@guidoism It's all in the implementation. The projects I've mentioned have been designed to occupy a lot of real estate on an #FPGA. I'm looking forward to seeing how compact this #VLIW design will become, and if I can reduce the size footprint by an estimated two orders of magnitude.

One of the first big insights from my #VLIW #GPU journey is the relative size of VLIW cores compared to a vector or SIMT implementation. If you look at the Vortex GPGPU from Georgia Tech, or the Nyuzi GPGPU from Jeff Bush, the core sizes of those projects encroach ~100,000 LUTs/logic elements. On the other hand, a VLIW-style core is closer to ~3000 LUTs/logic elements. This will hopefully enable a degree of rapid prototyping on #FPGA using open source tools.

Currently working on a VLIW-SIMD #GPU in support of @mntmn 's open-source hardware efforts. I'm also learning that the #VLIW architecture can solve some DSP-related problems in the open-source hardware space (i.e. cellular modems). Looking forward to seeing how this pans out. Wish me luck!

Could someone confirm/infirm that statement I saw recently:
"""
By design #VLIW processors like #TI #C6x #DSP or #HP #Intel #Itanium are immune to #Spectre
"""