Skip to content

Changes

Page history
Update ShiftAnalysis authored by Chris Cox's avatar Chris Cox
Show whitespace changes
Inline Side-by-side
ShiftAnalysis.md
View page @ 790ab0be
...@@ -165,19 +165,42 @@ number time per second test0 ...@@ -165,19 +165,42 @@ number time per second test0
14 "int64_t variable mask low" 16.45 sec 2918.53 M 11.20 14 "int64_t variable mask low" 16.45 sec 2918.53 M 11.20
15 "int64_t variable mask low by shift" 26.89 sec 1784.96 M 18.31 15 "int64_t variable mask low by shift" 26.89 sec 1784.96 M 18.31
``` ```
* Wow, the variable mask low operations are NOT being optimized correctly! * Wow, the variable mask low operations are not being optimized correctly!
* How the heck are the unsigned mask operations slower? * How the heck are the unsigned mask operations so much slower?
### Constant Mask High ### Constant Mask High
``` ```
test description absolute operations ratio with
number time per second test0
0 "uint8_t constant mask high" 1.06 sec 45481.12 M 1.00
1 "uint8_t constant mask high by shift" 1.05 sec 45609.90 M 1.00
2 "uint16_t constant mask high" 2.10 sec 22910.38 M 1.99
3 "uint16_t constant mask high by shift" 2.09 sec 22994.96 M 1.98
4 "uint32_t constant mask high" 5.86 sec 8186.49 M 5.56
5 "uint32_t constant mask high by shift" 5.86 sec 8186.69 M 5.56
6 "uint64_t constant mask high" 16.45 sec 2918.43 M 15.58
7 "uint64_t constant mask high by shift" 16.43 sec 2920.61 M 15.57
``` ```
* * These look good.
### Variable Mask High ### Variable Mask High
``` ```
test description absolute operations ratio with
number time per second test0
0 "uint8_t variable mask high" 1.06 sec 45269.22 M 1.00
1 "uint8_t variable mask high by shift" 24.66 sec 1946.68 M 23.25
2 "uint16_t variable mask high" 2.09 sec 22933.16 M 1.97
3 "uint16_t variable mask high by shift" 24.69 sec 1943.74 M 23.29
4 "uint32_t variable mask high" 5.88 sec 8166.26 M 5.54
5 "uint32_t variable mask high by shift" 25.43 sec 1887.30 M 23.99
6 "uint64_t variable mask high" 16.61 sec 2889.74 M 15.67
7 "uint64_t variable mask high by shift" 27.18 sec 1766.01 M 25.63
``` ```
* * And the variable mask high operations are unoptimized, just like the mask low case.
------- -------
...@@ -215,4 +238,3 @@ number time per second test0 ...@@ -215,4 +238,3 @@ number time per second test0