Messages in this thread | | | From | Denys Vlasenko <> | Subject | Re: [PATCH 0/1] vsprintf: optimize decimal conversion (again) | Date | Wed, 28 Mar 2012 13:20:07 +0200 |
| |
On Wednesday 28 March 2012 12:39, roma1390 wrote: > On 2012.03.28 13:33, Denys Vlasenko wrote: > > On Wednesday 28 March 2012 12:24, roma1390 wrote: > >> On 2012.03.28 13:13, Denys Vlasenko wrote: > >>> Second: run > >>> arm-linux-gnueabi-gcc -O2 -Wall test_{org,new}.c -S > >>> and email me resulting test_{org,new}.s files. > >> > >> test_{org,new}.s attached. > > > > > > Bingo. > > > > bl __aeabi_uidivmod > > > > Not good. Your gcc did not optimize division by constant. > > > > Can you add "noinline_for_stack": > > > > static noinline_for_stack<=== HERE > > char *put_dec(char *buf, unsigned long long n) > > > > amd regenerate and resend the test_new.s? > > > > Hello, > > Your requested asm are attached.
Asm looks good: there are only three long multiplies in the put_dec + put_dec_full4, and no divisions at all.
Please speed-test this version.
-- vda
| |