Algorithms/go-perfbook
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

more notes

Browse Source
This commit is contained in:
Damian Gryski 2018-01-04 08:04:51 -08:00
parent 9a32443804
commit cf2414f5c6
1 changed files with 14 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
performance.md
View File

@ -121,6 +121,9 @@ memory usage? What are you willing to give up in exchange for lower space?
Anything that can be measured can be optimized. Make sure you're measuring Anything that can be measured can be optimized. Make sure you're measuring
the right thing. Beware bad metrics. There are generally competing factors. the right thing. Beware bad metrics. There are generally competing factors.
Good performance work requires knowledge at many different levels, from
system design, networking, hardware (CPU, caches, storage).
This book is mostly going to talk about reducing CPU usage, reducing memory This book is mostly going to talk about reducing CPU usage, reducing memory
usage, and reducing latency. It's good to point out that you can very rarely usage, and reducing latency. It's good to point out that you can very rarely
do all three. Maybe CPU time is faster, but now your program uses more do all three. Maybe CPU time is faster, but now your program uses more
@ -156,6 +159,13 @@ Basic techniques:
But the engineering approach is correct: But the engineering approach is correct:
Benchmark. Analyze. Improve. Verify. Iterate. Benchmark. Analyze. Improve. Verify. Iterate.
Augment your data structure with more information:
- precomputed fields (size, etc)
- extra indexes for searching, "search fingers"
- limitations of when this is applicable:
must be cheap to keep updated
- all these fall under "do less work" (at the data structure level)
Trade space for time: Trade space for time:
- smaller data structures: pack things, compress data structures in memory - smaller data structures: pack things, compress data structures in memory
- precompute things you need (size of a linked list) - precompute things you need (size of a linked list)
@ -301,7 +311,7 @@ Techniques applicable to source code in general
* What causes heap allocations? * What causes heap allocations?
* Understanding escape analysis * Understanding escape analysis
* API design to limit allocations: allow passing in buffers so caller can reuse rather than forcing an allocation * API design to limit allocations: allow passing in buffers so caller can reuse rather than forcing an allocation
* reducing pointers * reducing pointers to reduce gc scan times
## Runtime ## Runtime
* cost of calls via interfaces (indirect calls on the CPU level) * cost of calls via interfaces (indirect calls on the CPU level)
@ -324,8 +334,8 @@ Techniques applicable to source code in general
## cgo ## cgo
* Performance characteristics of cgo calls * Performance characteristics of cgo calls
* Tricks to reduce the costs * Tricks to reduce the costs: batching
* Passing pointers between Go and C * Rules on passing pointers between Go and C
* syso files * syso files
## Assembly ## Assembly
@ -334,6 +344,7 @@ Techniques applicable to source code in general
* calling convention * calling convention
* using opcodes unsupported by the asm * using opcodes unsupported by the asm
* notes about why intrinsics are hard * notes about why intrinsics are hard
* all the tooling to make this easier: asmfmt, peachpy, c2goasm, ...
## Alternate implementations ## Alternate implementations
* Popular replacements for standard library packages: * Popular replacements for standard library packages: