daiduo/TurboPFor-Integer-Compression
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

TurboPFor: Benchmark app

Browse Source
This commit is contained in:
x
2023-03-31 13:34:54 +02:00
parent 6d534f8bff
commit 42c347c3f3
1 changed files with 78 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

120
lib/icapp.c
View File

@ -717,23 +717,32 @@ void meshdec(const uint8_t *in, unsigned inlen, float *out, unsigned nx, unsigne
#endif
#ifdef _BLOSC
unsigned blosccomp(unsigned char *in, unsigned inlen, unsigned char *out, unsigned outsize, unsigned compcode, int codlev, unsigned esize, int filter0, int filter1, int filter2) {
unsigned blosccomp(unsigned char *in, unsigned inlen, unsigned char *out, unsigned outsize, unsigned codid, int codlev, unsigned esize, int filter0, int filter1, int filter2) {
unsigned clevel = codid==ICC_ZSTD?((codlev+1)/2):codlev;
unsigned compcode = codid==ICC_LZ4?(clevel>9?BLOSC_LZ4HC:BLOSC_LZ4):BLOSC_ZSTD;
blosc2_schunk schunk;
schunk.typesize = esize;
blosc2_cparams cp = BLOSC2_CPARAMS_DEFAULTS;
cp.typesize = esize;
cp.compcode = compcode; //BLOSC_LZ4HC, BLOSC_LZ4, BLOSC_ZSTD, BLOSC_LZ4, BLOSC_BLOSCLZ
cp.clevel = codlev<1?1:(codlev<9?codlev:9); //blocksize=[1,32768[2,65536][3,131072][4,262144][5,262144][6,524288][7,524288][8,524288][9,1048576]
cp.clevel = clevel<1?1:(clevel<9?clevel:9); //blocksize=[1,32768[2,65536][3,131072][4,262144][5,262144][6,524288][7,524288][8,524288][9,1048576]
cp.nthreads = 1;
cp.schunk = &schunk;
cp.filters[BLOSC2_MAX_FILTERS - 1] = filter0; //BLOSC_NOFILTER, BLOSC_SHUFFLE, BLOSC_BITSHUFFLE
cp.filters[BLOSC2_MAX_FILTERS - 2] = filter1; //BLOSC_DELTA, BLOSC_FILTER_BYTEDELTA
cp.filters[BLOSC2_MAX_FILTERS - 3] = filter2; //BLOSC_TRUNC_PREC
blosc2_context *ctx = blosc2_create_cctx(cp);
int rc = blosc2_compress_ctx(ctx, in, (int)inlen, out, (int)outsize);
blosc2_free_ctx(ctx);
return rc;
}
unsigned bloscdecomp(unsigned char *in, unsigned inlen, unsigned char *out, unsigned outlen) {
unsigned bloscdecomp(unsigned char *in, unsigned inlen, unsigned char *out, unsigned outlen, unsigned esize) {
blosc2_schunk schunk;
schunk.typesize = esize;
blosc2_dparams dp = BLOSC2_DPARAMS_DEFAULTS;
dp.schunk = &schunk;
blosc2_context *ctx = blosc2_create_dctx(dp);
int rc = blosc2_decompress_ctx(ctx, in, inlen, out, outlen);
blosc2_free_ctx(ctx);
@ -1020,27 +1029,27 @@ unsigned char *bestr(unsigned id, unsigned b, unsigned char *s, char *prms, int
"%3d:128 ",
"%3d:129 ",
"%3d:streamvbyte StreamVByte SIMD ", //130
"%3d:streamvbyte StreamVByte SIMD ", //130
"%3d:streamvbyte delt StreamVByte delta ",
"%3d:streamvbyte zzag StreamVByte zigzag ",
"%3d:maskeydvbyte MasedVByte SIMD ",
"%3d:FastPFor FastPFor ",
"%3d:SimdFastPFor FastPFor SIMD ",
"%3d:SimdOptPFor FastPFor SIMD ",
"%3d:137 ",
"%3d:SPDP SPDP Floating Point ",
"%3d:138 ",
"%3d:139 ",
"%3d:SPDP SPDP Floating Point ", //140
"%3d:zfp zfp ",
"%3d:zfp zfp ", //140
"%3d:zfp zfp 2D ",
"%3d:zfp zfp 3D ",
"%3d:meshoptimizer 3D lz%s,%d ",
"%3d:meshoptimizer 3D lz%s,%d ",
"%3d:meshoptimizer 3D lz%s,%d ",
"%3d:blosc shuffle+%s,%d ",
"%3d:blosc shuffle delta+%s,%d ",
"%3d:fprazor Turbo razor ",
"%3d:blosc bitshuffle+%s,%d ",
"%3d:blosc bitshuffle+delta+%s,%d ",
"%3d:blosc bitshuffle+bytedelta+%s,%d ",
"%3d:blosc shuffle+%s,%d ",
"%3d:blosc shuffle+delta+%s,%d ",
"%3d:blosc shuffle+bytedelta+%s,%d ",
"%3d:fprazor Turbo Razor ",
"%3d:gb granular bitgroom ", //150
"%3d:bitgrooming bit grooming ",
@ -1053,7 +1062,7 @@ unsigned char *bestr(unsigned id, unsigned b, unsigned char *s, char *prms, int
"%3d:tp4z0enc Nibble transpose+zigzag ",
"%3d:tp4xenc Nibble tranp+xor integrated",
"%3d:tp4x0enc Nibble transpose+xor ",
"%3d:tp4x0enc Nibble transpose+xor ", //160
"%3d:161 speed test ",
"%3d:162 speed test ",
"%3d:163 speed test ",
@ -1063,6 +1072,18 @@ unsigned char *bestr(unsigned id, unsigned b, unsigned char *s, char *prms, int
"%3d:167 speed test ",
"%3d:168 speed test ",
"%3d:169 speed test ",
"%3d:meshoptimizer 3D lz%s,%d ", //170
"%3d:meshoptimizer 3D lz%s,%d ",
"%3d:meshoptimizer 3D lz%s,%d ",
"%3d:173 speed test ",
"%3d:174 speed test ",
"%3d:175 speed test ",
"%3d:176 speed test ",
"%3d:177 speed test ",
"%3d:178 speed test ",
"%3d:179 speed test ",
};
if(id < 80)
sprintf((char *)s,fmt[id], id, b, prms, prmi); // print bitsize
@ -1294,7 +1315,7 @@ unsigned bench8(unsigned char *in, unsigned n, unsigned char *out, unsigned char
TMBENCH( "", libmemcpy(out,in,n) ,n); l=n; pr(l,n); pr(n,n); TMBENCH2("120", libmemcpy( cpy,out,n) ,n); break;
#ifdef _SPDP
case 140: TMBENCH("",l=spdpenc(in,m*(USIZE),out,SPDPSIZE,codlev),n);pr(l,n); TMBENCH2("109",spdpdec( out, m*(USIZE), cpy,SPDPSIZE,codlev); ,n); break;
case 137: TMBENCH("",l=spdpenc(in,m*(USIZE),out,SPDPSIZE,codlev),n);pr(l,n); TMBENCH2("137",spdpdec( out, m*(USIZE), cpy,SPDPSIZE,codlev); ,n); break;
#endif
default: goto end;
}
@ -1724,37 +1745,33 @@ unsigned bench32(unsigned char *in, unsigned n, unsigned char *out, unsigned cha
case 135: TMBENCH("",l=FastPFore128v32(in, m, out,ns),n); pr(l,n); TMBENCH2("135", FastPFord128v32(out, m, cpy),n); break;
case 136: TMBENCH("",l=OptPFore128v32( in, m, out,ns),n); pr(l,n); TMBENCH2("136", OptPFord128v32( out, m, cpy),n); break;
#endif
#ifdef _SPDP
case 140: TMBENCH("",l=spdpenc(in,m*(USIZE),out,SPDPSIZE,codlev),n);pr(l,n); TMBENCH2("140",spdpdec( out, m*(USIZE), cpy,SPDPSIZE,codlev); ,n); break;
case 137: TMBENCH("",l=spdpenc(in,m*(USIZE),out,SPDPSIZE,codlev),n);pr(l,n); TMBENCH2("140",spdpdec( out, m*(USIZE), cpy,SPDPSIZE,codlev); ,n); break;
#endif
#ifdef _ZFP
case 141: {
TMBENCH("",l = zfpcompress(in,m,0,0,0, out, ns, zfp_type_float, zerrlim),n); pr(l,n); TMBENCH2("141",zfpdecompress(out, l, cpy,m,0,0,0, zfp_type_float, zerrlim),n);
case 140: {
TMBENCH("",l = zfpcompress(in,m,0,0,0, out, ns, zfp_type_float, zerrlim),n); pr(l,n); TMBENCH2("140",zfpdecompress(out, l, cpy,m,0,0,0, zfp_type_float, zerrlim),n);
if(zerrlim > DBL_EPSILON) { if(verbose) fpstat(in, m, cpy, -4); memcpy(cpy,in,n); } //lossy compression irreversible
} break;
case 142: if(ny<=0) goto end; { unsigned _ny = ny*(nz?nz:1)*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,_ny,0,0, out, ns, zfp_type_float, zerrlim),n); pr(l,n); TMBENCH2("142",zfpdecompress(out, l, cpy,nx,_ny,0,0, zfp_type_float, zerrlim),n);
case 141: if(ny<=0) goto end; { unsigned _ny = ny*(nz?nz:1)*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,_ny,0,0, out, ns, zfp_type_float, zerrlim),n); pr(l,n); TMBENCH2("141",zfpdecompress(out, l, cpy,nx,_ny,0,0, zfp_type_float, zerrlim),n);
if(zerrlim > DBL_EPSILON) { if(verbose) fpstat(in, m, cpy, -4); memcpy(cpy,in,n); } //lossy compression irreversible
} break;
case 143: if(nz<=0) goto end; { unsigned _nz = nz*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,ny,_nz,0, out, ns, zfp_type_float, zerrlim),n); pr(l,n); TMBENCH2("143",zfpdecompress(out, l, cpy,nx,ny,_nz,0, zfp_type_float, zerrlim),n);
case 142: if(nz<=0) goto end; { unsigned _nz = nz*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,ny,_nz,0, out, ns, zfp_type_float, zerrlim),n); pr(l,n); TMBENCH2("142",zfpdecompress(out, l, cpy,nx,ny,_nz,0, zfp_type_float, zerrlim),n);
if(zerrlim > DBL_EPSILON) { if(verbose) fpstat(in, m, cpy, -4); memcpy(cpy,in,n); } //lossy compression irreversible
} break;
#endif
#ifdef _MESHOPT
case 144: TMBENCH("", l = meshenc(in, m,0,0, out, ns, tmp,codid,codlev,codprm),n); pr(l,n); TMBENCH2("144", meshdec(out, l,cpy, m,0,0, tmp,codid,codlev,codprm),n); break;
case 145: if(ny<=0) goto end; { unsigned _ny = ny*(nz?nz:1)*(nw?nw:1);
TMBENCH("", l = meshenc(in, nx,_ny,0, out, ns, tmp,codid,codlev,codprm),n); pr(l,n); TMBENCH2("145", meshdec(out, l,cpy, nx,_ny,0, tmp,codid,codlev,codprm),n);
} break;
case 146: if(nz<=0) goto end; { unsigned _nz = nz*(nw?nw:1);
TMBENCH("", l = meshenc(in, nx,ny,_nz, out, ns, tmp,codid,codlev,codprm),n); pr(l,n); TMBENCH2("146", meshdec(out, l,cpy, nx,ny,_nz, tmp,codid,codlev,codprm),n);
} break;
#endif
#ifdef _BLOSC
case 147: { unsigned clev = codid==ICC_ZSTD?((codlev+1)/2):codlev; TMBENCH("",l = blosccomp(in, n, out, ns, codid==ICC_LZ4?(clev>9?BLOSC_LZ4HC:BLOSC_LZ4):BLOSC_ZSTD, clev, 4, BLOSC_SHUFFLE, 0, 0),n); pr(l,n); TMBENCH2("147", bloscdecomp(out, l, cpy, n),n); } break;
case 148: { unsigned clev = codid==ICC_ZSTD?((codlev+1)/2):codlev; TMBENCH("",l = blosccomp(in, n, out, ns, codid==ICC_LZ4?(clev>9?BLOSC_LZ4HC:BLOSC_LZ4):BLOSC_ZSTD, clev, 4, BLOSC_SHUFFLE, BLOSC_DELTA, 0),n); pr(l,n); TMBENCH2("148", bloscdecomp(out, l, cpy, n),n); } break;
//case 149: if(zerrlim>DBL_EPSILON) truncate_precision(6, 4, n, in, out);
// codlev = codid==ICC_ZSTD?((codlev+1)/2):codlev; TMBENCH("",l = blosccomp(in, n, out, ns, codid==ICC_LZ4?(codlev>9?BLOSC_LZ4HC:BLOSC_LZ4):BLOSC_ZSTD, codlev, 4, BLOSC_TRUNC_PREC, BLOSC_SHUFFLE, BLOSC_DELTA),n); pr(l,n); TMBENCH2("148", bloscdecomp(out, l, cpy, n),n); break;
case 143: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_BITSHUFFLE, 0, 0),n); pr(l,n); TMBENCH2("143", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 144: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_BITSHUFFLE, BLOSC_DELTA, 0),n); pr(l,n); TMBENCH2("144", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 145: { blosc2_init(); TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_BITSHUFFLE, BLOSC_FILTER_BYTEDELTA, 0),n); pr(l,n); TMBENCH2("145", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 146: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_SHUFFLE, 0, 0),n); pr(l,n); TMBENCH2("146", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 147: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_SHUFFLE, BLOSC_DELTA, 0),n); pr(l,n); TMBENCH2("147", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 148: { blosc2_init(); TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_SHUFFLE, BLOSC_FILTER_BYTEDELTA, 0),n); pr(l,n); TMBENCH2("148", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
#endif
// ----- speed test & lossy error bound analysis (with option -v1) -----------------------
case 149: TMBENCH("", fprazor32( in, m, out,zerrlim),n); l=n; pr(l,n); memcpy(cpy,in,n); if(verbose) fpstat(in, m, out, -4); break;
@ -1774,6 +1791,15 @@ unsigned bench32(unsigned char *in, unsigned n, unsigned char *out, unsigned cha
case 158: l = n; TMBENCH("", tp4z0enc(in, n, out, USIZE, tmp),n); pr(l,n); TMBENCH2("158", tp4z0dec(out, n,cpy, USIZE),n); break;
case 159: l = n; TMBENCH("", tp4xenc( in, n, out, USIZE),n); pr(l,n); TMBENCH2("159", tp4xdec( out, n,cpy, USIZE),n); break;
case 160: l = n; TMBENCH("", tp4x0enc(in, n, out, USIZE, tmp),n); pr(l,n); TMBENCH2("160", tp4x0dec(out, n,cpy, USIZE),n); break;
#ifdef _MESHOPT
case 170: TMBENCH("", l = meshenc(in, m,0,0, out, ns, tmp,codid,codlev,codprm),n); pr(l,n); TMBENCH2("170", meshdec(out, l,cpy, m,0,0, tmp,codid,codlev,codprm),n); break;
case 171: if(ny<=0) goto end; { unsigned _ny = ny*(nz?nz:1)*(nw?nw:1);
TMBENCH("", l = meshenc(in, nx,_ny,0, out, ns, tmp,codid,codlev,codprm),n); pr(l,n); TMBENCH2("171", meshdec(out, l,cpy, nx,_ny,0, tmp,codid,codlev,codprm),n);
} break;
case 172: if(nz<=0) goto end; { unsigned _nz = nz*(nw?nw:1);
TMBENCH("", l = meshenc(in, nx,ny,_nz, out, ns, tmp,codid,codlev,codprm),n); pr(l,n); TMBENCH2("172", meshdec(out, l,cpy, nx,ny,_nz, tmp,codid,codlev,codprm),n);
} break;
#endif
default: goto end;
}
if(l) {
@ -1912,22 +1938,32 @@ unsigned bench64(unsigned char *in, unsigned n, unsigned char *out, unsigned cha
TMBENCH( "", libmemcpy(out,in,n) ,n); l=n; pr(l,n); TMBENCH2("110", libmemcpy( cpy,out,n) ,n); break;
#ifdef _SPDP
case 140: TMBENCH("",l=spdpenc(in,m*(USIZE),out,SPDPSIZE,codlev),n);pr(l,n); TMBENCH2("109",spdpdec( out, m*(USIZE), cpy,SPDPSIZE,codlev); ,n); break;
case 137: TMBENCH("",l=spdpenc(in,m*(USIZE),out,SPDPSIZE,codlev),n);pr(l,n); TMBENCH2("109",spdpdec( out, m*(USIZE), cpy,SPDPSIZE,codlev); ,n); break;
#endif
#ifdef _ZFP
case 141: {
TMBENCH("",l = zfpcompress(in,m,0,0,0, out, ns, zfp_type_double, zerrlim),n); pr(l,n); TMBENCH2("141",zfpdecompress(out, l, cpy,m,0,0,0, zfp_type_double, zerrlim),n);
case 140: {
TMBENCH("",l = zfpcompress(in,m,0,0,0, out, ns, zfp_type_double, zerrlim),n); pr(l,n); TMBENCH2("140",zfpdecompress(out, l, cpy,m,0,0,0, zfp_type_double, zerrlim),n);
if(zerrlim > DBL_EPSILON) { if(verbose) fpstat(in, m, cpy, -8); memcpy(cpy,in,n); } //lossy compression irreversible
} break;
case 142: if(ny<=0) goto end; { unsigned _ny = ny*(nz?nz:1)*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,_ny,0,0, out, ns, zfp_type_double, zerrlim),n); pr(l,n); TMBENCH2("142",zfpdecompress(out, l, cpy,nx,_ny,0,0, zfp_type_double, zerrlim),n);
case 141: if(ny<=0) goto end; { unsigned _ny = ny*(nz?nz:1)*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,_ny,0,0, out, ns, zfp_type_double, zerrlim),n); pr(l,n); TMBENCH2("141",zfpdecompress(out, l, cpy,nx,_ny,0,0, zfp_type_double, zerrlim),n);
if(zerrlim > DBL_EPSILON) { if(verbose) fpstat(in, m, cpy, -8); memcpy(cpy,in,n); } //lossy compression irreversible
} break;
case 143: if(nz<=0) goto end; { unsigned _nz = nz*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,ny,_nz,0, out, ns, zfp_type_double, zerrlim),n); pr(l,n); TMBENCH2("143",zfpdecompress(out, l, cpy,nx,ny,_nz,0, zfp_type_double, zerrlim),n);
case 142: if(nz<=0) goto end; { unsigned _nz = nz*(nw?nw:1);
TMBENCH("",l = zfpcompress(in,nx,ny,_nz,0, out, ns, zfp_type_double, zerrlim),n); pr(l,n); TMBENCH2("142",zfpdecompress(out, l, cpy,nx,ny,_nz,0, zfp_type_double, zerrlim),n);
if(zerrlim > DBL_EPSILON) { if(verbose) fpstat(in, m, cpy, -8); memcpy(cpy,in,n); } //lossy compression irreversible
} break;
#endif
#ifdef _BLOSC
case 143: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_BITSHUFFLE, 0, 0),n); pr(l,n); TMBENCH2("143", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 144: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_BITSHUFFLE, BLOSC_DELTA, 0),n); pr(l,n); TMBENCH2("144", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 145: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_BITSHUFFLE, BLOSC_FILTER_BYTEDELTA, 0),n); pr(l,n); TMBENCH2("145", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 146: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_SHUFFLE, 0, 0),n); pr(l,n); TMBENCH2("146", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 147: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_SHUFFLE, BLOSC_DELTA, 0),n); pr(l,n); TMBENCH2("147", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
case 148: { TMBENCH("",l = blosccomp(in, n, out, ns, codid, codlev, USIZE, BLOSC_SHUFFLE, BLOSC_FILTER_BYTEDELTA, 0),n); pr(l,n); TMBENCH2("148", bloscdecomp(out, l, cpy, n,USIZE),n); } break;
#endif
// ----- speed test & lossy error bound analysis (with option -v1) -----------------------
case 149: TMBENCH("", fprazor64( in, m, out,zerrlim),n); l=n; pr(l,n); memcpy(cpy,in,n); if(verbose) fpstat(in, m, out, -8); break;
#ifdef _BITGROOMING