lattice/latticegpu.jl
1
0
Fork 0
mirror of https://igit.ific.uv.es/alramos/latticegpu.jl.git synced 2025-05-14 19:23:42 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Some errors

Browse source
This commit is contained in:
Fernando P.Panadero 2024-05-21 15:38:28 +02:00
parent 182fa82d13
commit abfd0bd72a
5 changed files with 135 additions and 109 deletions
Download patch file Download diff file Expand all files Collapse all files

15
src/Dirac/Diracfields.jl
View file

@ -120,7 +120,7 @@ end
Randomizes the SU2fund / SU3fund fermion field. If the argument t is present, it only randomizes that time-slice.
"""
function pfrandomize!(f::AbstractArray{Spinor{4, SU3fund{T}}}, lp::SpaceParm, t::Int64 = 0) where {T}
function pfrandomize!(f::AbstractArray{Spinor{4, SU3fund{T}}}, lp::SpaceParm{4,6,BC_PERIODIC,D}, t::Int64 = 0) where {T,D}
@timeit "Randomize pseudofermion field" begin
p = ntuple(i->CUDA.randn(T, lp.bsz, 3, lp.rsz,2),4) # complex generation not suported for Julia 1.5.4
@ -132,6 +132,19 @@ function pfrandomize!(f::AbstractArray{Spinor{4, SU3fund{T}}}, lp::SpaceParm, t:
return nothing
end
function pfrandomize!(f::AbstractArray{Spinor{4, SU3fund{T}}}, lp::Union{SpaceParm{4,6,BC_SF_ORBI,D},SpaceParm{4,6,BC_SF_AFWB,D},SpaceParm{4,6,BC_OPEN,D}}, t::Int64 = 0) where {T,D}
@timeit "Randomize pseudofermion field" begin
p = ntuple(i->CUDA.randn(T, lp.bsz, 3, lp.rsz,2),4) # complex generation not suported for Julia 1.5.4
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_assign_pf_su3!(f,p,lp,t)
end
end
SF_bndfix!(f,lp)
return nothing
end
function krnl_assign_pf_su3!(f::AbstractArray, p , lp::SpaceParm, t::Int64)
@inbounds begin

2
src/Dirac/Diracflow.jl
View file

@ -234,7 +234,7 @@ function krnl_Nablanabla(so, U, si, th, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
@inbounds begin
if ((point_time((b,r),lp) != 1) && (point_time((b,r),lp) != lp.iL[end])
if ((point_time((b,r),lp) != 1) && (point_time((b,r),lp) != lp.iL[end]))
so[b,r] = -4*si[b,r]

24
src/Dirac/Diracoper.jl
View file

@ -17,13 +17,13 @@ function Dw!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceParm{4,6
if abs(dpar.csw) > 1.0E-10
@timeit "Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_Dwimpr!(so, U, si, dws.csw, dpar.m0, dpar.tm, dpar.th, dpar.csw, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_Dwimpr!(so, U, si, dws.csw, dpar.m0, dpar.tm, dpar.th, dpar.csw, dpar.ct, lp)
end
end
else
@timeit "Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_Dw!(so, U, si, dpar.m0, dpar.tm, dpar.th, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_Dw!(so, U, si, dpar.m0, dpar.tm, dpar.th, dpar.ct, lp)
end
end
end
@ -32,7 +32,7 @@ function Dw!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceParm{4,6
return nothing
end
function krnl_Dwimpr!(so, U, si, Fcsw, m0, tm, th, csw, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
function krnl_Dwimpr!(so, U, si, Fcsw, m0, tm, th, csw, ct, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
# The field si is assumed to be zero at t = 0,T
b = Int64(CUDA.threadIdx().x); r = Int64(CUDA.blockIdx().x)
@ -67,7 +67,7 @@ function krnl_Dwimpr!(so, U, si, Fcsw, m0, tm, th, csw, lp::SpaceParm{4,6,BC_OPE
return nothing
end
function krnl_Dw!(so, U, si, m0, tm, th, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
function krnl_Dw!(so, U, si, m0, tm, th, ct, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
# The field si is assumed to be zero at t = 0,T
b = Int64(CUDA.threadIdx().x); r = Int64(CUDA.blockIdx().x)
@ -113,13 +113,13 @@ function g5Dw!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceParm{4
if abs(dpar.csw) > 1.0E-10
@timeit "g5Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dwimpr!(so, U, si, dws.csw, dpar.m0, dpar.tm, dpar.th, dpar.csw, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dwimpr!(so, U, si, dws.csw, dpar.m0, dpar.tm, dpar.th, dpar.csw, dpar.ct, lp)
end
end
else
@timeit "g5Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(so, U, si, dpar.m0, dpar.tm, dpar.th, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(so, U, si, dpar.m0, dpar.tm, dpar.th, dpar.ct, lp)
end
end
end
@ -128,7 +128,7 @@ function g5Dw!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceParm{4
return nothing
end
function krnl_g5Dwimpr!(so, U, si, Fcsw, m0, tm, th, csw, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
function krnl_g5Dwimpr!(so, U, si, Fcsw, m0, tm, th, csw, ct, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
# The field si is assumed to be zero at t = 0,T
b = Int64(CUDA.threadIdx().x); r = Int64(CUDA.blockIdx().x)
@ -166,7 +166,7 @@ function krnl_g5Dwimpr!(so, U, si, Fcsw, m0, tm, th, csw, lp::SpaceParm{4,6,BC_O
return nothing
end
function krnl_g5Dw!(so, U, si, m0, tm, th, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
function krnl_g5Dw!(so, U, si, m0, tm, th, ct, lp::SpaceParm{4,6,BC_OPEN,D}) where {D}
# The field si is assumed to be zero at t = 0,T
b = Int64(CUDA.threadIdx().x); r = Int64(CUDA.blockIdx().x)
@ -215,13 +215,13 @@ function DwdagDw!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpacePar
@timeit "g5Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dwimpr!(dws.st, U, si, dws.csw, dpar.m0, dpar.tm, dpar.th, dpar.csw, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dwimpr!(dws.st, U, si, dws.csw, dpar.m0, dpar.tm, dpar.th, dpar.csw, dpar.ct, lp)
end
end
SF_bndfix!(dws.st,lp)
@timeit "g5Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dwimpr!(so, U, dws.st, dws.csw, dpar.m0, -dpar.tm, dpar.th, dpar.csw, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dwimpr!(so, U, dws.st, dws.csw, dpar.m0, -dpar.tm, dpar.th, dpar.csw, dpar.ct, lp)
end
end
SF_bndfix!(so,lp)
@ -231,13 +231,13 @@ function DwdagDw!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpacePar
@timeit "g5Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(dws.st, U, si, dpar.m0, dpar.tm, dpar.th, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(dws.st, U, si, dpar.m0, dpar.tm, dpar.th, dpar.ct, lp)
end
end
SF_bndfix!(dws.st,lp)
@timeit "g5Dw" begin
CUDA.@sync begin
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(so, U, dws.st, dpar.m0, -dpar.tm, dpar.th, lp)
CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(so, U, dws.st, dpar.m0, -dpar.tm, dpar.th, dpar.ct, lp)
end
end
SF_bndfix!(so,lp)

4
src/YM/YMact.jl
View file

@ -11,7 +11,7 @@
##
## OPEN DEBERIA ESTAR BIEN LAS ACCIONES, CHECKEAR. LAS FUERZAS FALTAN
## OPEN
##
function krnl_impr!(plx, U::AbstractArray{T}, c0, c1, Ubnd::NTuple{NB,T}, cG, ztw, lp::SpaceParm{N,M,BC_OPEN,D}) where {T,NB,N,M,D}
@ -113,7 +113,7 @@ function krnl_plaq!(plx, U::AbstractArray{T}, Ubnd, cG, ztw, lp::SpaceParm{N,M,B
gt1 = U[bu1,id2,ru1]
if ( (it == lp.iL[end]) || (it == 1)) && !TOBC
S += 0.5*cG*(c0*tr(g2*ga/U[bu2,id1,ru2]))
S += 0.5*cG*(tr(U[b,id1,r]*gt1 / (U[b,id2,r]*U[bu2,id1,ru2])))
elseif (it == lp.iL[end]) && TOBC
nothing
else

19
src/YM/YMfields.jl
View file

@ -54,14 +54,28 @@ function krnl_assign_SU3!(frc::AbstractArray{T}, m, lp::SpaceParm{N,M,BC_PERIODI
return nothing
end
function krnl_assign_SU3!(frc::AbstractArray{T}, m, lp::SpaceParm{N,M,B,D}) where {T,N,M,B,D}
function krnl_assign_SU3!(frc::AbstractArray{T}, m, lp::SpaceParm{N,M,BC_OPEN,D}) where {T,N,M,D}
@inbounds begin
b = Int64(CUDA.threadIdx().x)
r = Int64(CUDA.blockIdx().x)
for id in 1:lp.ndim
frc[b,id,r] = SU3alg(m[b,id,1,r], m[b,id,2,r], m[b,id,3,r],
m[b,id,4,r], m[b,id,5,r], m[b,id,6,r],
m[b,id,7,r], m[b,id,8,r])
end
end
return nothing
end
function krnl_assign_SU3!(frc::AbstractArray{T}, m, lp::Union{SpaceParm{N,M,BC_SF_ORBI,D},SpaceParm{N,M,BC_SF_AFWB,D}}) where {T,N,M,D}
@inbounds begin
b = Int64(CUDA.threadIdx().x)
r = Int64(CUDA.blockIdx().x)
it = point_time((b,r), lp)
if ((B==BC_SF_AFWB)||(B==BC_SF_ORBI))
if it == 1
for id in 1:lp.ndim-1
frc[b,id,r] = zero(T)
@ -77,7 +91,6 @@ function krnl_assign_SU3!(frc::AbstractArray{T}, m, lp::SpaceParm{N,M,B,D}) wher
end
end
end
end
return nothing
end