This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Commit: RL78: Improve multiplication cost estimate
- From: Nick Clifton <nickc at redhat dot com>
- To: dj at redhat dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 06 Oct 2015 14:41:29 +0100
- Subject: Commit: RL78: Improve multiplication cost estimate
- Authentication-results: sourceware.org; auth=none
Hi DJ,
This is the patch that I have checked in.
Cheers
Nick
gcc/ChangeLog
2015-10-06 Nick Clifton <nickc@redhat.com>
* config/rl78/rl78.c (rl78_rtx_costs): Improve cost estimates for
multiplication.
Index: gcc/config/rl78/rl78.c
===================================================================
--- gcc/config/rl78/rl78.c (revision 228513)
+++ gcc/config/rl78/rl78.c (working copy)
@@ -4165,12 +4165,31 @@
return true;
}
+ if (mode == HImode)
+ {
+ if (code == MULT && ! speed)
+ {
+ * total = COSTS_N_INSNS (8);
+ return true;
+ }
+ return false;
+ }
+
if (mode == SImode)
{
switch (code)
{
case MULT:
- if (RL78_MUL_G14)
+ if (! speed)
+ /* If we are compiling for space then we do not want to use the
+ inline SImode multiplication patterns or shift sequences.
+ The cost is not set to 1 or 5 however as we have to allow for
+ the possibility that we might be converting a leaf function
+ into a non-leaf function. (There is no way to tell here).
+ A value of 13 seems to be a reasonable compromise for the
+ moment. */
+ * total = COSTS_N_INSNS (13);
+ else if (RL78_MUL_G14)
*total = COSTS_N_INSNS (14);
else if (RL78_MUL_G13)
*total = COSTS_N_INSNS (29);