madcow/scummvm-cursorfix
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity
Files
5b116987315fa80e8d88f1eedb4927173ffb1461
scummvm-cursorfix/engines/ags/lib/allegro/surface_avx2.cpp
Leon Krieg 5b11698731 Initial commit
2026-02-02 04:50:13 +01:00

1020 lines
52 KiB
C++
Raw Blame History

/* ScummVM - Graphic Adventure Engine
*
* ScummVM is the legal property of its developers, whose names
* are too numerous to list here. Please refer to the COPYRIGHT
* file distributed with this source distribution.
*
* This program 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, either version 3 of the License, or
* (at your option) any later version.
*
* This program 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 this program. If not, see <http://www.gnu.org/licenses/>.
*
*/
#include "ags/lib/allegro/gfx.h"
#include "ags/lib/allegro/color.h"
#include "ags/lib/allegro/flood.h"
#include "ags/ags.h"
#include "ags/globals.h"
#include "common/textconsole.h"
#include "graphics/screen.h"
#include <immintrin.h>
#if defined(__clang__)
#pragma clang attribute push (__attribute__((target("avx2"))), apply_to=function)
#elif defined(__GNUC__)
#pragma GCC push_options
#pragma GCC target("avx2")
#endif
namespace AGS3 {
class DrawInnerImpl_AVX2 {
static inline __m256i simd2BppTo4Bpp(__m256i pixels) {
__m128i x128 = _mm256_castsi256_si128(pixels);
__m256i x = _mm256_cvtepu16_epi32(x128);
// c is the extracted 5/6 bit color from the image
__m256i c = _mm256_srli_epi32(x, 11);
// We convert it back to normal by shifting it thrice over, naturally, and then using the 2 most
// sinificant bits in the original color for the least significant bits in the new one
__m256i r = _mm256_slli_epi32(_mm256_or_si256(_mm256_slli_epi32(c, 3), _mm256_srli_epi32(c, 2)), 16);
c = _mm256_srli_epi32(_mm256_and_si256(x, _mm256_set1_epi32(0x07e0)), 5);
__m256i g = _mm256_slli_epi32(_mm256_or_si256(_mm256_slli_epi32(c, 2), _mm256_srli_epi32(c, 4)), 8);
c = _mm256_and_si256(x, _mm256_set1_epi32(0x001f));
__m256i b = _mm256_or_si256(_mm256_slli_epi32(c, 3), _mm256_srli_epi32(c, 2));
// By default 2bpp to 4bpp makes the alpha channel 255
return _mm256_or_si256(_mm256_or_si256(_mm256_or_si256(r, g), b), _mm256_set1_epi32(0xff000000));
}
static inline __m256i simd4BppTo2Bpp(__m256i pixels) {
// x is the final 16 bit rgb pixel
__m256i x = _mm256_srli_epi32(_mm256_and_si256(pixels, _mm256_set1_epi32(0x000000ff)), 3);
x = _mm256_or_si256(x, _mm256_slli_epi32(_mm256_srli_epi32(_mm256_and_si256(pixels, _mm256_set1_epi32(0x0000ff00)), 8+2), 5));
x = _mm256_or_si256(x, _mm256_slli_epi32(_mm256_srli_epi32(_mm256_and_si256(pixels, _mm256_set1_epi32(0x00ff0000)), 16+3), 11));
x = _mm256_slli_epi32(x, 16);
x = _mm256_srai_epi32(x, 16);
x = _mm256_packs_epi32(x, _mm256_setzero_si256());
return _mm256_permute4x64_epi64(x, _MM_SHUFFLE(3, 1, 2, 0));
}
static inline __m256i rgbBlendSIMD2Bpp(__m256i srcCols, __m256i destCols, __m256i alphas) {
// Here we add 1 to alphas if its 0. This is what the original blender function did
alphas = _mm256_add_epi16(alphas, _mm256_and_si256(_mm256_cmpgt_epi16(alphas, _mm256_setzero_si256()), _mm256_set1_epi16(1)));
// Split the components into rgb
__m256i srcComps[] = {
_mm256_and_si256(srcCols, _mm256_set1_epi16(0x1f)), // B
_mm256_and_si256(_mm256_srli_epi16(srcCols, 5), _mm256_set1_epi16(0x3f)), // G
_mm256_srli_epi16(srcCols, 11), // R
}, destComps[] = {
_mm256_and_si256(destCols, _mm256_set1_epi16(0x1f)), // B
_mm256_and_si256(_mm256_srli_epi16(destCols, 5), _mm256_set1_epi16(0x3f)), // G
_mm256_srli_epi16(destCols, 11), // R
};
// Calculate the differences between the colors
__m256i diffs[] = {
_mm256_sub_epi16(srcComps[0], destComps[0]), // B
_mm256_sub_epi16(srcComps[1], destComps[1]), // G
_mm256_sub_epi16(srcComps[2], destComps[2]), // R
};
// Multiply by alpha and shift depth bits to the right
// pretty much the same as (int)(((float)component / 255.0f) * ((float)alpha / 255.0f) * 255.0f)
alphas = _mm256_srli_epi16(alphas, 2);
diffs[1] = _mm256_srli_epi16(_mm256_mullo_epi16(diffs[1], alphas), 6);
alphas = _mm256_srli_epi16(alphas, 1);
diffs[0] = _mm256_srli_epi16(_mm256_mullo_epi16(diffs[0], alphas), 5);
diffs[2] = _mm256_srli_epi16(_mm256_mullo_epi16(diffs[2], alphas), 5);
// Here we add the difference between the 2 colors times alpha onto the destination
diffs[0] = _mm256_and_si256(_mm256_add_epi16(diffs[0], destComps[0]), _mm256_set1_epi16(0x1f));
diffs[1] = _mm256_and_si256(_mm256_add_epi16(diffs[1], destComps[1]), _mm256_set1_epi16(0x3f));
diffs[2] = _mm256_and_si256(_mm256_add_epi16(diffs[2], destComps[2]), _mm256_set1_epi16(0x1f));
// We compile all the colors into diffs[0] as a 16 bit rgb pixel
diffs[0] = _mm256_or_si256(diffs[0], _mm256_slli_epi16(diffs[1], 5));
return _mm256_or_si256(diffs[0], _mm256_slli_epi16(diffs[2], 11));
}
// preserveAlpha:
// false => set destCols's alpha to 0
// true => keep destCols's alpha
static inline __m256i rgbBlendSIMD(__m256i srcCols, __m256i destCols, __m256i alphas, bool preserveAlpha) {
// Here we add 1 to alphas if its 0. This is what the original blender function did.
alphas = _mm256_add_epi32(alphas, _mm256_and_si256(_mm256_cmpgt_epi32(alphas, _mm256_setzero_si256()), _mm256_set1_epi32(1)));
// Get the alpha from the destination
__m256i alpha = _mm256_and_si256(destCols, _mm256_set1_epi32(0xff000000));
// Get red and blue components
__m256i srcColsCopy = srcCols;
srcColsCopy = _mm256_and_si256(srcColsCopy, _mm256_set1_epi32(0xff00ff));
__m256i destColsCopy = destCols;
destColsCopy = _mm256_and_si256(destColsCopy, _mm256_set1_epi32(0xff00ff));
// Compute the difference, then multiply by alpha and divide by 256
srcColsCopy = _mm256_sub_epi32(srcColsCopy, destColsCopy);
srcColsCopy = _mm256_mullo_epi32(srcColsCopy, alphas);
//srcColsCopy = _mm256_mul_epi32(srcColsCopy, alphas);
srcColsCopy = _mm256_srli_epi32(srcColsCopy, 8);
srcColsCopy = _mm256_add_epi32(srcColsCopy, destCols); // Add the new red/blue to the old red/blue
// Do the same for the green component
srcCols = _mm256_and_si256(srcCols, _mm256_set1_epi32(0xff00));
destCols = _mm256_and_si256(destCols, _mm256_set1_epi32(0xff00));
srcCols = _mm256_sub_epi32(srcCols, destCols);
srcCols = _mm256_mullo_epi32(srcCols, alphas);
//srcCols = _mm256_mul_epi32(srcCols, alphas);
srcCols = _mm256_srli_epi32(srcCols, 8);
srcCols = _mm256_add_epi32(srcCols, destCols); // Add the new green to the old green
// Keep values in 8bit range and glue red/blue and green together
srcColsCopy = _mm256_and_si256(srcColsCopy, _mm256_set1_epi32(0xff00ff));
srcCols = _mm256_and_si256(srcCols, _mm256_set1_epi32(0xff00));
srcCols = _mm256_or_si256(srcCols, srcColsCopy);
// Remember that alpha is not alphas, but rather the alpha of destcols
if (preserveAlpha) {
srcCols = _mm256_and_si256(srcCols, _mm256_set1_epi32(0x00ffffff));
srcCols = _mm256_or_si256(srcCols, alpha);
}
return srcCols;
}
static inline __m256i argbBlendSIMD(__m256i srcCols, __m256i destCols) {
__m256 srcA = _mm256_cvtepi32_ps(_mm256_srli_epi32(srcCols, 24));
srcA = _mm256_mul_ps(srcA, _mm256_set1_ps(1.0f / 255.0f));
__m256 srcR = _mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(srcCols, 16), _mm256_set1_epi32(0xff)));
__m256 srcG = _mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(srcCols, 8), _mm256_set1_epi32(0xff)));
__m256 srcB = _mm256_cvtepi32_ps(_mm256_and_si256(srcCols, _mm256_set1_epi32(0xff)));
__m256 destA = _mm256_cvtepi32_ps(_mm256_srli_epi32(destCols, 24));
destA = _mm256_mul_ps(destA, _mm256_set1_ps(1.0f / 255.0f));
__m256 destR = _mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(destCols, 16), _mm256_set1_epi32(0xff)));
__m256 destG = _mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(destCols, 8), _mm256_set1_epi32(0xff)));
__m256 destB = _mm256_cvtepi32_ps(_mm256_and_si256(destCols, _mm256_set1_epi32(0xff)));
// the destination alpha gets multiplied by 255 - source alpha
destA = _mm256_mul_ps(destA, _mm256_sub_ps(_mm256_set1_ps(1.0f), srcA));
// ((src * sAlpha) + (dest * dAlpha)) / (sAlpha + dAlpha)
__m256 combA = _mm256_add_ps(srcA, destA);
__m256 combArcp = _mm256_rcp_ps(combA);
destR = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(srcR, srcA), _mm256_mul_ps(destR, destA)), combArcp);
destG = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(srcG, srcA), _mm256_mul_ps(destG, destA)), combArcp);
destB = _mm256_mul_ps(_mm256_add_ps(_mm256_mul_ps(srcB, srcA), _mm256_mul_ps(destB, destA)), combArcp);
combA = _mm256_mul_ps(combA, _mm256_set1_ps(255.0));
// Now put it back together
return _mm256_or_si256(_mm256_slli_epi32(_mm256_cvtps_epi32(combA), 24),
_mm256_or_si256(_mm256_slli_epi32(_mm256_cvtps_epi32(destR), 16),
_mm256_or_si256(_mm256_slli_epi32(_mm256_cvtps_epi32(destG), 8),
_mm256_cvtps_epi32(destB))));
}
static inline __m256i blendTintSpriteSIMD(__m256i srcCols, __m256i destCols, __m256i alphas, bool light) {
// This function is NOT 1 to 1 with the original... It just approximates it
// It gets the value of the HSV of the dest color
// Then it gets the HSV of the srcCols
// how the values are transformed
// from 1 uint32x4_t srcCols with each lane being ARGB uint32
// srcCols[0] = A | R | G | B
// srcCols[1] = A | R | G | B
// srcCols[2] = A | R | G | B
// srcCols[3] = A | R | G | B
// ->
// to 4 float32x4_t's each being a separate channel with each lane
// corresponding to their respective srcCols lane
// dda = { A[0], A[1], A[2], A[3] }
// ddr = { R[0], R[1], R[2], R[3] }
// ddg = { G[0], G[1], G[2], G[3] }
// ddb = { B[0], B[1], B[2], B[3] }
// do the transformation (we don't actually need alpha at all)
__m256 ddr, ddg, ddb;
ddr = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(destCols, 16), _mm256_set1_epi32(0xff))), _mm256_set1_ps(1.0f / 255.0f));
ddg = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(destCols, 8), _mm256_set1_epi32(0xff))), _mm256_set1_ps(1.0f / 255.0f));
ddb = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(destCols, _mm256_set1_epi32(0xff))), _mm256_set1_ps(1.0f / 255.0f));
__m256 ssr, ssg, ssb;
ssr = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(srcCols, 16), _mm256_set1_epi32(0xff))), _mm256_set1_ps(1.0f / 255.0f));
ssg = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(_mm256_srli_epi32(srcCols, 8), _mm256_set1_epi32(0xff))), _mm256_set1_ps(1.0f / 255.0f));
ssb = _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_and_si256(srcCols, _mm256_set1_epi32(0xff))), _mm256_set1_ps(1.0f / 255.0f));
// Get the maxes and mins (needed for HSV->RGB and visa-versa)
__m256 dmaxes = _mm256_max_ps(ddr, _mm256_max_ps(ddg, ddb));
__m256 smaxes = _mm256_max_ps(ssr, _mm256_max_ps(ssg, ssb));
__m256 smins = _mm256_min_ps(ssr, _mm256_min_ps(ssg, ssb));
// This is here to stop from dividing by 0
const __m256 eplison0 = _mm256_set1_ps(0.0000001f);
__m256 chroma = _mm256_max_ps(_mm256_sub_ps(smaxes, smins), eplison0);
// RGB to HSV is a piecewise function, so we compute each part of the function first...
__m256 hr, hg, hb, hue;
hr = _mm256_div_ps(_mm256_sub_ps(ssg, ssb), chroma);
hr = _mm256_sub_ps(hr, _mm256_mul_ps(_mm256_cvtepi32_ps(_mm256_cvtps_epi32(_mm256_mul_ps(hr, _mm256_set1_ps(1.0f / 6.0f)))), _mm256_set1_ps(6.0f)));
hr = _mm256_add_ps(hr, _mm256_and_ps(_mm256_cmp_ps(hr, _mm256_setzero_ps(), _CMP_LT_OS), _mm256_set1_ps(6.0f)));
hg = _mm256_add_ps(_mm256_div_ps(_mm256_sub_ps(ssb, ssr), chroma), _mm256_set1_ps(2.0f));
hg = _mm256_max_ps(hg, _mm256_setzero_ps());
hb = _mm256_add_ps(_mm256_div_ps(_mm256_sub_ps(ssr, ssg), chroma), _mm256_set1_ps(4.0f));
hb = _mm256_max_ps(hb, _mm256_setzero_ps());
// And then compute which one will be used based on criteria
__m256 hrfactors = _mm256_and_ps(_mm256_and_ps(_mm256_cmp_ps(ssr, smaxes, _CMP_EQ_OS), _mm256_cmp_ps(ssr, ssb, _CMP_NEQ_OS)), _mm256_set1_ps(1.0f));
__m256 hgfactors = _mm256_and_ps(_mm256_and_ps(_mm256_cmp_ps(ssg, smaxes, _CMP_EQ_OS), _mm256_cmp_ps(ssg, ssr, _CMP_NEQ_OS)), _mm256_set1_ps(1.0f));
__m256 hbfactors = _mm256_and_ps(_mm256_and_ps(_mm256_cmp_ps(ssb, smaxes, _CMP_EQ_OS), _mm256_cmp_ps(ssb, ssg, _CMP_NEQ_OS)), _mm256_set1_ps(1.0f));
hue = _mm256_mul_ps(hr, hrfactors);
hue = _mm256_add_ps(hue, _mm256_mul_ps(hg, hgfactors));
hue = _mm256_add_ps(hue, _mm256_mul_ps(hb, hbfactors));
// Mess with the light like the original function
__m256 val = dmaxes;
if (light) {
val = _mm256_sub_ps(val, _mm256_sub_ps(_mm256_set1_ps(1.0f), _mm256_mul_ps(_mm256_cvtepi32_ps(alphas), _mm256_set1_ps(1.0f / 250.0f))));
val = _mm256_max_ps(val, _mm256_setzero_ps());
}
// then it stiches the HSV back together
// the hue and saturation come from the source (tint) color, and the value comes from
// the destinaion (real source) color
chroma = _mm256_mul_ps(val, _mm256_div_ps(_mm256_sub_ps(smaxes, smins), _mm256_add_ps(smaxes, eplison0)));
__m256 hprime_mod2 = _mm256_mul_ps(hue, _mm256_set1_ps(1.0f / 2.0f));
hprime_mod2 = _mm256_mul_ps(_mm256_sub_ps(hprime_mod2, _mm256_cvtepi32_ps(_mm256_cvtps_epi32(_mm256_sub_ps(hprime_mod2, _mm256_set1_ps(0.5))))), _mm256_set1_ps(2.0f));
__m256 x = _mm256_mul_ps(chroma, _mm256_sub_ps(_mm256_set1_ps(1), _mm256_and_ps(_mm256_castsi256_ps(_mm256_set1_epi32(0x7fffffff)), _mm256_sub_ps(hprime_mod2, _mm256_set1_ps(1)))));
//float32x4_t x = vmulq_f32(chroma, vsubq_f32(vmovq_n_f32(1.0f), vabsq_f32(vsubq_f32(hprime_mod2, vmovq_n_f32(1.0f)))));
__m256i hprime_rounded = _mm256_cvtps_epi32(_mm256_sub_ps(hue, _mm256_set1_ps(0.5)));
__m256i x_int = _mm256_cvtps_epi32(_mm256_mul_ps(x, _mm256_set1_ps(255.0f)));
__m256i c_int = _mm256_cvtps_epi32(_mm256_mul_ps(chroma, _mm256_set1_ps(255.0f)));
// Again HSV->RGB is also a piecewise function
__m256i val0 = _mm256_or_si256(_mm256_slli_epi32(x_int, 8), _mm256_slli_epi32(c_int, 16));
val0 = _mm256_and_si256(val0, _mm256_or_si256(_mm256_cmpeq_epi32(hprime_rounded, _mm256_set1_epi32(0)), _mm256_cmpeq_epi32(hprime_rounded, _mm256_set1_epi32(6))));
__m256i val1 = _mm256_or_si256(_mm256_slli_epi32(c_int, 8), _mm256_slli_epi32(x_int, 16));
val1 = _mm256_and_si256(val1, _mm256_cmpeq_epi32(hprime_rounded, _mm256_set1_epi32(1)));
__m256i val2 = _mm256_or_si256(_mm256_slli_epi32(c_int, 8), x_int);
val2 = _mm256_and_si256(val2, _mm256_cmpeq_epi32(hprime_rounded, _mm256_set1_epi32(2)));
__m256i val3 = _mm256_or_si256(_mm256_slli_epi32(x_int, 8), c_int);
val3 = _mm256_and_si256(val3, _mm256_cmpeq_epi32(hprime_rounded, _mm256_set1_epi32(3)));
__m256i val4 = _mm256_or_si256(_mm256_slli_epi32(x_int, 16), c_int);
val4 = _mm256_and_si256(val4, _mm256_cmpeq_epi32(hprime_rounded, _mm256_set1_epi32(4)));
__m256i val5 = _mm256_or_si256(_mm256_slli_epi32(c_int, 16), x_int);
val5 = _mm256_and_si256(val5, _mm256_cmpeq_epi32(hprime_rounded, _mm256_set1_epi32(5)));
// or the values together
__m256i final = _mm256_or_si256(val0, _mm256_or_si256(val1, _mm256_or_si256(val2, _mm256_or_si256(val3, _mm256_or_si256(val4, val5)))));
// add the minimums back in
__m256i val_add = _mm256_cvtps_epi32(_mm256_mul_ps(_mm256_sub_ps(val, chroma), _mm256_set1_ps(255.0f)));
val_add = _mm256_or_si256(val_add, _mm256_or_si256(_mm256_slli_epi32(val_add, 8), _mm256_or_si256(_mm256_slli_epi32(val_add, 16), _mm256_and_si256(destCols, _mm256_set1_epi32(0xff000000)))));
final = _mm256_add_epi32(final, val_add);
return final;
}
static inline __m256i mul32_as16(__m256i a, __m256i b) {
__m256i a16 = _mm256_packs_epi32(a, _mm256_setzero_si256());
__m256i b16 = _mm256_packs_epi32(b, _mm256_setzero_si256());
__m256i res = _mm256_mullo_epi16(a16, b16);
return _mm256_unpacklo_epi16(res, _mm256_setzero_si256());
}
static inline __m256i findmin32_as16(__m256i a, __m256i b) {
__m256i a16 = _mm256_packs_epi32(a, _mm256_setzero_si256());
__m256i b16 = _mm256_packs_epi32(b, _mm256_setzero_si256());
__m256i res = _mm256_min_epi16(a16, b16);
return _mm256_unpacklo_epi16(res, _mm256_setzero_si256());
}
static inline __m256i blendPixelSIMD(__m256i srcCols, __m256i destCols, __m256i alphas) {
__m256i srcAlphas, difAlphas, mask, ch1, ch2;
auto setupArgbAlphas = [&]() {
// This acts the same as this in the normal blender functions
// if (alpha == 0)
// alpha = aSrc;
// else
// alpha = aSrc * ((alpha & 0xff) + 1) / 256;
// where alpha is the alpha byte of the srcCols
srcAlphas = _mm256_srli_epi32(srcCols, 24);
difAlphas = _mm256_add_epi32(_mm256_and_si256(alphas, _mm256_set1_epi32(0xff)), _mm256_set1_epi32(1));
difAlphas = _mm256_srli_epi32(mul32_as16(srcAlphas, difAlphas), 8);
difAlphas = _mm256_slli_epi32(difAlphas, 24);
srcAlphas = _mm256_slli_epi32(srcAlphas, 24);
mask = _mm256_cmpeq_epi32(alphas, _mm256_setzero_si256());
srcAlphas = _mm256_and_si256(srcAlphas, mask);
difAlphas = _mm256_andnot_si256(mask, difAlphas);
srcCols = _mm256_and_si256(srcCols, _mm256_set1_epi32(0x00ffffff));
srcCols = _mm256_or_si256(srcCols, _mm256_or_si256(srcAlphas, difAlphas));
};
switch (_G(_blender_mode)) {
case kSourceAlphaBlender: // see BITMAP member function blendSourceAlpha
alphas = _mm256_srli_epi32(srcCols, 24);
return rgbBlendSIMD(srcCols, destCols, alphas, false);
case kArgbToArgbBlender: // see BITMAP member function blendArgbToArgb
setupArgbAlphas();
// only blend if alpha isn't 0, otherwise use destCols
mask = _mm256_cmpgt_epi32(_mm256_srli_epi32(srcCols, 24), _mm256_setzero_si256());
ch1 = _mm256_and_si256(argbBlendSIMD(srcCols, destCols), mask);
ch2 = _mm256_andnot_si256(mask, destCols);
return _mm256_or_si256(ch1, ch2);
case kArgbToRgbBlender: // see BITMAP member function blendArgbToRgb
setupArgbAlphas();
return rgbBlendSIMD(srcCols, destCols, _mm256_srli_epi32(srcCols, 24), false);
case kRgbToArgbBlender: // see BITMAP member function blendRgbToArgb
// if alpha is NOT 0 or 255
ch2 = _mm256_and_si256(srcCols, _mm256_set1_epi32(0x00ffffff));
ch2 = _mm256_or_si256(ch2, _mm256_slli_epi32(alphas, 24));
ch2 = argbBlendSIMD(ch2, destCols);
// if alpha is 0 or 255
ch1 = _mm256_or_si256(srcCols, _mm256_set1_epi32(0xff000000));
// mask and or them together
mask = _mm256_or_si256(_mm256_cmpeq_epi32(alphas, _mm256_setzero_si256()), _mm256_cmpeq_epi32(alphas, _mm256_set1_epi32(0xff)));
ch1 = _mm256_and_si256(ch1, mask);
ch2 = _mm256_andnot_si256(mask, ch2);
return _mm256_or_si256(ch1, ch2);
case kRgbToRgbBlender: // see BITMAP member function blendRgbToRgb
return rgbBlendSIMD(srcCols, destCols, alphas, false);
case kAlphaPreservedBlenderMode: // see BITMAP member function blendPreserveAlpha
return rgbBlendSIMD(srcCols, destCols, alphas, true);
case kOpaqueBlenderMode: // see BITMAP member function blendOpaque
return _mm256_or_si256(srcCols, _mm256_set1_epi32(0xff000000));
case kAdditiveBlenderMode: // see BITMAP member function blendAdditiveAlpha
srcAlphas = _mm256_add_epi32(_mm256_srli_epi32(srcCols, 24), _mm256_srli_epi32(destCols, 24));
srcAlphas = findmin32_as16(srcAlphas, _mm256_set1_epi32(0xff));
srcCols = _mm256_and_si256(srcCols, _mm256_set1_epi32(0x00ffffff));
return _mm256_or_si256(srcCols, _mm256_slli_epi32(srcAlphas, 24));
case kTintBlenderMode: // see BITMAP member function blendTintSprite
return blendTintSpriteSIMD(srcCols, destCols, alphas, false);
case kTintLightBlenderMode: // see BITMAP member function blendTintSprite
return blendTintSpriteSIMD(srcCols, destCols, alphas, true);
}
return _mm256_setzero_si256();
}
static inline __m256i blendPixelSIMD2Bpp(__m256i srcCols, __m256i destCols, __m256i alphas) {
__m256i mask, ch1, ch2;
switch (_G(_blender_mode)) {
case kSourceAlphaBlender:
case kOpaqueBlenderMode:
case kAdditiveBlenderMode:
return srcCols;
case kArgbToArgbBlender:
case kArgbToRgbBlender:
ch1 = _mm256_and_si256(_mm256_set1_epi16(0xff), _mm256_cmpeq_epi16(alphas, _mm256_setzero_si256()));
ch2 = _mm256_and_si256(alphas, _mm256_cmpgt_epi16(alphas, _mm256_setzero_si256()));
alphas = _mm256_or_si256(ch1, ch2);
// fall through
case kRgbToRgbBlender:
case kAlphaPreservedBlenderMode:
return rgbBlendSIMD2Bpp(srcCols, destCols, alphas);
case kRgbToArgbBlender:
mask = _mm256_or_si256(_mm256_cmpeq_epi16(alphas, _mm256_set1_epi16(0)), _mm256_cmpeq_epi16(alphas, _mm256_set1_epi16(255)));
ch1 = _mm256_and_si256(srcCols, mask);
ch2 = _mm256_andnot_si256(mask, rgbBlendSIMD2Bpp(srcCols, destCols, alphas));
return _mm256_or_si256(ch1, ch2);
case kTintBlenderMode:
case kTintLightBlenderMode:
__m256i srcColsLo = simd2BppTo4Bpp(srcCols);
__m256i srcColsHi = simd2BppTo4Bpp(_mm256_permute2x128_si256(srcCols, srcCols, _MM_SHUFFLE(2, 0, 0, 1)));
__m256i destColsLo = simd2BppTo4Bpp(destCols);
__m256i destColsHi = simd2BppTo4Bpp(_mm256_permute2x128_si256(destCols, destCols, _MM_SHUFFLE(2, 0, 0, 1)));
__m256i alphasLo = _mm256_unpacklo_epi16(alphas, _mm256_setzero_si256());
__m256i alphasHi = _mm256_unpacklo_epi16(_mm256_permute2x128_si256(alphas, alphas, _MM_SHUFFLE(2, 3, 0, 1)), _mm256_setzero_si256());
__m256i lo = simd4BppTo2Bpp(blendTintSpriteSIMD(srcColsLo, destColsLo, alphasLo, _G(_blender_mode) == kTintLightBlenderMode));
__m256i hi = simd4BppTo2Bpp(blendTintSpriteSIMD(srcColsHi, destColsHi, alphasHi, _G(_blender_mode) == kTintLightBlenderMode));
return _mm256_or_si256(lo, _mm256_permute2x128_si256(hi, hi, _MM_SHUFFLE(0, 0, 2, 0)));
}
return _mm256_setzero_si256();
}
template<int DestBytesPerPixel, int SrcBytesPerPixel>
static inline void drawPixelSIMD(byte *destPtr, const byte *srcP2, __m256i tint, __m256i alphas, __m256i maskedAlphas, __m256i transColors, int xDir, int xCtrBpp, int srcAlpha, int skipTrans, bool horizFlip, bool useTint, __m256i skipMask) {
__m256i srcCols, destCol;
if (DestBytesPerPixel == 4)
destCol = _mm256_loadu_si256((const __m256i *)destPtr);
else
destCol = simd2BppTo4Bpp(_mm256_castsi128_si256(_mm_loadu_si128((const __m128i *)destPtr)));
if (SrcBytesPerPixel == 4)
srcCols = _mm256_loadu_si256((const __m256i *)(srcP2 + xDir * xCtrBpp));
else
srcCols = simd2BppTo4Bpp(_mm256_castsi128_si256(_mm_loadu_si128((const __m128i *)(srcP2 + xDir * xCtrBpp))));
// we do this here because we need to check if we should skip the pixel before we blend it
__m256i mask1 = skipTrans ? _mm256_cmpeq_epi32(_mm256_and_si256(srcCols, maskedAlphas), transColors) : _mm256_setzero_si256();
mask1 = _mm256_or_si256(mask1, skipMask);
if (srcAlpha != -1) {
// take into account for useTint
if (useTint) {
srcCols = blendPixelSIMD(tint, srcCols, alphas);
} else {
srcCols = blendPixelSIMD(srcCols, destCol, alphas);
}
}
__m256i destCols2 = _mm256_and_si256(destCol, mask1);
__m256i srcCols2 = _mm256_andnot_si256(mask1, srcCols);
__m256i final = _mm256_or_si256(destCols2, srcCols2);
if (horizFlip) {
final = _mm256_shuffle_epi32(final, _MM_SHUFFLE(0, 1, 2, 3));
final = _mm256_permute2x128_si256(final, final, 0x01);
}
if (DestBytesPerPixel == 4) {
_mm256_storeu_si256((__m256i *)destPtr, final);
} else {
_mm_storeu_si128((__m128i *)destPtr, _mm256_extracti128_si256(simd4BppTo2Bpp(final), 0));
}
}
static inline void drawPixelSIMD2Bpp(byte *destPtr, const byte *srcP2, __m256i tint, __m256i alphas, __m256i transColors, int xDir, int xCtrBpp, int srcAlpha, int skipTrans, bool horizFlip, bool useTint, __m256i skipMask) {
__m256i destCol = _mm256_loadu_si256((const __m256i *)destPtr);
__m256i srcCols = _mm256_loadu_si256((const __m256i *)(srcP2 + xDir * xCtrBpp));
__m256i mask1 = skipTrans ? _mm256_cmpeq_epi16(srcCols, transColors) : _mm256_setzero_si256();
mask1 = _mm256_or_si256(mask1, skipMask);
if (srcAlpha != -1) {
// take into account for useTint
if (useTint) {
srcCols = blendPixelSIMD2Bpp(tint, srcCols, alphas);
} else {
srcCols = blendPixelSIMD2Bpp(srcCols, destCol, alphas);
}
}
__m256i destCols2 = _mm256_and_si256(destCol, mask1);
__m256i srcCols2 = _mm256_andnot_si256(mask1, srcCols);
__m256i final = _mm256_or_si256(destCols2, srcCols2);
if (horizFlip) {
final = _mm256_shufflelo_epi16(final, _MM_SHUFFLE(0, 1, 2, 3));
final = _mm256_shufflehi_epi16(final, _MM_SHUFFLE(0, 1, 2, 3));
final = _mm256_castpd_si256(_mm256_shuffle_pd(_mm256_castsi256_pd(final), _mm256_castsi256_pd(final), _MM_SHUFFLE2(0, 1)));
}
_mm256_storeu_si256((__m256i *)destPtr, final);
}
public:
// This template handles 2bpp and 4bpp, the other specializations handle 1bpp and format conversion blits
template<int DestBytesPerPixel, int SrcBytesPerPixel, bool Scale>
static void drawInner4BppWithConv(BITMAP::DrawInnerArgs &args) {
const int xDir = args.horizFlip ? -1 : 1;
byte rSrc, gSrc, bSrc, aSrc;
byte rDest = 0, gDest = 0, bDest = 0, aDest = 0;
__m256i tint = _mm256_slli_epi32(_mm256_set1_epi32(args.srcAlpha), 24);
tint = _mm256_or_si256(tint, _mm256_slli_epi32(_mm256_set1_epi32(args.tintRed), 16));
tint = _mm256_or_si256(tint, _mm256_slli_epi32(_mm256_set1_epi32(args.tintGreen), 8));
tint = _mm256_or_si256(tint, _mm256_set1_epi32(args.tintBlue));
__m256i maskedAlphas = _mm256_set1_epi32(args.alphaMask);
__m256i transColors = _mm256_set1_epi32(args.transColor);
__m256i alphas = _mm256_set1_epi32(args.srcAlpha);
// This is so that we can calculate what pixels to crop off in a vectorized way
__m256i addIndexes = _mm256_set_epi32(7, 6, 5, 4, 3, 2, 1, 0);
if (args.horizFlip) addIndexes = _mm256_set_epi32(0, 1, 2, 3, 4, 5, 6, 7);
// This is so that we can calculate in parralell the pixel indexes for scaled drawing
__m256i scaleAdds = _mm256_set_epi32((uint32)args.scaleX*7, (uint32)args.scaleX*6, (uint32)args.scaleX*5, (uint32)args.scaleX*4,
(uint32)args.scaleX*3, (uint32)args.scaleX*2, (uint32)args.scaleX, 0);
// Clip the bounds ahead of time (so we don't waste time checking if we are in bounds when
// we are in the inner loop)
int xCtrStart = 0, xCtrBppStart = 0, xCtrWidth = args.dstRect.width();
if (args.xStart + xCtrWidth > args.destArea.w) {
xCtrWidth = args.destArea.w - args.xStart;
}
if (args.xStart < 0) {
xCtrStart = -args.xStart;
xCtrBppStart = xCtrStart * SrcBytesPerPixel;
args.xStart = 0;
}
int destY = args.yStart, srcYCtr = 0, yCtr = 0, scaleYCtr = 0, yCtrHeight = args.dstRect.height();
if (Scale) yCtrHeight = args.dstRect.height();
if (args.yStart < 0) {
yCtr = -args.yStart;
destY = 0;
if (Scale) {
scaleYCtr = yCtr * args.scaleY;
srcYCtr = scaleYCtr / BITMAP::SCALE_THRESHOLD;
}
}
if (args.yStart + yCtrHeight > args.destArea.h) {
yCtrHeight = args.destArea.h - args.yStart;
}
/*if (xCtrWidth % 8 != 0) {
--yCtrHeight;
}*/
const int secondToLast = xCtrWidth - 8;
byte *destP = (byte *)args.destArea.getBasePtr(0, destY);
const byte *srcP = (const byte *)args.src.getBasePtr(
args.horizFlip ? args.srcArea.right - 8 : args.srcArea.left,
args.vertFlip ? args.srcArea.bottom - 1 - yCtr : args.srcArea.top + yCtr);
for (; yCtr < yCtrHeight; ++destY, ++yCtr, scaleYCtr += args.scaleY) {
__m256i xCtrWidthSIMD = _mm256_set1_epi32(xCtrWidth); // This is the width of the row
if (!Scale) {
// If we are not scaling the image
int xCtr = xCtrStart, xCtrBpp = xCtrBppStart, destX = args.xStart;
for (; xCtr < secondToLast; destX += 8, xCtr += 8, xCtrBpp += SrcBytesPerPixel*8) {
byte *destPtr = &destP[destX * DestBytesPerPixel];
drawPixelSIMD<DestBytesPerPixel, SrcBytesPerPixel>(destPtr, srcP, tint, alphas, maskedAlphas, transColors, xDir, xCtrBpp, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, _mm256_set1_epi32(0));
}
byte *destPtr = &destP[destX * DestBytesPerPixel];
__m256i srcCols = _mm256_setzero_si256();
__m256i destCols = _mm256_setzero_si256();
memcpy(&srcCols, srcP + xDir * xCtrBpp, (xCtrWidth - xCtr) * SrcBytesPerPixel);
memcpy(&destCols, destPtr, (xCtrWidth - xCtr) * DestBytesPerPixel);
// Skip pixels that are beyond the row
// __m256i skipMask = _mm256_cmpgt_epi32(_mm256_add_epi32(_mm256_add_epi32(_mm256_set1_epi32(xCtr), addIndexes), _mm256_set1_epi32(1)), xCtrWidthSIMD);
drawPixelSIMD<DestBytesPerPixel, SrcBytesPerPixel>((byte *)&destCols, (byte *)&srcCols, tint, alphas, maskedAlphas, transColors, xDir, 0, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, _mm256_set1_epi32(0));
memcpy(destPtr, &destCols, (xCtrWidth - xCtr) * DestBytesPerPixel);
// Goto next row in source and destination image
destP += args.destArea.pitch;
srcP += args.vertFlip ? -args.src.pitch : args.src.pitch;
} else {
// Here we are scaling the image
int newSrcYCtr = scaleYCtr / BITMAP::SCALE_THRESHOLD;
// Since the source yctr might not update every row of the destination, we have
// to see if we are on a new row...
if (srcYCtr != newSrcYCtr) {
int diffSrcYCtr = newSrcYCtr - srcYCtr; // Have we moved yet
srcP += args.src.pitch * diffSrcYCtr;
srcYCtr = newSrcYCtr;
}
// Now also since we might skip a pixel or 2 or duplicate one to reach the desired
// scaling size, we create a small dummy buffer that we copy the pixels into and then
// call the drawPixelsSIMD function
byte srcBuffer[4*8] = {0};
for (int xCtr = xCtrStart, xCtrBpp = xCtrBppStart, destX = args.xStart, scaleXCtr = xCtrStart * args.scaleX; xCtr < xCtrWidth; destX += 8, xCtr += 8, xCtrBpp += SrcBytesPerPixel*4) {
if (yCtr + 1 == yCtrHeight && xCtr + 4 > xCtrWidth) break; // Don't go past the last 4 pixels
__m256i indexes = _mm256_set1_epi32(scaleXCtr);
// Calculate in parallel the indexes of the pixels
if (SrcBytesPerPixel == 4)
indexes = _mm256_slli_epi32(_mm256_srli_epi32(_mm256_add_epi32(indexes, scaleAdds), BITMAP::SCALE_THRESHOLD_BITS), 2);
else
indexes = _mm256_slli_epi32(_mm256_srli_epi32(_mm256_add_epi32(indexes, scaleAdds), BITMAP::SCALE_THRESHOLD_BITS), 1);
// Simply memcpy them in. memcpy has no real performance overhead here
memcpy(&srcBuffer[0*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 0), SrcBytesPerPixel);
memcpy(&srcBuffer[1*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 1), SrcBytesPerPixel);
memcpy(&srcBuffer[2*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 2), SrcBytesPerPixel);
memcpy(&srcBuffer[3*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 3), SrcBytesPerPixel);
memcpy(&srcBuffer[4*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 4), SrcBytesPerPixel);
memcpy(&srcBuffer[5*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 5), SrcBytesPerPixel);
memcpy(&srcBuffer[6*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 6), SrcBytesPerPixel);
memcpy(&srcBuffer[7*(size_t)SrcBytesPerPixel], srcP + _mm256_extract_epi32(indexes, 7), SrcBytesPerPixel);
scaleXCtr += args.scaleX*8;
// Now this is pretty much the same as before with non-scaled code, except that we use
// our dummy source buffer instead of the actuall source bitmap
byte *destPtr = &destP[destX * (intptr_t)DestBytesPerPixel];
__m256i skipMask = _mm256_cmpgt_epi32(_mm256_add_epi32(_mm256_add_epi32(_mm256_set1_epi32(xCtr), addIndexes), _mm256_set1_epi32(1)), xCtrWidthSIMD);
drawPixelSIMD<DestBytesPerPixel, SrcBytesPerPixel>(destPtr, (const byte *)srcBuffer, tint, alphas, maskedAlphas, transColors, 1, 0, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, skipMask);
}
// We calculate every row here except the last (because then we need to
// check for if we fall off the edge of the row)
// The only exception here is scaling drawing this is because:
// 1) if statements are costly, and the less we do the faster this loop is
// 2) with this, the only branch in the normal drawing loop is the width check
// 3) the scaling code will actually draw the until the last 4 pixels of the image
// and do the extra if checks because the scaling code is already much slower
// than the normal drawing loop, and the less duplicate code helps here.
if (yCtr + 1 != yCtrHeight) destP += args.destArea.pitch;
}
}
// Get the last x values of the last row
int xCtr = xCtrStart, xCtrBpp = xCtrBppStart, destX = args.xStart;
// We have a picture that is a multiple of 8, so no extra pixels to draw
/*if (xCtrWidth % 8 == 0)*/ return;
// Drawing the last few not scaled pixels here.
// Same as the loop above but now we check if we are going to overflow,
// and thus we don't need to mask out pixels that go over the row.
if (!Scale) {
for (; xCtr + 8 < xCtrWidth; destX += 8, xCtr += 8, xCtrBpp += SrcBytesPerPixel*8) {
byte *destPtr = &destP[(ptrdiff_t)destX * DestBytesPerPixel];
drawPixelSIMD<DestBytesPerPixel, SrcBytesPerPixel>(destPtr, srcP, tint, alphas, maskedAlphas, transColors, xDir, xCtrBpp, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, _mm256_setzero_si256());
}
// Because we move in 8 pixel units, and horizFlip moves in 1, we have to move
// 1 pixel past the last pixel we did not blit, meaning going forward 7 pixels.
if (args.horizFlip) srcP += SrcBytesPerPixel * 7;
} else {
// So if we are scaling, set up the xCtr to what it was before (AKA the last 8 or so pixels of the image)
xCtr = xCtrWidth - xCtrWidth % 8;
xCtrBpp = xCtr * SrcBytesPerPixel;
destX = args.xStart+xCtr;
}
// For the last 8 pixels, we just do them in serial, nothing special
for (; xCtr < xCtrWidth; ++destX, ++xCtr, xCtrBpp += SrcBytesPerPixel) {
const byte *srcColPtr = (const byte *)(srcP + xDir * xCtrBpp);
if (Scale) {
srcColPtr = (const byte *)(srcP + (xCtr * args.scaleX) / BITMAP::SCALE_THRESHOLD * SrcBytesPerPixel);
}
byte *destVal = (byte *)&destP[destX * DestBytesPerPixel];
uint32 srcCol = args.dstBitmap.getColor(srcColPtr, SrcBytesPerPixel);
// Check if this is a transparent color we should skip
if (args.skipTrans && ((srcCol & args.alphaMask) == args.transColor))
continue;
args.src.format.colorToARGB(srcCol, aSrc, rSrc, gSrc, bSrc);
if (args.srcAlpha != -1) {
if (args.useTint) {
rDest = rSrc;
gDest = gSrc;
bDest = bSrc;
aDest = aSrc;
rSrc = args.tintRed;
gSrc = args.tintGreen;
bSrc = args.tintBlue;
aSrc = args.srcAlpha;
}
args.dstBitmap.blendPixel(aSrc, rSrc, gSrc, bSrc, aDest, rDest, gDest, bDest, args.srcAlpha, args.useTint, destVal);
srcCol = args.dstBitmap.format.ARGBToColor(aDest, rDest, gDest, bDest);
} else {
srcCol = args.dstBitmap.format.ARGBToColor(aSrc, rSrc, gSrc, bSrc);
}
if (DestBytesPerPixel == 4)
*(uint32 *)destVal = srcCol;
else
*(uint16 *)destVal = srcCol;
}
}
template<bool Scale>
static void drawInner2Bpp(BITMAP::DrawInnerArgs &args) {
const int xDir = args.horizFlip ? -1 : 1;
byte rSrc, gSrc, bSrc, aSrc;
byte rDest = 0, gDest = 0, bDest = 0, aDest = 0;
__m256i tint = _mm256_set1_epi16(args.src.format.ARGBToColor(args.srcAlpha, args.tintRed, args.tintGreen, args.tintBlue));
__m256i transColors = _mm256_set1_epi16(args.transColor);
__m256i alphas = _mm256_set1_epi16(args.srcAlpha);
// This is so that we can calculate what pixels to crop off in a vectorized way
__m256i addIndexes = _mm256_set_epi16(15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0);
// This is so that we can calculate in parralell the pixel indexes for scaled drawing
if (args.horizFlip) addIndexes = _mm256_set_epi16(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15);
__m256i scaleAdds = _mm256_set_epi32((uint32)args.scaleX*7, (uint32)args.scaleX*6, (uint32)args.scaleX*5, (uint32)args.scaleX*4, (uint32)args.scaleX*3, (uint32)args.scaleX*2, (uint32)args.scaleX, 0);
__m256i scaleAdds2 = _mm256_set_epi32((uint32)args.scaleX*15, (uint32)args.scaleX*14, (uint32)args.scaleX*13, (uint32)args.scaleX*12, (uint32)args.scaleX*11, (uint32)args.scaleX*10, (uint32)args.scaleX*9, (uint32)args.scaleX*8);
// Clip the bounds ahead of time (so we don't waste time checking if we are in bounds when
// we are in the inner loop)
int xCtrStart = 0, xCtrBppStart = 0, xCtrWidth = args.dstRect.width();
if (args.xStart + xCtrWidth > args.destArea.w) {
xCtrWidth = args.destArea.w - args.xStart;
}
if (args.xStart < 0) {
xCtrStart = -args.xStart;
xCtrBppStart = xCtrStart * 2;
args.xStart = 0;
}
int destY = args.yStart, yCtr = 0, srcYCtr = 0, scaleYCtr = 0, yCtrHeight = args.dstRect.height();
if (Scale) yCtrHeight = args.dstRect.height();
if (args.yStart < 0) {
yCtr = -args.yStart;
destY = 0;
if (Scale) {
scaleYCtr = yCtr * args.scaleY;
srcYCtr = scaleYCtr / BITMAP::SCALE_THRESHOLD;
}
}
if (args.yStart + yCtrHeight > args.destArea.h) {
yCtrHeight = args.destArea.h - args.yStart;
}
/*if (!Scale && xCtrWidth % 16 != 0) {
--yCtrHeight;
}*/
const int secondToLast = xCtrWidth - 16;
byte *destP = (byte *)args.destArea.getBasePtr(0, destY);
const byte *srcP = (const byte *)args.src.getBasePtr(
args.horizFlip ? args.srcArea.right - 16 : args.srcArea.left,
args.vertFlip ? args.srcArea.bottom - 1 - yCtr : args.srcArea.top + yCtr);
for (; yCtr < yCtrHeight; ++destY, ++yCtr, scaleYCtr += args.scaleY) {
__m256i xCtrWidthSIMD = _mm256_set1_epi16(xCtrWidth); // This is the width of the row
if (!Scale) {
// If we are not scaling the image
int xCtr = xCtrStart, xCtrBpp = xCtrBppStart, destX = args.xStart;
for (; xCtr < secondToLast; destX += 16, xCtr += 16, xCtrBpp += 32) {
byte *destPtr = &destP[destX * 2];
drawPixelSIMD2Bpp(destPtr, srcP, tint, alphas, transColors, xDir, xCtrBpp, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, _mm256_set1_epi32(0));
}
byte *destPtr = &destP[destX * 2];
__m256i srcCols = _mm256_setzero_si256();
__m256i destCols = _mm256_setzero_si256();
const int copySize = (xCtrWidth - xCtr) * 2;
memcpy(&srcCols, srcP + xDir * xCtrBpp, copySize);
memcpy(&destCols, destPtr, copySize);
// Skip pixels that are beyond the row
// __m256i skipMask = _mm256_cmpgt_epi16(_mm256_add_epi16(_mm256_add_epi16(_mm256_set1_epi16(xCtr), addIndexes), _mm256_set1_epi16(1)), xCtrWidthSIMD);
drawPixelSIMD2Bpp((byte *)&destCols, (byte *)&srcCols, tint, alphas, transColors, xDir, 0, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, _mm256_set1_epi32(0));
memcpy(destPtr, &destCols, copySize);
// Goto next row in source and destination image
destP += args.destArea.pitch;
srcP += args.vertFlip ? -args.src.pitch : args.src.pitch;
} else {
// Here we are scaling the image
int newSrcYCtr = scaleYCtr / BITMAP::SCALE_THRESHOLD;
// Since the source yctr might not update every row of the destination, we have
// to see if we are on a new row...
if (srcYCtr != newSrcYCtr) {
int diffSrcYCtr = newSrcYCtr - srcYCtr;
srcP += args.src.pitch * diffSrcYCtr;
srcYCtr = newSrcYCtr;
}
// Now also since we might skip a pixel or 2 or duplicate one to reach the desired
// scaling size, we create a small dummy buffer that we copy the pixels into and then
// call the drawPixelsSIMD function
uint16 srcBuffer[16];
for (int xCtr = xCtrStart, xCtrBpp = xCtrBppStart, destX = args.xStart, scaleXCtr = xCtrStart * args.scaleX; xCtr < xCtrWidth; destX += 16, xCtr += 16, xCtrBpp += 32) {
if (yCtr + 1 == yCtrHeight && xCtr + 8 > xCtrWidth) break;
__m256i indexes = _mm256_set1_epi32(scaleXCtr), indexes2 = _mm256_set1_epi32(scaleXCtr);
// Calculate in parallel the indexes of the pixels
indexes = _mm256_slli_epi32(_mm256_srli_epi32(_mm256_add_epi32(indexes, scaleAdds), BITMAP::SCALE_THRESHOLD_BITS), 1);
indexes2 = _mm256_slli_epi32(_mm256_srli_epi32(_mm256_add_epi32(indexes2, scaleAdds2), BITMAP::SCALE_THRESHOLD_BITS), 1);
// Simply memcpy them in. memcpy has no real performance overhead here
srcBuffer[0] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 0));
srcBuffer[1] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 1));
srcBuffer[2] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 2));
srcBuffer[3] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 3));
srcBuffer[4] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 4));
srcBuffer[5] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 5));
srcBuffer[6] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 6));
srcBuffer[7] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes, 7));
srcBuffer[8] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 0));
srcBuffer[9] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 1));
srcBuffer[10] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 2));
srcBuffer[11] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 3));
srcBuffer[12] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 4));
srcBuffer[13] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 5));
srcBuffer[14] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 6));
srcBuffer[15] = *(const uint16 *)(srcP + _mm256_extract_epi32(indexes2, 7));
scaleXCtr += args.scaleX*16;
// Now this is pretty much the same as before with non-scaled code, except that we use
// our dummy source buffer instead of the actuall source bitmap
byte *destPtr = &destP[destX * 2];
__m256i skipMask = _mm256_cmpgt_epi16(_mm256_add_epi16(_mm256_add_epi16(_mm256_set1_epi16(xCtr), addIndexes), _mm256_set1_epi16(1)), xCtrWidthSIMD);
drawPixelSIMD2Bpp(destPtr, (const byte *)srcBuffer, tint, alphas, transColors, 1, 0, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, skipMask);
}
// We calculate every row here except the last (because then we need to
// check for if we fall off the edge of the row)
// The only exception here is scaling drawing this is because:
// 1) if statements are costly, and the less we do the faster this loop is
// 2) with this, the only branch in the normal drawing loop is the width check
// 3) the scaling code will actually draw the until the last 4 pixels of the image
// and do the extra if checks because the scaling code is already much slower
// than the normal drawing loop, and the less duplicate code helps here.
if (yCtr + 1 != yCtrHeight) destP += args.destArea.pitch;
}
}
// We have a picture that is a multiple of 16, so no extra pixels to draw
/*if (xCtrWidth % 16 == 0)*/ return;
// Get the last x values of the last row
int xCtr = xCtrStart, xCtrBpp = xCtrBppStart, destX = args.xStart;
// Drawing the last few not scaled pixels here.
// Same as the loop above but now we check if we are going to overflow,
// and thus we don't need to mask out pixels that go over the row.
if (!Scale) {
for (; xCtr + 16 < xCtrWidth; destX += 16, xCtr += 16, xCtrBpp += 32) {
byte *destPtr = &destP[destX * 2];
drawPixelSIMD2Bpp(destPtr, srcP, tint, alphas, transColors, xDir, xCtrBpp, args.srcAlpha, args.skipTrans, args.horizFlip, args.useTint, _mm256_setzero_si256());
}
// Because we move in 16 pixel units, and horizFlip moves in 1, we have to move
// 1 pixel past the last pixel we did not blit, meaning going forward 15 pixels.
if (args.horizFlip) srcP += 2 * 15;
} else {
// So if we are scaling, set up the xCtr to what it was before (AKA the last 16 or so pixels of the image)
xCtr = xCtrWidth - xCtrWidth % 16;
xCtrBpp = xCtr * 2;
destX = args.xStart+xCtr;
}
// For the last 16 pixels, we just do them in serial, nothing special
for (; xCtr < xCtrWidth; ++destX, ++xCtr, xCtrBpp += 2) {
const byte *srcColPtr = (const byte *)(srcP + xDir * xCtrBpp);
if (Scale) {
srcColPtr = (const byte *)(srcP + (xCtr * args.scaleX) / BITMAP::SCALE_THRESHOLD * 2);
}
byte *destVal = (byte *)&destP[destX * 2];
uint32 srcCol = (uint32)(*(const uint16 *)srcColPtr);
// Check if this is a transparent color we should skip
if (args.skipTrans && srcCol == args.transColor)
continue;
args.src.format.colorToARGB(srcCol, aSrc, rSrc, gSrc, bSrc);
if (args.srcAlpha != -1) {
if (args.useTint) {
rDest = rSrc;
gDest = gSrc;
bDest = bSrc;
aDest = aSrc;
rSrc = args.tintRed;
gSrc = args.tintGreen;
bSrc = args.tintBlue;
aSrc = args.srcAlpha;
}/* else {
format.colorToARGB((uint32)(*(uint16 *)destVal), aDest, rDest, gDest, bDest);
}*/
args.dstBitmap.blendPixel(aSrc, rSrc, gSrc, bSrc, aDest, rDest, gDest, bDest, args.srcAlpha, args.useTint, destVal);
srcCol = args.dstBitmap.format.ARGBToColor(aDest, rDest, gDest, bDest);
} else {
srcCol = args.dstBitmap.format.ARGBToColor(aSrc, rSrc, gSrc, bSrc);
}
*(uint16 *)destVal = srcCol;
}
}
template<bool Scale>
static void drawInner1Bpp(BITMAP::DrawInnerArgs &args) {
const int xDir = args.horizFlip ? -1 : 1;
__m256i transColors = _mm256_set1_epi16(args.transColor | (args.transColor << 8));
// This is so that we can calculate in parralell the pixel indexes for scaled drawing
__m256i scaleAdds1 = _mm256_set_epi32((uint32)args.scaleX*7, (uint32)args.scaleX*6, (uint32)args.scaleX*5, (uint32)args.scaleX*4, (uint32)args.scaleX*3, (uint32)args.scaleX*2, (uint32)args.scaleX, 0);
__m256i scaleAdds2 = _mm256_set_epi32((uint32)args.scaleX*15, (uint32)args.scaleX*14, (uint32)args.scaleX*13, (uint32)args.scaleX*12, (uint32)args.scaleX*11, (uint32)args.scaleX*10, (uint32)args.scaleX*9, (uint32)args.scaleX*8);
__m256i scaleAdds3 = _mm256_set_epi32((uint32)args.scaleX*23, (uint32)args.scaleX*22, (uint32)args.scaleX*21, (uint32)args.scaleX*20, (uint32)args.scaleX*19, (uint32)args.scaleX*18, (uint32)args.scaleX*17, (uint32)args.scaleX*16);
__m256i scaleAdds4 = _mm256_set_epi32((uint32)args.scaleX*31, (uint32)args.scaleX*30, (uint32)args.scaleX*29, (uint32)args.scaleX*28, (uint32)args.scaleX*27, (uint32)args.scaleX*26, (uint32)args.scaleX*25, (uint32)args.scaleX*24);
// Clip the bounds ahead of time (so we don't waste time checking if we are in bounds when
// we are in the inner loop)
int xCtrStart = 0, xCtrWidth = args.dstRect.width();
if (args.xStart + xCtrWidth > args.destArea.w) {
xCtrWidth = args.destArea.w - args.xStart;
}
if (args.xStart < 0) {
xCtrStart = -args.xStart;
args.xStart = 0;
}
int destY = args.yStart, yCtr = 0, srcYCtr = 0, scaleYCtr = 0, yCtrHeight = args.dstRect.height();
if (Scale) yCtrHeight = args.dstRect.height();
if (args.yStart < 0) {
yCtr = -args.yStart;
destY = 0;
if (Scale) {
scaleYCtr = yCtr * args.scaleY;
srcYCtr = scaleYCtr / BITMAP::SCALE_THRESHOLD;
}
}
if (args.yStart + yCtrHeight > args.destArea.h) {
yCtrHeight = args.destArea.h - args.yStart;
}
byte *destP = (byte *)args.destArea.getBasePtr(0, destY);
const byte *srcP = (const byte *)args.src.getBasePtr(
args.horizFlip ? args.srcArea.right - 32 : args.srcArea.left,
args.vertFlip ? args.srcArea.bottom - 1 - yCtr : args.srcArea.top + yCtr);
for (; yCtr < yCtrHeight; ++destY, ++yCtr, scaleYCtr += args.scaleY) {
if (Scale) {
// So here we update the srcYCtr differently due to this being for
// scaling
int newSrcYCtr = scaleYCtr / BITMAP::SCALE_THRESHOLD;
if (srcYCtr != newSrcYCtr) {
// Since the source yctr might not update every row of the destination, we have
// to see if we are on a new row...
int diffSrcYCtr = newSrcYCtr - srcYCtr;
srcP += args.src.pitch * diffSrcYCtr;
srcYCtr = newSrcYCtr;
}
}
int xCtr = xCtrStart, destX = args.xStart, scaleXCtr = xCtrStart * args.scaleX;
for (; xCtr + 32 < xCtrWidth; destX += 32, xCtr += 32) {
byte *destPtr = &destP[destX];
// Here we dont use the drawPixelSIMD function because 1bpp bitmaps in allegro
// can't have any blending applied to them
__m256i destCols = _mm256_loadu_si256((const __m256i *)destPtr);
__m256i srcCols = _mm256_loadu_si256((const __m256i *)(srcP + xDir * xCtr));
if (Scale) {
// If we are scaling, we have to set each pixel individually
__m256i indexes1 = _mm256_set1_epi32(scaleXCtr), indexes2 = _mm256_set1_epi32(scaleXCtr);
__m256i indexes3 = _mm256_set1_epi32(scaleXCtr), indexes4 = _mm256_set1_epi32(scaleXCtr);
// Calculate in parallel the indexes of the pixels
indexes1 = _mm256_srli_epi32(_mm256_add_epi32(indexes1, scaleAdds1), BITMAP::SCALE_THRESHOLD_BITS);
indexes2 = _mm256_srli_epi32(_mm256_add_epi32(indexes2, scaleAdds2), BITMAP::SCALE_THRESHOLD_BITS);
indexes3 = _mm256_srli_epi32(_mm256_add_epi32(indexes3, scaleAdds3), BITMAP::SCALE_THRESHOLD_BITS);
indexes4 = _mm256_srli_epi32(_mm256_add_epi32(indexes4, scaleAdds4), BITMAP::SCALE_THRESHOLD_BITS);
srcCols = _mm256_set_epi8(
srcP[_mm256_extract_epi32(indexes4, 7)],
srcP[_mm256_extract_epi32(indexes4, 6)],
srcP[_mm256_extract_epi32(indexes4, 5)],
srcP[_mm256_extract_epi32(indexes4, 4)],
srcP[_mm256_extract_epi32(indexes4, 3)],
srcP[_mm256_extract_epi32(indexes4, 2)],
srcP[_mm256_extract_epi32(indexes4, 1)],
srcP[_mm256_extract_epi32(indexes4, 0)],
srcP[_mm256_extract_epi32(indexes3, 7)],
srcP[_mm256_extract_epi32(indexes3, 6)],
srcP[_mm256_extract_epi32(indexes3, 5)],
srcP[_mm256_extract_epi32(indexes3, 4)],
srcP[_mm256_extract_epi32(indexes3, 3)],
srcP[_mm256_extract_epi32(indexes3, 2)],
srcP[_mm256_extract_epi32(indexes3, 1)],
srcP[_mm256_extract_epi32(indexes3, 0)],
srcP[_mm256_extract_epi32(indexes2, 7)],
srcP[_mm256_extract_epi32(indexes2, 6)],
srcP[_mm256_extract_epi32(indexes2, 5)],
srcP[_mm256_extract_epi32(indexes2, 4)],
srcP[_mm256_extract_epi32(indexes2, 3)],
srcP[_mm256_extract_epi32(indexes2, 2)],
srcP[_mm256_extract_epi32(indexes2, 1)],
srcP[_mm256_extract_epi32(indexes2, 0)],
srcP[_mm256_extract_epi32(indexes1, 7)],
srcP[_mm256_extract_epi32(indexes1, 6)],
srcP[_mm256_extract_epi32(indexes1, 5)],
srcP[_mm256_extract_epi32(indexes1, 4)],
srcP[_mm256_extract_epi32(indexes1, 3)],
srcP[_mm256_extract_epi32(indexes1, 2)],
srcP[_mm256_extract_epi32(indexes1, 1)],
srcP[_mm256_extract_epi32(indexes1, 0)]);
scaleXCtr += args.scaleX*16;
}
// Mask out transparent pixels
__m256i mask1 = args.skipTrans ? _mm256_cmpeq_epi8(srcCols, transColors) : _mm256_setzero_si256();
__m256i final = _mm256_or_si256(_mm256_andnot_si256(mask1, srcCols), _mm256_and_si256(destCols, mask1));
if (args.horizFlip) {
__m256i final_swap16 = _mm256_srli_epi16(final, 8);
final_swap16 = _mm256_or_si256(final_swap16, _mm256_slli_epi16(_mm256_and_si256(final, _mm256_set1_epi16(0xff)), 8));
final_swap16 = _mm256_shufflelo_epi16(final_swap16, _MM_SHUFFLE(0, 1, 2, 3));
final_swap16 = _mm256_shufflehi_epi16(final_swap16, _MM_SHUFFLE(0, 1, 2, 3));
final = _mm256_castpd_si256(_mm256_shuffle_pd(_mm256_castsi256_pd(final_swap16), _mm256_castsi256_pd(final_swap16), _MM_SHUFFLE2(0, 1)));
}
_mm256_storeu_si256((__m256i *)destPtr, final);
}
// Get the last x values
// Because we move in 16 pixel units, and horizFlip moves in 1, we have to move
// 1 pixel past the last pixel we did not blit, meaning going forward 15 pixels.
if (args.horizFlip) srcP += 15;
for (; xCtr < xCtrWidth; ++destX, ++xCtr, scaleXCtr += args.scaleX) {
const byte *srcCol = (const byte *)(srcP + xDir * xCtr);
if (Scale) {
srcCol = (const byte *)(srcP + scaleXCtr / BITMAP::SCALE_THRESHOLD);
}
// Check if this is a transparent color we should skip
if (args.skipTrans && *srcCol == args.transColor)
continue;
byte *destVal = (byte *)&destP[destX];
*destVal = *srcCol;
}
if (args.horizFlip) srcP -= 15; // Undo what we did up there
destP += args.destArea.pitch; // Goto next row
// Only advance the src row by 1 every time like this if we don't scale
if (!Scale) srcP += args.vertFlip ? -args.src.pitch : args.src.pitch;
}
}
}; // end of class DrawInnerImpl_AVX2
template<bool Scale>
void BITMAP::drawAVX2(DrawInnerArgs &args) {
if (args.sameFormat) {
switch (format.bytesPerPixel) {
case 1: DrawInnerImpl_AVX2::drawInner1Bpp<Scale>(args); break;
case 2: DrawInnerImpl_AVX2::drawInner2Bpp<Scale>(args); break;
case 4: DrawInnerImpl_AVX2::drawInner4BppWithConv<4, 4, Scale>(args); break;
}
} else if (format.bytesPerPixel == 4 && args.src.format.bytesPerPixel == 2) {
DrawInnerImpl_AVX2::drawInner4BppWithConv<4, 2, Scale>(args);
} else if (format.bytesPerPixel == 2 && args.src.format.bytesPerPixel == 4) {
DrawInnerImpl_AVX2::drawInner4BppWithConv<2, 4, Scale>(args);
}
}
template void BITMAP::drawAVX2<false>(DrawInnerArgs &);
template void BITMAP::drawAVX2<true>(DrawInnerArgs &);
} // namespace AGS3
#if defined(__clang__)
#pragma clang attribute pop
#elif defined(__GNUC__)
#pragma GCC pop_options
#endif
Reference in New Issue View Git Blame Copy Permalink