Return the sum of the squares of the differences of the two input matrices

this is only necessary when no initialization from within a kernel blob happens

nonstandard API: *this is emptied. Also see "A Standard flat_map" https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p0429r9.pdf

identify the local size values by the store to it

Attempts to find x in node's dependency graph

if the kernel contained a loop, it is possible, that values inside a single subcfg don't dominate their uses inside the same subcfg. This function ide

* Convert a 32-bit floating-point number in IEEE single-precision format to a 16-bit floating-point number in * IEEE half-precision format, in bit re

* Convert a 16-bit floating-point number in IEEE half-precision format, in bit representation, to * a 32-bit floating-point number in IEEE single-pre

Generate a function with the following definition: void kernel(args) { if(__acpp_sscp_is_device) original_kernel(args) else { /* Perform host kernel

create the actual while loop around the subcfgs and the switch instruction to select the next subCFG based on the value in \a LastBarrierIdStorage

look through geps and bitcasts to find the alloca

Partially taken from CGCUDANV.cpp

searches for llvm.var.annotation and returns the value that is annotated by it, as well the annotation instruction

calculate uniformity analysis