portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-11-01 04:24:32 +00:00
Code Releases Activity

change loop counts to local variables so clang compiler doesn't complain

Browse Source
This commit is contained in:
Meifeng Lin
2022-12-20 10:09:50 -08:00
parent f6661ce29b
commit 1df8669898
2 changed files with 22 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
Grid/qcd/action/fermion/CloverHelpers.h
View File

@@ -120,8 +120,8 @@ public:
int DimRep = Impl::Dimension; int DimRep = Impl::Dimension;
autoView(in_v, in, AcceleratorWrite); autoView(in_v, in, AcceleratorWrite);
int osites=in.Grid()->oSites();
accelerator_for(ss, in.Grid()->oSites(), 1, { accelerator_for(ss, osites, 1, {
for (int sa=0; sa<Ns; sa++) for (int sa=0; sa<Ns; sa++)
for (int ca=0; ca<DimRep; ca++) for (int ca=0; ca<DimRep; ca++)
in_v[ss]()(sa,sa)(ca,ca) = c; in_v[ss]()(sa,sa)(ca,ca) = c;
@@ -245,8 +245,8 @@ public:
int DimRep = Impl::Dimension; int DimRep = Impl::Dimension;
autoView(in_v, in, AcceleratorWrite); autoView(in_v, in, AcceleratorWrite);
int osites=in.Grid()->oSites();
accelerator_for(ss, in.Grid()->oSites(), 1, { accelerator_for(ss, osites, 1, {
for (int sa=0; sa<Ns; sa++) for (int sa=0; sa<Ns; sa++)
for (int ca=0; ca<DimRep; ca++) for (int ca=0; ca<DimRep; ca++)
in_v[ss]()(sa,sa)(ca,ca) = c; in_v[ss]()(sa,sa)(ca,ca) = c;

27
Grid/qcd/action/fermion/WilsonCloverHelpers.h
View File

@@ -85,7 +85,8 @@ public:
T = Zero(); T = Zero();
autoView(T_v,T,AcceleratorWrite); autoView(T_v,T,AcceleratorWrite);
autoView(F_v,F,AcceleratorRead); autoView(F_v,F,AcceleratorRead);
accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(), int osites=T.Grid()->oSites();
accelerator_for(i, osites,CloverField::vector_type::Nsimd(),
{ {
coalescedWrite(T_v[i]()(0, 1), coalescedRead(timesMinusI(F_v[i]()()))); coalescedWrite(T_v[i]()(0, 1), coalescedRead(timesMinusI(F_v[i]()())));
coalescedWrite(T_v[i]()(1, 0), coalescedRead(timesMinusI(F_v[i]()()))); coalescedWrite(T_v[i]()(1, 0), coalescedRead(timesMinusI(F_v[i]()())));
@@ -103,7 +104,8 @@ public:
autoView(T_v, T,AcceleratorWrite); autoView(T_v, T,AcceleratorWrite);
autoView(F_v, F,AcceleratorRead); autoView(F_v, F,AcceleratorRead);
accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(), int osites=T.Grid()->oSites();
accelerator_for(i, osites,CloverField::vector_type::Nsimd(),
{ {
coalescedWrite(T_v[i]()(0, 1), coalescedRead(-F_v[i]()())); coalescedWrite(T_v[i]()(0, 1), coalescedRead(-F_v[i]()()));
coalescedWrite(T_v[i]()(1, 0), coalescedRead(F_v[i]()())); coalescedWrite(T_v[i]()(1, 0), coalescedRead(F_v[i]()()));
@@ -121,7 +123,8 @@ public:
autoView(T_v,T,AcceleratorWrite); autoView(T_v,T,AcceleratorWrite);
autoView(F_v,F,AcceleratorRead); autoView(F_v,F,AcceleratorRead);
accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(), int osites=T.Grid()->oSites();
accelerator_for(i, osites,CloverField::vector_type::Nsimd(),
{ {
coalescedWrite(T_v[i]()(0, 0), coalescedRead(timesMinusI(F_v[i]()()))); coalescedWrite(T_v[i]()(0, 0), coalescedRead(timesMinusI(F_v[i]()())));
coalescedWrite(T_v[i]()(1, 1), coalescedRead(timesI(F_v[i]()()))); coalescedWrite(T_v[i]()(1, 1), coalescedRead(timesI(F_v[i]()())));
@@ -139,7 +142,8 @@ public:
autoView( T_v , T, AcceleratorWrite); autoView( T_v , T, AcceleratorWrite);
autoView( F_v , F, AcceleratorRead); autoView( F_v , F, AcceleratorRead);
accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(), int osites=T.Grid()->oSites();
accelerator_for(i, osites,CloverField::vector_type::Nsimd(),
{ {
coalescedWrite(T_v[i]()(0, 1), coalescedRead(timesI(F_v[i]()()))); coalescedWrite(T_v[i]()(0, 1), coalescedRead(timesI(F_v[i]()())));
coalescedWrite(T_v[i]()(1, 0), coalescedRead(timesI(F_v[i]()()))); coalescedWrite(T_v[i]()(1, 0), coalescedRead(timesI(F_v[i]()())));
@@ -157,7 +161,8 @@ public:
autoView( T_v ,T,AcceleratorWrite); autoView( T_v ,T,AcceleratorWrite);
autoView( F_v ,F,AcceleratorRead); autoView( F_v ,F,AcceleratorRead);
accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(), int osites=T.Grid()->oSites();
accelerator_for(i, osites,CloverField::vector_type::Nsimd(),
{ {
coalescedWrite(T_v[i]()(0, 1), coalescedRead(-(F_v[i]()()))); coalescedWrite(T_v[i]()(0, 1), coalescedRead(-(F_v[i]()())));
coalescedWrite(T_v[i]()(1, 0), coalescedRead((F_v[i]()()))); coalescedWrite(T_v[i]()(1, 0), coalescedRead((F_v[i]()())));
@@ -176,7 +181,8 @@ public:
autoView( T_v , T,AcceleratorWrite); autoView( T_v , T,AcceleratorWrite);
autoView( F_v , F,AcceleratorRead); autoView( F_v , F,AcceleratorRead);
accelerator_for(i, T.Grid()->oSites(),CloverField::vector_type::Nsimd(), int osites=T.Grid()->oSites();
accelerator_for(i, osites,CloverField::vector_type::Nsimd(),
{ {
coalescedWrite(T_v[i]()(0, 0), coalescedRead(timesI(F_v[i]()()))); coalescedWrite(T_v[i]()(0, 0), coalescedRead(timesI(F_v[i]()())));
coalescedWrite(T_v[i]()(1, 1), coalescedRead(timesMinusI(F_v[i]()()))); coalescedWrite(T_v[i]()(1, 1), coalescedRead(timesMinusI(F_v[i]()())));
@@ -200,7 +206,8 @@ public:
autoView(phi_v, phi, AcceleratorRead); autoView(phi_v, phi, AcceleratorRead);
autoView(C_v, C, AcceleratorRead); autoView(C_v, C, AcceleratorRead);
typedef decltype(coalescedRead(out_v[0])) calcSpinor; typedef decltype(coalescedRead(out_v[0])) calcSpinor;
accelerator_for(sss,out.Grid()->oSites(),Nsimd,{ int osites=out.Grid()->oSites();
accelerator_for(sss,osites,Nsimd,{
calcSpinor tmp; calcSpinor tmp;
multClover(tmp,C_v[sss],phi_v(sss)); multClover(tmp,C_v[sss],phi_v(sss));
coalescedWrite(out_v[sss],tmp); coalescedWrite(out_v[sss],tmp);
@@ -610,7 +617,8 @@ public:
autoView(triangle_v, triangle, AcceleratorWrite); autoView(triangle_v, triangle, AcceleratorWrite);
// NOTE: this function cannot be 'private' since nvcc forbids this for kernels // NOTE: this function cannot be 'private' since nvcc forbids this for kernels
accelerator_for(ss, full.Grid()->oSites(), 1, { int osites=full.Grid()->oSites();
accelerator_for(ss, osites, 1, {
for(int s_row = 0; s_row < Ns; s_row++) { for(int s_row = 0; s_row < Ns; s_row++) {
for(int s_col = 0; s_col < Ns; s_col++) { for(int s_col = 0; s_col < Ns; s_col++) {
if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue; if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue;
@@ -650,7 +658,8 @@ public:
autoView(full_v, full, AcceleratorWrite); autoView(full_v, full, AcceleratorWrite);
// NOTE: this function cannot be 'private' since nvcc forbids this for kernels // NOTE: this function cannot be 'private' since nvcc forbids this for kernels
accelerator_for(ss, full.Grid()->oSites(), 1, { int osites=full.Grid()->oSites();
accelerator_for(ss, osites, 1, {
for(int s_row = 0; s_row < Ns; s_row++) { for(int s_row = 0; s_row < Ns; s_row++) {
for(int s_col = 0; s_col < Ns; s_col++) { for(int s_col = 0; s_col < Ns; s_col++) {
if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue; if(abs(s_row - s_col) > 1 || s_row + s_col == 3) continue;