Re: why not inline all functions?

I worked on compilers for Prime and Apollo in the 80's, and implemented
procedure inlining for both. For testing, I tried to inline
everything. I recall one program that seemingly "hung" the compiler.
Being persistent, I let it compile overnight. It finally did, and
executed properly. On inspection, the call graph for the code resembled
a binary tree, and quickly resulted in a huge program that took hours to
compile. The final released product tried to determine good candidates
for inlining (another topic). As I recall, we didn't expand recursive
routines, though we could have inlined a few times.


Large programs not only take a long time to compile, but may incurr
excessive paging, as there may be insufficient memory to hold the
executable. The big win obtained from inlining is due to optimization.
Inlining a procedure exposes the procedure's body to the surrounding
code, allowing for classical optimizations to take place. One of the
benchmarks in our test suite terminated with a divide-by-zero exception
after inlining. A matrix-multiplication function was inlined. The
optimizer could then determine that there was no use of the calculation,
so the code was eliminated. Thus, it took zero time to do the
calculation. The divide-by-zero error occurred when dividing elapsed
time into a constant. At any rate, we had qualms about reporting
benchmark figures for such a result.
--

Return to the comp.compilers page.
Search the comp.compilers archives again.