Test for Dslash

src/Dirac/Diracflw.jl

@@ -93,6 +93,33 @@ function Nablanabla!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::Space
 end
 
 
+function krnl_g5Dslsh!(so, U, si, m0, th, lp::SpaceParm{4,6,B,D}) where {B,D}
+
+    b = Int64(CUDA.threadIdx().x);  r = Int64(CUDA.blockIdx().x)
+
+    @inbounds begin 
+        
+	        bu1, ru1 = up((b,r), 1, lp)
+            bd1, rd1 = dw((b,r), 1, lp)
+            bu2, ru2 = up((b,r), 2, lp)
+            bd2, rd2 = dw((b,r), 2, lp)
+            bu3, ru3 = up((b,r), 3, lp)
+            bd3, rd3 = dw((b,r), 3, lp)
+            bu4, ru4 = up((b,r), 4, lp)
+            bd4, rd4 = dw((b,r), 4, lp)
+
+
+        so[b,r] = -0.5*( th[1]*U[b,1,r]*dmul(Gamma{1},si[bu1,ru1]) +conj(th[1])*dag(U[bd1,1,rd1])*dmul(Gamma{1},si[bd1,rd1]) +
+                         th[2]*U[b,2,r]*dmul(Gamma{2},si[bu1,ru1]) +conj(th[2])*dag(U[bd1,2,rd1])*dmul(Gamma{2},si[bd1,rd1]) +
+                         th[3]*U[b,3,r]*dmul(Gamma{3},si[bu1,ru1]) +conj(th[3])*dag(U[bd1,3,rd1])*dmul(Gamma{3},si[bd1,rd1]) + 
+                         th[4]*U[b,4,r]*dmul(Gamma{4},si[bu1,ru1]) +conj(th[4])*dag(U[bd1,4,rd1])*dmul(Gamma{4},si[bd1,rd1]) )
+
+
+        so[b,r] = dmul(Gamma{5}, so[b,r])
+    end
+        
+    return nothing
+end
 
 function Dslash_sq!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceParm{4,6,B,D}) where {B,D}
 
@@ -105,9 +132,9 @@ function Dslash_sq!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceP
             end
             end
 
-            @timeit "g5Dw" begin
+            @timeit "g5Dslsh" begin
             CUDA.@sync begin
-                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(dws.st, U, si,-4.0, dpar.th, lp)
+                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh!(dws.st, U, si,-4.0, dpar.th, lp)
             end
             end
 
@@ -120,15 +147,15 @@ function Dslash_sq!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceP
             if abs(dpar.csw) > 1.0E-10
                 @timeit "Dw_improvement" begin
                         CUDA.@sync begin
-                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw_impr!(dws.st, dws.csw, dpar.csw, si,  lp)
+                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh_impr!(dws.st, dws.csw, dpar.csw, si,  lp)
                         end
                 end
             end
 
 
-            @timeit "g5Dw" begin
+            @timeit "g5Dslsh" begin
             CUDA.@sync begin
-                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(so, U, dws.st, -4.0, dpar.th, lp)
+                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh!(so, U, dws.st, -4.0, dpar.th, lp)
             end
             end
 
@@ -142,7 +169,7 @@ function Dslash_sq!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceP
             if abs(dpar.csw) > 1.0E-10
                 @timeit "Dw_improvement" begin
                         CUDA.@sync begin
-                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw_impr!(so, dws.csw, dpar.csw, dws.st,  lp)
+                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh_impr!(so, dws.csw, dpar.csw, dws.st,  lp)
                         end
                 end
             end               
@@ -153,25 +180,25 @@ function Dslash_sq!(so, U, si, dpar::DiracParam, dws::DiracWorkspace, lp::SpaceP
     else
         @timeit "DwdagDw" begin
             CUDA.@sync begin
-                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(dws.st, U, si, -4.0, dpar.th, lp)
+                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh!(dws.st, U, si, -4.0, dpar.th, lp)
             end
 
             if abs(dpar.csw) > 1.0E-10
                 @timeit "Dw_improvement" begin
                         CUDA.@sync begin
-                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw_impr!(dws.st, dws.csw, dpar.csw, si,  lp)
+                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh_impr!(dws.st, dws.csw, dpar.csw, si,  lp)
                         end
                 end
             end               
 
             CUDA.@sync begin
-                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw!(so, U, dws.st, -4.0, dpar.th, lp)
+                CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh!(so, U, dws.st, -4.0, dpar.th, lp)
             end
 
             if abs(dpar.csw) > 1.0E-10
                 @timeit "Dw_improvement" begin
                         CUDA.@sync begin
-                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dw_impr!(so, dws.csw, dpar.csw, dws.st,  lp)
+                            CUDA.@cuda threads=lp.bsz blocks=lp.rsz krnl_g5Dslsh_impr!(so, dws.csw, dpar.csw, dws.st,  lp)
                         end
                 end
             end