Optimize software decoder and improve OAuth authentication UX

.gitignore

@@ -8,4 +8,6 @@
 *.m4a*
 *.webm*
 *.s
+!source/network_decoder/yuv_converter.s
+!source/system/test.s
 *.dmp

source/network_decoder/converter.cpp

@@ -129,44 +129,15 @@ Result_with_string Util_converter_yuv422_to_yuv420p(u8 *yuv422, u8 **yuv420p, in
 }
 
 Result_with_string Util_converter_yuv420p_to_bgr565(u8 *yuv420p, u8 **bgr565, int width, int height) {
-	int index = 0;
-	u8 *ybase = yuv420p;
-	u8 *ubase = yuv420p + width * height;
-	u8 *vbase = yuv420p + width * height + width * height / 4;
-	Result_with_string result;
-
-	*bgr565 = (u8 *)malloc(width * height * 2);
-	if (*bgr565 == NULL) {
-		result.code = DEF_ERR_OUT_OF_MEMORY;
-		result.string = DEF_ERR_OUT_OF_MEMORY_STR;
-		return result;
-	}
-
-	u8 Y[4], U, V, r[4], g[4], b[4];
-	for (int y = 0; y < height; y++) {
-		for (int x = 0; x < width; x++) {
-			// YYYYYYYYUUVV
-			Y[0] = ybase[x + y * width];
-			U = ubase[y / 2 * width / 2 + (x / 2)];
-			V = vbase[y / 2 * width / 2 + (x / 2)];
-			b[0] = YUV2B(Y[0], U);
-			g[0] = YUV2G(Y[0], U, V);
-			r[0] = YUV2R(Y[0], V);
-			b[0] = b[0] >> 3;
-			g[0] = g[0] >> 2;
-			r[0] = r[0] >> 3;
-			*(*bgr565 + index++) = (g[0] & 0b00000111) << 5 | b[0];
-			*(*bgr565 + index++) = (g[0] & 0b00111000) >> 3 | (r[0] & 0b00011111) << 3;
-		}
-	}
-	return result;
+	return Util_converter_yuv420p_to_bgr565_asm(yuv420p, bgr565, width, height);
 }
 
 Result_with_string Util_converter_yuv420p_to_bgr565_asm(u8 *yuv420p, u8 **bgr565, int width, int height) {
 	Result_with_string result;
 
 	*bgr565 = (u8 *)malloc(width * height * 2);
 	if (*bgr565 == NULL) {
+		logger.warning("converter", "bgr565 allocation failed for size: " + std::to_string(width * height * 2));
 		result.code = DEF_ERR_OUT_OF_MEMORY;
 		result.string = DEF_ERR_OUT_OF_MEMORY_STR;
 		return result;
@@ -177,40 +148,15 @@ Result_with_string Util_converter_yuv420p_to_bgr565_asm(u8 *yuv420p, u8 **bgr565
 }
 
 Result_with_string Util_converter_yuv420p_to_bgr888(u8 *yuv420p, u8 **bgr888, int width, int height) {
-	int index = 0;
-	u8 *ybase = yuv420p;
-	u8 *ubase = yuv420p + width * height;
-	u8 *vbase = yuv420p + width * height + width * height / 4;
-	Result_with_string result;
-
-	*bgr888 = (u8 *)malloc(width * height * 3);
-	if (*bgr888 == NULL) {
-		result.code = DEF_ERR_OUT_OF_MEMORY;
-		result.string = DEF_ERR_OUT_OF_MEMORY_STR;
-		return result;
-	}
-
-	u8 Y[4], U, V;
-	for (int y = 0; y < height; y++) {
-		for (int x = 0; x < width; x++) {
-			// YYYYYYYYUUVV
-			Y[0] = *ybase++;
-			U = ubase[y / 2 * width / 2 + (x / 2)];
-			V = vbase[y / 2 * width / 2 + (x / 2)];
-
-			*(*bgr888 + index++) = YUV2B(Y[0], U);
-			*(*bgr888 + index++) = YUV2G(Y[0], U, V);
-			*(*bgr888 + index++) = YUV2R(Y[0], V);
-		}
-	}
-	return result;
+	return Util_converter_yuv420p_to_bgr888_asm(yuv420p, bgr888, width, height);
 }
 
 Result_with_string Util_converter_yuv420p_to_bgr888_asm(u8 *yuv420p, u8 **bgr888, int width, int height) {
 	Result_with_string result;
 
 	*bgr888 = (u8 *)malloc(width * height * 3);
 	if (*bgr888 == NULL) {
+		logger.warning("converter", "bgr888 allocation failed for size: " + std::to_string(width * height * 3));
 		result.code = DEF_ERR_OUT_OF_MEMORY;
 		result.string = DEF_ERR_OUT_OF_MEMORY_STR;
 		return result;

source/network_decoder/network_decoder.cpp

@@ -661,6 +661,11 @@ Result_with_string NetworkDecoder::init_decoder(int type) {
 		decoder_context[type]->lowres = 0;
 		decoder_context[type]->flags = AV_CODEC_FLAG_OUTPUT_CORRUPT;
 
+		if (!var_is_new3ds) {
+			decoder_context[type]->skip_frame = AVDISCARD_NONREF;
+			decoder_context[type]->skip_loop_filter = AVDISCARD_ALL;
+		}
+
 		if (codec[type]->capabilities & AV_CODEC_CAP_FRAME_THREADS) {
 			decoder_context[type]->thread_type = FF_THREAD_FRAME;
 		} else if (codec[type]->capabilities & AV_CODEC_CAP_SLICE_THREADS) {

source/network_decoder/yuv_converter.s

@@ -0,0 +1,278 @@
+// YUV420p to BGR565/BGR888 converter
+
+.data
+.align 4
+.global memcpy_asm
+.global memcpy_asm_4b
+.global yuv420p_to_bgr565_asm
+.global yuv420p_to_bgr888_asm
+.type memcpy_asm, %function
+.type memcpy_asm_4b, %function
+.type yuv420p_to_bgr565_asm, %function
+.type yuv420p_to_bgr888_asm, %function
+
+// ITU-R BT.709 coefficients
+coef_y:  .int 298
+coef_ub: .int 516
+coef_vr: .int 409
+coef_ug: .int 100
+coef_vg: .int 208
+
+.text
+
+// r0 = dest, r1 = src, r2 = size
+memcpy_asm:
+	push {r4-r11}
+	cmp r2, #64
+	ble simple_copy_path
+
+	mov r11, r0
+	add r11, r2
+	sub r11, #1
+
+	ands r3, r0, #31
+	beq already_aligned
+	rsb r3, r3, #32
+	sub r2, r2, r3
+align_bytes:
+	subs r3, r3, #1
+	blt already_aligned
+	ldrb r4, [r1], #1
+	strb r4, [r0], #1
+	b align_bytes
+
+already_aligned:
+	pld [r1, #32]
+	pld [r1, #64]
+	pld [r1, #96]
+
+main_copy_loop:
+	cmp r0, r11
+	bgt copy_finished
+	ldm r1!, {r3-r10}
+	pld [r1, #96]
+	stm r0!, {r3-r10}
+	ldm r1!, {r3-r10}
+	pld [r1, #96]
+	stm r0!, {r3-r10}
+	b main_copy_loop
+
+simple_copy_path:
+	cmp r2, #0
+	ble copy_finished
+byte_by_byte_loop:
+	subs r2, r2, #1
+	blt copy_finished
+	ldrb r3, [r1], #1
+	strb r3, [r0], #1
+	b byte_by_byte_loop
+
+copy_finished:
+	pop {r4-r11}
+	bx lr
+
+memcpy_asm_4b:
+	ldr r2, [r1]
+	str r2, [r0]
+	bx lr
+
+// r0 = Y plane, r1 = output, r2 = width, r3 = height
+yuv420p_to_bgr565_asm:
+	push {r4-r11, lr}
+	sub sp, sp, #16
+
+	mul r4, r2, r3
+	add r5, r0, r4
+	add r6, r5, r4, lsr #2
+	lsr r4, r2, #1
+	str r4, [sp, #0]
+	str r2, [sp, #4]
+	str r3, [sp, #8]
+
+	ldr r7, =coef_y
+	ldr r7, [r7]
+	ldr r8, =coef_ub
+	ldr r8, [r8]
+	ldr r9, =coef_vr
+	ldr r9, [r9]
+
+	mov r10, #0
+
+process_row:
+	ldr r2, [sp, #8]
+	cmp r10, r2
+	bge conversion_done
+	mov r11, #0
+
+process_column:
+	ldr r2, [sp, #4]
+	cmp r11, r2
+	bge next_row_565
+
+	lsr r4, r10, #1
+	ldr r2, [sp, #0]
+	mul r4, r4, r2
+	lsr r2, r11, #1
+	add r4, r4, r2
+
+	ldrb r2, [r5, r4]
+	ldrb r3, [r6, r4]
+	sub r2, #128
+	sub r3, #128
+
+	mul r12, r2, r8
+	mul r14, r3, r9
+	mov r4, #100
+	mul r4, r2, r4
+	mov r2, #208
+	mla r4, r3, r2, r4
+	str r4, [sp, #12]
+
+	ldr r2, [sp, #4]
+	mul r3, r10, r2
+	add r3, r11
+	ldrb r4, [r0, r3]
+	sub r4, #16
+	mul r4, r7
+
+	add r2, r4, r12
+	add r2, #128
+	asr r2, #8
+	usat r2, #8, r2
+	lsr r2, #3
+
+	add r3, r4, r14
+	add r3, #128
+	asr r3, #8
+	usat r3, #8, r3
+	lsr r3, #3
+
+	ldr r4, [sp, #12]
+	rsb r4, r4, #0
+	ldr r12, [sp, #4]
+	mul r14, r10, r12
+	add r14, r11
+	ldrb r12, [r0, r14]
+	sub r12, #16
+	mul r12, r7
+	add r4, r12, r4
+	add r4, #128
+	asr r4, #8
+	usat r4, #8, r4
+	lsr r4, #2
+
+	orr r2, r2, r4, lsl #5
+	orr r2, r2, r3, lsl #11
+	lsl r3, r14, #1
+	strh r2, [r1, r3]
+
+	add r11, #1
+	b process_column
+
+next_row_565:
+	add r10, #1
+	b process_row
+
+conversion_done:
+	add sp, sp, #16
+	pop {r4-r11, pc}
+
+// r0 = Y plane, r1 = output, r2 = width, r3 = height
+yuv420p_to_bgr888_asm:
+	push {r4-r11, lr}
+	sub sp, sp, #16
+
+	mul r4, r2, r3
+	add r5, r0, r4
+	add r6, r5, r4, lsr #2
+	lsr r4, r2, #1
+	str r4, [sp, #0]
+	str r2, [sp, #4]
+	str r3, [sp, #8]
+
+	ldr r7, =coef_y
+	ldr r7, [r7]
+	ldr r8, =coef_ub
+	ldr r8, [r8]
+	ldr r9, =coef_vr
+	ldr r9, [r9]
+
+	mov r10, #0
+
+process_row_888:
+	ldr r2, [sp, #8]
+	cmp r10, r2
+	bge conversion_done_888
+	mov r11, #0
+
+process_column_888:
+	ldr r2, [sp, #4]
+	cmp r11, r2
+	bge next_row_888
+
+	lsr r4, r10, #1
+	ldr r2, [sp, #0]
+	mul r4, r4, r2
+	lsr r2, r11, #1
+	add r4, r4, r2
+
+	ldrb r2, [r5, r4]
+	ldrb r3, [r6, r4]
+	sub r2, #128
+	sub r3, #128
+
+	mul r12, r2, r8
+	mul r14, r3, r9
+	mov r4, #100
+	mul r4, r2, r4
+	mov r2, #208
+	mla r4, r3, r2, r4
+	str r4, [sp, #12]
+
+	ldr r2, [sp, #4]
+	mul r3, r10, r2
+	add r3, r11
+	ldrb r4, [r0, r3]
+	sub r4, #16
+	mul r4, r7
+
+	add r2, r4, r12
+	add r2, #128
+	asr r2, #8
+	usat r2, #8, r2
+
+	add r3, r4, r14
+	add r3, #128
+	asr r3, #8
+	usat r3, #8, r3
+
+	ldr r4, [sp, #12]
+	rsb r4, r4, #0
+	ldr r12, [sp, #4]
+	mul r14, r10, r12
+	add r14, r11
+	ldrb r12, [r0, r14]
+	sub r12, #16
+	mul r12, r7
+	add r4, r12, r4
+	add r4, #128
+	asr r4, #8
+	usat r4, #8, r4
+
+	add r12, r14, r14, lsl #1
+	strb r2, [r1, r12]
+	add r12, #1
+	strb r4, [r1, r12]
+	add r12, #1
+	strb r3, [r1, r12]
+
+	add r11, #1
+	b process_column_888
+
+next_row_888:
+	add r10, #1
+	b process_row_888
+
+conversion_done_888:
+	add sp, sp, #16
+	pop {r4-r11, pc}