Merging CPU cores

I have an idea to increase single-thread performance.
CPU cores should be merged into one macrocore which give any core any available resources.
Example: you want to multiply 100 vectors in one non-multithreaded function.
Normally:one core will execute the operations using its own resources.
With macrocore:the multiplication will be assigned to all available SIMD/FPU/ALU units in the macrocore(all core units).

>>23
Interesting, what if >>7 is combined with >>23 and >>1
1.perfect clock gating(minimum TDP)
2.asynchronous CPU running at the speed of hardware.
3.Use all cores as unified resources.
4.30 Instructions per cycle -> as many instructions "per cycle" as hardware can handle
5.No register renaming.