ARM: NEON assembly optimization for function BlitRGBtoRGBPixelAlpha

src/video/SDL_blit_A.c

@@ -421,6 +421,23 @@ BlitRGBtoRGBPixelAlphaARMSIMD(SDL_BlitInfo * info)
 }
 #endif
 
+#if SDL_ARM_NEON_BLITTERS
+void BlitRGBtoRGBPixelAlphaARMNEONAsm(int32_t w, int32_t h, uint32_t *dst, int32_t dst_stride, uint32_t *src, int32_t src_stride);
+
+static void
+BlitRGBtoRGBPixelAlphaARMNEON(SDL_BlitInfo * info)
+{
+	int32_t width = info->dst_w;
+	int32_t height = info->dst_h;
+	uint32_t *dstp = (uint32_t *)info->dst;
+	int32_t dststride = width + (info->dst_skip >> 2);
+	uint32_t *srcp = (uint32_t *)info->src;
+	int32_t srcstride = width + (info->src_skip >> 2);
+
+	BlitRGBtoRGBPixelAlphaARMNEONAsm(width, height, dstp, dststride, srcp, srcstride);
+}
+#endif
+
 /* fast RGB888->(A)RGB888 blending with surface alpha=128 special case */
 static void
 BlitRGBtoRGBSurfaceAlpha128(SDL_BlitInfo * info)
@@ -1356,6 +1373,10 @@ SDL_CalculateBlitA(SDL_Surface * surface)
                 }
 #endif /* __MMX__ || __3dNOW__ */
                 if (sf->Amask == 0xff000000) {
+#if SDL_ARM_NEON_BLITTERS
+                    if (SDL_HasNEON())
+                        return BlitRGBtoRGBPixelAlphaARMNEON;
+#endif
 #if SDL_ARM_SIMD_BLITTERS
                     if (SDL_HasARMSIMD())
                         return BlitRGBtoRGBPixelAlphaARMSIMD;

src/video/arm/pixman-arm-neon-asm.S

@@ -0,0 +1,159 @@
+/*
+ * Copyright © 2009 Nokia Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * Author:  Siarhei Siamashka (siarhei.siamashka@nokia.com)
+ */
+
+/*
+ * Copyright (c) 2018 RISC OS Open Ltd
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* Prevent the stack from becoming executable for no reason... */
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif
+
+    .text
+    .fpu neon
+    .arch armv7a
+    .object_arch armv4
+    .eabi_attribute 10, 0 /* suppress Tag_FP_arch */
+    .eabi_attribute 12, 0 /* suppress Tag_Advanced_SIMD_arch */
+    .arm
+    .altmacro
+    .p2align 2
+
+#include "pixman-arm-asm.h"
+#include "pixman-arm-neon-asm.h"
+
+/* Global configuration options and preferences */
+
+/*
+ * The code can optionally make use of unaligned memory accesses to improve
+ * performance of handling leading/trailing pixels for each scanline.
+ * Configuration variable RESPECT_STRICT_ALIGNMENT can be set to 0 for
+ * example in linux if unaligned memory accesses are not configured to
+ * generate.exceptions.
+ */
+.set RESPECT_STRICT_ALIGNMENT, 1
+
+/*
+ * Set default prefetch type. There is a choice between the following options:
+ *
+ * PREFETCH_TYPE_NONE (may be useful for the ARM cores where PLD is set to work
+ * as NOP to workaround some HW bugs or for whatever other reason)
+ *
+ * PREFETCH_TYPE_SIMPLE (may be useful for simple single-issue ARM cores where
+ * advanced prefetch intruduces heavy overhead)
+ *
+ * PREFETCH_TYPE_ADVANCED (useful for superscalar cores such as ARM Cortex-A8
+ * which can run ARM and NEON instructions simultaneously so that extra ARM
+ * instructions do not add (many) extra cycles, but improve prefetch efficiency)
+ *
+ * Note: some types of function can't support advanced prefetch and fallback
+ *       to simple one (those which handle 24bpp pixels)
+ */
+.set PREFETCH_TYPE_DEFAULT, PREFETCH_TYPE_ADVANCED
+
+/* Prefetch distance in pixels for simple prefetch */
+.set PREFETCH_DISTANCE_SIMPLE, 64
+
+/******************************************************************************/
+
+.macro RGBtoRGBPixelAlpha_process_pixblock_head
+    vmvn        d30, d3  /* get inverted source alpha */
+    vmov        d31, d7  /* dest alpha is always unchanged */
+    vmull.u8    q14, d0, d3
+    vmlal.u8    q14, d4, d30
+    vmull.u8    q0, d1, d3
+    vmlal.u8    q0, d5, d30
+    vmull.u8    q1, d2, d3
+    vmlal.u8    q1, d6, d30
+    vrshr.u16   q2, q14, #8
+    vrshr.u16   q3, q0, #8
+    vraddhn.u16 d28, q14, q2
+    vrshr.u16   q2, q1, #8
+    vraddhn.u16 d29, q0, q3
+    vraddhn.u16 d30, q1, q2
+.endm
+
+.macro RGBtoRGBPixelAlpha_process_pixblock_tail
+    /* nothing */
+.endm
+
+.macro RGBtoRGBPixelAlpha_process_pixblock_tail_head
+    vld4.8      {d0-d3}, [SRC]!
+                                    PF add PF_X, PF_X, #8
+        vst4.8      {d28-d31}, [DST_W :128]!
+                                    PF tst PF_CTL, #0xF
+    vld4.8      {d4-d7}, [DST_R :128]!
+                                    PF addne PF_X, PF_X, #8
+    vmvn        d30, d3  /* get inverted source alpha */
+    vmov        d31, d7  /* dest alpha is always unchanged */
+    vmull.u8    q14, d0, d3
+                                    PF subne PF_CTL, PF_CTL, #1
+    vmlal.u8    q14, d4, d30
+                                    PF cmp PF_X, ORIG_W
+    vmull.u8    q0, d1, d3
+                                    PF pld, [PF_SRC, PF_X, lsl #src_bpp_shift]
+    vmlal.u8    q0, d5, d30
+                                    PF pld, [PF_DST, PF_X, lsl #dst_bpp_shift]
+    vmull.u8    q1, d2, d3
+                                    PF subge PF_X, PF_X, ORIG_W
+    vmlal.u8    q1, d6, d30
+                                    PF subges PF_CTL, PF_CTL, #0x10
+    vrshr.u16   q2, q14, #8
+                                    PF ldrgeb DUMMY, [PF_SRC, SRC_STRIDE, lsl #src_bpp_shift]!
+    vrshr.u16   q3, q0, #8
+                                    PF ldrgeb DUMMY, [PF_DST, DST_STRIDE, lsl #dst_bpp_shift]!
+    vraddhn.u16 d28, q14, q2
+    vrshr.u16   q2, q1, #8
+    vraddhn.u16 d29, q0, q3
+    vraddhn.u16 d30, q1, q2
+.endm
+
+generate_composite_function \
+    BlitRGBtoRGBPixelAlphaARMNEONAsm, 32, 0, 32, \
+    FLAG_DST_READWRITE | FLAG_DEINTERLEAVE_32BPP, \
+    8, /* number of pixels, processed in a single block */ \
+    5, /* prefetch distance */ \
+    default_init, \
+    default_cleanup, \
+    RGBtoRGBPixelAlpha_process_pixblock_head, \
+    RGBtoRGBPixelAlpha_process_pixblock_tail, \
+    RGBtoRGBPixelAlpha_process_pixblock_tail_head