Messages in this thread | | | Date | Thu, 29 Mar 2012 08:23:44 +0300 | From | roma1390 <> | Subject | Re: [PATCH 0/1] vsprintf: optimize decimal conversion (again) |
| |
On 2012.03.28 14:23, Denys Vlasenko wrote: > It looks ok - it is a bit faster that original code; > but I expect algorithm #2 will do better. > > BTW, please always run test of original code too when you measure speed, > and post both results.
And the test results are:
./test_new-org-with-noinline_for_stack Conversions per second: 8:5188000 123:4340000 123456:3536000 12345678:3164000 123456789:2036000 2^32:2032000 2^64:1544000 Conversions per second: 8:5016000 123:4320000 123456:3536000 12345678:3272000 123456789:2032000 2^32:1964000 2^64:1540000 Conversions per second: 8:5192000 123:4336000 123456:3420000 12345678:3268000 123456789:2036000 2^32:2032000 2^64:1544000 Conversions per second: 8:5164000 123:4336000 123456:3536000 12345678:3272000 123456789:1968000 2^32:2032000 2^64:1544000 Tested 524288 ^C
# ./test_new-org-with-noinline_for_stack_if-LONG_MAX_header_arm Conversions per second: 8:5020000 123:4332000 123456:3536000 12345678:3272000 123456789:2828000 2^32:2260000 2^64:1536000 Conversions per second: 8:5192000 123:4332000 123456:3416000 12345678:3252000 123456789:2828000 2^32:2336000 2^64:1540000 Conversions per second: 8:5172000 123:4336000 123456:3540000 12345678:3272000 123456789:2736000 2^32:2336000 2^64:1536000 Conversions per second: 8:5192000 123:4192000 123456:3520000 12345678:3276000 123456789:2828000 2^32:2340000 2^64:1484000 Tested 524288 ^C
roma1390
| |