ryan/mozjpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

AltiVec SIMD implementation of slow integer inverse DCT

Browse Source 
git-svn-id: svn://svn.code.sf.net/p/libjpeg-turbo/code/trunk@1461 632fc199-4ca6-4c93-a231-07263d6284db
This commit is contained in:
DRC
2014-12-20 01:17:39 +00:00
parent 63c1674ebc
commit 0d435698f4
3 changed files with 291 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
simd/jsimd.h
View File

@@ -646,6 +646,10 @@ EXTERN(void) jsimd_idct_islow_mips_dspr2
(void * dct_table, JCOEFPTR coef_block, int * output_buf,
JSAMPLE * output_col);
EXTERN(void) jsimd_idct_islow_altivec
(void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col);
/* Fast Integer Inverse DCT */
EXTERN(void) jsimd_idct_ifast_mmx
(void * dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,

13
simd/jsimd_powerpc.c
View File

@@ -333,6 +333,17 @@ jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
GLOBAL(int)
jsimd_can_idct_islow (void)
{
init_simd();
/* The code is optimised for these values only */
if (DCTSIZE != 8)
return 0;
if (sizeof(DCTELEM) != 2)
return 0;
if (simd_support & JSIMD_ALTIVEC)
return 1;
return 0;
}
@@ -364,6 +375,8 @@ jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
JCOEFPTR coef_block, JSAMPARRAY output_buf,
JDIMENSION output_col)
{
jsimd_idct_islow_altivec(compptr->dct_table, coef_block, output_buf,
output_col);
}
GLOBAL(void)

278
simd/jsimd_powerpc_altivec.c
View File

@@ -235,6 +235,7 @@ jsimd_fdct_islow_altivec (DCTELEM *data)
out1l, out1h, out2l, out2h, out3l, out3h, out5l, out5h, out6l, out6h,
out7l, out7h;
/* Constants */
__vector short
PW_F130_F054 = { __4X2(ISLOW_F_0_541 + ISLOW_F_0_765, ISLOW_F_0_541) },
PW_F054_MF130 = { __4X2(ISLOW_F_0_541, ISLOW_F_0_541 - ISLOW_F_1_847) },
@@ -245,8 +246,6 @@ jsimd_fdct_islow_altivec (DCTELEM *data)
PW_MF050_MF256 = { __4X2(ISLOW_F_2_053 - ISLOW_F_2_562, -ISLOW_F_2_562) },
PW_MF256_F050 = { __4X2(-ISLOW_F_2_562, ISLOW_F_3_072 - ISLOW_F_2_562) },
PW_DESCALE_P2X = { __8X(1 << (ISLOW_PASS1_BITS - 1)) };
/* Constants */
__vector unsigned short PASS1_BITS = { __8X(ISLOW_PASS1_BITS) };
__vector int zero = { __4X(0) },
PD_DESCALE_P1 = { __4X(1 << (ISLOW_DESCALE_P1 - 1)) },
@@ -439,6 +438,278 @@ jsimd_fdct_ifast_altivec (DCTELEM *data)
}
/* SLOW INTEGER INVERSE DCT */
#define DO_IDCT_ISLOW(in, PASS) \
{ \
/* Even part */ \
\
in##26l = vec_mergeh(in##2, in##6); \
in##26h = vec_mergel(in##2, in##6); \
\
tmp3l = vec_msums(in##26l, PW_F130_F054, zero32); \
tmp3h = vec_msums(in##26h, PW_F130_F054, zero32); \
tmp2l = vec_msums(in##26l, PW_F054_MF130, zero32); \
tmp2h = vec_msums(in##26h, PW_F054_MF130, zero32); \
\
tmp0 = vec_add(in##0, in##4); \
tmp1 = vec_sub(in##0, in##4); \
\
tmp0l = vec_unpackh(tmp0); \
tmp0h = vec_unpackl(tmp0); \
tmp0l = vec_sl(tmp0l, CONST_BITS); \
tmp0h = vec_sl(tmp0h, CONST_BITS); \
\
tmp10l = vec_add(tmp0l, tmp3l); \
tmp10h = vec_add(tmp0h, tmp3h); \
tmp13l = vec_sub(tmp0l, tmp3l); \
tmp13h = vec_sub(tmp0h, tmp3h); \
\
tmp1l = vec_unpackh(tmp1); \
tmp1h = vec_unpackl(tmp1); \
tmp1l = vec_sl(tmp1l, CONST_BITS); \
tmp1h = vec_sl(tmp1h, CONST_BITS); \
\
tmp11l = vec_add(tmp1l, tmp2l); \
tmp11h = vec_add(tmp1h, tmp2h); \
tmp12l = vec_sub(tmp1l, tmp2l); \
tmp12h = vec_sub(tmp1h, tmp2h); \
\
/* Odd part */ \
\
z3 = vec_add(in##3, in##7); \
z4 = vec_add(in##1, in##5); \
\
z34l = vec_mergeh(z3, z4); \
z34h = vec_mergel(z3, z4); \
\
z3l = vec_msums(z34l, PW_MF078_F117, zero32); \
z3h = vec_msums(z34h, PW_MF078_F117, zero32); \
z4l = vec_msums(z34l, PW_F117_F078, zero32); \
z4h = vec_msums(z34h, PW_F117_F078, zero32); \
\
in##71l = vec_mergeh(in##7, in##1); \
in##71h = vec_mergel(in##7, in##1); \
\
tmp0l = vec_msums(in##71l, PW_MF060_MF089, z3l); \
tmp0h = vec_msums(in##71h, PW_MF060_MF089, z3h); \
tmp3l = vec_msums(in##71l, PW_MF089_F060, z4l); \
tmp3h = vec_msums(in##71h, PW_MF089_F060, z4h); \
\
in##53l = vec_mergeh(in##5, in##3); \
in##53h = vec_mergel(in##5, in##3); \
\
tmp1l = vec_msums(in##53l, PW_MF050_MF256, z4l); \
tmp1h = vec_msums(in##53h, PW_MF050_MF256, z4h); \
tmp2l = vec_msums(in##53l, PW_MF256_F050, z3l); \
tmp2h = vec_msums(in##53h, PW_MF256_F050, z3h); \
\
/* Final output stage */ \
\
out0l = vec_add(tmp10l, tmp3l); \
out0h = vec_add(tmp10h, tmp3h); \
out7l = vec_sub(tmp10l, tmp3l); \
out7h = vec_sub(tmp10h, tmp3h); \
\
out0l = vec_add(out0l, PD_DESCALE_P##PASS); \
out0h = vec_add(out0h, PD_DESCALE_P##PASS); \
out0l = vec_sra(out0l, DESCALE_P##PASS); \
out0h = vec_sra(out0h, DESCALE_P##PASS); \
\
out7l = vec_add(out7l, PD_DESCALE_P##PASS); \
out7h = vec_add(out7h, PD_DESCALE_P##PASS); \
out7l = vec_sra(out7l, DESCALE_P##PASS); \
out7h = vec_sra(out7h, DESCALE_P##PASS); \
\
out0 = vec_pack(out0l, out0h); \
out7 = vec_pack(out7l, out7h); \
\
out1l = vec_add(tmp11l, tmp2l); \
out1h = vec_add(tmp11h, tmp2h); \
out6l = vec_sub(tmp11l, tmp2l); \
out6h = vec_sub(tmp11h, tmp2h); \
\
out1l = vec_add(out1l, PD_DESCALE_P##PASS); \
out1h = vec_add(out1h, PD_DESCALE_P##PASS); \
out1l = vec_sra(out1l, DESCALE_P##PASS); \
out1h = vec_sra(out1h, DESCALE_P##PASS); \
\
out6l = vec_add(out6l, PD_DESCALE_P##PASS); \
out6h = vec_add(out6h, PD_DESCALE_P##PASS); \
out6l = vec_sra(out6l, DESCALE_P##PASS); \
out6h = vec_sra(out6h, DESCALE_P##PASS); \
\
out1 = vec_pack(out1l, out1h); \
out6 = vec_pack(out6l, out6h); \
\
out2l = vec_add(tmp12l, tmp1l); \
out2h = vec_add(tmp12h, tmp1h); \
out5l = vec_sub(tmp12l, tmp1l); \
out5h = vec_sub(tmp12h, tmp1h); \
\
out2l = vec_add(out2l, PD_DESCALE_P##PASS); \
out2h = vec_add(out2h, PD_DESCALE_P##PASS); \
out2l = vec_sra(out2l, DESCALE_P##PASS); \
out2h = vec_sra(out2h, DESCALE_P##PASS); \
\
out5l = vec_add(out5l, PD_DESCALE_P##PASS); \
out5h = vec_add(out5h, PD_DESCALE_P##PASS); \
out5l = vec_sra(out5l, DESCALE_P##PASS); \
out5h = vec_sra(out5h, DESCALE_P##PASS); \
\
out2 = vec_pack(out2l, out2h); \
out5 = vec_pack(out5l, out5h); \
\
out3l = vec_add(tmp13l, tmp0l); \
out3h = vec_add(tmp13h, tmp0h); \
out4l = vec_sub(tmp13l, tmp0l); \
out4h = vec_sub(tmp13h, tmp0h); \
\
out3l = vec_add(out3l, PD_DESCALE_P##PASS); \
out3h = vec_add(out3h, PD_DESCALE_P##PASS); \
out3l = vec_sra(out3l, DESCALE_P##PASS); \
out3h = vec_sra(out3h, DESCALE_P##PASS); \
\
out4l = vec_add(out4l, PD_DESCALE_P##PASS); \
out4h = vec_add(out4h, PD_DESCALE_P##PASS); \
out4l = vec_sra(out4l, DESCALE_P##PASS); \
out4h = vec_sra(out4h, DESCALE_P##PASS); \
\
out3 = vec_pack(out3l, out3h); \
out4 = vec_pack(out4l, out4h); \
}
void
jsimd_idct_islow_altivec (void * dct_table_, JCOEFPTR coef_block,
JSAMPARRAY output_buf, JDIMENSION output_col)
{
short *dct_table = (short *)dct_table_;
__vector short row0, row1, row2, row3, row4, row5, row6, row7,
col0, col1, col2, col3, col4, col5, col6, col7,
quant0, quant1, quant2, quant3, quant4, quant5, quant6, quant7,
tmp0, tmp1, tmp2, tmp3, z3, z4,
z34l, z34h, col71l, col71h, col26l, col26h, col53l, col53h,
row71l, row71h, row26l, row26h, row53l, row53h,
out0, out1, out2, out3, out4, out5, out6, out7;
__vector int tmp0l, tmp0h, tmp1l, tmp1h, tmp2l, tmp2h, tmp3l, tmp3h,
tmp10l, tmp10h, tmp11l, tmp11h, tmp12l, tmp12h, tmp13l, tmp13h,
z3l, z3h, z4l, z4h,
out0l, out0h, out1l, out1h, out2l, out2h, out3l, out3h, out4l, out4h,
out5l, out5h, out6l, out6h, out7l, out7h;
__vector signed char outb;
long long *outptr, *outbptr = (long long *)(&outb);
/* Constants */
__vector short zero16 = { __8X(0) },
PW_F130_F054 = { __4X2(ISLOW_F_0_541 + ISLOW_F_0_765, ISLOW_F_0_541) },
PW_F054_MF130 = { __4X2(ISLOW_F_0_541, ISLOW_F_0_541 - ISLOW_F_1_847) },
PW_MF078_F117 = { __4X2(ISLOW_F_1_175 - ISLOW_F_1_961, ISLOW_F_1_175) },
PW_F117_F078 = { __4X2(ISLOW_F_1_175, ISLOW_F_1_175 - ISLOW_F_0_390) },
PW_MF060_MF089 = { __4X2(ISLOW_F_0_298 - ISLOW_F_0_899, -ISLOW_F_0_899) },
PW_MF089_F060 = { __4X2(-ISLOW_F_0_899, ISLOW_F_1_501 - ISLOW_F_0_899) },
PW_MF050_MF256 = { __4X2(ISLOW_F_2_053 - ISLOW_F_2_562, -ISLOW_F_2_562) },
PW_MF256_F050 = { __4X2(-ISLOW_F_2_562, ISLOW_F_3_072 - ISLOW_F_2_562) };
__vector unsigned short PASS1_BITS = { __8X(ISLOW_PASS1_BITS) };
__vector int zero32 = { __4X(0) },
PD_DESCALE_P1 = { __4X(1 << (ISLOW_DESCALE_P1 - 1)) },
PD_DESCALE_P2 = { __4X(1 << (ISLOW_DESCALE_P2 + 3 - 1)) };
__vector unsigned int DESCALE_P1 = { __4X(ISLOW_DESCALE_P1) },
DESCALE_P2 = { __4X(ISLOW_DESCALE_P2 + 3) },
CONST_BITS = { __4X(ISLOW_CONST_BITS) };
__vector signed char PB_CENTERJSAMP = { __16X(CENTERJSAMPLE) };
/* Pass 1: process columns. */
col0 = *(__vector short *)&coef_block[0];
col1 = *(__vector short *)&coef_block[8];
col2 = *(__vector short *)&coef_block[16];
col3 = *(__vector short *)&coef_block[24];
col4 = *(__vector short *)&coef_block[32];
col5 = *(__vector short *)&coef_block[40];
col6 = *(__vector short *)&coef_block[48];
col7 = *(__vector short *)&coef_block[56];
tmp1 = vec_or(col1, col2);
tmp2 = vec_or(col3, col4);
tmp1 = vec_or(tmp1, tmp2);
tmp3 = vec_or(col5, col6);
tmp3 = vec_or(tmp3, col7);
tmp1 = vec_or(tmp1, tmp3);
quant0 = *(__vector short *)&dct_table[0];
col0 = vec_mladd(col0, quant0, zero16);
if (vec_all_eq(tmp1, zero16)) {
/* AC terms all zero */
col0 = vec_sl(col0, PASS1_BITS);
row0 = vec_splat(col0, 0);
row1 = vec_splat(col0, 1);
row2 = vec_splat(col0, 2);
row3 = vec_splat(col0, 3);
row4 = vec_splat(col0, 4);
row5 = vec_splat(col0, 5);
row6 = vec_splat(col0, 6);
row7 = vec_splat(col0, 7);
} else {
quant1 = *(__vector short *)&dct_table[8];
quant2 = *(__vector short *)&dct_table[16];
quant3 = *(__vector short *)&dct_table[24];
quant4 = *(__vector short *)&dct_table[32];
quant5 = *(__vector short *)&dct_table[40];
quant6 = *(__vector short *)&dct_table[48];
quant7 = *(__vector short *)&dct_table[56];
col1 = vec_mladd(col1, quant1, zero16);
col2 = vec_mladd(col2, quant2, zero16);
col3 = vec_mladd(col3, quant3, zero16);
col4 = vec_mladd(col4, quant4, zero16);
col5 = vec_mladd(col5, quant5, zero16);
col6 = vec_mladd(col6, quant6, zero16);
col7 = vec_mladd(col7, quant7, zero16);
DO_IDCT_ISLOW(col, 1);
TRANSPOSE(out, row);
}
DO_IDCT_ISLOW(row, 2);
TRANSPOSE(out, col);
outb = vec_packs(col0, col1);
outb = vec_add(outb, PB_CENTERJSAMP);
outptr = (long long *)(output_buf[0] + output_col);
*outptr = outbptr[0];
outptr = (long long *)(output_buf[1] + output_col);
*outptr = outbptr[1];
outb = vec_packs(col2, col3);
outb = vec_add(outb, PB_CENTERJSAMP);
outptr = (long long *)(output_buf[2] + output_col);
*outptr = outbptr[0];
outptr = (long long *)(output_buf[3] + output_col);
*outptr = outbptr[1];
outb = vec_packs(col4, col5);
outb = vec_add(outb, PB_CENTERJSAMP);
outptr = (long long *)(output_buf[4] + output_col);
*outptr = outbptr[0];
outptr = (long long *)(output_buf[5] + output_col);
*outptr = outbptr[1];
outb = vec_packs(col6, col7);
outb = vec_add(outb, PB_CENTERJSAMP);
outptr = (long long *)(output_buf[6] + output_col);
*outptr = outbptr[0];
outptr = (long long *)(output_buf[7] + output_col);
*outptr = outbptr[1];
}
/* FAST INTEGER INVERSE DCT
*
* This is similar to the SSE2 implementation, except that we left-shift the
@@ -455,7 +726,6 @@ jsimd_fdct_ifast_altivec (DCTELEM *data)
#define IFAST_F_1_613 (IFAST_F_2_613 - 256) /* FIX(2.613125930) - FIX(1) */
#define IFAST_PASS1_BITS 2
#define IFAST_CENTERJSAMPLE 128
#define DO_IDCT_IFAST(in) \
{ \
@@ -536,7 +806,7 @@ jsimd_idct_ifast_altivec (void * dct_table_, JCOEFPTR coef_block,
__vector unsigned short
PRE_MULTIPLY_SCALE_BITS = { __8X(IFAST_PRE_MULTIPLY_SCALE_BITS) },
PASS1_BITS3 = { __8X(IFAST_PASS1_BITS + 3) };
__vector signed char PB_CENTERJSAMP = { __16X(IFAST_CENTERJSAMPLE) };
__vector signed char PB_CENTERJSAMP = { __16X(CENTERJSAMPLE) };
/* Pass 1: process columns. */