2 * This file is part of OpenTTD.
3 * 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.
4 * 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.
5 * 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/>.
8 /** @file 32bpp_sse_func.hpp Functions related to SSE 32 bpp blitter. */
10 #ifndef BLITTER_32BPP_SSE_FUNC_HPP
11 #define BLITTER_32BPP_SSE_FUNC_HPP
14 * This file is compiled multiple times with different defines for SSE_VERSION and MARGIN_NORMAL_THRESHOLD.
15 * Be careful when declaring things with external linkage.
16 * Use internal linkage instead, i.e. "static".
18 #define INTERNAL_LINKAGE static
22 GNU_TARGET(SSE_TARGET
)
23 INTERNAL_LINKAGE
inline void InsertFirstUint32(const uint32_t value
, __m128i
&into
)
25 #if (SSE_VERSION >= 4)
26 into
= _mm_insert_epi32(into
, value
, 0);
28 into
= _mm_insert_epi16(into
, value
, 0);
29 into
= _mm_insert_epi16(into
, value
>> 16, 1);
33 GNU_TARGET(SSE_TARGET
)
34 INTERNAL_LINKAGE
inline void InsertSecondUint32(const uint32_t value
, __m128i
&into
)
36 #if (SSE_VERSION >= 4)
37 into
= _mm_insert_epi32(into
, value
, 1);
39 into
= _mm_insert_epi16(into
, value
, 2);
40 into
= _mm_insert_epi16(into
, value
>> 16, 3);
44 GNU_TARGET(SSE_TARGET
)
45 INTERNAL_LINKAGE
inline void LoadUint64(const uint64_t value
, __m128i
&into
)
47 #ifdef POINTER_IS_64BIT
48 into
= _mm_cvtsi64_si128(value
);
50 #if (SSE_VERSION >= 4)
51 into
= _mm_cvtsi32_si128(value
);
52 InsertSecondUint32(value
>> 32, into
);
54 (*(um128i
*) &into
).m128i_u64
[0] = value
;
59 GNU_TARGET(SSE_TARGET
)
60 INTERNAL_LINKAGE
inline __m128i
PackUnsaturated(__m128i from
, const __m128i
&mask
)
62 #if (SSE_VERSION == 2)
63 from
= _mm_and_si128(from
, mask
); // PAND, wipe high bytes to keep low bytes when packing
64 return _mm_packus_epi16(from
, from
); // PACKUSWB, pack 2 colours (with saturation)
66 return _mm_shuffle_epi8(from
, mask
);
70 GNU_TARGET(SSE_TARGET
)
71 INTERNAL_LINKAGE
inline __m128i
DistributeAlpha(const __m128i from
, const __m128i
&mask
)
73 #if (SSE_VERSION == 2)
74 __m128i alphaAB
= _mm_shufflelo_epi16(from
, 0x3F); // PSHUFLW, put alpha1 in front of each rgb1
75 alphaAB
= _mm_shufflehi_epi16(alphaAB
, 0x3F); // PSHUFHW, put alpha2 in front of each rgb2
76 return _mm_andnot_si128(mask
, alphaAB
); // PANDN, set alpha fields to 0
78 return _mm_shuffle_epi8(from
, mask
);
82 GNU_TARGET(SSE_TARGET
)
83 INTERNAL_LINKAGE
inline __m128i
AlphaBlendTwoPixels(__m128i src
, __m128i dst
, const __m128i
&distribution_mask
, const __m128i
&pack_mask
, const __m128i
&alpha_mask
)
85 __m128i srcAB
= _mm_unpacklo_epi8(src
, _mm_setzero_si128()); // PUNPCKLBW, expand each uint8_t into uint16
86 __m128i dstAB
= _mm_unpacklo_epi8(dst
, _mm_setzero_si128());
88 __m128i alphaMaskAB
= _mm_cmpgt_epi16(srcAB
, _mm_setzero_si128()); // PCMPGTW (alpha > 0) ? 0xFFFF : 0
89 __m128i alphaAB
= _mm_sub_epi16(srcAB
, alphaMaskAB
); // if (alpha > 0) a++;
90 alphaAB
= DistributeAlpha(alphaAB
, distribution_mask
);
92 srcAB
= _mm_sub_epi16(srcAB
, dstAB
); // PSUBW, (r - Cr)
93 srcAB
= _mm_mullo_epi16(srcAB
, alphaAB
); // PMULLW, a*(r - Cr)
94 srcAB
= _mm_srli_epi16(srcAB
, 8); // PSRLW, a*(r - Cr)/256
95 srcAB
= _mm_add_epi16(srcAB
, dstAB
); // PADDW, a*(r - Cr)/256 + Cr
97 alphaMaskAB
= _mm_and_si128(alphaMaskAB
, alpha_mask
); // PAND, set non alpha fields to 0
98 srcAB
= _mm_or_si128(srcAB
, alphaMaskAB
); // POR, set alpha fields to 0xFFFF is src alpha was > 0
100 return PackUnsaturated(srcAB
, pack_mask
);
104 * rgb = rgb * ((256/4) * 4 - (alpha/4)) / ((256/4) * 4)
106 GNU_TARGET(SSE_TARGET
)
107 INTERNAL_LINKAGE
inline __m128i
DarkenTwoPixels(__m128i src
, __m128i dst
, const __m128i
&distribution_mask
, const __m128i
&tr_nom_base
)
109 __m128i srcAB
= _mm_unpacklo_epi8(src
, _mm_setzero_si128());
110 __m128i dstAB
= _mm_unpacklo_epi8(dst
, _mm_setzero_si128());
111 __m128i alphaAB
= DistributeAlpha(srcAB
, distribution_mask
);
112 alphaAB
= _mm_srli_epi16(alphaAB
, 2); // Reduce to 64 levels of shades so the max value fits in 16 bits.
113 __m128i nom
= _mm_sub_epi16(tr_nom_base
, alphaAB
);
114 dstAB
= _mm_mullo_epi16(dstAB
, nom
);
115 dstAB
= _mm_srli_epi16(dstAB
, 8);
116 return _mm_packus_epi16(dstAB
, dstAB
);
119 IGNORE_UNINITIALIZED_WARNING_START
120 GNU_TARGET(SSE_TARGET
)
121 INTERNAL_LINKAGE Colour
ReallyAdjustBrightness(Colour colour
, uint8_t brightness
)
123 uint64_t c16
= colour
.b
| (uint64_t) colour
.g
<< 16 | (uint64_t) colour
.r
<< 32;
125 uint64_t c16_ob
= c16
; // Helps out of order execution.
126 c16
/= Blitter_32bppBase::DEFAULT_BRIGHTNESS
;
127 c16
&= 0x01FF01FF01FFULL
;
129 /* Sum overbright (maximum for each rgb is 508, 9 bits, -255 is changed in -256 so we just have to take the 8 lower bits into account). */
130 c16_ob
= (((c16_ob
>> (8 + 7)) & 0x0100010001ULL
) * 0xFF) & c16
;
131 const uint ob
= ((uint16_t) c16_ob
+ (uint16_t) (c16_ob
>> 16) + (uint16_t) (c16_ob
>> 32)) / 2;
133 const uint32_t alpha32
= colour
.data
& 0xFF000000;
135 LoadUint64(c16
, ret
);
137 __m128i ob128
= _mm_cvtsi32_si128(ob
);
138 ob128
= _mm_shufflelo_epi16(ob128
, 0xC0);
139 __m128i white
= OVERBRIGHT_VALUE_MASK
;
141 ret
= _mm_subs_epu16(white
, c128
); // PSUBUSW, (255 - rgb)
142 ret
= _mm_mullo_epi16(ret
, ob128
); // PMULLW, ob*(255 - rgb)
143 ret
= _mm_srli_epi16(ret
, 8); // PSRLW, ob*(255 - rgb)/256
144 ret
= _mm_add_epi16(ret
, c128
); // PADDW, ob*(255 - rgb)/256 + rgb
147 ret
= _mm_packus_epi16(ret
, ret
); // PACKUSWB, saturate and pack.
148 return alpha32
| _mm_cvtsi128_si32(ret
);
150 IGNORE_UNINITIALIZED_WARNING_STOP
152 /** ReallyAdjustBrightness() is not called that often.
153 * Inlining this function implies a far jump, which has a huge latency.
155 INTERNAL_LINKAGE
inline Colour
AdjustBrightneSSE(Colour colour
, uint8_t brightness
)
157 /* Shortcut for normal brightness. */
158 if (brightness
== Blitter_32bppBase::DEFAULT_BRIGHTNESS
) return colour
;
160 return ReallyAdjustBrightness(colour
, brightness
);
163 GNU_TARGET(SSE_TARGET
)
164 INTERNAL_LINKAGE
inline __m128i
AdjustBrightnessOfTwoPixels([[maybe_unused
]] __m128i from
, [[maybe_unused
]] uint32_t brightness
)
166 #if (SSE_VERSION < 3)
169 /* The following dataflow differs from the one of AdjustBrightness() only for alpha.
170 * In order to keep alpha in colAB, insert a 1 in a unused brightness byte (a*1->a).
171 * OK, not a 1 but DEFAULT_BRIGHTNESS to compensate the div.
173 brightness
&= 0xFF00FF00;
174 brightness
+= Blitter_32bppBase::DEFAULT_BRIGHTNESS
;
176 __m128i colAB
= _mm_unpacklo_epi8(from
, _mm_setzero_si128());
177 __m128i briAB
= _mm_cvtsi32_si128(brightness
);
178 briAB
= _mm_shuffle_epi8(briAB
, BRIGHTNESS_LOW_CONTROL_MASK
); // DEFAULT_BRIGHTNESS in 0, 0x00 in 2.
179 colAB
= _mm_mullo_epi16(colAB
, briAB
);
180 __m128i colAB_ob
= _mm_srli_epi16(colAB
, 8 + 7);
181 colAB
= _mm_srli_epi16(colAB
, 7);
184 * Maximum for each rgb is 508 => 9 bits. The highest bit tells if there is overbright.
185 * -255 is changed in -256 so we just have to take the 8 lower bits into account.
187 colAB
= _mm_and_si128(colAB
, BRIGHTNESS_DIV_CLEANER
);
188 colAB_ob
= _mm_and_si128(colAB_ob
, OVERBRIGHT_PRESENCE_MASK
);
189 colAB_ob
= _mm_mullo_epi16(colAB_ob
, OVERBRIGHT_VALUE_MASK
);
190 colAB_ob
= _mm_and_si128(colAB_ob
, colAB
);
191 __m128i obAB
= _mm_hadd_epi16(_mm_hadd_epi16(colAB_ob
, _mm_setzero_si128()), _mm_setzero_si128());
193 obAB
= _mm_srli_epi16(obAB
, 1); // Reduce overbright strength.
194 obAB
= _mm_shuffle_epi8(obAB
, OVERBRIGHT_CONTROL_MASK
);
195 __m128i retAB
= OVERBRIGHT_VALUE_MASK
; // ob_mask is equal to white.
196 retAB
= _mm_subs_epu16(retAB
, colAB
); // (255 - rgb)
197 retAB
= _mm_mullo_epi16(retAB
, obAB
); // ob*(255 - rgb)
198 retAB
= _mm_srli_epi16(retAB
, 8); // ob*(255 - rgb)/256
199 retAB
= _mm_add_epi16(retAB
, colAB
); // ob*(255 - rgb)/256 + rgb
201 return _mm_packus_epi16(retAB
, retAB
);
205 #if FULL_ANIMATION == 0
207 * Draws a sprite to a (screen) buffer. It is templated to allow faster operation.
209 * @tparam mode blitter mode
210 * @param bp further blitting parameters
211 * @param zoom zoom level at which we are drawing
213 IGNORE_UNINITIALIZED_WARNING_START
214 template <BlitterMode mode
, Blitter_32bppSSE2::ReadMode read_mode
, Blitter_32bppSSE2::BlockType bt_last
, bool translucent
>
215 GNU_TARGET(SSE_TARGET
)
216 #if (SSE_VERSION == 2)
217 inline void Blitter_32bppSSE2::Draw(const Blitter::BlitterParams
*bp
, ZoomLevel zoom
)
218 #elif (SSE_VERSION == 3)
219 inline void Blitter_32bppSSSE3::Draw(const Blitter::BlitterParams
*bp
, ZoomLevel zoom
)
220 #elif (SSE_VERSION == 4)
221 inline void Blitter_32bppSSE4::Draw(const Blitter::BlitterParams
*bp
, ZoomLevel zoom
)
224 const uint8_t * const remap
= bp
->remap
;
225 Colour
*dst_line
= (Colour
*) bp
->dst
+ bp
->top
* bp
->pitch
+ bp
->left
;
226 int effective_width
= bp
->width
;
228 /* Find where to start reading in the source sprite. */
229 const SpriteData
* const sd
= (const SpriteData
*) bp
->sprite
;
230 const SpriteInfo
* const si
= &sd
->infos
[zoom
];
231 const MapValue
*src_mv_line
= (const MapValue
*) &sd
->data
[si
->mv_offset
] + bp
->skip_top
* si
->sprite_width
;
232 const Colour
*src_rgba_line
= (const Colour
*) ((const uint8_t *) &sd
->data
[si
->sprite_offset
] + bp
->skip_top
* si
->sprite_line_size
);
234 if (read_mode
!= RM_WITH_MARGIN
) {
235 src_rgba_line
+= bp
->skip_left
;
236 src_mv_line
+= bp
->skip_left
;
238 const MapValue
*src_mv
= src_mv_line
;
240 /* Load these variables into register before loop. */
241 const __m128i alpha_and
= ALPHA_AND_MASK
;
242 #define ALPHA_BLEND_PARAM_3 alpha_and
243 #if (SSE_VERSION == 2)
244 const __m128i clear_hi
= CLEAR_HIGH_BYTE_MASK
;
245 #define ALPHA_BLEND_PARAM_1 alpha_and
246 #define ALPHA_BLEND_PARAM_2 clear_hi
247 #define DARKEN_PARAM_1 tr_nom_base
248 #define DARKEN_PARAM_2 tr_nom_base
250 const __m128i a_cm
= ALPHA_CONTROL_MASK
;
251 const __m128i pack_low_cm
= PACK_LOW_CONTROL_MASK
;
252 #define ALPHA_BLEND_PARAM_1 a_cm
253 #define ALPHA_BLEND_PARAM_2 pack_low_cm
254 #define DARKEN_PARAM_1 a_cm
255 #define DARKEN_PARAM_2 tr_nom_base
257 const __m128i tr_nom_base
= TRANSPARENT_NOM_BASE
;
259 for (int y
= bp
->height
; y
!= 0; y
--) {
260 Colour
*dst
= dst_line
;
261 const Colour
*src
= src_rgba_line
+ META_LENGTH
;
262 if (mode
== BM_COLOUR_REMAP
|| mode
== BM_CRASH_REMAP
) src_mv
= src_mv_line
;
264 if (read_mode
== RM_WITH_MARGIN
) {
265 assert(bt_last
== BT_NONE
); // or you must ensure block type is preserved
266 src
+= src_rgba_line
[0].data
;
267 dst
+= src_rgba_line
[0].data
;
268 if (mode
== BM_COLOUR_REMAP
|| mode
== BM_CRASH_REMAP
) src_mv
+= src_rgba_line
[0].data
;
269 const int width_diff
= si
->sprite_width
- bp
->width
;
270 effective_width
= bp
->width
- (int) src_rgba_line
[0].data
;
271 const int delta_diff
= (int) src_rgba_line
[1].data
- width_diff
;
272 const int new_width
= effective_width
- delta_diff
;
273 effective_width
= delta_diff
> 0 ? new_width
: effective_width
;
274 if (effective_width
<= 0) goto next_line
;
280 for (uint x
= (uint
) effective_width
; x
> 0; x
--) {
281 if (src
->a
) *dst
= *src
;
288 for (uint x
= (uint
) effective_width
/ 2; x
> 0; x
--) {
289 __m128i srcABCD
= _mm_loadl_epi64((const __m128i
*) src
);
290 __m128i dstABCD
= _mm_loadl_epi64((__m128i
*) dst
);
291 _mm_storel_epi64((__m128i
*) dst
, AlphaBlendTwoPixels(srcABCD
, dstABCD
, ALPHA_BLEND_PARAM_1
, ALPHA_BLEND_PARAM_2
, ALPHA_BLEND_PARAM_3
));
296 if ((bt_last
== BT_NONE
&& effective_width
& 1) || bt_last
== BT_ODD
) {
297 __m128i srcABCD
= _mm_cvtsi32_si128(src
->data
);
298 __m128i dstABCD
= _mm_cvtsi32_si128(dst
->data
);
299 dst
->data
= _mm_cvtsi128_si32(AlphaBlendTwoPixels(srcABCD
, dstABCD
, ALPHA_BLEND_PARAM_1
, ALPHA_BLEND_PARAM_2
, ALPHA_BLEND_PARAM_3
));
303 case BM_COLOUR_REMAP
:
304 #if (SSE_VERSION >= 3)
305 for (uint x
= (uint
) effective_width
/ 2; x
> 0; x
--) {
306 __m128i srcABCD
= _mm_loadl_epi64((const __m128i
*) src
);
307 __m128i dstABCD
= _mm_loadl_epi64((__m128i
*) dst
);
308 uint32_t mvX2
= *((uint32_t *) const_cast<MapValue
*>(src_mv
));
311 if (mvX2
& 0x00FF00FF) {
312 #define CMOV_REMAP(m_colour, m_colour_init, m_src, m_m) \
313 /* Written so the compiler uses CMOV. */ \
314 Colour m_colour = m_colour_init; \
316 const Colour srcm = (Colour) (m_src); \
317 const uint m = (uint8_t) (m_m); \
318 const uint r = remap[m]; \
319 const Colour cmap = (this->LookupColourInPalette(r).data & 0x00FFFFFF) | (srcm.data & 0xFF000000); \
320 m_colour = r == 0 ? m_colour : cmap; \
321 m_colour = m != 0 ? m_colour : srcm; \
323 #ifdef POINTER_IS_64BIT
324 uint64_t srcs
= _mm_cvtsi128_si64(srcABCD
);
325 uint64_t remapped_src
= 0;
326 CMOV_REMAP(c0
, 0, srcs
, mvX2
);
327 remapped_src
= c0
.data
;
328 CMOV_REMAP(c1
, 0, srcs
>> 32, mvX2
>> 16);
329 remapped_src
|= (uint64_t) c1
.data
<< 32;
330 srcABCD
= _mm_cvtsi64_si128(remapped_src
);
332 Colour remapped_src
[2];
333 CMOV_REMAP(c0
, 0, _mm_cvtsi128_si32(srcABCD
), mvX2
);
334 remapped_src
[0] = c0
.data
;
335 CMOV_REMAP(c1
, 0, src
[1], mvX2
>> 16);
336 remapped_src
[1] = c1
.data
;
337 srcABCD
= _mm_loadl_epi64((__m128i
*) &remapped_src
);
340 if ((mvX2
& 0xFF00FF00) != 0x80008000) srcABCD
= AdjustBrightnessOfTwoPixels(srcABCD
, mvX2
);
344 _mm_storel_epi64((__m128i
*) dst
, AlphaBlendTwoPixels(srcABCD
, dstABCD
, ALPHA_BLEND_PARAM_1
, ALPHA_BLEND_PARAM_2
, ALPHA_BLEND_PARAM_3
));
350 if ((bt_last
== BT_NONE
&& effective_width
& 1) || bt_last
== BT_ODD
) {
352 for (uint x
= (uint
) effective_width
; x
> 0; x
--) {
354 /* In case the m-channel is zero, do not remap this pixel in any way. */
357 const uint r
= remap
[src_mv
->m
];
359 Colour remapped_colour
= AdjustBrightneSSE(this->LookupColourInPalette(r
), src_mv
->v
);
361 *dst
= remapped_colour
;
363 remapped_colour
.a
= src
->a
;
364 srcABCD
= _mm_cvtsi32_si128(remapped_colour
.data
);
365 goto bmcr_alpha_blend_single
;
369 srcABCD
= _mm_cvtsi32_si128(src
->data
);
371 bmcr_alpha_blend_single
:
372 __m128i dstABCD
= _mm_cvtsi32_si128(dst
->data
);
373 srcABCD
= AlphaBlendTwoPixels(srcABCD
, dstABCD
, ALPHA_BLEND_PARAM_1
, ALPHA_BLEND_PARAM_2
, ALPHA_BLEND_PARAM_3
);
375 dst
->data
= _mm_cvtsi128_si32(srcABCD
);
377 #if (SSE_VERSION == 2)
386 /* Make the current colour a bit more black, so it looks like this image is transparent. */
387 for (uint x
= (uint
) bp
->width
/ 2; x
> 0; x
--) {
388 __m128i srcABCD
= _mm_loadl_epi64((const __m128i
*) src
);
389 __m128i dstABCD
= _mm_loadl_epi64((__m128i
*) dst
);
390 _mm_storel_epi64((__m128i
*) dst
, DarkenTwoPixels(srcABCD
, dstABCD
, DARKEN_PARAM_1
, DARKEN_PARAM_2
));
395 if ((bt_last
== BT_NONE
&& bp
->width
& 1) || bt_last
== BT_ODD
) {
396 __m128i srcABCD
= _mm_cvtsi32_si128(src
->data
);
397 __m128i dstABCD
= _mm_cvtsi32_si128(dst
->data
);
398 dst
->data
= _mm_cvtsi128_si32(DarkenTwoPixels(srcABCD
, dstABCD
, DARKEN_PARAM_1
, DARKEN_PARAM_2
));
402 case BM_TRANSPARENT_REMAP
:
403 /* Apply custom transparency remap. */
404 for (uint x
= (uint
) bp
->width
; x
> 0; x
--) {
406 *dst
= this->LookupColourInPalette(remap
[GetNearestColourIndex(*dst
)]);
415 for (uint x
= (uint
) bp
->width
; x
> 0; x
--) {
416 if (src_mv
->m
== 0) {
418 uint8_t g
= MakeDark(src
->r
, src
->g
, src
->b
);
419 *dst
= ComposeColourRGBA(g
, g
, g
, src
->a
, *dst
);
422 uint r
= remap
[src_mv
->m
];
423 if (r
!= 0) *dst
= ComposeColourPANoCheck(this->AdjustBrightness(this->LookupColourInPalette(r
), src_mv
->v
), src
->a
, *dst
);
432 for (uint x
= (uint
) bp
->width
; x
> 0; x
--) {
434 *dst
= Colour(0, 0, 0);
444 if (mode
== BM_COLOUR_REMAP
|| mode
== BM_CRASH_REMAP
) src_mv_line
+= si
->sprite_width
;
445 src_rgba_line
= (const Colour
*) ((const uint8_t*) src_rgba_line
+ si
->sprite_line_size
);
446 dst_line
+= bp
->pitch
;
449 IGNORE_UNINITIALIZED_WARNING_STOP
452 * Draws a sprite to a (screen) buffer. Calls adequate templated function.
454 * @param bp further blitting parameters
455 * @param mode blitter mode
456 * @param zoom zoom level at which we are drawing
458 #if (SSE_VERSION == 2)
459 void Blitter_32bppSSE2::Draw(Blitter::BlitterParams
*bp
, BlitterMode mode
, ZoomLevel zoom
)
460 #elif (SSE_VERSION == 3)
461 void Blitter_32bppSSSE3::Draw(Blitter::BlitterParams
*bp
, BlitterMode mode
, ZoomLevel zoom
)
462 #elif (SSE_VERSION == 4)
463 void Blitter_32bppSSE4::Draw(Blitter::BlitterParams
*bp
, BlitterMode mode
, ZoomLevel zoom
)
468 if (bp
->skip_left
!= 0 || bp
->width
<= MARGIN_NORMAL_THRESHOLD
) {
470 const BlockType bt_last
= (BlockType
) (bp
->width
& 1);
472 default: Draw
<BM_NORMAL
, RM_WITH_SKIP
, BT_EVEN
, true>(bp
, zoom
); return;
473 case BT_ODD
: Draw
<BM_NORMAL
, RM_WITH_SKIP
, BT_ODD
, true>(bp
, zoom
); return;
476 if (((const Blitter_32bppSSE_Base::SpriteData
*) bp
->sprite
)->flags
& SF_TRANSLUCENT
) {
477 Draw
<BM_NORMAL
, RM_WITH_MARGIN
, BT_NONE
, true>(bp
, zoom
);
479 Draw
<BM_NORMAL
, RM_WITH_MARGIN
, BT_NONE
, false>(bp
, zoom
);
485 case BM_COLOUR_REMAP
:
486 if (((const Blitter_32bppSSE_Base::SpriteData
*) bp
->sprite
)->flags
& SF_NO_REMAP
) goto bm_normal
;
487 if (bp
->skip_left
!= 0 || bp
->width
<= MARGIN_REMAP_THRESHOLD
) {
488 Draw
<BM_COLOUR_REMAP
, RM_WITH_SKIP
, BT_NONE
, true>(bp
, zoom
); return;
490 Draw
<BM_COLOUR_REMAP
, RM_WITH_MARGIN
, BT_NONE
, true>(bp
, zoom
); return;
492 case BM_TRANSPARENT
: Draw
<BM_TRANSPARENT
, RM_NONE
, BT_NONE
, true>(bp
, zoom
); return;
493 case BM_TRANSPARENT_REMAP
: Draw
<BM_TRANSPARENT_REMAP
, RM_NONE
, BT_NONE
, true>(bp
, zoom
); return;
494 case BM_CRASH_REMAP
: Draw
<BM_CRASH_REMAP
, RM_NONE
, BT_NONE
, true>(bp
, zoom
); return;
495 case BM_BLACK_REMAP
: Draw
<BM_BLACK_REMAP
, RM_NONE
, BT_NONE
, true>(bp
, zoom
); return;
498 #endif /* FULL_ANIMATION */
500 #endif /* WITH_SSE */
501 #endif /* BLITTER_32BPP_SSE_FUNC_HPP */