mirrors/foot
1
0
Fork 0
mirror of https://codeberg.org/dnkl/foot.git synced 2026-03-22 05:33:45 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

install: re-write description of recommended flags, and add size comparison

Browse source
This commit is contained in:
Daniel Eklöf 2020-11-22 16:11:01 +01:00
parent e23ab4a7a3
commit af6d1aa522
No known key found for this signature in database
GPG key ID: 5BBD4992C116573F
1 changed files with 27 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

37
INSTALL.md
View file

@ -130,13 +130,20 @@ mkdir -p bld/release && cd bld/release
export CFLAGS="$CFLAGS -O3 -march=native" export CFLAGS="$CFLAGS -O3 -march=native"
meson --buildtype=release --prefix=/usr -Db_lto=true ../.. meson --buildtype=release --prefix=/usr -Db_lto=true ../..
``` ```
Both `-O3` and `-Db_lto=true` are **highly** recommended.
For performance reasons, I strongly recommend doing a To optimize for performance, both `-O3`, `-Db_lto=true` and doing a
[PGO](#profile-guided-optimization) (Profile Guided Optimization) [PGO](#profile-guided-optimization) (Profile Guided Optimization)
build. build is recommended.
If you do not want this, just build: If you instead want to optimize for size, use `-Os` instead of `-O3`.
In general, `-Os` results in the smallest (and slowest) binary. A full
PGO build will be slightly larger than a `-Os` build, but smaller than
a `-O3` build. A partial PGO build will be somewhere in between a full
PGO build and a `-O3` build. See the [size
comparison](#size-comparison).
If you are doing a non-PGO build, just build:
```sh ```sh
ninja ninja
@ -151,13 +158,8 @@ for example, uses `CFLAGS` to specify the default set of flags.
Thus, we do `export CFLAGS+="..."` to at least not throw away those Thus, we do `export CFLAGS+="..."` to at least not throw away those
flags. flags.
When packaging, you may want to use the default `CFLAGS` only, but
note this: foot is a performance critical application that relies on
compiler optimizations to perform well.
In particular, with GCC 10.1, it is **very** important `-O3` is used In particular, with GCC 10.1, it is **very** important `-O3` is used
(and not e.g. `-O2`) when doing a [PGO](#profile-guided-optimization) **when doing a [PGO](#profile-guided-optimization) build**.
build.
#### Profile Guided Optimization #### Profile Guided Optimization
@ -175,6 +177,12 @@ meson configure -Db_pgo=generate
ninja ninja
``` ```
Remainder: make sure `-O3` is being used.
To get **maximum** performance, you can also add
`-fprofile-partial-training`, but do note that this will result in a
noticeably larger binary (see [size comparison](#size-comparison))
Next, we need to actually generate the profiling data. Next, we need to actually generate the profiling data.
There are two ways to do this: a [partial PGO build using a PGO There are two ways to do this: a [partial PGO build using a PGO
@ -266,6 +274,15 @@ ninja
Continue reading in [Running the new build](#running-the-new-build) Continue reading in [Running the new build](#running-the-new-build)
##### Size comparison
| PGO | default | `-fprofile-partial-training`|
|-----------------|--------:|----------------------------:|
| none (`-O3`) | 340K | - |
| none (`-Os`) | 260K | - |
| partial | 328K | 376K |
| full | 292K | 352K |
### Debug build ### Debug build
```sh ```sh