diff --git a/src/Dirac/Dirac.jl b/src/Dirac/Dirac.jl index 2cf902b..c2d1381 100644 --- a/src/Dirac/Dirac.jl +++ b/src/Dirac/Dirac.jl @@ -64,7 +64,7 @@ end function krnl_Dw!(so, U, si, m0, th, lp::SpaceParm{4,6,B,D}) where {B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() # For SF: # - cttilde affects mass term at x0 = a, T-a @@ -85,7 +85,7 @@ end function krnl_g5Dw!(so, U, si, m0, th, lp::SpaceParm{4,6,B,D}) where {B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() @inbounds begin so[b,r] = (4+m0)*si[b,r] diff --git a/src/LatticeGPU.jl b/src/LatticeGPU.jl index 72886b6..cc3725f 100644 --- a/src/LatticeGPU.jl +++ b/src/LatticeGPU.jl @@ -23,7 +23,7 @@ include("Space/Space.jl") using .Space export SpaceParm -export up, dw, updw, point_coord, point_index, point_time +export up, dw, updw, point_coord, point_index, point_time, assign_thx export BC_PERIODIC, BC_OPEN, BC_SF_AFWB, BC_SF_ORBI include("Fields/Fields.jl") diff --git a/src/Space/Space.jl b/src/Space/Space.jl index 67204b4..9061c6b 100644 --- a/src/Space/Space.jl +++ b/src/Space/Space.jl @@ -333,7 +333,10 @@ Returns the sum of the cartesian coordinates of the point p=(b,r). return s end -export up, dw, updw, point_index, point_coord, point_time + +@inline assign_thx() = convert(Int64, CUDA.threadIdx().x), convert(Int64, CUDA.blockIdx().x) + +export up, dw, updw, point_index, point_coord, point_time, assign_thx export BC_PERIODIC, BC_OPEN, BC_SF_AFWB, BC_SF_ORBI end diff --git a/src/YM/YMact.jl b/src/YM/YMact.jl index 0b26490..03b68a4 100644 --- a/src/YM/YMact.jl +++ b/src/YM/YMact.jl @@ -11,7 +11,7 @@ function krnl_impr!(plx, U::AbstractArray{T}, c0, c1, Ubnd::NTuple{NB,T}, cG, ztw, lp::SpaceParm{N,M,B,D}) where {T,NB,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b, r), lp) Ush = @cuStaticSharedMem(T, (D,2)) @@ -121,7 +121,7 @@ end function krnl_plaq!(plx, U::AbstractArray{T}, Ubnd, cG, ztw, lp::SpaceParm{N,M,B,D}) where {T,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b, r), lp) Ush = @cuStaticSharedMem(T, (D,2)) @@ -172,7 +172,7 @@ end function krnl_force_wilson_pln!(frc1, frc2, U::AbstractArray{T}, Ubnd, cG, ztw, ipl, lp::SpaceParm{N,M,B,D}) where {T,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b,r), lp) Ush = @cuStaticSharedMem(T, (D,2)) @@ -233,7 +233,7 @@ end function krnl_force_impr_pln!(frc1, frc2, U::AbstractArray{T}, c0, c1, Ubnd, cG, ztw, ipl, lp::SpaceParm{N,M,B,D}) where {T,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b, r), lp) Ush = @cuStaticSharedMem(T, (D,2)) diff --git a/src/YM/YMfields.jl b/src/YM/YMfields.jl index f5e043b..63a11a8 100644 --- a/src/YM/YMfields.jl +++ b/src/YM/YMfields.jl @@ -36,7 +36,7 @@ end function krnl_assign_SU3!(frc::AbstractArray{T}, m, lp::SpaceParm{N,M,BC_PERIODIC,D}) where {T,N,M,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() 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], @@ -47,7 +47,7 @@ end function krnl_assign_SU3!(frc::AbstractArray{T}, m, lp::SpaceParm{N,M,B,D}) where {T,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b,r), lp) if ((B==BC_SF_AFWB)||(B==BC_SF_ORBI)) @@ -72,7 +72,7 @@ end function krnl_assign_SU2!(frc, m, lp::SpaceParm{N,M,BC_PERIODIC,D}) where {N,M,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() for id in 1:lp.ndim frc[b,id,r] = SU2alg(m[b,id,1,r], m[b,id,2,r], m[b,id,3,r]) end diff --git a/src/YM/YMflow.jl b/src/YM/YMflow.jl index 241c6b5..58a1043 100644 --- a/src/YM/YMflow.jl +++ b/src/YM/YMflow.jl @@ -89,7 +89,7 @@ end function krnl_add_zth!(frc, frc2::AbstractArray{TA}, U::AbstractArray{TG}, lp::SpaceParm{N,M,B,D}) where {TA,TG,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b, r), lp) SFBC = ((B == BC_SF_AFWB) || (B == BC_SF_ORBI) ) @@ -261,7 +261,7 @@ Eoft_plaq(U, gp::GaugeParm{T,G,NN}, lp::SpaceParm{N,M,B,D}, ymws::YMworkspace) w function krnl_plaq_pln!(plx, U::AbstractArray{T}, Ubnd, ztw, ipl, lp::SpaceParm{N,M,B,D}) where {T,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() id1, id2 = lp.plidx[ipl] SFBC = ((B == BC_SF_AFWB) || (B == BC_SF_ORBI)) && (id1 == lp.iL[end]) @@ -387,7 +387,7 @@ Eoft_clover(U, gp::GaugeParm, lp::SpaceParm{N,M,B,D}, ymws::YMworkspace{T}) wher function krnl_add_et!(rm, op, frc1, U, lp::SpaceParm{4,M,B,D}) where {M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() X1 = (frc1[b,1,r]+frc1[b,2,r]+frc1[b,3,r]+frc1[b,4,r]) @@ -399,7 +399,7 @@ end function krnl_add_qd!(rm, op, frc1, frc2, U, lp::SpaceParm{4,M,B,D}) where {M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() I = point_coord((b,r), lp) rm[I] += op(dot( (frc1[b,1,r]+frc1[b,2,r]+frc1[b,3,r]+frc1[b,4,r]), @@ -409,7 +409,7 @@ end function krnl_field_tensor!(frc1::AbstractArray{TA}, frc2, U::AbstractArray{T}, Ubnd, ipl1, ipl2, ztw1, ztw2, lp::SpaceParm{4,M,B,D}) where {TA,T,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b,r), lp) SFBC = ((B == BC_SF_AFWB) || (B == BC_SF_ORBI) ) diff --git a/src/YM/YMsf.jl b/src/YM/YMsf.jl index 7f47d68..2a3e3c5 100644 --- a/src/YM/YMsf.jl +++ b/src/YM/YMsf.jl @@ -44,7 +44,7 @@ end function krnl_sfcoupling!(rm, U::AbstractArray{T}, Ubnd, lp::SpaceParm{N,M,B,D}) where {T,N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() I = point_coord((b,r), lp) it = I[N] @@ -99,7 +99,7 @@ end function krnl_setbnd_it0!(U, phi1, phi2, lp::SpaceParm{N,M,B,D}) where {N,M,B,D} - b, r = CUDA.threadIdx().x, CUDA.blockIdx().x + b, r = assign_thx() it = point_time((b,r), lp) SFBC = (B == BC_SF_AFWB) || (B == BC_SF_ORBI)