ShuriZma
/
duckstation
mirror of https://github.com/stenzek/duckstation
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Build: Fix Android armv8 performance regression

This commit is contained in:
Connor McLaughlin 2020-11-22 02:00:16 +10:00
parent e528f3f273
commit b49067d165
2 changed files with 3 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
android/app/build.gradle
View File

@ -30,7 +30,7 @@ android {
defaultConfig { defaultConfig {
externalNativeBuild { externalNativeBuild {
cmake { cmake {
arguments "-DCMAKE_BUILD_TYPE=Release -DANDROID_ARM_NEON=ON" arguments "-DCMAKE_BUILD_TYPE=Release"
abiFilters "arm64-v8a", "armeabi-v7a" abiFilters "arm64-v8a", "armeabi-v7a"
} }
} }

6
src/core/gpu_sw.cpp
View File

@ -11,8 +11,6 @@ Log_SetChannel(GPU_SW);
#if defined(CPU_X64) #if defined(CPU_X64)
#include <emmintrin.h> #include <emmintrin.h>
#elif defined(CPU_AARCH32)
#include <arm_neon.h>
#elif defined(CPU_AARCH64) #elif defined(CPU_AARCH64)
#ifdef _MSC_VER #ifdef _MSC_VER
#include <arm64_neon.h> #include <arm64_neon.h>
@ -157,7 +155,7 @@ ALWAYS_INLINE void CopyOutRow16<HostDisplayPixelFormat::RGBA5551, u16>(const u16
_mm_storeu_si128(reinterpret_cast<__m128i*>(dst_ptr), value); _mm_storeu_si128(reinterpret_cast<__m128i*>(dst_ptr), value);
dst_ptr += 8; dst_ptr += 8;
} }
#elif defined(CPU_AARCH32) || defined(CPU_AARCH64) #elif defined(CPU_AARCH64)
const u32 aligned_width = Common::AlignDownPow2(width, 8); const u32 aligned_width = Common::AlignDownPow2(width, 8);
for (; col < aligned_width; col += 8) for (; col < aligned_width; col += 8)
{ {
@ -197,7 +195,7 @@ ALWAYS_INLINE void CopyOutRow16<HostDisplayPixelFormat::RGB565, u16>(const u16*
_mm_storeu_si128(reinterpret_cast<__m128i*>(dst_ptr), value); _mm_storeu_si128(reinterpret_cast<__m128i*>(dst_ptr), value);
dst_ptr += 8; dst_ptr += 8;
} }
#elif defined(CPU_AARCH32) || defined(CPU_AARCH64) #elif defined(CPU_AARCH64)
const u32 aligned_width = Common::AlignDownPow2(width, 8); const u32 aligned_width = Common::AlignDownPow2(width, 8);
const uint16x8_t single_mask = vdupq_n_u16(0x1F); const uint16x8_t single_mask = vdupq_n_u16(0x1F);
for (; col < aligned_width; col += 8) for (; col < aligned_width; col += 8)