq_process->C == 1) ||
|
||||
((r==harq_process->C-1) && (skipped_last==0))) { // last segment with odd number of segments
|
||||
|
||||
start_meas(dlsch_turbo_decoding_stats);
|
||||
ret = tc
|
||||
(&harq_process->d[r][96],
|
||||
harq_process->c[r],
|
||||
Kr,
|
||||
f1f2mat_old[iind*2],
|
||||
f1f2mat_old[(iind*2)+1],
|
||||
dlsch->max_turbo_iterations,
|
||||
crc_type,
|
||||
(r==0) ? harq_process->F : 0,
|
||||
&phy_vars_ue->dlsch_tc_init_stats,
|
||||
&phy_vars_ue->dlsch_tc_alpha_stats,
|
||||
&phy_vars_ue->dlsch_tc_beta_stats,
|
||||
&phy_vars_ue->dlsch_tc_gamma_stats,
|
||||
&phy_vars_ue->dlsch_tc_ext_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl1_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl2_stats); //(is_crnti==0)?harq_pid:harq_pid+1);
|
||||
stop_meas(dlsch_turbo_decoding_stats);
|
||||
// printf("single decode, exit\n");
|
||||
// exit(-1);
|
||||
}
|
||||
else {
|
||||
// we can merge code segments
|
||||
if ((skipped_last == 0) && (r<harq_process->C-1)) {
|
||||
skipped_last = 1;
|
||||
Kr_last = Kr;
|
||||
}
|
||||
else {
|
||||
skipped_last=0;
|
||||
|
||||
if (Kr_last == Kr) { // decode 2 code segments with AVX2 version
|
||||
#ifdef DEBUG_DLSCH_DECODING
|
||||
printf("single decoding segment %d (%p)\n",r-1,&harq_process->d[r-1][96]);
|
||||
#endif
|
||||
start_meas(dlsch_turbo_decoding_stats);
|
||||
#ifdef DEBUG_DLSCH_DECODING
|
||||
printf("double decoding segments %d,%d (%p,%p)\n",r-1,r,&harq_process->d[r-1][96],&harq_process->d[r][96]);
|
||||
#endif
|
||||
ret = tc_2cw
|
||||
(&harq_process->d[r-1][96],
|
||||
&harq_process->d[r][96],
|
||||
harq_process->c[r-1],
|
||||
harq_process->c[r],
|
||||
Kr,
|
||||
f1f2mat_old[iind*2],
|
||||
f1f2mat_old[(iind*2)+1],
|
||||
dlsch->max_turbo_iterations,
|
||||
crc_type,
|
||||
(r==0) ? harq_process->F : 0,
|
||||
&phy_vars_ue->dlsch_tc_init_stats,
|
||||
&phy_vars_ue->dlsch_tc_alpha_stats,
|
||||
&phy_vars_ue->dlsch_tc_beta_stats,
|
||||
&phy_vars_ue->dlsch_tc_gamma_stats,
|
||||
&phy_vars_ue->dlsch_tc_ext_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl1_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl2_stats); //(is_crnti==0)?harq_pid:harq_pid+1);
|
||||
/*
|
||||
ret = tc
|
||||
(&harq_process->d[r-1][96],
|
||||
harq_process->c[r-1],
|
||||
Kr_last,
|
||||
f1f2mat_old[iind*2],
|
||||
f1f2mat_old[(iind*2)+1],
|
||||
dlsch->max_turbo_iterations,
|
||||
crc_type,
|
||||
(r==0) ? harq_process->F : 0,
|
||||
&phy_vars_ue->dlsch_tc_init_stats,
|
||||
&phy_vars_ue->dlsch_tc_alpha_stats,
|
||||
&phy_vars_ue->dlsch_tc_beta_stats,
|
||||
&phy_vars_ue->dlsch_tc_gamma_stats,
|
||||
&phy_vars_ue->dlsch_tc_ext_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl1_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl2_stats); //(is_crnti==0)?harq_pid:harq_pid+1);
|
||||
|
||||
exit(-1);*/
|
||||
stop_meas(dlsch_turbo_decoding_stats);
|
||||
}
|
||||
else { // Kr_last != Kr
|
||||
start_meas(dlsch_turbo_decoding_stats);
|
||||
ret = tc
|
||||
(&harq_process->d[r-1][96],
|
||||
harq_process->c[r-1],
|
||||
Kr_last,
|
||||
f1f2mat_old[iind*2],
|
||||
f1f2mat_old[(iind*2)+1],
|
||||
dlsch->max_turbo_iterations,
|
||||
crc_type,
|
||||
(r==0) ? harq_process->F : 0,
|
||||
&phy_vars_ue->dlsch_tc_init_stats,
|
||||
&phy_vars_ue->dlsch_tc_alpha_stats,
|
||||
&phy_vars_ue->dlsch_tc_beta_stats,
|
||||
&phy_vars_ue->dlsch_tc_gamma_stats,
|
||||
&phy_vars_ue->dlsch_tc_ext_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl1_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl2_stats); //(is_crnti==0)?harq_pid:harq_pid+1);
|
||||
stop_meas(dlsch_turbo_decoding_stats);
|
||||
|
||||
start_meas(dlsch_turbo_decoding_stats);
|
||||
ret = tc
|
||||
(&harq_process->d[r][96],
|
||||
harq_process->c[r],
|
||||
Kr,
|
||||
f1f2mat_old[iind*2],
|
||||
f1f2mat_old[(iind*2)+1],
|
||||
dlsch->max_turbo_iterations,
|
||||
crc_type,
|
||||
(r==0) ? harq_process->F : 0,
|
||||
&phy_vars_ue->dlsch_tc_init_stats,
|
||||
&phy_vars_ue->dlsch_tc_alpha_stats,
|
||||
&phy_vars_ue->dlsch_tc_beta_stats,
|
||||
&phy_vars_ue->dlsch_tc_gamma_stats,
|
||||
&phy_vars_ue->dlsch_tc_ext_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl1_stats,
|
||||
&phy_vars_ue->dlsch_tc_intl2_stats); //(is_crnti==0)?harq_pid:harq_pid+1);
|
||||
stop_meas(dlsch_turbo_decoding_stats);
|
||||
|
||||
}
|
||||
}
|
||||
}
|
||||
#endif
|
||||
|
||||
|
||||
if ((err_flag == 0) && (ret>=(1+dlsch->max_turbo_iterations))) {// a Code segment is in error so break;
|
||||
|
||||
@@ -79,9 +79,9 @@ extern int exit_openair;
|
||||
//extern void do_OFDM_mod(mod_sym_t **txdataF, int32_t **txdata, uint32_t frame, uint16_t next_slot, LTE_DL_FRAME_PARMS *frame_parms);
|
||||
|
||||
|
||||
unsigned char dlsch_input_buffer[2700] __attribute__ ((aligned(16)));
|
||||
int eNB_sync_buffer0[640*6] __attribute__ ((aligned(16)));
|
||||
int eNB_sync_buffer1[640*6] __attribute__ ((aligned(16)));
|
||||
unsigned char dlsch_input_buffer[2700] __attribute__ ((aligned(32)));
|
||||
int eNB_sync_buffer0[640*6] __attribute__ ((aligned(32)));
|
||||
int eNB_sync_buffer1[640*6] __attribute__ ((aligned(32)));
|
||||
int *eNB_sync_buffer[2] = {eNB_sync_buffer0, eNB_sync_buffer1};
|
||||
|
||||
extern uint16_t hundred_times_log10_NPRB[100];
|
||||
|
||||
Reference in New Issue
Block a user