<!-- Strictly, the latency-hiding is a feature of the zero-overhead scheduling implemented by modern GPUs... this might or might not be considered to be a property of 'SIMT' itself -->
A downside of SIMT execution is the fact that controlthread-specific control-flow has to be simulatedperformed using "masking:", whenleading to poor utilisation where control-flow is not coherent for all threads of a processor. hitsFor instance, to handle an ''if''-''then''-''else'' block, anditswhere various threads of a processor execute the different paths though the block, all threads must actually passprocess throughboth paths (as all threads of thea blockprocessor always execute in lock-step), but for"masking" processorsis thatused hitto thedisable ''if''and partenable the ''else''various partthreads isas "maskedappropriate. This out"masking" strategy is what distinguishes SIMT from ordinary SIMD, and vicehas versa. Athe benefit of itinexpensive thissynchronization isbetween inexpensivethe synchronizationthreads of a processor.<ref name="spp">{{cite book |author1=Michael McCool |author2=James Reinders |author3=Arch Robison |title=Structured Parallel Programming: Patterns for Efficient Computation |publisher=Elsevier |year=2013 |pages=209 ff.}}</ref>
