maki/openairinterface5g
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

First-level of support for new aarch64 machines

Browse Source 
Testing on Neoverse N1 (Ampere 3GHz). Tested only on a subset of phy-simulators.

- Changes:
- use SIMDE consistently
- adaptations of LDPC decoder generator for ARMv8 performance
- SIMDe modifications of Intel CRC to allow for aarch64 build.
  optimizations for 128-bit to improve performance of LDPC encode/decode
  on aarch64 (Neoverse 1)
- added BG2 files for 128-bit ldpc encoder (aarch64)
- testing on Xeon
- testing on x86
- minor changes to build/run on x86
- change in crc.h after returning to aarch64
- removed some warning in ldpc decoder generator for x86_64
- char
- Delete irrelevant constants
- Correctly declare variables
- Define stdbool for all architectures
- Remove definition of _MM_SHUFFLE and use SIMDE_MM_SHUFFLE
- Remove commented code
- Fix CMakeLists.txt
- Include SIMDE avx2 functions in tools_defs.h
This commit is contained in:
Raymond Knopp
2023-09-13 17:19:20 +02:00
committed by Robert Schmidt
parent 79fd37b489
commit a7bc282302
168 changed files with 30595 additions and 21476 deletions
Download Patch File Download Diff File Expand all files Collapse all files
openair1/PHY/LTE_ESTIMATION/freq_equalization.c
+13 -30
View File
@@ -23,8 +23,8 @@
#include "PHY/sse_intrin.h"
#include "PHY/LTE_ESTIMATION/lte_estimation.h"
// This is 4096/(1:4096) in __m128i format
__m128i inv_ch[4096];/* = {512,512,512,512,512,512,512,512,
// This is 4096/(1:4096) in simde__m128i format
simde__m128i inv_ch[4096];/* = {512,512,512,512,512,512,512,512,
256,256,256,256,256,256,256,256,
170,170,170,170,170,170,170,170,
128,128,128,128,128,128,128,128,
@@ -283,7 +283,7 @@ __m128i inv_ch[4096];/* = {512,512,512,512,512,512,512,512,
};*/
void init_fde() {
for (int i=1;i<4096;i++) inv_ch[i] = _mm_set1_epi16(4096/i);
for (int i=1;i<4096;i++) inv_ch[i] = simde_mm_set1_epi16(4096/i);
}
void freq_equalization(LTE_DL_FRAME_PARMS *frame_parms,
@@ -296,17 +296,10 @@ void freq_equalization(LTE_DL_FRAME_PARMS *frame_parms,
{
uint16_t re;
int16_t amp;
#if defined(__x86_64__) || defined(__i386__)
__m128i *ul_ch_mag128,*ul_ch_magb128,*rxdataF_comp128;
rxdataF_comp128 = (__m128i *)&rxdataF_comp[0][symbol*frame_parms->N_RB_DL*12];
ul_ch_mag128 = (__m128i *)&ul_ch_mag[0][symbol*frame_parms->N_RB_DL*12];
ul_ch_magb128 = (__m128i *)&ul_ch_magb[0][symbol*frame_parms->N_RB_DL*12];
#elif defined(__arm__) || defined(__aarch64__)
int16x8_t *ul_ch_mag128,*ul_ch_magb128,*rxdataF_comp128;
rxdataF_comp128 = (int16x8_t*)&rxdataF_comp[0][symbol*frame_parms->N_RB_DL*12];
ul_ch_mag128 = (int16x8_t*)&ul_ch_mag[0][symbol*frame_parms->N_RB_DL*12];
ul_ch_magb128 = (int16x8_t*)&ul_ch_magb[0][symbol*frame_parms->N_RB_DL*12];
#endif
simde__m128i *ul_ch_mag128,*ul_ch_magb128,*rxdataF_comp128;
rxdataF_comp128 = (simde__m128i *)&rxdataF_comp[0][symbol*frame_parms->N_RB_DL*12];
ul_ch_mag128 = (simde__m128i *)&ul_ch_mag[0][symbol*frame_parms->N_RB_DL*12];
ul_ch_magb128 = (simde__m128i *)&ul_ch_magb[0][symbol*frame_parms->N_RB_DL*12];
AssertFatal(symbol<frame_parms->symbols_per_tti,"symbol %d >= %d\n",
symbol,frame_parms->symbols_per_tti);
@@ -320,26 +313,16 @@ void freq_equalization(LTE_DL_FRAME_PARMS *frame_parms,
if (amp>4095)
amp=4095;
//printf("freq_eq: symbol %d re %d => mag %d,amp %d,inv %d, prod %d (%d,%d)\n",symbol,re,*((int16_t*)(&ul_ch_mag128[re])),amp,_mm_extract_epi16(inv_ch[amp],0),(*((int16_t*)(&ul_ch_mag128[re]))*_mm_extract_epi16(inv_ch[amp],0))>>3,*(int16_t*)&(rxdataF_comp128[re]),*(1+(int16_t*)&(rxdataF_comp128[re])));
#if defined(__x86_64__) || defined(__i386__)
rxdataF_comp128[re] = _mm_srai_epi16(_mm_mullo_epi16(rxdataF_comp128[re],inv_ch[amp]),3);
// printf("freq_eq: symbol %d re %d => mag %d,amp %d,inv %d, prod %d
// (%d,%d)\n",symbol,re,*((int16_t*)(&ul_ch_mag128[re])),amp,simde_mm_extract_epi16(inv_ch[amp],0),(*((int16_t*)(&ul_ch_mag128[re]))*_mm_extract_epi16(inv_ch[amp],0))>>3,*(int16_t*)&(rxdataF_comp128[re]),*(1+(int16_t*)&(rxdataF_comp128[re])));
rxdataF_comp128[re] = simde_mm_srai_epi16(simde_mm_mullo_epi16(rxdataF_comp128[re],inv_ch[amp]),3);
if (Qm==4)
ul_ch_mag128[re] = _mm_set1_epi16(324); // this is 512*2/sqrt(10)
ul_ch_mag128[re] = simde_mm_set1_epi16(324); // this is 512*2/sqrt(10)
else {
ul_ch_mag128[re] = _mm_set1_epi16(316); // this is 512*4/sqrt(42)
ul_ch_magb128[re] = _mm_set1_epi16(158); // this is 512*2/sqrt(42)
ul_ch_mag128[re] = simde_mm_set1_epi16(316); // this is 512*4/sqrt(42)
ul_ch_magb128[re] = simde_mm_set1_epi16(158); // this is 512*2/sqrt(42)
}
#elif defined(__arm__) || defined(__aarch64__)
rxdataF_comp128[re] = vmulq_s16(rxdataF_comp128[re],inv_ch[amp]);
if (Qm==4)
ul_ch_mag128[re] = vdupq_n_s16(324); // this is 512*2/sqrt(10)
else {
ul_ch_mag128[re] = vdupq_n_s16(316); // this is 512*4/sqrt(42)
ul_ch_magb128[re] = vdupq_n_s16(158); // this is 512*2/sqrt(42)
}
#endif
// printf("(%d,%d)\n",*(int16_t*)&(rxdataF_comp128[re]),*(1+(int16_t*)&(rxdataF_comp128[re])));
}