Using new homogeneous LFs in the FUV and in the FIR Herschel/PEP and Herschel/HerMES, we study the evolution of the dust attenuation with redshift. With this information in hand, we are able to estimate the redshift evolution of the total (FUV + FIR) star formation rate density SFRD_TOT. By integrating SFRD_TOT, we follow the mass building and analyze the redshift evolution of the stellar mass density (SMD). This letter aims at providing a complete view of star formation from the local universe to z = 4 and, using assumptions on earlier star formation history, compares this evolution to what was known before in an attempt to draw a homogeneous picture of the global evolution of star formation in galaxies. The main conclusions of this letter are: 1) the dust attenuation A_FUV is found to increase from z = 0 to z sim 1.2 and then starts to decrease up to our last data point at z = 3.6; 2) the estimated SFRD confirms published results up to z = 2. At z > 2, we observe either a plateau or a small increase up to z = 3 and then a likely decrease up to z = 3.6; 3) the peak of A_FUV is delayed with respect to the plateau of SFRD_TOT and a likely origin might be found in the evolution of the bright ends of the FUV and FIR LFs; 4) using assumptions (namely exponential rise and linear rise with time) for the evolution of the star formation density from z = 3.6 to z_form = 10, we integrate SFRD_TOT and find a good agreement with the published SMDs.