Jan/WLED_MM_Infinity
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

usre faster math for _S builds

Browse Source 
similar to -ffast-math, but still supports NaN and infinites (needed for sunrise/sunset calculations)

10%-20% faster on -S3
This commit is contained in:
Frank
2024-10-27 19:02:05 +01:00
parent e652bd7064
commit 422fdb134a
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
platformio.ini
View File

@@ -1270,7 +1270,9 @@ build_flags =
build_flags = build_flags =
-O2 ;; optimize for performance instead of size -O2 ;; optimize for performance instead of size
;-ffast-math ;; gives a few (2-5) percent speedup on ESP32-S3, but causes slight slowdown on classic ESP32 ;-ffast-math ;; gives a few (2-5) percent speedup on ESP32-S3, but causes slight slowdown on classic ESP32
-mtarget-align -free -fipa-pta ;; these are very useful, too -funsafe-math-optimizations -fsingle-precision-constant ;; less dangerous than -ffast-math; still allows the compiler to exploit FMA and reciprocals (10% faster on -S3)
;; -mtarget-align ;; this one is included in -O2, so its not necessary to explicitly add it
-free -fipa-pta ;; these are very useful, too
-fno-jump-tables -fno-tree-switch-conversion ;; needed -fno-jump-tables -fno-tree-switch-conversion ;; needed
-freorder-blocks -Wwrite-strings -fstrict-volatile-bitfields ;; needed -freorder-blocks -Wwrite-strings -fstrict-volatile-bitfields ;; needed
build_unflags = build_unflags =