avg_0[0] = avg_0[0] + avg_0[1];
  avg_1[0] = avg_1[0] + avg_1[1];
  avg_0[0] = max (avg_0[0], avg_1[0]);
  avg_1[0] = avg_0[0];

avg_1[0] = max(avg_1[0],avg_1[1]);

avg_1[0] = max(avg_0[1],avg_1[1]); taking this into account in correlation.

2) Fixing bug in mrc_TM4 - removing loop over tx antennas.

Fixing bug in passing arguments to dlsch_dual_stream_correlation, introducing dlsch_dual_stream_correlation_opposite

Revert "Experimanetal change in dlsch_channel_commensation_TM34. Will not work for more than 2 antennas."

This reverts commit cbd30e62.

This reverts commit 83324945, reversing
changes made to cbd30e62.

Testing.

Changing names of output files in dlsim respectiively.

Conflicts:
	openair1/PHY/TOOLS/cdot_prod.c
	targets/PROJECTS/GENERIC-LTE-EPC/CONF/enb.band7.usrpb210.conf

1. Fixing a bug in dlsch_channel_level_TM56.
avg128D = _mm_setzero_si128() must be done inside the rx_antennas loop and not before

2. Cleaning up code in dlsch_demodulation.c

3. Offsets for TM5 and TM6 still must be validated.

4. Right now IA and I-UA receivers for TM4 have the same offset -
validate if any additional penalty/gain is needed.

5. Shortened file names in dlsim.c since matlab can read only 63-characters file names.

6. Tested: TM4 U2 -gS perfect CE works fine for EESM and MIESM abstraction (MCS 4 MSE= 0.007 dB on both streams).

Validating mcs-dependent shift.

Passing 2 avrage channel lvel and not max how it was done before.
Continuing with calibration.

TM4 and verify if any additional gain/penatly is needed.

Going back to fixed shift in log2_approx for TM4. Playing with computing the channel average in order to find optimal shift.