ryan/mozjpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
4f51f36eb347e27b51da84b511f25f8b9c8672fe
mozjpeg/simd/powerpc/jfdctint-altivec.c
DRC 6e632af9f6 Demote "fast" [I]DCT algorithms to legacy status 
- Refer to the "slow" [I]DCT algorithms as "accurate" instead, since
  they are not slow under libjpeg-turbo.
- Adjust documentation claims to reflect the fact that the "slow" and
  "fast" algorithms produce about the same performance on AVX2-equipped
  CPUs (because of the dual-lane nature of AVX2, it was not possible to
  accelerate the "fast" algorithm beyond what was achievable with SSE2.)
  Also adjust the claims to reflect the fact that the "fast" algorithm
  tends to be ~5-15% faster than the "slow" algorithm on
  non-AVX2-equipped CPUs, regardless of the use of the libjpeg-turbo
  SIMD extensions.
- Indicate the legacy status of the "fast" and float algorithms in the
  documentation and cjpeg/djpeg usage info.
- Remove obsolete paragraph in the djpeg man page that suggested that
  the float algorithm could be faster than the "fast" algorithm on some
  CPUs.
2020-11-05 15:59:31 -06:00

259 lines
8.4 KiB
C
Raw Blame History

/*
* AltiVec optimizations for libjpeg-turbo
*
* Copyright (C) 2014, 2020, D. R. Commander. All Rights Reserved.
*
* This software is provided 'as-is', without any express or implied
* warranty. In no event will the authors be held liable for any damages
* arising from the use of this software.
*
* Permission is granted to anyone to use this software for any purpose,
* including commercial applications, and to alter it and redistribute it
* freely, subject to the following restrictions:
*
* 1. The origin of this software must not be misrepresented; you must not
* claim that you wrote the original software. If you use this software
* in a product, an acknowledgment in the product documentation would be
* appreciated but is not required.
* 2. Altered source versions must be plainly marked as such, and must not be
* misrepresented as being the original software.
* 3. This notice may not be removed or altered from any source distribution.
*/
/* ACCURATE INTEGER FORWARD DCT */
#include "jsimd_altivec.h"
#define F_0_298 2446 /* FIX(0.298631336) */
#define F_0_390 3196 /* FIX(0.390180644) */
#define F_0_541 4433 /* FIX(0.541196100) */
#define F_0_765 6270 /* FIX(0.765366865) */
#define F_0_899 7373 /* FIX(0.899976223) */
#define F_1_175 9633 /* FIX(1.175875602) */
#define F_1_501 12299 /* FIX(1.501321110) */
#define F_1_847 15137 /* FIX(1.847759065) */
#define F_1_961 16069 /* FIX(1.961570560) */
#define F_2_053 16819 /* FIX(2.053119869) */
#define F_2_562 20995 /* FIX(2.562915447) */
#define F_3_072 25172 /* FIX(3.072711026) */
#define CONST_BITS 13
#define PASS1_BITS 2
#define DESCALE_P1 (CONST_BITS - PASS1_BITS)
#define DESCALE_P2 (CONST_BITS + PASS1_BITS)
#define DO_FDCT_COMMON(PASS) { \
/* (Original) \
* z1 = (tmp12 + tmp13) * 0.541196100; \
* data2 = z1 + tmp13 * 0.765366865; \
* data6 = z1 + tmp12 * -1.847759065; \
* \
* (This implementation) \
* data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100; \
* data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065); \
*/ \
\
tmp1312l = vec_mergeh(tmp13, tmp12); \
tmp1312h = vec_mergel(tmp13, tmp12); \
\
out2l = vec_msums(tmp1312l, pw_f130_f054, pd_descale_p##PASS); \
out2h = vec_msums(tmp1312h, pw_f130_f054, pd_descale_p##PASS); \
out6l = vec_msums(tmp1312l, pw_f054_mf130, pd_descale_p##PASS); \
out6h = vec_msums(tmp1312h, pw_f054_mf130, pd_descale_p##PASS); \
\
out2l = vec_sra(out2l, descale_p##PASS); \
out2h = vec_sra(out2h, descale_p##PASS); \
out6l = vec_sra(out6l, descale_p##PASS); \
out6h = vec_sra(out6h, descale_p##PASS); \
\
out2 = vec_pack(out2l, out2h); \
out6 = vec_pack(out6l, out6h); \
\
/* Odd part */ \
\
z3 = vec_add(tmp4, tmp6); \
z4 = vec_add(tmp5, tmp7); \
\
/* (Original) \
* z5 = (z3 + z4) * 1.175875602; \
* z3 = z3 * -1.961570560; z4 = z4 * -0.390180644; \
* z3 += z5; z4 += z5; \
* \
* (This implementation) \
* z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602; \
* z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644); \
*/ \
\
z34l = vec_mergeh(z3, z4); \
z34h = vec_mergel(z3, z4); \
\
z3l = vec_msums(z34l, pw_mf078_f117, pd_descale_p##PASS); \
z3h = vec_msums(z34h, pw_mf078_f117, pd_descale_p##PASS); \
z4l = vec_msums(z34l, pw_f117_f078, pd_descale_p##PASS); \
z4h = vec_msums(z34h, pw_f117_f078, pd_descale_p##PASS); \
\
/* (Original) \
* z1 = tmp4 + tmp7; z2 = tmp5 + tmp6; \
* tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869; \
* tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110; \
* z1 = z1 * -0.899976223; z2 = z2 * -2.562915447; \
* data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4; \
* data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4; \
* \
* (This implementation) \
* tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223; \
* tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447; \
* tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447); \
* tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223); \
* data7 = tmp4 + z3; data5 = tmp5 + z4; \
* data3 = tmp6 + z3; data1 = tmp7 + z4; \
*/ \
\
tmp47l = vec_mergeh(tmp4, tmp7); \
tmp47h = vec_mergel(tmp4, tmp7); \
\
out7l = vec_msums(tmp47l, pw_mf060_mf089, z3l); \
out7h = vec_msums(tmp47h, pw_mf060_mf089, z3h); \
out1l = vec_msums(tmp47l, pw_mf089_f060, z4l); \
out1h = vec_msums(tmp47h, pw_mf089_f060, z4h); \
\
out7l = vec_sra(out7l, descale_p##PASS); \
out7h = vec_sra(out7h, descale_p##PASS); \
out1l = vec_sra(out1l, descale_p##PASS); \
out1h = vec_sra(out1h, descale_p##PASS); \
\
out7 = vec_pack(out7l, out7h); \
out1 = vec_pack(out1l, out1h); \
\
tmp56l = vec_mergeh(tmp5, tmp6); \
tmp56h = vec_mergel(tmp5, tmp6); \
\
out5l = vec_msums(tmp56l, pw_mf050_mf256, z4l); \
out5h = vec_msums(tmp56h, pw_mf050_mf256, z4h); \
out3l = vec_msums(tmp56l, pw_mf256_f050, z3l); \
out3h = vec_msums(tmp56h, pw_mf256_f050, z3h); \
\
out5l = vec_sra(out5l, descale_p##PASS); \
out5h = vec_sra(out5h, descale_p##PASS); \
out3l = vec_sra(out3l, descale_p##PASS); \
out3h = vec_sra(out3h, descale_p##PASS); \
\
out5 = vec_pack(out5l, out5h); \
out3 = vec_pack(out3l, out3h); \
}
#define DO_FDCT_PASS1() { \
/* Even part */ \
\
tmp10 = vec_add(tmp0, tmp3); \
tmp13 = vec_sub(tmp0, tmp3); \
tmp11 = vec_add(tmp1, tmp2); \
tmp12 = vec_sub(tmp1, tmp2); \
\
out0 = vec_add(tmp10, tmp11); \
out0 = vec_sl(out0, pass1_bits); \
out4 = vec_sub(tmp10, tmp11); \
out4 = vec_sl(out4, pass1_bits); \
\
DO_FDCT_COMMON(1); \
}
#define DO_FDCT_PASS2() { \
/* Even part */ \
\
tmp10 = vec_add(tmp0, tmp3); \
tmp13 = vec_sub(tmp0, tmp3); \
tmp11 = vec_add(tmp1, tmp2); \
tmp12 = vec_sub(tmp1, tmp2); \
\
out0 = vec_add(tmp10, tmp11); \
out0 = vec_add(out0, pw_descale_p2x); \
out0 = vec_sra(out0, pass1_bits); \
out4 = vec_sub(tmp10, tmp11); \
out4 = vec_add(out4, pw_descale_p2x); \
out4 = vec_sra(out4, pass1_bits); \
\
DO_FDCT_COMMON(2); \
}
void jsimd_fdct_islow_altivec(DCTELEM *data)
{
__vector short row0, row1, row2, row3, row4, row5, row6, row7,
col0, col1, col2, col3, col4, col5, col6, col7,
tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
tmp47l, tmp47h, tmp56l, tmp56h, tmp1312l, tmp1312h,
z3, z4, z34l, z34h,
out0, out1, out2, out3, out4, out5, out6, out7;
__vector int z3l, z3h, z4l, z4h,
out1l, out1h, out2l, out2h, out3l, out3h, out5l, out5h, out6l, out6h,
out7l, out7h;
/* Constants */
__vector short
pw_f130_f054 = { __4X2(F_0_541 + F_0_765, F_0_541) },
pw_f054_mf130 = { __4X2(F_0_541, F_0_541 - F_1_847) },
pw_mf078_f117 = { __4X2(F_1_175 - F_1_961, F_1_175) },
pw_f117_f078 = { __4X2(F_1_175, F_1_175 - F_0_390) },
pw_mf060_mf089 = { __4X2(F_0_298 - F_0_899, -F_0_899) },
pw_mf089_f060 = { __4X2(-F_0_899, F_1_501 - F_0_899) },
pw_mf050_mf256 = { __4X2(F_2_053 - F_2_562, -F_2_562) },
pw_mf256_f050 = { __4X2(-F_2_562, F_3_072 - F_2_562) },
pw_descale_p2x = { __8X(1 << (PASS1_BITS - 1)) };
__vector unsigned short pass1_bits = { __8X(PASS1_BITS) };
__vector int pd_descale_p1 = { __4X(1 << (DESCALE_P1 - 1)) },
pd_descale_p2 = { __4X(1 << (DESCALE_P2 - 1)) };
__vector unsigned int descale_p1 = { __4X(DESCALE_P1) },
descale_p2 = { __4X(DESCALE_P2) };
/* Pass 1: process rows */
row0 = vec_ld(0, data);
row1 = vec_ld(16, data);
row2 = vec_ld(32, data);
row3 = vec_ld(48, data);
row4 = vec_ld(64, data);
row5 = vec_ld(80, data);
row6 = vec_ld(96, data);
row7 = vec_ld(112, data);
TRANSPOSE(row, col);
tmp0 = vec_add(col0, col7);
tmp7 = vec_sub(col0, col7);
tmp1 = vec_add(col1, col6);
tmp6 = vec_sub(col1, col6);
tmp2 = vec_add(col2, col5);
tmp5 = vec_sub(col2, col5);
tmp3 = vec_add(col3, col4);
tmp4 = vec_sub(col3, col4);
DO_FDCT_PASS1();
/* Pass 2: process columns */
TRANSPOSE(out, row);
tmp0 = vec_add(row0, row7);
tmp7 = vec_sub(row0, row7);
tmp1 = vec_add(row1, row6);
tmp6 = vec_sub(row1, row6);
tmp2 = vec_add(row2, row5);
tmp5 = vec_sub(row2, row5);
tmp3 = vec_add(row3, row4);
tmp4 = vec_sub(row3, row4);
DO_FDCT_PASS2();
vec_st(out0, 0, data);
vec_st(out1, 16, data);
vec_st(out2, 32, data);
vec_st(out3, 48, data);
vec_st(out4, 64, data);
vec_st(out5, 80, data);
vec_st(out6, 96, data);
vec_st(out7, 112, data);
}
Reference in New Issue View Git Blame Copy Permalink