V100 Execution unit configuration

[Moiralex]

Hello,
In the QV100 and GV100 tested configurations it is mentioned that: "Volta GV100 has 4 SP SIMD units, 4 SFU units, 4 DP units per core, 4 Tensor core units". Knowing that in GPGPU-sim those are modeled as 32-wide SIMD units they correspond to 128 scalar units of each type per SM. However in the Volta whitepaper it is mentioned that each SM features 64 int32, 64 fp32 and 32 fp64 cores. In deprecated models such as the GTX480 you mention that you halve the clock speed in order to mimic the half-warp-wide real execution units, however this does not seem to be the case here. Additionally in the provided configuration each SM has an equal number of SP and DP units despite the whitepaper clearly stating there exist half as many of the latter. What am i getting wrong here?
Thanks in advance for your answer!