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

IcBench App: External libs

Browse Source
This commit is contained in:
x
2019-07-15 17:49:41 +02:00
parent 1766abccdf
commit d8dc46a95d
4 changed files with 13 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
ext/beplug_.h
View File

@ -90,7 +90,7 @@ extern "C" {
} }
#endif #endif
#if C_SIMDCOMP #if C_SIMDCOMP128
#undef SIMDBITPACKING_H_ #undef SIMDBITPACKING_H_
#include "vabyte.h" // Standard Variable Byte #include "vabyte.h" // Standard Variable Byte
#include "simdcomp/include/simdcomp.h" #include "simdcomp/include/simdcomp.h"
@ -112,7 +112,9 @@ unsigned char *simdfor_selectx( unsigned char *__restrict in, unsigned n, unsign
unsigned i; unsigned i;
for(i=0; i < n;i++) out[i] = simdselectFOR(start, (const __m128i *)in, b, i); return in + simdpackFOR_compressedbytes(n, b); for(i=0; i < n;i++) out[i] = simdselectFOR(start, (const __m128i *)in, b, i); return in + simdpackFOR_compressedbytes(n, b);
} }
#ifdef __AVX2__ #endif
#if C_SIMDCOMP256
#undef SIMDBITPACKING_H_
unsigned char *avxpackwithoutmaskn(uint32_t *in, uint32_t n, uint32_t *out, uint32_t b) { unsigned char *avxpackwithoutmaskn(uint32_t *in, uint32_t n, uint32_t *out, uint32_t b) {
uint32_t *ip; uint32_t *ip;
for(ip = in; ip != in+(n&~(256-1)); ip += 256,out += 8 * b) for(ip = in; ip != in+(n&~(256-1)); ip += 256,out += 8 * b)
@ -125,7 +127,6 @@ unsigned char *avxunpackn(uint32_t *in, uint32_t n, uint32_t *out, uint32_t b) {
avxunpack((__m256i *)in, op, b); avxunpack((__m256i *)in, op, b);
return (unsigned char *)simdunpack_shortlength((__m128i *)in, n & (256-1), op, b); return (unsigned char *)simdunpack_shortlength((__m128i *)in, n & (256-1), op, b);
} }
#endif
#endif #endif
#if C_BITSHUFFLE #if C_BITSHUFFLE

9
ext/beplugc_.c
View File

@ -82,16 +82,15 @@
case P_QMX: { JASS::compress_integer_qmx_improved qmx; unsigned r=qmx.encode(out+4, outsize, (uint32_t *)in, n); ctou32(out)=r; return out+4+r; } case P_QMX: { JASS::compress_integer_qmx_improved qmx; unsigned r=qmx.encode(out+4, outsize, (uint32_t *)in, n); ctou32(out)=r; return out+4+r; }
#endif #endif
#if C_SIMDCOMP #if C_SIMDCOMP128
//case SC_PACK: if(b < 0) { b = maxbits_length(in,n); *out++ = b; } return fastpackwithoutmask32(in, n, (uint32_t *)out, b); //case SC_PACK: if(b < 0) { b = maxbits_length(in,n); *out++ = b; } return fastpackwithoutmask32(in, n, (uint32_t *)out, b);
case SC_FOR: case SC_FOR:
case SC_FORDA: case SC_FORDA:
case SC_SIMDPACK128:if(b < 0) b = maxbits(in), *out++ = b; return (unsigned char *)simdpack_length(in, n, (__m128i *)out, b); case SC_SIMDPACK128:if(b < 0) b = maxbits(in), *out++ = b; return (unsigned char *)simdpack_length(in, n, (__m128i *)out, b);
#ifdef __AVX2__
case SC_SIMDPACK256:if(b < 0) b = avxmaxbits(in), *out++ = b; return (unsigned char *)avxpackwithoutmaskn(in, n, (unsigned *)out, b);
#endif
#endif #endif
#if C_SIMDCOMP256
case SC_SIMDPACK256:if(b < 0) b = avxmaxbits(in), *out++ = b; return (unsigned char *)avxpackwithoutmaskn(in, n, (unsigned *)out, b);
#endif
#if C_STREAMVBYTE #if C_STREAMVBYTE
case P_STREAMVBYTE: return out + streamvbyte_encode(in, n, out); case P_STREAMVBYTE: return out + streamvbyte_encode(in, n, out);
#endif #endif

6
ext/beplugd_.c
View File

@ -81,14 +81,14 @@
case PC_OPTPFD : return optpfddec32( in, n, out); //if(n < 128) return vbytedec(in, n, out); else { unsigned all_array[2048]; return (unsigned char *)detailed_p4_decode(out, (unsigned *)in, all_array); } case PC_OPTPFD : return optpfddec32( in, n, out); //if(n < 128) return vbytedec(in, n, out); else { unsigned all_array[2048]; return (unsigned char *)detailed_p4_decode(out, (unsigned *)in, all_array); }
#endif #endif
#if C_SIMDCOMP #if C_SIMDCOMP128
//case SC_PACK: if(b < 0) b = *in++; return fastunpack32((uint32_t *)in, n, out, b); //case SC_PACK: if(b < 0) b = *in++; return fastunpack32((uint32_t *)in, n, out, b);
case SC_FOR: case SC_FOR:
case SC_SIMDPACK128: if(b < 0) b = *in++; return (unsigned char *)simdunpack_length( (__m128i *)in, n, out, b); case SC_SIMDPACK128: if(b < 0) b = *in++; return (unsigned char *)simdunpack_length( (__m128i *)in, n, out, b);
case SC_FORDA: if(b < 0) b = *in++; return simdfor_selectx(in, n, out, 0, b); case SC_FORDA: if(b < 0) b = *in++; return simdfor_selectx(in, n, out, 0, b);
#ifdef __AVX2__ #endif
#if C_SIMDCOMP256
case SC_SIMDPACK256: if(b < 0) b = *in++; return (unsigned char *)avxunpackn( (unsigned *)in, n, out, b); case SC_SIMDPACK256: if(b < 0) b = *in++; return (unsigned char *)avxunpackn( (unsigned *)in, n, out, b);
#endif
#endif #endif
#if C_STREAMVBYTE #if C_STREAMVBYTE

2
ext/beplugr_.h
View File

@ -22,7 +22,7 @@
{ FP_GROUPSIMPLE, "FP_GROUPSIMPLE", C_FASTPFOR128, 0, 0,"","Group Simple" }, { FP_GROUPSIMPLE, "FP_GROUPSIMPLE", C_FASTPFOR128, 0, 0,"","Group Simple" },
{ SC_SIMDPACK128, "SC_SIMDPack128", C_SIMDCOMP128, BLK_V128,0,"","Bit packing (SSE4.1)"}, { SC_SIMDPACK128, "SC_SIMDPack128", C_SIMDCOMP128, BLK_V128,0,"","Bit packing (SSE4.1)"},
{ SC_SIMDPACK256, "SC_SIMDPack256", C_SIMDCOMP128, BLK_V256,0,"","Bit packing (SSE4.1)"}, { SC_SIMDPACK256, "SC_SIMDPack256", C_SIMDCOMP256, BLK_V256,0,"","Bit packing (SSE4.1)"},
{ SC_FOR, "SC_For", C_SIMDCOMP128, BLK_V128,0,"","For (SSE4.1)"}, { SC_FOR, "SC_For", C_SIMDCOMP128, BLK_V128,0,"","For (SSE4.1)"},
{ SC_FORDA, "SC_ForDA", C_SIMDCOMP128, BLK_V128,0,"","For direct access (SSE4.1)"}, { SC_FORDA, "SC_ForDA", C_SIMDCOMP128, BLK_V128,0,"","For direct access (SSE4.1)"},