cuberite/libdeflate
1
0
Fork 0
mirror of https://github.com/cuberite/libdeflate.git synced 2025-08-05 18:57:01 -04:00
Code Issues Packages Projects Releases Wiki Activity

Assume fast unaligned access on WebAssembly

Browse Source 
I saw this tweet claiming this flag makes libdeflate run 20% faster on
WebAssembly: https://twitter.com/Algunenano/status/1317098341377900550.

Indeed, when tried even in a complex PNG compression benchmark I've
observed 10-15% improvement when this flag is enabled.

Even though WebAssembly might be running on top of a variety of
underlying platforms, the spec requires it to support unaligned access,
and on majority of platforms it will translate to a faster code.

Hence, I think it makes sense to enable this flag by default.
This commit is contained in:
Ingvar Stepanyan 2021-01-13 13:51:23 +00:00 committed by Eric Biggers
parent e1803dc145
commit a07ed5824a
1 changed files with 10 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
common/compiler_gcc.h
View File

@ -175,7 +175,16 @@ typedef unsigned char __v32qu __attribute__((__vector_size__(32)));
# define bswap64 __builtin_bswap64 # define bswap64 __builtin_bswap64
#endif #endif
#if defined(__x86_64__) || defined(__i386__) || defined(__ARM_FEATURE_UNALIGNED) || defined(__powerpc64__) #if defined(__x86_64__) || defined(__i386__) || \
defined(__ARM_FEATURE_UNALIGNED) || defined(__powerpc64__) || \
/*
* For all compilation purposes, WebAssembly behaves like any other CPU
* instruction set. Even though WebAssembly engine might be running on top
* of different actual CPU architectures, the WebAssembly spec itself
* permits unaligned access and it will be fast on most of those platforms,
* and simulated at the engine level on others, so it's worth treating it
* as a CPU architecture with fast unaligned access.
*/ defined(__wasm__)
# define UNALIGNED_ACCESS_IS_FAST 1 # define UNALIGNED_ACCESS_IS_FAST 1
#endif #endif