feat(draw/sw): add neon 64bit support for rgb565 blending operations (#8590)

docs/src/details/integration/chip/arm.rst

@@ -52,6 +52,13 @@ To add Arm2D to your project, follow these steps:
 Neon Acceleration
 -----------------
 
-Several Cortex-A microprocessors support the `Neon SIMD <https://www.arm.com/technologies/neon>`__ instruction set. LVGL has built-in support to improve the performance of software rendering by utilizing Neon instructions. To enable Neon acceleration, set ``LV_USE_DRAW_SW_ASM`` to ``LV_DRAW_SW_ASM_NEON`` in ``lv_conf.h``.
+Some ARM Cortex-A and Cortex-R processors with the ARMv7 architecture and every ARM Cortex-A and Cortex-R processor from the ARMv8 architecture support the `Neon SIMD <https://www.arm.com/technologies/neon>` instruction set.
+LVGL has built-in support to improve the performance of software rendering by utilizing Neon instructions.
 
+Architecture Support
+--------------------
 
+Both 32-bit and 64-bit ARM architectures are supported. Simply set ``LV_USE_DRAW_SW_ASM`` to ``LV_DRAW_SW_ASM_NEON`` in ``lv_conf``.
+
+.. note::
+   All ARM Cortex-A and Cortex-R 64-bit processors include Neon support as a mandatory feature starting with the ARMv8 architecture specification. This makes Neon acceleration universally available on all 64-bit ARM platforms, including current and future ARM architectures.

scripts/lv_conf_internal_gen.py

@@ -63,10 +63,10 @@ fout.write(
 #define LV_STDLIB_RTTHREAD          3
 #define LV_STDLIB_CUSTOM            255
 
-#define LV_DRAW_SW_ASM_NONE         0
+#define LV_DRAW_SW_ASM_NONE             0
-#define LV_DRAW_SW_ASM_NEON         1
+#define LV_DRAW_SW_ASM_NEON             1
-#define LV_DRAW_SW_ASM_HELIUM       2
+#define LV_DRAW_SW_ASM_HELIUM           2
-#define LV_DRAW_SW_ASM_CUSTOM       255
+#define LV_DRAW_SW_ASM_CUSTOM           255
 
 #define LV_NEMA_HAL_CUSTOM          0
 #define LV_NEMA_HAL_STM32           1

src/draw/sw/blend/lv_draw_sw_blend_to_rgb565.c

@@ -155,6 +155,22 @@ static inline void * /* LV_ATTRIBUTE_FAST_MEM */ drawbuf_next_row(const void * b
     #define LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(...)      LV_RESULT_INVALID
 #endif
 
+#ifndef LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565
+    #define LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565(...)                   LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_WITH_OPA
+    #define LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_WITH_OPA(...)          LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_WITH_MASK
+    #define LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_WITH_MASK(...)         LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA
+    #define LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(...)      LV_RESULT_INVALID
+#endif
+
 #ifndef LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565
     #define LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565(...)                   LV_RESULT_INVALID
 #endif
@@ -933,7 +949,7 @@ static void LV_ATTRIBUTE_FAST_MEM rgb565_swapped_image_blend(lv_draw_sw_blend_im
 
     if(dsc->blend_mode == LV_BLEND_MODE_NORMAL) {
         if(mask_buf == NULL && opa >= LV_OPA_MAX) {
-            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565(dsc)) {
+            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565(dsc)) {
                 uint32_t line_in_bytes = w * 2;
                 for(y = 0; y < h; y++) {
                     lv_memcpy(dest_buf_u16, src_buf_u16, line_in_bytes);
@@ -944,7 +960,7 @@ static void LV_ATTRIBUTE_FAST_MEM rgb565_swapped_image_blend(lv_draw_sw_blend_im
             }
         }
         else if(mask_buf == NULL && opa < LV_OPA_MAX) {
-            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_WITH_OPA(dsc)) {
+            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_WITH_OPA(dsc)) {
                 for(y = 0; y < h; y++) {
                     for(x = 0; x < w; x++) {
                         dest_buf_u16[x] = lv_color_16_16_mix(lv_color_swap_16(src_buf_u16[x]), dest_buf_u16[x], opa);
@@ -955,7 +971,7 @@ static void LV_ATTRIBUTE_FAST_MEM rgb565_swapped_image_blend(lv_draw_sw_blend_im
             }
         }
         else if(mask_buf && opa >= LV_OPA_MAX) {
-            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_WITH_MASK(dsc)) {
+            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_WITH_MASK(dsc)) {
                 for(y = 0; y < h; y++) {
                     for(x = 0; x < w; x++) {
                         dest_buf_u16[x] = lv_color_16_16_mix(lv_color_swap_16(src_buf_u16[x]), dest_buf_u16[x], mask_buf[x]);
@@ -967,7 +983,7 @@ static void LV_ATTRIBUTE_FAST_MEM rgb565_swapped_image_blend(lv_draw_sw_blend_im
             }
         }
         else {
-            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(dsc)) {
+            if(LV_RESULT_INVALID == LV_DRAW_SW_RGB565_SWAPPED_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(dsc)) {
                 for(y = 0; y < h; y++) {
                     for(x = 0; x < w; x++) {
                         dest_buf_u16[x] = lv_color_16_16_mix(lv_color_swap_16(src_buf_u16[x]), dest_buf_u16[x], LV_OPA_MIX2(mask_buf[x], opa));

src/draw/sw/blend/neon/lv_draw_sw_blend_neon_to_rgb565.h

@@ -0,0 +1,190 @@
+/**
+ * @file lv_draw_sw_blend_neon_to_rgb565.h
+ *
+ */
+
+#ifndef LV_DRAW_SW_BLEND_NEON_TO_RGB565_H
+#define LV_DRAW_SW_BLEND_NEON_TO_RGB565_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*********************
+ *      INCLUDES
+ *********************/
+
+#include "../../../../lv_conf_internal.h"
+#if LV_USE_DRAW_SW_ASM == LV_DRAW_SW_ASM_NEON
+
+#include "../../../../misc/lv_types.h"
+
+/*********************
+ *      DEFINES
+ *********************/
+
+#ifndef LV_DRAW_SW_COLOR_BLEND_TO_RGB565
+#define LV_DRAW_SW_COLOR_BLEND_TO_RGB565(dsc) lv_draw_sw_blend_neon_color_to_rgb565(dsc)
+
+#endif
+
+#ifndef LV_DRAW_SW_COLOR_BLEND_TO_RGB565_WITH_OPA
+#define LV_DRAW_SW_COLOR_BLEND_TO_RGB565_WITH_OPA(dsc) lv_draw_sw_blend_neon_color_to_rgb565_with_opa(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_COLOR_BLEND_TO_RGB565_WITH_MASK
+#define LV_DRAW_SW_COLOR_BLEND_TO_RGB565_WITH_MASK(dsc) lv_draw_sw_blend_neon_color_to_rgb565_with_mask(dsc)
+
+#endif
+
+#ifndef LV_DRAW_SW_COLOR_BLEND_TO_RGB565_MIX_MASK_OPA
+#define LV_DRAW_SW_COLOR_BLEND_TO_RGB565_MIX_MASK_OPA(dsc) lv_draw_sw_blend_neon_color_to_rgb565_with_opa_mask(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565
+#define LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565(dsc) lv_draw_sw_blend_neon_l8_to_rgb565(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565_WITH_OPA
+#define LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565_WITH_OPA(...) LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565_WITH_MASK
+#define LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565_WITH_MASK(...) LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA
+#define LV_DRAW_SW_L8_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(...) LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565
+#define LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565(dsc) lv_draw_sw_blend_neon_al88_to_rgb565(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565_WITH_OPA
+#define LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565_WITH_OPA(dsc) lv_draw_sw_blend_neon_al88_to_rgb565_with_opa(dsc)
+#endif
+
+#if 0 /* Disabled as it's not tested */
+#ifndef LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565_WITH_MASK
+#define LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565_WITH_MASK(dsc) lv_draw_sw_blend_neon_al88_to_rgb565_with_mask(dsc)
+
+#endif
+
+#ifndef LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA
+#define LV_DRAW_SW_AL88_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(dsc) lv_draw_sw_blend_neon_al88_to_rgb565_with_opa_mask(dsc)
+#endif
+#endif
+
+#ifndef LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565
+#define LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565(dsc) lv_draw_sw_blend_neon_rgb565_to_rgb565(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_WITH_OPA
+#define LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_WITH_OPA(dsc) lv_draw_sw_blend_neon_rgb565_to_rgb565_with_opa(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_WITH_MASK
+#define LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_WITH_MASK(dsc) lv_draw_sw_blend_neon_rgb565_to_rgb565_with_mask(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA
+#define LV_DRAW_SW_RGB565_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(dsc) lv_draw_sw_blend_neon_rgb565_to_rgb565_with_opa_mask(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565
+#define LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565(dsc, src_px_size) lv_draw_sw_blend_neon_rgb888_to_rgb565(dsc, src_px_size)
+#endif
+
+#ifndef LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565_WITH_OPA
+#define LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565_WITH_OPA(dsc, src_px_size) lv_draw_sw_blend_neon_rgb888_to_rgb565_with_opa(dsc, src_px_size)
+#endif
+
+#ifndef LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565_WITH_MASK
+#define LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565_WITH_MASK(dsc, src_px_size) lv_draw_sw_blend_neon_rgb888_to_rgb565_with_mask(dsc, src_px_size)
+#endif
+
+#ifndef LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA
+#define LV_DRAW_SW_RGB888_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(dsc, src_px_size) lv_draw_sw_blend_neon_rgb888_to_rgb565_with_opa_mask(dsc, src_px_size)
+#endif
+
+#ifndef LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565
+#define LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565(dsc) lv_draw_sw_blend_neon_argb888_to_rgb565(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565_WITH_OPA
+#define LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565_WITH_OPA(dsc) lv_draw_sw_blend_neon_argb888_to_rgb565_with_opa(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565_WITH_MASK
+#define LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565_WITH_MASK(dsc) lv_draw_sw_blend_neon_argb888_to_rgb565_with_mask(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA
+#define LV_DRAW_SW_ARGB8888_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(dsc) lv_draw_sw_blend_neon_argb888_to_rgb565_with_opa_mask(dsc)
+#endif
+
+#ifndef LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565
+#define LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565(dsc) lv_draw_sw_blend_neon_argb888_premultiplied_to_rgb565(dsc)
+#endif
+
+/*
+ * Bleding operations with premultiplied argb8888 require division.
+ * As division is not supported for integer values in neon we don't define these functions
+ */
+#ifndef LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565_WITH_OPA
+#define LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565_WITH_OPA(...) LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565_WITH_MASK
+#define LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565_WITH_MASK(...) LV_RESULT_INVALID
+#endif
+
+#ifndef LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA
+#define LV_DRAW_SW_ARGB8888_PREMULTIPLIED_BLEND_NORMAL_TO_RGB565_MIX_MASK_OPA(...) LV_RESULT_INVALID
+#endif
+
+/**********************
+ *      TYPEDEFS
+ **********************/
+
+/**********************
+ * GLOBAL PROTOTYPES
+ **********************/
+
+lv_result_t lv_draw_sw_blend_neon_color_to_rgb565(lv_draw_sw_blend_fill_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_color_to_rgb565_with_opa(lv_draw_sw_blend_fill_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_color_to_rgb565_with_mask(lv_draw_sw_blend_fill_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_color_to_rgb565_with_opa_mask(lv_draw_sw_blend_fill_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_l8_to_rgb565(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_al88_to_rgb565(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_al88_to_rgb565_with_opa(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_al88_to_rgb565_with_mask(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_al88_to_rgb565_with_opa_mask(lv_draw_sw_blend_image_dsc_t * dsc);
+
+lv_result_t lv_draw_sw_blend_neon_rgb565_to_rgb565(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_rgb565_to_rgb565_with_opa(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_rgb565_to_rgb565_with_mask(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_rgb565_to_rgb565_with_opa_mask(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_rgb888_to_rgb565(lv_draw_sw_blend_image_dsc_t * dsc, uint8_t src_px_size);
+lv_result_t lv_draw_sw_blend_neon_rgb888_to_rgb565_with_opa(lv_draw_sw_blend_image_dsc_t * dsc, uint8_t src_px_size);
+lv_result_t lv_draw_sw_blend_neon_rgb888_to_rgb565_with_mask(lv_draw_sw_blend_image_dsc_t * dsc, uint8_t src_px_size);
+lv_result_t lv_draw_sw_blend_neon_rgb888_to_rgb565_with_opa_mask(lv_draw_sw_blend_image_dsc_t * dsc,
+                                                                 uint8_t src_px_size);
+lv_result_t lv_draw_sw_blend_neon_argb888_to_rgb565(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_argb888_premultiplied_to_rgb565(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_argb888_to_rgb565_with_opa(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_argb888_to_rgb565_with_mask(lv_draw_sw_blend_image_dsc_t * dsc);
+lv_result_t lv_draw_sw_blend_neon_argb888_to_rgb565_with_opa_mask(lv_draw_sw_blend_image_dsc_t * dsc);
+
+/**********************
+ *      MACROS
+ **********************/
+
+#endif /* LV_USE_DRAW_SW_ASM == LV_DRAW_SW_ASM_NEON */
+
+#ifdef __cplusplus
+} /*extern "C"*/
+#endif
+
+#endif /*LV_DRAW_SW_BLEND_NEON_TO_RGB565_H*/

src/lv_conf_internal.h

@@ -25,10 +25,10 @@
 #define LV_STDLIB_RTTHREAD          3
 #define LV_STDLIB_CUSTOM            255
 
-#define LV_DRAW_SW_ASM_NONE         0
+#define LV_DRAW_SW_ASM_NONE             0
-#define LV_DRAW_SW_ASM_NEON         1
+#define LV_DRAW_SW_ASM_NEON             1
-#define LV_DRAW_SW_ASM_HELIUM       2
+#define LV_DRAW_SW_ASM_HELIUM           2
-#define LV_DRAW_SW_ASM_CUSTOM       255
+#define LV_DRAW_SW_ASM_CUSTOM           255
 
 #define LV_NEMA_HAL_CUSTOM          0
 #define LV_NEMA_HAL_STM32           1