// Copyright 2023 The Chromium Authors // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. #ifndef THIRD_PARTY_BLINK_RENDERER_PLATFORM_GRAPHICS_CPU_LOONGARCH64_WEBGL_IMAGE_CONVERSION_LSX_H_ #define THIRD_PARTY_BLINK_RENDERER_PLATFORM_GRAPHICS_CPU_LOONGARCH64_WEBGL_IMAGE_CONVERSION_LSX_H_ #include "build/build_config.h" #if defined(ARCH_CPU_LOONGARCH_FAMILY) #include <lsxintrin.h> namespace blink { namespace simd { ALWAYS_INLINE void UnpackOneRowOfRGBA4444LittleToRGBA8( const uint16_t*& source, uint8_t*& destination, unsigned& pixels_per_row) { v8u16 immediate0x0f = __lsx_vreplgr2vr_h(0x0F); unsigned pixels_per_row_trunc = (pixels_per_row / 8) * 8; for (unsigned i = 0; i < pixels_per_row_trunc; i += 8) { v8u16 packed_value = *((v8u16*)(source)); __m128i r = __lsx_vsrli_h(packed_value, 12); __m128i g = __lsx_vand_v(__lsx_vsrli_h(packed_value, 8), immediate0x0f); __m128i b = __lsx_vand_v(__lsx_vsrli_h(packed_value, 4), immediate0x0f); __m128i a = __lsx_vand_v(packed_value, immediate0x0f); __m128i component_r = __lsx_vor_v(__lsx_vslli_b(r, 4), r); __m128i component_g = __lsx_vor_v(__lsx_vslli_b(g, 4), g); __m128i component_b = __lsx_vor_v(__lsx_vslli_b(b, 4), b); __m128i component_a = __lsx_vor_v(__lsx_vslli_b(a, 4), a); __m128i component_rb = __lsx_vpackev_b(component_b, component_r); __m128i component_ga = __lsx_vpackev_b(component_a, component_g); __m128i component_rgba1 = __lsx_vilvl_b(component_ga, component_rb); __m128i component_rgba2 = __lsx_vilvh_b(component_ga, component_rb); __lsx_vst(component_rgba1, (void*)destination, 0); __lsx_vst(component_rgba2, (void*)destination, 16); source += 8; destination += 32; } pixels_per_row -= pixels_per_row_trunc; } ALWAYS_INLINE void UnpackOneRowOfRGBA5551LittleToRGBA8( const uint16_t*& source, uint8_t*& destination, unsigned& pixels_per_row) { __m128i immediate0x1f = __lsx_vreplgr2vr_h(0x1F); __m128i immediate0x7 = __lsx_vreplgr2vr_h(0x7); __m128i immediate0x1 = __lsx_vreplgr2vr_h(0x1); unsigned pixels_per_row_trunc = (pixels_per_row / 8) * 8; for (unsigned i = 0; i < pixels_per_row_trunc; i += 8) { v8u16 packed_value = *((v8u16*)(source)); __m128i r = __lsx_vsrli_h(packed_value, 11); __m128i g = __lsx_vand_v(__lsx_vsrli_h(packed_value, 6), immediate0x1f); __m128i b = __lsx_vand_v(__lsx_vsrli_h(packed_value, 1), immediate0x1f); __m128i component_r = __lsx_vor_v(__lsx_vslli_b(r, 3), __lsx_vand_v(r, immediate0x7)); __m128i component_g = __lsx_vor_v(__lsx_vslli_b(g, 3), __lsx_vand_v(g, immediate0x7)); __m128i component_b = __lsx_vor_v(__lsx_vslli_b(b, 3), __lsx_vand_v(b, immediate0x7)); __m128i component_a = __lsx_vseq_h( __lsx_vand_v(packed_value, immediate0x1), immediate0x1); __m128i component_rb = __lsx_vpackev_b(component_b, component_r); __m128i component_ga = __lsx_vpackev_b(component_a, component_g); __m128i component_rgba1 = __lsx_vilvl_b(component_ga, component_rb); __m128i component_rgba2 = __lsx_vilvh_b(component_ga, component_rb); __lsx_vst(component_rgba1, (void*)destination, 0); __lsx_vst(component_rgba2, (void*)destination, 16); source += 8; destination += 32; } pixels_per_row -= pixels_per_row_trunc; } ALWAYS_INLINE void PackOneRowOfRGBA8LittleToRA8(const uint8_t*& source, uint8_t*& destination, unsigned& pixels_per_row) { unsigned pixels_per_row_trunc = (pixels_per_row / 4) * 4; v16u8 mask_zero = __lsx_vldi(0); v4u32 mask_lalpha = __lsx_vreplgr2vr_w(0x0ff); v4f32 mask_falpha = __lsx_vffint_s_w(mask_lalpha); v16u8 ra_index = {0,19, 4,23, 8,27, 12,31}; for (unsigned i = 0; i < pixels_per_row_trunc; i += 4) { v16u8 bgra = *((__m128i*)(source)); //if A !=0, A=0; else A=0xFF v4f32 alpha_factor = __lsx_vseq_b(bgra, mask_zero); //if A!=0, A=A; else A=0xFF alpha_factor = __lsx_vor_v(bgra, alpha_factor); alpha_factor = __lsx_vsrli_w(alpha_factor, 24); alpha_factor = __lsx_vffint_s_w(alpha_factor); alpha_factor = __lsx_vfdiv_s(mask_falpha, alpha_factor); v16u8 component_r = __lsx_vand_v(bgra, mask_lalpha); component_r = __lsx_vffint_s_w(component_r); component_r = __lsx_vfmul_s(component_r, alpha_factor); component_r = __lsx_vftintrz_w_s(component_r); v2u64 ra = __lsx_vshuf_b(bgra, component_r, ra_index); __lsx_vstelm_d(ra, destination, 0, 0); source += 16; destination += 8; } pixels_per_row -= pixels_per_row_trunc; } ALWAYS_INLINE void UnpackOneRowOfBGRA8LittleToRGBA8(const uint32_t*& source, uint32_t*& destination, unsigned& pixels_per_row) { __m128i bgra, rgba; unsigned pixels_per_row_trunc = (pixels_per_row / 4) * 4; for (unsigned i = 0; i < pixels_per_row_trunc; i += 4) { bgra = *((__m128i*)(source)); rgba = __lsx_vshuf4i_b(bgra, 0xc6); __lsx_vst(rgba, destination, 0); source += 4; destination += 4; } pixels_per_row -= pixels_per_row_trunc; } ALWAYS_INLINE void PackOneRowOfRGBA8LittleToR8(const uint8_t*& source, uint8_t*& destination, unsigned& pixels_per_row) { unsigned pixels_per_row_trunc = (pixels_per_row / 4) * 4; v16u8 mask_zero = __lsx_vldi(0); v4u32 mask_lalpha = __lsx_vreplgr2vr_w(0x0ff); v4f32 mask_falpha = __lsx_vffint_s_w(mask_lalpha); for (unsigned i = 0; i < pixels_per_row_trunc; i += 4) { v16u8 bgra = *((__m128i*)(source)); //if A !=0, A=0; else A=0xFF v4f32 alpha_factor = __lsx_vseq_b(bgra, mask_zero); //if A!=0, A=A; else A=0xFF alpha_factor = __lsx_vor_v(bgra, alpha_factor); alpha_factor = __lsx_vsrli_w(alpha_factor, 24); alpha_factor = __lsx_vffint_s_w(alpha_factor); alpha_factor = __lsx_vfdiv_s(mask_falpha, alpha_factor); v16u8 component_r = __lsx_vand_v(bgra, mask_lalpha); component_r = __lsx_vffint_s_w(component_r); component_r = __lsx_vfmul_s(component_r, alpha_factor); component_r = __lsx_vftintrz_w_s(component_r); component_r = __lsx_vpickev_b(component_r, component_r); component_r = __lsx_vpickev_b(component_r, component_r); __lsx_vstelm_w(component_r, destination, 0, 0); source += 16; destination += 4; } pixels_per_row -= pixels_per_row_trunc; } ALWAYS_INLINE void PackOneRowOfRGBA8LittleToRGBA8(const uint8_t*& source, uint8_t*& destination, unsigned& pixels_per_row) { unsigned pixels_per_row_trunc = (pixels_per_row / 4) * 4; v16u8 mask_zero = __lsx_vldi(0); v4u32 mask_lalpha = __lsx_vreplgr2vr_w(0x0ff); v4f32 mask_falpha = __lsx_vffint_s_w(mask_lalpha); v16u8 rgba_index = {0,1,2,19, 4,5,6,23, 8,9,10,27, 12,13,14,31}; for (unsigned i = 0; i < pixels_per_row_trunc; i += 4) { v16u8 bgra = *((__m128i*)(source)); //if A !=0, A=0; else A=0xFF v4f32 alpha_factor = __lsx_vseq_b(bgra, mask_zero); //if A!=0, A=A; else A=0xFF alpha_factor = __lsx_vor_v(bgra, alpha_factor); alpha_factor = __lsx_vsrli_w(alpha_factor, 24); alpha_factor = __lsx_vffint_s_w(alpha_factor); alpha_factor = __lsx_vfdiv_s(mask_falpha, alpha_factor); v16u8 bgra_01 = __lsx_vilvl_b(mask_zero, bgra); v16u8 bgra_23 = __lsx_vilvh_b(mask_zero, bgra); v16u8 bgra_0 = __lsx_vilvl_b(mask_zero, bgra_01); v16u8 bgra_1 = __lsx_vilvh_b(mask_zero, bgra_01); v16u8 bgra_2 = __lsx_vilvl_b(mask_zero, bgra_23); v16u8 bgra_3 = __lsx_vilvh_b(mask_zero, bgra_23); bgra_0 = __lsx_vffint_s_w(bgra_0); bgra_1 = __lsx_vffint_s_w(bgra_1); bgra_2 = __lsx_vffint_s_w(bgra_2); bgra_3 = __lsx_vffint_s_w(bgra_3); v4f32 alpha_factor_0 = __lsx_vreplvei_w(alpha_factor, 0); v4f32 alpha_factor_1 = __lsx_vreplvei_w(alpha_factor, 1); v4f32 alpha_factor_2 = __lsx_vreplvei_w(alpha_factor, 2); v4f32 alpha_factor_3 = __lsx_vreplvei_w(alpha_factor, 3); bgra_0 = __lsx_vfmul_s(alpha_factor_0, bgra_0); bgra_1 = __lsx_vfmul_s(alpha_factor_1, bgra_1); bgra_2 = __lsx_vfmul_s(alpha_factor_2, bgra_2); bgra_3 = __lsx_vfmul_s(alpha_factor_3, bgra_3); bgra_0 = __lsx_vftintrz_w_s(bgra_0); bgra_1 = __lsx_vftintrz_w_s(bgra_1); bgra_2 = __lsx_vftintrz_w_s(bgra_2); bgra_3 = __lsx_vftintrz_w_s(bgra_3); bgra_01 = __lsx_vpickev_b(bgra_1, bgra_0); bgra_23 = __lsx_vpickev_b(bgra_3, bgra_2); v4u32 rgba = __lsx_vpickev_b(bgra_23, bgra_01); rgba = __lsx_vshuf_b(bgra, rgba, rgba_index); __lsx_vst(rgba, destination, 0); source += 16; destination += 16; } pixels_per_row -= pixels_per_row_trunc; } } // namespace simd } // namespace blink #endif // ARCH_CPU_LOONGARCH_FAMILY #endif // THIRD_PARTY_BLINK_RENDERER_PLATFORM_GRAPHICS_CPU_LOONGARCH64_WEBGL_IMAGE_CONVERSION_LSX_H_
Codebase Browser
chromium