Conceptually, the concept of "target-independent RTL" seems weird to me.
But I guess you expect these to work because the backend never gets
involved trying to recognize insns or addressing modes? I think a
comment to that effect somewhere near the C code to load these tests
would be good.
I think this potentially suggests a missing feature: the dump should
indicate whether the reader should leave it alone or ask the target to
verify it by recognizing the insns.
Also,
+ (edge-from entry (flags "FALLTHRU"))
+ (cinsn 1 (set (reg:DI %2)
+ (lshiftrt:DI (reg:DI %0)
+ (const_int 32)))
+ "../../src/gcc/testsuite/gcc.dg/asr_div1.c":14
Weren't we going to make file/line information optional?
Otherwise this is probably going to be OK since I got past my "this is
all wrong" initial reaction. You'll want to Cc target maintainers for
the other two patches.
Bernd
