portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-11-04 05:54:32 +00:00
Code Releases Activity

Formatting emacs compliant

Browse Source
This commit is contained in:
paboyle
2018-01-12 23:25:02 +00:00
parent b815f5f764
commit bd15c38ae8
6 changed files with 730 additions and 730 deletions
Download Patch File Download Diff File Expand all files Collapse all files

66
lib/simd/Intel512avx.h
View File

@@ -1,4 +1,4 @@
/*************************************************************************************
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
*************************************************************************************/
/* END LEGAL */
#ifndef GRID_ASM_AV512_H
#define GRID_ASM_AV512_H
@@ -44,46 +44,46 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define ZENDf(Criir,Ciirr, tmp) ZEND1f(Criir,Ciirr, tmp) ZEND2f(Criir,Ciirr, tmp)
#define ZENDd(Criir,Ciirr, tmp) ZEND1d(Criir,Ciirr, tmp) ZEND2d(Criir,Ciirr, tmp)
#define ZMULMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
VSHUFMEMf(O,P,tmp) \
VMULMEMf(O,P,B,Biirr) \
VMULMEMf(O,P,C,Ciirr) \
VMULf(tmp,B,Briir) \
#define ZMULMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMf(O,P,tmp) \
VMULMEMf(O,P,B,Biirr) \
VMULMEMf(O,P,C,Ciirr) \
VMULf(tmp,B,Briir) \
VMULf(tmp,C,Criir)
#define ZMULMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
VSHUFMEMd(O,P,tmp) \
VMULMEMd(O,P,B,Biirr) \
VMULMEMd(O,P,C,Ciirr) \
VMULd(tmp,B,Briir) \
#define ZMULMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMd(O,P,tmp) \
VMULMEMd(O,P,B,Biirr) \
VMULMEMd(O,P,C,Ciirr) \
VMULd(tmp,B,Briir) \
VMULd(tmp,C,Criir)
#define ZMADDMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
VSHUFMEMf(O,P,tmp) \
VMADDMEMf(O,P,B,Biirr) \
VMADDMEMf(O,P,C,Ciirr) \
VMADDf(tmp,B,Briir) \
#define ZMADDMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMf(O,P,tmp) \
VMADDMEMf(O,P,B,Biirr) \
VMADDMEMf(O,P,C,Ciirr) \
VMADDf(tmp,B,Briir) \
VMADDf(tmp,C,Criir)
#define ZMADDMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMd(O,P,tmp) \
VMADDMEMd(O,P,B,Biirr) \
VMADDMEMd(O,P,C,Ciirr) \
VMADDd(tmp,B,Briir) \
VSHUFMEMd(O,P,tmp) \
VMADDMEMd(O,P,B,Biirr) \
VMADDMEMd(O,P,C,Ciirr) \
VMADDd(tmp,B,Briir) \
VMADDd(tmp,C,Criir)
// Merges accumulation for complex dot chain; less efficient under avx512
#define ZEND1f(Criir,Ciirr, tmp) "vshufps $0xb1," #Criir "," #Criir "," #tmp ";\n"\
"vaddps " #tmp "," #Criir "," #Criir"{%k6}" ";\n"
#define ZEND1f(Criir,Ciirr, tmp) "vshufps $0xb1," #Criir "," #Criir "," #tmp ";\n" \
"vaddps " #tmp "," #Criir "," #Criir"{%k6}" ";\n"
#define ZEND2f(Criir,Ciirr, tmp) "vshufps $0xb1," #Ciirr "," #Ciirr "," #tmp ";\n"\
"vsubps " #tmp "," #Ciirr "," #Criir"{%k7}" ";\n"
#define ZEND2f(Criir,Ciirr, tmp) "vshufps $0xb1," #Ciirr "," #Ciirr "," #tmp ";\n" \
"vsubps " #tmp "," #Ciirr "," #Criir"{%k7}" ";\n"
#define ZEND1d(Criir,Ciirr, tmp) "vshufpd $0x55," #Criir "," #Criir "," #tmp ";\n"\
"vaddps " #tmp "," #Criir "," #Criir"{%k6}" ";\n"
"vaddps " #tmp "," #Criir "," #Criir"{%k6}" ";\n"
#define ZEND2d(Criir,Ciirr, tmp) "vshufpd $0x55," #Ciirr "," #Ciirr "," #tmp ";\n"\
"vsubpd " #tmp "," #Ciirr "," #Criir"{%k7};\n" // ri+ir ; ri+ir,rr-ii
#define ZEND2d(Criir,Ciirr, tmp) "vshufpd $0x55," #Ciirr "," #Ciirr "," #tmp ";\n" \
"vsubpd " #tmp "," #Ciirr "," #Criir"{%k7};\n" // ri+ir ; ri+ir,rr-ii
#define VMOVRDUPd(OFF,A,DEST) "vpshufd $0x44," #OFF "*64(" #A ")," #DEST ";\n" // 32 bit level: 1,0,3,2
#define VMOVIDUPd(OFF,A,DEST) "vpshufd $0xee," #OFF "*64(" #A ")," #DEST ";\n" // 32 bit level: 3,2,3,2
@@ -123,10 +123,10 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define VMADDSUBIDUPd(O,P,B,accum) "vfmaddsub231pd (" #O"*16+8)("#P "){1to8},"#B "," #accum ";\n"
#define VMULRDUPd(O,P,B,accum) "vmulpd (" #O"*16+0)("#P "){1to8},"#B "," #accum ";\n"
#define VMULIDUPd(O,P,B,accum) "vmulpd (" #O"*16+8)("#P "){1to8},"#B "," #accum ";\n"
/*
* TimesI is used only in the XP recon
* Could zero the regs and use RECON_ACCUM
*/
/*
* TimesI is used only in the XP recon
* Could zero the regs and use RECON_ACCUM
*/
#define VTIMESI0f(A,DEST, Z) VSHUFf(A,DEST)
#define VTIMESI1f(A,DEST, Z) "vaddps " #DEST "," #Z "," #DEST"{%k6}" ";\n"

66
lib/simd/Intel512common.h
View File

@@ -1,4 +1,4 @@
/*************************************************************************************
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
*************************************************************************************/
/* END LEGAL */
#ifndef GRID_ASM_INTEL_COMMON_512_H
#define GRID_ASM_INTEL_COMMON_512_H
@@ -36,10 +36,10 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
////////////////////////////////////////////////////////////////////////////////////////////////////
// Opcodes common
////////////////////////////////////////////////////////////////////////////////////////////////////
#define MASK_REGS \
__asm__ ("mov $0xAAAA, %%eax \n"\
"kmovw %%eax, %%k6 \n"\
"mov $0x5555, %%eax \n"\
#define MASK_REGS \
__asm__ ("mov $0xAAAA, %%eax \n" \
"kmovw %%eax, %%k6 \n" \
"mov $0x5555, %%eax \n" \
"kmovw %%eax, %%k7 \n" : : : "%eax");
//#define label(B) __asm__ ( __func__ _LINE__ #B ":\n" );
@@ -47,44 +47,44 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define VZEROf(A) "vpxorq " #A "," #A "," #A ";\n"
#define VZEROd(A) "vpxorq " #A "," #A "," #A ";\n"
#define VTIMESIf(A,DEST, Z) \
VTIMESI0f(A,DEST, Z) \
VTIMESI1f(A,DEST, Z) \
#define VTIMESIf(A,DEST, Z) \
VTIMESI0f(A,DEST, Z) \
VTIMESI1f(A,DEST, Z) \
VTIMESI2f(A,DEST, Z)
#define VTIMESId(A,DEST, Z) \
VTIMESI0d(A,DEST, Z) \
VTIMESI1d(A,DEST, Z) \
#define VTIMESId(A,DEST, Z) \
VTIMESI0d(A,DEST, Z) \
VTIMESI1d(A,DEST, Z) \
VTIMESI2d(A,DEST, Z)
#define VTIMESMINUSIf(A,DEST, Z) \
VTIMESMINUSI0f(A,DEST, Z) \
VTIMESMINUSI1f(A,DEST, Z) \
VTIMESMINUSI2f(A,DEST, Z)
#define VTIMESMINUSIf(A,DEST, Z) \
VTIMESMINUSI0f(A,DEST, Z) \
VTIMESMINUSI1f(A,DEST, Z) \
VTIMESMINUSI2f(A,DEST, Z)
#define VTIMESMINUSId(A,DEST, Z) \
VTIMESMINUSI0d(A,DEST, Z) \
VTIMESMINUSI1d(A,DEST, Z) \
VTIMESMINUSI2d(A,DEST, Z)
#define VTIMESMINUSId(A,DEST, Z) \
VTIMESMINUSI0d(A,DEST, Z) \
VTIMESMINUSI1d(A,DEST, Z) \
VTIMESMINUSI2d(A,DEST, Z)
#define VACCTIMESIf(A,ACC,tmp) \
VACCTIMESI0f(A,ACC,tmp) \
VACCTIMESI1f(A,ACC,tmp) \
VACCTIMESI2f(A,ACC,tmp)
VACCTIMESI0f(A,ACC,tmp) \
VACCTIMESI1f(A,ACC,tmp) \
VACCTIMESI2f(A,ACC,tmp)
#define VACCTIMESId(A,ACC,tmp) \
VACCTIMESI0d(A,ACC,tmp) \
VACCTIMESI1d(A,ACC,tmp) \
VACCTIMESI2d(A,ACC,tmp)
VACCTIMESI0d(A,ACC,tmp) \
VACCTIMESI1d(A,ACC,tmp) \
VACCTIMESI2d(A,ACC,tmp)
#define VACCTIMESMINUSIf(A,ACC,tmp) \
VACCTIMESMINUSI0f(A,ACC,tmp) \
VACCTIMESMINUSI1f(A,ACC,tmp) \
#define VACCTIMESMINUSIf(A,ACC,tmp) \
VACCTIMESMINUSI0f(A,ACC,tmp) \
VACCTIMESMINUSI1f(A,ACC,tmp) \
VACCTIMESMINUSI2f(A,ACC,tmp)
#define VACCTIMESMINUSId(A,ACC,tmp) \
VACCTIMESMINUSI0d(A,ACC,tmp) \
VACCTIMESMINUSI1d(A,ACC,tmp) \
#define VACCTIMESMINUSId(A,ACC,tmp) \
VACCTIMESMINUSI0d(A,ACC,tmp) \
VACCTIMESMINUSI1d(A,ACC,tmp) \
VACCTIMESMINUSI2d(A,ACC,tmp)
#define LOAD64a(A,ptr) "movq %0, %" #A : : "r"(ptr) : #A

6
lib/simd/Intel512double.h
View File

@@ -1,4 +1,4 @@
/*************************************************************************************
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
*************************************************************************************/
/* END LEGAL */
// No guard can be multiply included as undef clearage
#undef VZERO
#undef VMOV

46
lib/simd/Intel512imci.h
View File

@@ -1,4 +1,4 @@
/*************************************************************************************
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
*************************************************************************************/
/* END LEGAL */
#ifndef GRID_ASM_AV512_H
#define GRID_ASM_AV512_H
@@ -44,32 +44,32 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define ZENDf(Criir,Ciirr, tmp) ZEND1f(Criir,Ciirr, tmp) ZEND2f(Criir,Ciirr, tmp)
#define ZENDd(Criir,Ciirr, tmp) ZEND1d(Criir,Ciirr, tmp) ZEND2d(Criir,Ciirr, tmp)
#define ZMULMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
VSHUFMEMf(O,P,tmp) \
VMULMEMf(O,P,B,Biirr) \
VMULMEMf(O,P,C,Ciirr) \
VMULf(tmp,B,Briir) \
#define ZMULMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMf(O,P,tmp) \
VMULMEMf(O,P,B,Biirr) \
VMULMEMf(O,P,C,Ciirr) \
VMULf(tmp,B,Briir) \
VMULf(tmp,C,Criir)
#define ZMULMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
VSHUFMEMd(O,P,tmp) \
#define ZMULMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMd(O,P,tmp) \
VMULMEMd(O,P,B,Biirr) \
VMULMEMd(O,P,C,Ciirr) \
VMULd(tmp,B,Briir) \
VMULd(tmp,C,Criir)
VMULMEMd(O,P,C,Ciirr) \
VMULd(tmp,B,Briir) \
VMULd(tmp,C,Criir)
#define ZMADDMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr)\
VSHUFMEMf(O,P,tmp) \
VMADDMEMf(O,P,B,Biirr) \
VMADDMEMf(O,P,C,Ciirr) \
VMADDf(tmp,B,Briir) \
#define ZMADDMEM2SPf(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMf(O,P,tmp) \
VMADDMEMf(O,P,B,Biirr) \
VMADDMEMf(O,P,C,Ciirr) \
VMADDf(tmp,B,Briir) \
VMADDf(tmp,C,Criir)
#define ZMADDMEM2SPd(O,P,tmp,B,C,Briir,Biirr,Criir,Ciirr) \
VSHUFMEMd(O,P,tmp) \
VMADDMEMd(O,P,B,Biirr) \
VMADDMEMd(O,P,C,Ciirr) \
VMADDd(tmp,B,Briir) \
VSHUFMEMd(O,P,tmp) \
VMADDMEMd(O,P,B,Biirr) \
VMADDMEMd(O,P,C,Ciirr) \
VMADDd(tmp,B,Briir) \
VMADDd(tmp,C,Criir)
#define ZEND1d(Criir,Ciirr, tmp) "vaddpd " #Criir "{cdab} ," #Criir "," #Criir"{%k6}" ";\n"
@@ -106,7 +106,7 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
#define VACCTIMESMINUSI1f(A,ACC,tmp) "vsubps " #A "{cdab}," #ACC "," #ACC"{%k7}" ";\n"
#define VACCTIMESMINUSI2f(A,ACC,tmp) "vaddps " #A "{cdab}," #ACC "," #ACC"{%k6}" ";\n"
// Acc = Acc - i A
// Acc = Acc - i A
#define VACCTIMESMINUSI0d(A,ACC,tmp)
#define VACCTIMESMINUSI1d(A,ACC,tmp) "vsubpd " #A "{cdab}," #ACC "," #ACC"{%k7}" ";\n"
#define VACCTIMESMINUSI2d(A,ACC,tmp) "vaddpd " #A "{cdab}," #ACC "," #ACC"{%k6}" ";\n"

6
lib/simd/Intel512single.h
View File

@@ -1,4 +1,4 @@
/*************************************************************************************
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
@@ -23,8 +23,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
*************************************************************************************/
/* END LEGAL */
// No guard can be multiply included as undef clearge of macros
#undef VZERO
#undef VMOV

1270
lib/simd/Intel512wilson.h
View File

File diff suppressed because it is too large Load Diff