WIP: studies on CMS DY with phase space optimizations

epochX/cudacpp/CODEGEN/PLUGIN/CUDACPP_SA_OUTPUT/MG5aMC_patches/PROD/patch.P1

@@ -1,8 +1,141 @@
+diff --git b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig.f a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig.f
+index bc9bcfeb9..0c1962d3e 100644
+--- b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig.f
++++ a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig.f
+@@ -315,8 +315,10 @@ C     ENDDO
+
+ C     set the running scale 
+ C     and update the couplings accordingly
++      CALL COUNTERS_START_COUNTER( 5, VECSIZE_USED ) ! FortranUpdateScaleCouplings=5
+       CALL UPDATE_SCALE_COUPLING_VEC(ALL_P, ALL_WGT, ALL_Q2FACT,
+      $  VECSIZE_USED)
++      CALL COUNTERS_STOP_COUNTER( 5 ) ! FortranUpdateScaleCouplings=5
+
+       IF(GROUPED_MC_GRID_STATUS.EQ.0) THEN
+ C       If we were in the initialization phase of the grid for MC over
+diff --git b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig1.f a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig1.f
+index db3c284ca..f1cd4e976 100644
+--- b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig1.f
++++ a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig1.f
+@@ -127,6 +127,7 @@ C     Continue only if IMODE is 0, 4 or 5
+       IF(IMODE.NE.0.AND.IMODE.NE.4.AND.IMODE.NE.5) RETURN
+
+
++      CALL COUNTERS_START_COUNTER( 4, 1 ) ! FortranPDFs=4
+       IF (ABS(LPP(IB(1))).GE.1) THEN
+ C       LP=SIGN(1,LPP(IB(1)))
+         IF (DSQRT(Q2FACT(IB(1))).EQ.0D0) THEN
+@@ -148,6 +149,7 @@ C       LP=SIGN(1,LPP(IB(2)))
+         ENDIF
+         G2=PDG2PDF(LPP(IB(2)),0, IB(2),XBK(IB(2)), QSCALE)
+       ENDIF
++      CALL COUNTERS_STOP_COUNTER( 4 ) ! FortranPDFs=2
+       PD(0) = 0D0
+       IPROC = 0
+       IPROC=IPROC+1  ! g g > t t~
+@@ -186,7 +188,9 @@ C     Select a flavor combination (need to do here for right sign)
+         R=R-DABS(PD(IPSEL))/PD(0)
+       ENDDO
+
++      CALL COUNTERS_START_COUNTER( 6, 1 ) ! FortranReweight=6
+       DSIGUU=DSIGUU*REWGT(PP,1)
++      CALL COUNTERS_STOP_COUNTER( 6 ) ! FortranReweight=6
+
+ C     Apply the bias weight specified in the run card (default is 1.0)
+       DSIGUU=DSIGUU*CUSTOM_BIAS(PP,DSIGUU,1,1)
+@@ -360,6 +364,7 @@ C     Continue only if IMODE is 0, 4 or 5
+         STOP
+       ENDIF
+
++      CALL COUNTERS_START_COUNTER( 4, VECSIZE_USED ) ! FortranPDFs=2
+       DO CURR_WARP=1, NB_WARP_USED
+         IF(IMIRROR_VEC(CURR_WARP).EQ.1)THEN
+           IB(1) = 1
+@@ -382,6 +387,7 @@ C           LP=SIGN(1,LPP(IB(2)))
+           ENDIF
+         ENDDO  ! IWARP LOOP
+       ENDDO  ! CURRWARP LOOP
++      CALL COUNTERS_STOP_COUNTER( 4 ) ! FortranPDFs=2
+       ALL_PD(0,:) = 0D0
+       IPROC = 0
+       IPROC=IPROC+1  ! g g > t t~
+@@ -426,7 +432,9 @@ C         Select a flavor combination (need to do here for right sign)
+           CHANNEL = SUBDIAG(1)
+
+
++          CALL COUNTERS_START_COUNTER( 6, 1 ) ! FortranReweight=6
+           ALL_RWGT(IVEC) = REWGT(ALL_PP(0,1,IVEC), IVEC)
++          CALL COUNTERS_STOP_COUNTER( 6 ) ! FortranReweight=6
+
+           IF(FRAME_ID.NE.6)THEN
+             CALL BOOST_TO_FRAME(ALL_PP(0,1,IVEC), FRAME_ID, P_MULTI(0
+@@ -482,11 +490,13 @@ C         Set sign of dsig based on sign of PDF and matrix element
+           ALL_OUT(IVEC)=0D0
+         ENDIF
+ C       Generate events only if IMODE is 0.
++        CALL COUNTERS_START_COUNTER( 7, 1 ) ! FortranUnweight=7
+         IF(IMODE.EQ.0.AND.DABS(ALL_OUT(IVEC)).GT.0D0)THEN
+ C         Call UNWGT to unweight and store events
+           CALL UNWGT(ALL_PP(0,1,IVEC), ALL_OUT(IVEC)*ALL_WGT(IVEC),1,
+      $      SELECTED_HEL(IVEC), SELECTED_COL(IVEC), IVEC)
+         ENDIF
++        CALL COUNTERS_STOP_COUNTER( 7 ) ! FortranUnweight=7
+       ENDDO
+
+       END
+@@ -555,7 +565,7 @@ C         Call UNWGT to unweight and store events
+
+       IF( FBRIDGE_MODE .LE. 0 ) THEN  ! (FortranOnly=0 or BothQuiet=-1 or BothDebug=-2)
+ #endif
+-        CALL COUNTERS_SMATRIX1MULTI_START( -1, VECSIZE_USED )  ! fortranMEs=-1
++        CALL COUNTERS_START_COUNTER( 9, VECSIZE_USED ) ! FortranMEs=9
+         DO IVEC=1, VECSIZE_USED
+           CALL SMATRIX1(P_MULTI(0,1,IVEC),
+      &	                         hel_rand(IVEC),
+@@ -571,7 +581,7 @@ C       ======================================================
+ C       *START* Included from CUDACPP template smatrix_multi.f
+ C       (into function smatrix$i_multi in auto_dsig$i.f)
+ C       ======================================================
+-        CALL COUNTERS_SMATRIX1MULTI_STOP( -1 )  ! fortranMEs=-1
++        CALL COUNTERS_STOP_COUNTER( 9 ) ! FortranMEs=9
+ #ifdef MG5AMC_MEEXPORTER_CUDACPP
+       ENDIF
+
+@@ -581,7 +591,7 @@ C       ======================================================
+           STOP
+         ENDIF
+         IF ( FIRST ) THEN  ! exclude first pass (helicity filtering) from timers (#461)
+-          CALL COUNTERS_SMATRIX1MULTI_START( 1, VECSIZE_USED )  ! cudacppHEL=1
++          CALL COUNTERS_START_COUNTER( 11, 0 ) ! 11=CudaCpp-Initialise (was CudaCpp-HEL; counter set to 1 on bridge creation, do not increment it further)
+           CALL FBRIDGESEQUENCE_NOMULTICHANNEL( FBRIDGE_PBRIDGE,  ! multi channel disabled for helicity filtering
+      &      P_MULTI, ALL_G, HEL_RAND, COL_RAND, OUT2,
+      &      SELECTED_HEL2, SELECTED_COL2, .TRUE.)  ! quit after computing helicities
+@@ -602,9 +612,9 @@ C         ENDIF
+           ENDIF
+           WRITE (6,*) 'NGOODHEL =', NGOODHEL
+           WRITE (6,*) 'NCOMB =', NCOMB
+-          CALL COUNTERS_SMATRIX1MULTI_STOP( 1 )  ! cudacppHEL=1
++          CALL COUNTERS_STOP_COUNTER( 11 ) ! 11=CudaCpp-Initialise (was CudaCpp-HEL)
+         ENDIF
+-        CALL COUNTERS_SMATRIX1MULTI_START( 0, VECSIZE_USED )  ! cudacppMEs=0
++        CALL COUNTERS_START_COUNTER( 19, VECSIZE_USED ) ! CudaCppMEs=19
+         IF ( .NOT. MULTI_CHANNEL ) THEN
+           CALL FBRIDGESEQUENCE_NOMULTICHANNEL( FBRIDGE_PBRIDGE,  ! multi channel disabled
+      &      P_MULTI, ALL_G, HEL_RAND, COL_RAND, OUT2,
+@@ -618,7 +628,7 @@ C         ENDIF
+      &      HEL_RAND, COL_RAND, CHANNELS, OUT2,
+      &      SELECTED_HEL2, SELECTED_COL2, .FALSE.)  ! do not quit after computing helicities
+         ENDIF
+-        CALL COUNTERS_SMATRIX1MULTI_STOP( 0 )  ! cudacppMEs=0
++        CALL COUNTERS_STOP_COUNTER( 19 ) ! CudaCppMEs=19
+       ENDIF
+
+       IF( FBRIDGE_MODE .LT. 0 ) THEN  ! (BothQuiet=-1 or BothDebug=-2)
 diff --git b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/driver.f a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/driver.f
-index 1124a9164..27a6e4674 100644
+index ecd11b239..4650934b2 100644
 --- b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/driver.f
 +++ a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/driver.f
-@@ -74,13 +74,77 @@ c      common/to_colstats/ncols,ncolflow,ncolalt,ic
+@@ -76,16 +76,95 @@ c      common/to_colstats/ncols,ncolflow,ncolalt,ic
 
        include 'coupl.inc' ! needs VECSIZE_MEMMAX (defined in vector.inc)
        INTEGER VECSIZE_USED
@@ -27,7 +160,19 @@ index 1124a9164..27a6e4674 100644
 +      CALL OMPNUMTHREADS_NOT_SET_MEANS_ONE_THREAD()
 +#endif
 +      CALL COUNTERS_INITIALISE()
-+
++c Use null-terminated C-string in COUNTERS_REGISTER_COUNTER calls (maybe it is not needed, but it does not harm)
++      CALL COUNTERS_REGISTER_COUNTER( 1, 'Fortran Initialise(I/O)'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 3, 'Fortran PhaseSpaceSampling'//char(0) ) ! uniform [0,1] + vegas to [0,1] + map to momenta
++      CALL COUNTERS_REGISTER_COUNTER( 4, 'Fortran PDFs'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 5, 'Fortran UpdateScaleCouplings'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 6, 'Fortran Reweight'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 7, 'Fortran Unweight(LHE-I/O)'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 8, 'Fortran SamplePutPoint'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 9, 'Fortran MEs'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 11, 'CudaCpp Initialise'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 12, 'CudaCpp Finalise'//char(0) )
++      CALL COUNTERS_REGISTER_COUNTER( 19, 'CudaCpp MEs'//char(0) )
++c     CALL COUNTERS_REGISTER_COUNTER( 21, 'TEST    SampleGetX'//char(0) )
 +#ifdef MG5AMC_MEEXPORTER_CUDACPP
 +      fbridge_mode = 1 ! CppOnly=1, default for CUDACPP
 +#else
@@ -71,24 +216,39 @@ index 1124a9164..27a6e4674 100644
 +      endif
 +
 +#ifdef MG5AMC_MEEXPORTER_CUDACPP
++      CALL COUNTERS_START_COUNTER( 11, 1 ) ! 11=CudaCpp-Initialise
 +      CALL FBRIDGECREATE(FBRIDGE_PBRIDGE, VECSIZE_USED, NEXTERNAL, 4) ! this must be at the beginning as it initialises the CUDA device
 +      FBRIDGE_NCBYF1 = 0
 +      FBRIDGE_CBYF1SUM = 0
 +      FBRIDGE_CBYF1SUM2 = 0
 +      FBRIDGE_CBYF1MAX = -1D100
 +      FBRIDGE_CBYF1MIN = 1D100
++      CALL COUNTERS_STOP_COUNTER( 11 ) ! 11=CudaCpp-Initialise
 +#endif
  c
  c     Read process number
  c
-@@ -208,8 +272,33 @@ c      call sample_result(xsec,xerr)
++      CALL COUNTERS_START_COUNTER( 1, 1 ) ! FortranInitialise=1
+       call open_file(lun+1, 'dname.mg', fopened)
+       if (.not.fopened)then
+          goto 11
+@@ -156,6 +235,7 @@ c   If CKKW-type matching, read IS Sudakov grid
+           print *,'Running CKKW as lower mult sample'
+         endif
+       endif
++      CALL COUNTERS_STOP_COUNTER( 1 ) ! FortranInitialise=1
+
+ c     
+ c     Get user input
+@@ -216,8 +296,35 @@ c      call sample_result(xsec,xerr)
  c      write(*,*) 'Final xsec: ',xsec
 
        rewind(lun)
 -
        close(lun)
 +
 +#ifdef MG5AMC_MEEXPORTER_CUDACPP
++      CALL COUNTERS_START_COUNTER( 12, 1 ) ! 12=CudaCpp-Finalise
 +      CALL FBRIDGEDELETE(FBRIDGE_PBRIDGE) ! this must be at the end as it shuts down the CUDA device
 +      IF( FBRIDGE_MODE .LE. -1 ) THEN ! (BothQuiet=-1 or BothDebug=-2)
 +        WRITE(*,'(a,f10.8,a,e8.2)')
@@ -111,12 +271,13 @@ index 1124a9164..27a6e4674 100644
 +     &    FBRIDGE_CBYF1SUM / FBRIDGE_NCBYF1, ' +- ',
 +     &    SQRT( FBRIDGE_CBYF1SUM2 ) / FBRIDGE_NCBYF1 ! ~standard error
 +      ENDIF
++      CALL COUNTERS_STOP_COUNTER( 12 ) ! 12=CudaCpp-Finalise
 +#endif
 +      CALL COUNTERS_FINALISE()
        end
 
  c     $B$ get_user_params $B$ ! tag for MadWeight
-@@ -387,7 +476,7 @@ c
+@@ -400,7 +507,7 @@ c
        fopened=.false.
        tempname=filename 	 
        fine=index(tempname,' ') 	 
@@ -126,7 +287,7 @@ index 1124a9164..27a6e4674 100644
        open(unit=lun,file=tempname,status='old',ERR=20)
        fopened=.true.
 diff --git b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/matrix1.f a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/matrix1.f
-index 1acba8200..069c74ef4 100644
+index bf488e4b0..707ea4032 100644
 --- b/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/matrix1.f
 +++ a/epochX/cudacpp/gg_tt.mad/SubProcesses/P1_gg_ttx/matrix1.f
 @@ -71,7 +71,10 @@ C
@@ -141,7 +302,7 @@ index 1acba8200..069c74ef4 100644
  C     
  C     This is just to temporarily store the reference grid for
  C      helicity of the DiscreteSampler so as to obtain its number of
-@@ -211,6 +214,17 @@ C     ----------
+@@ -224,6 +227,17 @@ C            update.
            ENDIF
            IF(NTRY(1).EQ.MAXTRIES)THEN
              ISHEL=MIN(ISUM_HEL,NGOOD)