https://gcc.gnu.org/g:6753df43a0251cffbfe2fca85e978807f976fb01
commit r15-9221-g6753df43a0251cffbfe2fca85e978807f976fb01
Author: Georg-Johann Lay <a...@gjlay.de>
Date: Fri Apr 4 22:01:43 2025 +0200
AVR: Speed up __umulhisi3 for small devices with MUL.
__umulhisi3 had an "rcall 1f" to save 6 bytes, which is an unreasonable
size gain vs. cycle cost. Just use the same code on all devices with MUL,
irrespective of program memory size.
libgcc/
* config/avr/lib1funcs.S (__umulhisi3) [Have MUL]: Reduce call
depth by 1.
Diff:
---
libgcc/config/avr/lib1funcs.S | 9 +--------
1 file changed, 1 insertion(+), 8 deletions(-)
diff --git a/libgcc/config/avr/lib1funcs.S b/libgcc/config/avr/lib1funcs.S
index 96f20ca8e646..01208927c6b0 100644
--- a/libgcc/config/avr/lib1funcs.S
+++ b/libgcc/config/avr/lib1funcs.S
@@ -581,19 +581,12 @@ DEFUN __umulhisi3
mul A1, B1
movw C2, r0
mul A0, B1
-#ifdef __AVR_HAVE_JMP_CALL__
- ;; This function is used by many other routines, often multiple times.
- ;; Therefore, if the flash size is not too limited, avoid the RCALL
- ;; and inverst 6 Bytes to speed things up.
add C1, r0
adc C2, r1
clr __zero_reg__
adc C3, __zero_reg__
-#else
- rcall 1f
-#endif
mul A1, B0
-1: add C1, r0
+ add C1, r0
adc C2, r1
clr __zero_reg__
adc C3, __zero_reg__
