control-flowThe loop body branches

Reported by --explain after a verdict.

mettle explain control-flow

The body has a nested loop or a data-dependent if. Lanes execute in lockstep, so a branch that goes one way for lane 3 and the other way for lane 4 has no single instruction to be.

Note the compiler already if-converts short branches into masked arithmetic where it can (see if-converted). This code fires when the branch was too big or too effectful for that.

Fixes

• hoist the condition out when it does not depend on the index

• replace a small if with arithmetic: x = c * a + (1 - c) * b

• split one loop into two, each with a straight-line body

Related

• call-in-bodyThe loop body calls a function
• extern-call-in-bodyThe loop body calls an external function
• indirect-callThe loop body calls through a function pointer
• alloc-in-bodyThe loop body allocates
• inline-asmThe loop body contains inline assembly
• early-exitThe loop can leave before its trip count
• int16-elements16-bit integer elements have no kernel
• int64-elements64-bit integer elements have no kernel