mirror of
https://github.com/paboyle/Grid.git
synced 2025-06-18 15:57:05 +01:00
Compiles GPU and CPU, still gives good performance on CPU
This commit is contained in:
@ -90,15 +90,7 @@ public:
|
||||
auto UU = coalescedRead(U(mu));
|
||||
mult(&phi(), &UU, &chi());
|
||||
}
|
||||
|
||||
#ifdef GPU_VEC
|
||||
static accelerator_inline void copyLinkGpu(int lane,
|
||||
SiteDoubledGaugeField & UU,
|
||||
const SiteDoubledGaugeField &U)
|
||||
{
|
||||
auto U_l = extractLane(lane,U);
|
||||
insertLane(lane,UU,U_l);
|
||||
}
|
||||
|
||||
static accelerator_inline void multLinkGpu(int lane,
|
||||
typename SiteHalfSpinor::scalar_object &phi,
|
||||
const SiteDoubledGaugeField &U,
|
||||
@ -108,17 +100,6 @@ public:
|
||||
auto U_l = extractLane(lane,U(mu));
|
||||
phi() = U_l * chi();
|
||||
}
|
||||
#else
|
||||
static accelerator_inline void multLinkGpu(int lane,
|
||||
SiteHalfSpinor &phi,
|
||||
const SiteDoubledGaugeField &U,
|
||||
const SiteHalfSpinor &chi,
|
||||
int mu)
|
||||
{
|
||||
auto U_l = U(mu);
|
||||
phi() = U_l * chi();
|
||||
}
|
||||
#endif
|
||||
|
||||
static accelerator_inline void multLinkProp(SitePropagator &phi,
|
||||
const SiteDoubledGaugeField &U,
|
||||
|
Reference in New Issue
Block a user