Hi,
this patch updates costs of integer divides to match actual latencies (the
scheduler model already does the right thing). It is essentially no-op, since
we end up expanding idiv for all sensible constants, so this only may end
up disabling vectorization in some cases, but I did not find any such examples.
However in general it is better ot have actual latencies than random numbers.
Bootstrapped/regtested x86_64-linux, commited.
Honza
gcc/ChangeLog:
2021-03-18 Jan Hubicka <hubi...@ucw.cz>
* config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
integer divides1.
diff --git a/gcc/config/i386/x86-tune-costs.h b/gcc/config/i386/x86-tune-costs.h
index db03738313e..58b3b81985b 100644
--- a/gcc/config/i386/x86-tune-costs.h
+++ b/gcc/config/i386/x86-tune-costs.h
@@ -1741,13 +1741,11 @@ struct processor_costs znver3_cost = {
COSTS_N_INSNS (3)}, /* other. */
0, /* cost of multiply per each bit
set. */
- /* Depending on parameters, idiv can get faster on ryzen. This is upper
- bound. */
- {COSTS_N_INSNS (16), /* cost of a divide/mod for QI. */
- COSTS_N_INSNS (22), /* HI. */
- COSTS_N_INSNS (30), /* SI. */
- COSTS_N_INSNS (45), /* DI. */
- COSTS_N_INSNS (45)}, /*
other. */
+ {COSTS_N_INSNS (9), /* cost of a divide/mod for QI. */
+ COSTS_N_INSNS (10), /* HI. */
+ COSTS_N_INSNS (12), /* SI. */
+ COSTS_N_INSNS (17), /* DI. */
+ COSTS_N_INSNS (17)}, /*
other. */
COSTS_N_INSNS (1), /* cost of movsx. */
COSTS_N_INSNS (1), /* cost of movzx. */
8, /* "large" insn. */
