Skip to content

Enable max-mixed multicg_offset_QUDA in specific situations. #93

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
SaltyChiang wants to merge 1 commit into
base: develop
Choose a base branch
from
+12 −1
Open

Enable max-mixed multicg_offset_QUDA in specific situations. #93

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 12 additions & 1 deletion generic_ks/ks_multicg_offset_quda.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,6 +136,7 @@ int ks_multicg_offset_field_gpu(
double* final_residual = (double*)malloc(num_offsets*sizeof(double));
double* final_relative_residual = (double*)malloc(num_offsets*sizeof(double));

double max_residual = 0.0, min_residual = __DBL_MAX__;
for(i=0; i<num_offsets; ++i){

residual[i] = qic[i].resid;
Expand All @@ -145,6 +146,8 @@ int ks_multicg_offset_field_gpu(
// scale the shifted residual relative to the residue
residual[i] = fabs(residue[0] / residue[i]) * residual[0];
if (residual[i] < 1e-14) residual[i] = 1e-14;
if (residual[i] > max_residual) max_residual = residual[i];
if (residual[i] < min_residual) min_residual = residual[i];
} else {
residual[i] = qic[i].resid; // for a mixed-precision solver use residual for higher shifts
}
Expand All @@ -160,7 +163,15 @@ int ks_multicg_offset_field_gpu(
}

inv_args.max_iter = qic[0].max*qic[0].nrestart;
#if defined(MAX_MIXED) || defined(HALF_MIXED) // never do half precision with multi-shift solver
#if defined(MAX_MIXED)
if (residual[0] > 3e-5 && min_residual / max_residual > 4e-3) {
// 3e-5 ~ 2**-15 (Machine epsilon of QUDA's half precision)
// 4e-3 ~ 2**-8 (Refinement will not use too many iterations)
inv_args.mixed_precision = 2;
} else {
inv_args.mixed_precision = 1;
}
#elif defined(HALF_MIXED)
inv_args.mixed_precision = 1;
#else
inv_args.mixed_precision = 0;
Expand Down