summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorrubidium <rubidium@openttd.org>2014-01-02 23:52:13 +0000
committerrubidium <rubidium@openttd.org>2014-01-02 23:52:13 +0000
commit4c84d13454414b84b72540265b793aeec938cdd3 (patch)
treea69300b5060379854a96a10bbe1fa786011403dc
parent78df732a7b7e987b66c85eeb243ce5e0dd59d3a6 (diff)
downloadopenttd-4c84d13454414b84b72540265b793aeec938cdd3.tar.xz
(svn r26214) -Add: specialised animated SSE4 blitter (MJP)
With 32bpp base set about 15-20% faster in the Draw function (slower with 8bpp base set). Overall, with 32bpp base set, about 5% faster.
-rw-r--r--projects/openttd_vs100.vcxproj2
-rw-r--r--projects/openttd_vs100.vcxproj.filters6
-rw-r--r--projects/openttd_vs80.vcproj8
-rw-r--r--projects/openttd_vs90.vcproj8
-rw-r--r--source.list4
-rw-r--r--src/blitter/32bpp_anim.hpp4
-rw-r--r--src/blitter/32bpp_anim_sse4.cpp410
-rw-r--r--src/blitter/32bpp_anim_sse4.hpp47
-rw-r--r--src/blitter/32bpp_sse2.cpp12
-rw-r--r--src/blitter/32bpp_sse2.hpp23
-rw-r--r--src/blitter/32bpp_sse4.cpp4
-rw-r--r--src/blitter/32bpp_sse4.hpp4
-rw-r--r--src/blitter/32bpp_ssse3.hpp2
13 files changed, 516 insertions, 18 deletions
diff --git a/projects/openttd_vs100.vcxproj b/projects/openttd_vs100.vcxproj
index dccd6adb2..127077c77 100644
--- a/projects/openttd_vs100.vcxproj
+++ b/projects/openttd_vs100.vcxproj
@@ -1115,6 +1115,8 @@
<ClCompile Include="..\src\script\api\script_window.cpp" />
<ClCompile Include="..\src\blitter\32bpp_anim.cpp" />
<ClInclude Include="..\src\blitter\32bpp_anim.hpp" />
+ <ClCompile Include="..\src\blitter\32bpp_anim_sse4.cpp" />
+ <ClInclude Include="..\src\blitter\32bpp_anim_sse4.hpp" />
<ClCompile Include="..\src\blitter\32bpp_base.cpp" />
<ClInclude Include="..\src\blitter\32bpp_base.hpp" />
<ClCompile Include="..\src\blitter\32bpp_optimized.cpp" />
diff --git a/projects/openttd_vs100.vcxproj.filters b/projects/openttd_vs100.vcxproj.filters
index d40f97328..7ecffde58 100644
--- a/projects/openttd_vs100.vcxproj.filters
+++ b/projects/openttd_vs100.vcxproj.filters
@@ -2574,6 +2574,12 @@
<ClInclude Include="..\src\blitter\32bpp_anim.hpp">
<Filter>Blitters</Filter>
</ClInclude>
+ <ClCompile Include="..\src\blitter\32bpp_anim_sse4.cpp">
+ <Filter>Blitters</Filter>
+ </ClCompile>
+ <ClInclude Include="..\src\blitter\32bpp_anim_sse4.hpp">
+ <Filter>Blitters</Filter>
+ </ClInclude>
<ClCompile Include="..\src\blitter\32bpp_base.cpp">
<Filter>Blitters</Filter>
</ClCompile>
diff --git a/projects/openttd_vs80.vcproj b/projects/openttd_vs80.vcproj
index cadc6a966..a343c5acf 100644
--- a/projects/openttd_vs80.vcproj
+++ b/projects/openttd_vs80.vcproj
@@ -3803,6 +3803,14 @@
>
</File>
<File
+ RelativePath=".\..\src\blitter\32bpp_anim_sse4.cpp"
+ >
+ </File>
+ <File
+ RelativePath=".\..\src\blitter\32bpp_anim_sse4.hpp"
+ >
+ </File>
+ <File
RelativePath=".\..\src\blitter\32bpp_base.cpp"
>
</File>
diff --git a/projects/openttd_vs90.vcproj b/projects/openttd_vs90.vcproj
index 04f1eec2d..0faf59fbf 100644
--- a/projects/openttd_vs90.vcproj
+++ b/projects/openttd_vs90.vcproj
@@ -3800,6 +3800,14 @@
>
</File>
<File
+ RelativePath=".\..\src\blitter\32bpp_anim_sse4.cpp"
+ >
+ </File>
+ <File
+ RelativePath=".\..\src\blitter\32bpp_anim_sse4.hpp"
+ >
+ </File>
+ <File
RelativePath=".\..\src\blitter\32bpp_base.cpp"
>
</File>
diff --git a/source.list b/source.list
index d237247dc..4e6e2e795 100644
--- a/source.list
+++ b/source.list
@@ -901,6 +901,10 @@ script/api/script_window.cpp
#else
blitter/32bpp_anim.cpp
blitter/32bpp_anim.hpp
+#if WITH_SSE
+blitter/32bpp_anim_sse4.cpp
+blitter/32bpp_anim_sse4.hpp
+#end
blitter/32bpp_base.cpp
blitter/32bpp_base.hpp
blitter/32bpp_optimized.cpp
diff --git a/src/blitter/32bpp_anim.hpp b/src/blitter/32bpp_anim.hpp
index db5d614a0..e707c4437 100644
--- a/src/blitter/32bpp_anim.hpp
+++ b/src/blitter/32bpp_anim.hpp
@@ -15,8 +15,8 @@
#include "32bpp_optimized.hpp"
/** The optimised 32 bpp blitter with palette animation. */
-class Blitter_32bppAnim FINAL : public Blitter_32bppOptimized {
-private:
+class Blitter_32bppAnim : public Blitter_32bppOptimized {
+protected:
uint16 *anim_buf; ///< In this buffer we keep track of the 8bpp indexes so we can do palette animation
int anim_buf_width; ///< The width of the animation buffer.
int anim_buf_height; ///< The height of the animation buffer.
diff --git a/src/blitter/32bpp_anim_sse4.cpp b/src/blitter/32bpp_anim_sse4.cpp
new file mode 100644
index 000000000..c6008b007
--- /dev/null
+++ b/src/blitter/32bpp_anim_sse4.cpp
@@ -0,0 +1,410 @@
+/* $Id$ */
+
+/*
+ * This file is part of OpenTTD.
+ * OpenTTD 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, version 2.
+ * OpenTTD 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 OpenTTD. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/** @file 32bpp_sse4_anim.cpp Implementation of the SSE4 32 bpp blitter with animation support. */
+
+#ifdef WITH_SSE
+
+#include "../stdafx.h"
+#include "../video/video_driver.hpp"
+#include "../table/sprites.h"
+#include "32bpp_anim_sse4.hpp"
+
+/** Instantiation of the SSE4 32bpp blitter factory. */
+static FBlitter_32bppSSE4_Anim iFBlitter_32bppSSE4_Anim;
+
+#if defined(__GNUC__)
+ #pragma GCC diagnostic push
+ #pragma GCC diagnostic ignored "-Wunused-variable"
+#endif
+/**
+ * Draws a sprite to a (screen) buffer. It is templated to allow faster operation.
+ *
+ * @tparam mode blitter mode
+ * @param bp further blitting parameters
+ * @param zoom zoom level at which we are drawing
+ */
+template <BlitterMode mode, Blitter_32bppSSE2::ReadMode read_mode, Blitter_32bppSSE2::BlockType bt_last>
+inline void Blitter_32bppSSE4_Anim::Draw(const Blitter::BlitterParams *bp, ZoomLevel zoom)
+{
+ Colour *dst_line = (Colour *) bp->dst + bp->top * bp->pitch + bp->left;
+ uint16 *anim_line = this->anim_buf + ((uint32 *)bp->dst - (uint32 *)_screen.dst_ptr) + bp->top * this->anim_buf_width + bp->left;
+ int effective_width = bp->width;
+
+ /* Find where to start reading in the source sprite. */
+ const byte * const remap = bp->remap;
+ const Blitter_32bppSSE_Base::SpriteData * const sd = (const Blitter_32bppSSE_Base::SpriteData *) bp->sprite;
+ const SpriteInfo * const si = &sd->infos[zoom];
+ const MapValue *src_mv_line = (const MapValue *) &sd->data[si->mv_offset] + bp->skip_top * si->sprite_width;
+ const Colour *src_rgba_line = (const Colour *) ((const byte *) &sd->data[si->sprite_offset] + bp->skip_top * si->sprite_line_size);
+
+ if (read_mode != RM_WITH_MARGIN) {
+ src_rgba_line += bp->skip_left;
+ src_mv_line += bp->skip_left;
+ }
+
+ /* Load these variables into register before loop. */
+ const __m128i a_cm = ALPHA_CONTROL_MASK;
+ const __m128i pack_low_cm = PACK_LOW_CONTROL_MASK;
+ const __m128i briAB_cm = BRIGHTNESS_LOW_CONTROL_MASK;
+ const __m128i div_cleaner = BRIGHTNESS_DIV_CLEANER;
+ const __m128i ob_check = OVERBRIGHT_PRESENCE_MASK;
+ const __m128i ob_mask = OVERBRIGHT_VALUE_MASK;
+ const __m128i ob_cm = OVERBRIGHT_CONTROL_MASK;
+ const __m128i tr_nom_base = TRANSPARENT_NOM_BASE;
+
+ for (int y = bp->height; y != 0; y--) {
+ Colour *dst = dst_line;
+ const Colour *src = src_rgba_line + META_LENGTH;
+ const MapValue *src_mv = src_mv_line;
+ uint16 *anim = anim_line;
+
+ switch (mode) {
+ default: {
+ switch (read_mode) {
+ case RM_WITH_MARGIN: {
+ src += src_rgba_line[0].data;
+ dst += src_rgba_line[0].data;
+ src_mv += src_rgba_line[0].data;
+ anim += src_rgba_line[0].data;
+ const int width_diff = si->sprite_width - bp->width;
+ effective_width = bp->width - (int) src_rgba_line[0].data;
+ const int delta_diff = (int) src_rgba_line[1].data - width_diff;
+ const int new_width = effective_width - (delta_diff & ~1);
+ effective_width = delta_diff > 0 ? new_width : effective_width;
+ if (effective_width <= 0) break;
+ /* FALLTHROUGH */
+ }
+
+ case RM_WITH_SKIP: {
+ uint32 mvX2 = *((uint32 *) const_cast<MapValue *>(src_mv));
+ __m128i srcABCD = _mm_loadu_si128((const __m128i*) src);
+ __m128i dstABCD = _mm_loadu_si128((__m128i*) dst);
+ for (uint x = (uint) effective_width/2; x != 0; x--) {
+ /* Remap colours. */
+ const byte m0 = mvX2;
+ if (m0 >= PALETTE_ANIM_START) {
+ const Colour c0 = (this->LookupColourInPalette(m0).data & 0x00FFFFFF) | (src[0].data & 0xFF000000);
+ INSR32(AdjustBrightness(c0, (byte) (mvX2 >> 8)).data, srcABCD, 0);
+ }
+ const byte m1 = mvX2 >> 16;
+ if (m1 >= PALETTE_ANIM_START) {
+ const Colour c1 = (this->LookupColourInPalette(m1).data & 0x00FFFFFF) | (src[1].data & 0xFF000000);
+ INSR32(AdjustBrightness(c1, (byte) (mvX2 >> 24)).data, srcABCD, 1);
+ }
+
+ /* Update anim buffer. */
+ const byte a0 = src[0].a;
+ const byte a1 = src[1].a;
+ uint32 anim01 = 0;
+ if (a0 == 255) {
+ if (a1 == 255) {
+ *(uint32*) anim = mvX2;
+ goto bmno_full_opacity;
+ }
+ anim01 = (uint16) mvX2;
+ } else if (a0 == 0) {
+ if (a1 == 0) {
+ goto bmno_full_transparency;
+ } else {
+ if (a1 == 255) anim[1] = (uint16) (mvX2 >> 16);
+ goto bmno_alpha_blend;
+ }
+ }
+ if (a1 > 0) {
+ if (a1 == 255) anim01 |= mvX2 & 0xFFFF0000;
+ *(uint32*) anim = anim01;
+ } else {
+ anim[0] = (uint16) anim01;
+ }
+
+ /* Blend colours. */
+bmno_alpha_blend:
+ ALPHA_BLEND_2(pack_low_cm);
+bmno_full_opacity:
+ srcABCD = _mm_blend_epi16(srcABCD, dstABCD, 0xF0);
+
+ src_mv += 2;
+ mvX2 = *((uint32 *) const_cast<MapValue *>(src_mv));
+ src += 2;
+ anim += 2;
+ dstABCD = _mm_loadu_si128((__m128i*) (dst+2));
+ _mm_storeu_si128((__m128i *) dst, srcABCD);
+ srcABCD = _mm_loadu_si128((const __m128i*) src);
+ dst += 2;
+ continue;
+
+bmno_full_transparency:
+ src_mv += 2;
+ mvX2 = *((uint32 *) const_cast<MapValue *>(src_mv));
+ dst += 2;
+ src += 2;
+ anim += 2;
+ dstABCD = _mm_loadu_si128((__m128i*) dst);
+ srcABCD = _mm_loadu_si128((const __m128i*) src);
+ }
+
+ if (bt_last == BT_ODD) {
+ if (src->a == 0) {
+ } else if (src->a == 255) {
+ *anim = (uint16) mvX2;
+ *dst = ((byte) mvX2 >= PALETTE_ANIM_START) ? AdjustBrightness(LookupColourInPalette((byte) mvX2), (byte) (mvX2 >> 8)) : *src;
+ } else {
+ *anim = 0;
+ if ((byte) mvX2 >= PALETTE_ANIM_START) {
+ ALIGN(16) Colour colour = AdjustBrightness(LookupColourInPalette((byte) mvX2), (byte) (mvX2 >> 8));
+ colour.a = src->a;
+ srcABCD = _mm_load_si128((__m128i*) &colour);
+ }
+ ALPHA_BLEND_2(pack_low_cm);
+ (*dst).data = EXTR32(srcABCD, 0);
+ }
+ }
+ break;
+ }
+
+ default: NOT_REACHED();
+ }
+ break;
+ }
+
+ case BM_COLOUR_REMAP: {
+ switch (read_mode) {
+ case RM_WITH_MARGIN: {
+ src += src_rgba_line[0].data;
+ src_mv += src_rgba_line[0].data;
+ dst += src_rgba_line[0].data;
+ anim += src_rgba_line[0].data;
+ const int width_diff = si->sprite_width - bp->width;
+ effective_width = bp->width - (int) src_rgba_line[0].data;
+ const int delta_diff = (int) src_rgba_line[1].data - width_diff;
+ const int nd = effective_width - delta_diff;
+ effective_width = delta_diff > 0 ? nd : effective_width;
+ if (effective_width <= 0) break;
+ /* FALLTHROUGH */
+ }
+
+ case RM_WITH_SKIP: {
+ uint32 mvX2 = *((uint32 *) const_cast<MapValue *>(src_mv));
+ __m128i srcABCD = _mm_loadu_si128((const __m128i*) src);
+ __m128i dstABCD = _mm_loadu_si128((__m128i*) dst);
+
+ for (uint x = (uint) effective_width / 2; x != 0; x--) {
+ /* Remap colours. */
+ const uint m0 = (byte) mvX2;
+ const uint r0 = remap[m0];
+ const uint m1 = (byte) (mvX2 >> 16);
+ const uint r1 = remap[m1];
+ if (mvX2 & 0x00FF00FF) {
+ /* Written so the compiler uses CMOV. */
+ const Colour src0 = src[0];
+ const Colour c0map = (this->LookupColourInPalette(r0).data & 0x00FFFFFF) | (src0.data & 0xFF000000);
+ Colour c0 = dst[0];
+ c0 = r0 == 0 ? c0 : c0map;
+ c0 = m0 != 0 ? c0 : src0;
+ INSR32(c0.data, srcABCD, 0);
+
+ const Colour src1 = src[1];
+ const Colour c1map = (this->LookupColourInPalette(r1).data & 0x00FFFFFF) | (src1.data & 0xFF000000);
+ Colour c1 = dst[1];
+ c1 = r1 == 0 ? c1 : c1map;
+ c1 = m1 != 0 ? c1 : src1;
+ INSR32(c1.data, srcABCD, 1);
+
+ if ((mvX2 & 0xFF00FF00) != 0x80008000) {
+ ADJUST_BRIGHTNESS_2(srcABCD, mvX2);
+ }
+ }
+
+ /* Update anim buffer. */
+ const byte a0 = src[0].a;
+ const byte a1 = src[1].a;
+ uint32 anim01 = mvX2 & 0xFF00FF00;
+ if (a0 == 255) {
+ anim01 |= r0;
+ if (a1 == 255) {
+ *(uint32*) anim = anim01 | (r1 << 16);
+ goto bmcr_full_opacity;
+ }
+ } else if (a0 == 0) {
+ if (a1 == 0) {
+ goto bmcr_full_transparency;
+ } else {
+ if (a1 == 255) {
+ anim[1] = r1 | (anim01 >> 16);
+ }
+ goto bmcr_alpha_blend;
+ }
+ }
+ if (a1 > 0) {
+ if (a1 == 255) anim01 |= r1 << 16;
+ *(uint32*) anim = anim01;
+ } else {
+ anim[0] = (uint16) anim01;
+ }
+
+ /* Blend colours. */
+bmcr_alpha_blend:
+ ALPHA_BLEND_2(pack_low_cm);
+bmcr_full_opacity:
+ srcABCD = _mm_blend_epi16(srcABCD, dstABCD, 0xF0);
+
+ src += 2;
+ src_mv += 2;
+ anim += 2;
+ mvX2 = *((uint32 *) const_cast<MapValue *>(src_mv));
+ dstABCD = _mm_loadu_si128((__m128i*) (dst+2));
+ _mm_storeu_si128((__m128i *) dst, srcABCD);
+ srcABCD = _mm_loadu_si128((const __m128i*) src);
+ dst += 2;
+ continue;
+
+bmcr_full_transparency:
+ src_mv += 2;
+ mvX2 = *((uint32 *) const_cast<MapValue *>(src_mv));
+ dst += 2;
+ src += 2;
+ anim += 2;
+ dstABCD = _mm_loadu_si128((__m128i*) dst);
+ srcABCD = _mm_loadu_si128((const __m128i*) src);
+ }
+
+ if (effective_width & 1) {
+ /* In case the m-channel is zero, do not remap this pixel in any way. */
+ if (src->a == 0) {
+ } else if ((byte) mvX2 != 0) {
+ const uint r = remap[(byte) mvX2];
+ *anim = (src->a == 255) ? (r | ((uint16) mvX2 & 0xFF00)) : 0;
+ if (r != 0) {
+ Colour remapped_colour = AdjustBrightness(LookupColourInPalette(r), (byte) (mvX2 >> 8));
+ if (src->a == 255) {
+ *dst = remapped_colour;
+ } else {
+ remapped_colour.a = src->a;
+ INSR32(remapped_colour.data, srcABCD, 0);
+ goto bmcr_alpha_blend_single;
+ }
+ }
+ } else {
+ *anim = 0;
+ if (src->a == 255) {
+ *dst = *src;
+ } else {
+bmcr_alpha_blend_single:
+ ALPHA_BLEND_2(pack_low_cm);
+ (*dst).data = EXTR32(srcABCD, 0);
+ }
+ }
+ }
+ break;
+ }
+
+ default: NOT_REACHED();
+ }
+ break;
+ }
+
+ case BM_TRANSPARENT: {
+ /* Make the current colour a bit more black, so it looks like this image is transparent. */
+ __m128i srcABCD = _mm_loadu_si128((const __m128i*) src);
+ __m128i dstABCD = _mm_loadu_si128((__m128i*) dst);
+ for (uint x = (uint) bp->width / 2; x > 0; x--) {
+ __m128i srcAB = _mm_unpacklo_epi8(srcABCD, _mm_setzero_si128());
+ __m128i dstAB = _mm_unpacklo_epi8(dstABCD, _mm_setzero_si128());
+ __m128i dstCD = _mm_unpackhi_epi8(dstABCD, _mm_setzero_si128());
+ __m128i alphaAB = _mm_shuffle_epi8(srcAB, a_cm);
+ alphaAB = _mm_srli_epi16(alphaAB, 2); // Reduce to 64 levels of shades so the max value fits in 16 bits.
+ __m128i nom = _mm_sub_epi16(tr_nom_base, alphaAB);
+ dstAB = _mm_mullo_epi16(dstAB, nom);
+ dstAB = _mm_srli_epi16(dstAB, 8);
+ dstAB = _mm_packus_epi16(dstAB, dstCD);
+ Colour *old_dst = dst;
+ src += 2;
+ dst += 2;
+ anim += 2;
+ dstABCD = _mm_loadu_si128((__m128i*) dst);
+ _mm_storeu_si128((__m128i *) old_dst, dstAB);
+ srcABCD = _mm_loadu_si128((const __m128i*) src);
+ if (src[-2].a) anim[-2] = 0;
+ if (src[-1].a) anim[-1] = 0;
+ }
+ if (bp->width & 1) {
+ __m128i srcAB = _mm_unpacklo_epi8(srcABCD, _mm_setzero_si128());
+ __m128i dstAB = _mm_unpacklo_epi8(dstABCD, _mm_setzero_si128());
+ __m128i alphaAB = _mm_shuffle_epi8(srcAB, a_cm);
+ alphaAB = _mm_srli_epi16(alphaAB, 2);
+ __m128i nom = _mm_sub_epi16(tr_nom_base, alphaAB);
+ dstAB = _mm_mullo_epi16(dstAB, nom);
+ dstAB = _mm_srli_epi16(dstAB, 8);
+ dstAB = _mm_packus_epi16(dstAB, dstAB);
+ (*dst).data = EXTR32(dstAB, 0);
+ if (src[0].a) anim[0] = 0;
+ }
+ break;
+ }
+ }
+
+ src_mv_line += si->sprite_width;
+ src_rgba_line = (const Colour*) ((const byte*) src_rgba_line + si->sprite_line_size);
+ dst_line += bp->pitch;
+ anim_line += this->anim_buf_width;
+ }
+}
+#if defined(__GNUC__)
+ #pragma GCC diagnostic pop
+#endif
+
+/**
+ * Draws a sprite to a (screen) buffer. Calls adequate templated function.
+ *
+ * @param bp further blitting parameters
+ * @param mode blitter mode
+ * @param zoom zoom level at which we are drawing
+ */
+void Blitter_32bppSSE4_Anim::Draw(Blitter::BlitterParams *bp, BlitterMode mode, ZoomLevel zoom)
+{
+ switch (mode) {
+ case BM_NORMAL: {
+ const BlockType bt_last = (BlockType) (bp->width & 1);
+ if (bp->skip_left != 0 || bp->width <= MARGIN_THRESHOLD) {
+ switch (bt_last) {
+ case BT_EVEN: Draw<BM_NORMAL, RM_WITH_SKIP, BT_EVEN>(bp, zoom); return;
+ case BT_ODD: Draw<BM_NORMAL, RM_WITH_SKIP, BT_ODD>(bp, zoom); return;
+ default: NOT_REACHED();
+ }
+ } else {
+ switch (bt_last) {
+ case BT_EVEN: Draw<BM_NORMAL, RM_WITH_MARGIN, BT_EVEN>(bp, zoom); return;
+ case BT_ODD: Draw<BM_NORMAL, RM_WITH_MARGIN, BT_ODD>(bp, zoom); return;
+ default: NOT_REACHED();
+ }
+ }
+ break;
+ }
+ case BM_COLOUR_REMAP:
+ if (bp->skip_left != 0 || bp->width <= MARGIN_THRESHOLD) {
+ Draw<BM_COLOUR_REMAP, RM_WITH_SKIP, BT_NONE>(bp, zoom); return;
+ } else {
+ Draw<BM_COLOUR_REMAP, RM_WITH_MARGIN, BT_NONE>(bp, zoom); return;
+ }
+ case BM_TRANSPARENT: Draw<BM_TRANSPARENT, RM_NONE, BT_NONE>(bp, zoom); return;
+ default: NOT_REACHED();
+ }
+}
+
+/** Same code as seen in 32bpp_sse2.cpp but some macros are not the same. */
+inline Colour Blitter_32bppSSE4_Anim::AdjustBrightness(Colour colour, uint8 brightness)
+{
+ /* Shortcut for normal brightness. */
+ if (brightness == DEFAULT_BRIGHTNESS) return colour;
+
+ return Blitter_32bppSSE4::ReallyAdjustBrightness(colour, brightness);
+}
+
+#endif /* WITH_SSE */
diff --git a/src/blitter/32bpp_anim_sse4.hpp b/src/blitter/32bpp_anim_sse4.hpp
new file mode 100644
index 000000000..90aec22d8
--- /dev/null
+++ b/src/blitter/32bpp_anim_sse4.hpp
@@ -0,0 +1,47 @@
+/* $Id$ */
+
+/*
+ * This file is part of OpenTTD.
+ * OpenTTD 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, version 2.
+ * OpenTTD 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 OpenTTD. If not, see <http://www.gnu.org/licenses/>.
+ */
+
+/** @file 32bpp_sse4_anim.hpp A SSE4 32 bpp blitter with animation support. */
+
+#ifndef BLITTER_32BPP_SSE4_ANIM_HPP
+#define BLITTER_32BPP_SSE4_ANIM_HPP
+
+#ifdef WITH_SSE
+
+#include "32bpp_anim.hpp"
+#include "32bpp_sse4.hpp"
+
+#undef MARGIN_THRESHOLD
+#define MARGIN_THRESHOLD 4
+
+/** The SSE4 32 bpp blitter with palette animation. */
+class Blitter_32bppSSE4_Anim FINAL : public Blitter_32bppAnim, public Blitter_32bppSSE_Base {
+private:
+
+public:
+ template <BlitterMode mode, Blitter_32bppSSE_Base::ReadMode read_mode, Blitter_32bppSSE_Base::BlockType bt_last>
+ /* virtual */ void Draw(const Blitter::BlitterParams *bp, ZoomLevel zoom);
+ /* virtual */ void Draw(Blitter::BlitterParams *bp, BlitterMode mode, ZoomLevel zoom);
+ /* virtual */ Colour AdjustBrightness(Colour colour, uint8 brightness);
+ /* virtual */ Sprite *Encode(const SpriteLoader::Sprite *sprite, AllocatorProc *allocator) {
+ return Blitter_32bppSSE_Base::Encode(sprite, allocator);
+ }
+
+ /* virtual */ const char *GetName() { return "32bpp-sse4-anim"; }
+};
+
+/** Factory for the SSE4 32 bpp blitter (with palette animation). */
+class FBlitter_32bppSSE4_Anim: public BlitterFactory {
+public:
+ FBlitter_32bppSSE4_Anim() : BlitterFactory("32bpp-sse4-anim", "SSE4 Blitter (palette animation)", HasCPUIDFlag(1, 2, 19)) {}
+ /* virtual */ Blitter *CreateInstance() { return new Blitter_32bppSSE4_Anim(); }
+};
+
+#endif /* WITH_SSE */
+#endif /* BLITTER_32BPP_SSE4_ANIM_HPP */
diff --git a/src/blitter/32bpp_sse2.cpp b/src/blitter/32bpp_sse2.cpp
index 09c0ffa8a..52f37d690 100644
--- a/src/blitter/32bpp_sse2.cpp
+++ b/src/blitter/32bpp_sse2.cpp
@@ -200,7 +200,7 @@ void Blitter_32bppSSE2::Draw(Blitter::BlitterParams *bp, BlitterMode mode, ZoomL
}
}
-Sprite *Blitter_32bppSSE2::Encode(const SpriteLoader::Sprite *sprite, AllocatorProc *allocator)
+Sprite *Blitter_32bppSSE_Base::Encode(const SpriteLoader::Sprite *sprite, AllocatorProc *allocator)
{
/* First uint32 of a line = ~1 & the number of transparent pixels from the left.
* Second uint32 of a line = the number of transparent pixels from the right.
@@ -252,10 +252,10 @@ Sprite *Blitter_32bppSSE2::Encode(const SpriteLoader::Sprite *sprite, AllocatorP
if (src->m != 0) {
/* Get brightest value (or default brightness if it's a black pixel). */
const uint8 rgb_max = max(src->r, max(src->g, src->b));
- dst_mv->v = (rgb_max == 0) ? DEFAULT_BRIGHTNESS : rgb_max;
+ dst_mv->v = (rgb_max == 0) ? Blitter_32bppBase::DEFAULT_BRIGHTNESS : rgb_max;
/* Pre-convert the mapping channel to a RGB value. */
- const Colour colour = AdjustBrightness(LookupColourInPalette(src->m), dst_mv->v);
+ const Colour colour = AdjustBrightness(Blitter_32bppBase::LookupColourInPalette(src->m), dst_mv->v);
dst_rgba->r = colour.r;
dst_rgba->g = colour.g;
dst_rgba->b = colour.b;
@@ -263,7 +263,7 @@ Sprite *Blitter_32bppSSE2::Encode(const SpriteLoader::Sprite *sprite, AllocatorP
dst_rgba->r = src->r;
dst_rgba->g = src->g;
dst_rgba->b = src->b;
- dst_mv->v = DEFAULT_BRIGHTNESS;
+ dst_mv->v = Blitter_32bppBase::DEFAULT_BRIGHTNESS;
}
} else {
dst_rgba->data = 0;
@@ -310,10 +310,10 @@ inline Colour Blitter_32bppSSE2::AdjustBrightness(Colour colour, uint8 brightnes
/* Shortcut for normal brightness. */
if (brightness == DEFAULT_BRIGHTNESS) return colour;
- return this->ReallyAdjustBrightness(colour, brightness);
+ return Blitter_32bppSSE2::ReallyAdjustBrightness(colour, brightness);
}
-Colour Blitter_32bppSSE2::ReallyAdjustBrightness(Colour colour, uint8 brightness)
+/* static */ Colour Blitter_32bppSSE2::ReallyAdjustBrightness(Colour colour, uint8 brightness)
{
ALIGN(16) uint64 c16 = colour.b | (uint64) colour.g << 16 | (uint64) colour.r << 32;
c16 *= brightness;
diff --git a/src/blitter/32bpp_sse2.hpp b/src/blitter/32bpp_sse2.hpp
index 0e11510cf..a0ed74cdb 100644
--- a/src/blitter/32bpp_sse2.hpp
+++ b/src/blitter/32bpp_sse2.hpp
@@ -73,9 +73,11 @@ typedef union ALIGN(16) um128i {
srcABCD = _mm_packus_epi16(srcAB, srcAB); /* PACKUSWB, pack 2 colours (with saturation) */ \
}
-/** The SSE2 32 bpp blitter (without palette animation). */
-class Blitter_32bppSSE2 : public Blitter_32bppSimple {
+/** Base methods for 32bpp SSE blitters. */
+class Blitter_32bppSSE_Base {
public:
+ virtual ~Blitter_32bppSSE_Base() {}
+
struct MapValue {
uint8 m;
uint8 v;
@@ -108,12 +110,23 @@ public:
byte data[]; ///< Data, all zoomlevels.
};
+ Sprite *Encode(const SpriteLoader::Sprite *sprite, AllocatorProc *allocator);
+ virtual Colour AdjustBrightness(Colour colour, uint8 brightness) = 0;
+};
+
+/** The SSE2 32 bpp blitter (without palette animation). */
+class Blitter_32bppSSE2 : public Blitter_32bppSimple, public Blitter_32bppSSE_Base {
+public:
virtual Colour AdjustBrightness(Colour colour, uint8 brightness);
- Colour ReallyAdjustBrightness(Colour colour, uint8 brightness);
+ static Colour ReallyAdjustBrightness(Colour colour, uint8 brightness);
/* virtual */ void Draw(Blitter::BlitterParams *bp, BlitterMode mode, ZoomLevel zoom);
- template <BlitterMode mode, ReadMode read_mode, BlockType bt_last>
+ template <BlitterMode mode, Blitter_32bppSSE_Base::ReadMode read_mode, Blitter_32bppSSE_Base::BlockType bt_last>
void Draw(const Blitter::BlitterParams *bp, ZoomLevel zoom);
- /* virtual */ Sprite *Encode(const SpriteLoader::Sprite *sprite, AllocatorProc *allocator);
+
+ /* virtual */ Sprite *Encode(const SpriteLoader::Sprite *sprite, AllocatorProc *allocator) {
+ return Blitter_32bppSSE_Base::Encode(sprite, allocator);
+ }
+
/* virtual */ const char *GetName() { return "32bpp-sse2"; }
};
diff --git a/src/blitter/32bpp_sse4.cpp b/src/blitter/32bpp_sse4.cpp
index 7243f754b..7d2bc91aa 100644
--- a/src/blitter/32bpp_sse4.cpp
+++ b/src/blitter/32bpp_sse4.cpp
@@ -289,10 +289,10 @@ inline Colour Blitter_32bppSSE4::AdjustBrightness(Colour colour, uint8 brightnes
/* Shortcut for normal brightness. */
if (brightness == DEFAULT_BRIGHTNESS) return colour;
- return this->ReallyAdjustBrightness(colour, brightness);
+ return Blitter_32bppSSE4::ReallyAdjustBrightness(colour, brightness);
}
-Colour Blitter_32bppSSE4::ReallyAdjustBrightness(Colour colour, uint8 brightness)
+/* static */ Colour Blitter_32bppSSE4::ReallyAdjustBrightness(Colour colour, uint8 brightness)
{
ALIGN(16) uint64 c16 = colour.b | (uint64) colour.g << 16 | (uint64) colour.r << 32;
c16 *= brightness;
diff --git a/src/blitter/32bpp_sse4.hpp b/src/blitter/32bpp_sse4.hpp
index 1930404b6..6545c1692 100644
--- a/src/blitter/32bpp_sse4.hpp
+++ b/src/blitter/32bpp_sse4.hpp
@@ -48,10 +48,10 @@
class Blitter_32bppSSE4 : public Blitter_32bppSSSE3 {
public:
Colour AdjustBrightness(Colour colour, uint8 brightness);
- Colour ReallyAdjustBrightness(Colour colour, uint8 brightness);
+ static Colour ReallyAdjustBrightness(Colour colour, uint8 brightness);
/* virtual */ void Draw(Blitter::BlitterParams *bp, BlitterMode mode, ZoomLevel zoom);
- template <BlitterMode mode, ReadMode read_mode, BlockType bt_last>
+ template <BlitterMode mode, Blitter_32bppSSE_Base::ReadMode read_mode, Blitter_32bppSSE_Base::BlockType bt_last>
void Draw(const Blitter::BlitterParams *bp, ZoomLevel zoom);
/* virtual */ const char *GetName() { return "32bpp-sse4"; }
};
diff --git a/src/blitter/32bpp_ssse3.hpp b/src/blitter/32bpp_ssse3.hpp
index 7ea1fbd8b..cede4185d 100644
--- a/src/blitter/32bpp_ssse3.hpp
+++ b/src/blitter/32bpp_ssse3.hpp
@@ -78,7 +78,7 @@
class Blitter_32bppSSSE3 : public Blitter_32bppSSE2 {
public:
/* virtual */ void Draw(Blitter::BlitterParams *bp, BlitterMode mode, ZoomLevel zoom);
- template <BlitterMode mode, ReadMode read_mode, BlockType bt_last>
+ template <BlitterMode mode, Blitter_32bppSSE_Base::ReadMode read_mode, Blitter_32bppSSE_Base::BlockType bt_last>
void Draw(const Blitter::BlitterParams *bp, ZoomLevel zoom);
/* virtual */ const char *GetName() { return "32bpp-ssse3"; }
};