Fix typo in src/GUI/PageInput.h

CMakeLists.txt

@@ -11,12 +11,16 @@ project(simplescreenrecorder VERSION 0.4.4)
 
 if(CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64|amd64|i386|i686")
 	set(PROCESSOR_IS_X86 TRUE)
+elseif(CMAKE_SYSTEM_PROCESSOR MATCHES "loongarch")
+	set(PROCESSOR_IS_LOONGARCH TRUE)
 else()
 	set(PROCESSOR_IS_X86 FALSE)
+	set(PROCESSOR_IS_LOONGARCH FALSE)
 endif()
 
 option(ENABLE_32BIT_GLINJECT "Build the 32-bit version of 'libssr-glinject' on 64-bit systems (in addition to the 64-bit version). Required for OpenGL recording of 32-bit applications on 64-bit systems." FALSE)
 option(ENABLE_X86_ASM "Allow x86/x64 assembly or intrinsics." ${PROCESSOR_IS_X86})
+option(ENABLE_LOONGARCH_ASM "Allow loongarch assembly or intrinsics." ${PROCESSOR_IS_LOONGARCH})
 option(ENABLE_FFMPEG_VERSIONS "Use FFmpeg version numbers for feature support tests. Enable when using FFmpeg, disable when using Libav." TRUE)
 option(ENABLE_JACK_METADATA "Use the JACK metadata API. May not work with very old JACK versions." TRUE)
 option(WITH_OPENGL_RECORDING "Build with OpenGL recording support." TRUE)
@@ -28,6 +32,7 @@ option(WITH_JACK "Build with JACK support." TRUE)
 option(WITH_QT5 "Build with Qt5 (instead of Qt4)." FALSE)
 option(WITH_SIMPLESCREENRECORDER "Build the 'simplescreenrecorder' executable." TRUE)
 option(WITH_GLINJECT "Build the 'libssr-glinject' library. Required for OpenGL recording." TRUE)
+option(WITH_SIMDE "Build with simde support." ${PROCESSOR_IS_LOONGARCH})
 
 set(CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/cmake)
 

cmake/FindSIMDE.cmake

@@ -0,0 +1,13 @@
+# rules for finding the SIMDE library
+
+find_package(PkgConfig REQUIRED)
+pkg_check_modules(PC_SIMDE simde)
+
+find_path(SIMDE_INCLUDE_DIR simde/x86/sse2.h simde/x86/ssse3.h HINTS ${PC_SIMDE_INCLUDEDIR} ${PC_SIMDE_INCLUDE_DIRS})
+
+include(FindPackageHandleStandardArgs)
+find_package_handle_standard_args(SIMDE DEFAULT_MSG SIMDE_INCLUDE_DIR)
+
+mark_as_advanced(SIMDE_INCLUDE_DIR)
+
+set(SIMDE_INCLUDE_DIRS ${SIMDE_INCLUDE_DIR})

src/AV/FastResampler.cpp

@@ -105,13 +105,22 @@ FastResampler::FastResampler(unsigned int channels, float gain) {
 			default: m_firfilter2_ptr = &FastResampler_FirFilter2_Cn_SSE2; break;
 		}
 	} else {
+#endif
+#if SSR_USE_LOONGARCH_ASM
+	if(CPUFeatures::HasLSX()) {
+		switch(m_channels) {
+			case 1:  m_firfilter2_ptr = &FastResampler_FirFilter2_C1_LSX; break;
+			case 2:  m_firfilter2_ptr = &FastResampler_FirFilter2_C2_LSX; break;
+			default: m_firfilter2_ptr = &FastResampler_FirFilter2_Cn_LSX; break;
+		}
+	} else {
 #endif
 		switch(m_channels) {
 			case 1:  m_firfilter2_ptr = &FastResampler_FirFilter2_C1_Fallback; break;
 			case 2:  m_firfilter2_ptr = &FastResampler_FirFilter2_C2_Fallback; break;
 			default: m_firfilter2_ptr = &FastResampler_FirFilter2_Cn_Fallback; break;
 		}
-#if SSR_USE_X86_ASM
+#if SSR_USE_X86_ASM || SSR_USE_LOONGARCH_ASM
 	}
 #endif
 

src/AV/FastResampler_FirFilter.h

@@ -31,3 +31,9 @@ void FastResampler_FirFilter2_C1_SSE2(unsigned int channels, unsigned int filter
 void FastResampler_FirFilter2_C2_SSE2(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output);
 void FastResampler_FirFilter2_Cn_SSE2(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output);
 #endif
+
+#if SSR_USE_LOONGARCH_ASM
+void FastResampler_FirFilter2_C1_LSX(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output);
+void FastResampler_FirFilter2_C2_LSX(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output);
+void FastResampler_FirFilter2_Cn_LSX(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output);
+#endif

src/AV/FastResampler_FirFilter_LSX.cpp

@@ -0,0 +1,90 @@
+/*
+Copyright (c) 2012-2024 Maarten Baert <[email protected]>
+
+This file is part of SimpleScreenRecorder.
+
+SimpleScreenRecorder is free software: you can redistribute it and/or modify
+it under the terms of the GNU General Public License as published by
+the Free Software Foundation, either version 3 of the License, or
+(at your option) any later version.
+
+SimpleScreenRecorder is distributed in the hope that it will be useful,
+but WITHOUT ANY WARRANTY; without even the implied warranty of
+MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+GNU General Public License for more details.
+
+You should have received a copy of the GNU General Public License
+along with SimpleScreenRecorder.  If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#include "FastResampler_FirFilter.h"
+
+#if SSR_USE_LOONGARCH_ASM
+
+#ifndef SIMDE_ENABLE_NATIVE_ALIASES
+#define SIMDE_ENABLE_NATIVE_ALIASES
+#endif
+//using simde to translate SSE2 to LSX(loongarch 128-bit simd)
+#include<simde/x86/sse2.h>
+
+//void FastResampler_FirFilter2_C1_SSE2(...)
+void FastResampler_FirFilter2_C1_LSX(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output) {
+	Q_UNUSED(channels);
+	__m128 sum = _mm_setzero_ps();
+	__m128 v_frac = _mm_set1_ps(frac);
+	for(unsigned int i = 0; i < filter_length / 4; ++i) {
+		__m128 v_coef1 = _mm_load_ps(coef1), v_coef2 = _mm_load_ps(coef2);
+		coef1 += 4; coef2 += 4;
+		__m128 filter_value = _mm_add_ps(v_coef1, _mm_mul_ps(_mm_sub_ps(v_coef2, v_coef1), v_frac));
+		__m128 v_input = _mm_loadu_ps(input);
+		input += 4;
+		sum = _mm_add_ps(sum, _mm_mul_ps(v_input, filter_value));
+	}
+	__m128 sum2 = _mm_add_ps(sum, _mm_shuffle_ps(sum, sum, 0x0e));
+	__m128 sum3 = _mm_add_ss(sum2, _mm_shuffle_ps(sum2, sum2, 0x01));
+	_mm_store_ss(output, sum3);
+}
+
+//void FastResampler_FirFilter2_C2_SSE2(...)
+void FastResampler_FirFilter2_C2_LSX(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output) {
+	Q_UNUSED(channels);
+	__m128 sum = _mm_setzero_ps();
+	__m128 v_frac = _mm_set1_ps(frac);
+	for(unsigned int i = 0; i < filter_length / 4; ++i) {
+		__m128 v_coef1 = _mm_load_ps(coef1), v_coef2 = _mm_load_ps(coef2);
+		coef1 += 4; coef2 += 4;
+		__m128 filter_value = _mm_add_ps(v_coef1, _mm_mul_ps(_mm_sub_ps(v_coef2, v_coef1), v_frac));
+		__m128 v_input1 = _mm_loadu_ps(input), v_input2 = _mm_loadu_ps(input + 4);
+		input += 8;
+		sum = _mm_add_ps(sum, _mm_mul_ps(v_input1, _mm_unpacklo_ps(filter_value, filter_value)));
+		sum = _mm_add_ps(sum, _mm_mul_ps(v_input2, _mm_unpackhi_ps(filter_value, filter_value)));
+	}
+	__m128 sum2 = _mm_add_ps(sum, _mm_shuffle_ps(sum, sum, 0xee));
+	_mm_store_sd((double*) output, _mm_castps_pd(sum2));
+}
+
+//void FastResampler_FirFilter2_Cn_SSE2(...)
+void FastResampler_FirFilter2_Cn_LSX(unsigned int channels, unsigned int filter_length, float* coef1, float* coef2, float frac, float* input, float* output) {
+	Q_UNUSED(channels);
+	for(unsigned int c = 0; c < channels; ++c) {
+		__m128 sum = _mm_setzero_ps();
+		__m128 v_frac = _mm_set1_ps(frac);
+		float *input2 = input + c;
+		for(unsigned int i = 0; i < filter_length / 4; ++i) {
+			__m128 v_coef1 = _mm_load_ps(coef1), v_coef2 = _mm_load_ps(coef2);
+			coef1 += 4; coef2 += 4;
+			__m128 filter_value = _mm_add_ps(v_coef1, _mm_mul_ps(_mm_sub_ps(v_coef2, v_coef1), v_frac));
+			__m128 v_input1 = _mm_load_ss(input2); input2 += channels;
+			__m128 v_input2 = _mm_load_ss(input2); input2 += channels;
+			__m128 v_input3 = _mm_load_ss(input2); input2 += channels;
+			__m128 v_input4 = _mm_load_ss(input2); input2 += channels;
+			__m128 v_input = _mm_movelh_ps(_mm_unpacklo_ps(v_input1, v_input2), _mm_unpacklo_ps(v_input3, v_input4));
+			sum = _mm_add_ps(sum, _mm_mul_ps(v_input, filter_value));
+		}
+		__m128 sum2 = _mm_add_ps(sum, _mm_shuffle_ps(sum, sum, 0x0e));
+		__m128 sum3 = _mm_add_ss(sum2, _mm_shuffle_ps(sum2, sum2, 0x01));
+		_mm_store_ss(output + c, sum3);
+	}
+}
+
+#endif

src/AV/FastScaler.cpp

@@ -165,6 +165,14 @@ void FastScaler::Convert_BGRA_YUV444(unsigned int width, unsigned int height, co
 		return;
 	}
 #endif
+#if SSR_USE_LOONGARCH_ASM
+	if(CPUFeatures::HasLSX()) {
+		Convert_BGRA_YUV444_LSX(width, height, in_data, in_stride, out_data, out_stride);
+	} else {
+		Convert_BGRA_YUV444_Fallback(width, height, in_data, in_stride, out_data, out_stride);
+	}
+	return;
+#endif
 
 	Convert_BGRA_YUV444_Fallback(width, height, in_data, in_stride, out_data, out_stride);
 
@@ -190,6 +198,14 @@ void FastScaler::Convert_BGRA_YUV422(unsigned int width, unsigned int height, co
 		return;
 	}
 #endif
+#if SSR_USE_LOONGARCH_ASM
+	if(CPUFeatures::HasLSX()) {
+		Convert_BGRA_YUV422_LSX(width, height, in_data, in_stride, out_data, out_stride);
+	} else {
+		Convert_BGRA_YUV422_Fallback(width, height, in_data, in_stride, out_data, out_stride);
+	}
+	return;
+#endif
 
 	Convert_BGRA_YUV422_Fallback(width, height, in_data, in_stride, out_data, out_stride);
 
@@ -215,6 +231,14 @@ void FastScaler::Convert_BGRA_YUV420(unsigned int width, unsigned int height, co
 		return;
 	}
 #endif
+#if SSR_USE_LOONGARCH_ASM
+	if(CPUFeatures::HasLSX()) {
+		Convert_BGRA_YUV420_LSX(width, height, in_data, in_stride, out_data, out_stride);
+	} else {
+		Convert_BGRA_YUV420_Fallback(width, height, in_data, in_stride, out_data, out_stride);
+	}
+	return;
+#endif
 
 	Convert_BGRA_YUV420_Fallback(width, height, in_data, in_stride, out_data, out_stride);
 
@@ -239,6 +263,14 @@ void FastScaler::Convert_BGRA_NV12(unsigned int width, unsigned int height, cons
 		return;
 	}
 #endif
+#if SSR_USE_LOONGARCH_ASM
+	if(CPUFeatures::HasLSX()) {
+		Convert_BGRA_NV12_LSX(width, height, in_data, in_stride, out_data, out_stride);
+	} else {
+		Convert_BGRA_NV12_Fallback(width, height, in_data, in_stride, out_data, out_stride);
+	}
+	return;
+#endif
 
 	Convert_BGRA_NV12_Fallback(width, height, in_data, in_stride, out_data, out_stride);
 
@@ -261,6 +293,14 @@ void FastScaler::Convert_BGRA_BGR(unsigned int width, unsigned int height, const
 		return;
 	}
 #endif
+#if SSR_USE_LOONGARCH_ASM
+	if(CPUFeatures::HasLSX()) {
+		Convert_BGRA_BGR_LSX(width, height, in_data, in_stride, out_data, out_stride);
+	} else {
+		Convert_BGRA_BGR_Fallback(width, height, in_data, in_stride, out_data, out_stride);
+	}
+	return;
+#endif
 
 	Convert_BGRA_BGR_Fallback(width, height, in_data, in_stride, out_data, out_stride);
 
@@ -284,6 +324,14 @@ void FastScaler::Scale_BGRA(unsigned int in_width, unsigned int in_height, const
 		return;
 	}
 #endif
+#if SSR_USE_LOONGARCH_ASM
+	if(CPUFeatures::HasLSX()) {
+		Scale_BGRA_LSX(in_width, in_height, in_data, in_stride, out_width, out_height, out_data, out_stride);
+	} else {
+		Scale_BGRA_Fallback(in_width, in_height, in_data, in_stride, out_width, out_height, out_data, out_stride);
+	}
+	return;
+#endif
 
 	Scale_BGRA_Fallback(in_width, in_height, in_data, in_stride, out_width, out_height, out_data, out_stride);
 

src/AV/FastScaler_Convert.h

@@ -33,3 +33,11 @@ void Convert_BGRA_YUV420_SSSE3(unsigned int w, unsigned int h, const uint8_t* in
 void Convert_BGRA_NV12_SSSE3(unsigned int w, unsigned int h, const uint8_t* in_data, int in_stride, uint8_t* const out_data[2], const int out_stride[2]);
 void Convert_BGRA_BGR_SSSE3(unsigned int w, unsigned int h, const uint8_t* in_data, int in_stride, uint8_t* out_data, int out_stride);
 #endif
+
+#if SSR_USE_LOONGARCH_ASM
+void Convert_BGRA_YUV444_LSX(unsigned int w, unsigned int h, const uint8_t* in_data, int in_stride, uint8_t* const out_data[3], const int out_stride[3]);
+void Convert_BGRA_YUV422_LSX(unsigned int w, unsigned int h, const uint8_t* in_data, int in_stride, uint8_t* const out_data[3], const int out_stride[3]);
+void Convert_BGRA_YUV420_LSX(unsigned int w, unsigned int h, const uint8_t* in_data, int in_stride, uint8_t* const out_data[3], const int out_stride[3]);
+void Convert_BGRA_NV12_LSX(unsigned int w, unsigned int h, const uint8_t* in_data, int in_stride, uint8_t* const out_data[2], const int out_stride[2]);
+void Convert_BGRA_BGR_LSX(unsigned int w, unsigned int h, const uint8_t* in_data, int in_stride, uint8_t* out_data, int out_stride);
+#endif