remove unnecessary first pass on all bridge events in cudacpp helicity calculation from madevent (and improve timers)

.github/workflows/c-cpp.yml

@@ -38,7 +38,7 @@ jobs:
   CPU_MAC:
     runs-on: macos-latest
     env:
-      FC: gfortran-11
+      FC: gfortran-14 # see #971
     strategy:
       matrix:
         folder: [ epochX/cudacpp/ee_mumu.mad/SubProcesses/P1_epem_mupmum, epochX/cudacpp/gg_ttgg.mad/SubProcesses/P1_gg_ttxgg ]

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

@@ -1,8 +1,8 @@
 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 4fbb8e6ba..f9e2335de 100644
+index 4fbb8e6ba..d5accb9fb 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
-@@ -484,23 +484,140 @@ C
+@@ -484,23 +484,142 @@ C
        INTEGER VECSIZE_USED
 
        INTEGER IVEC
@@ -40,7 +40,7 @@ index 4fbb8e6ba..f9e2335de 100644
 +      
 +      IF( FBRIDGE_MODE .LE. 0 ) THEN ! (FortranOnly=0 or BothQuiet=-1 or BothDebug=-2)
 +#endif
-+        call counters_smatrix1multi_start( -1, VECSIZE_USED ) ! fortran=-1
++        call counters_smatrix1multi_start( -1, VECSIZE_USED ) ! fortranMEs=-1
  !$OMP PARALLEL
  !$OMP DO
 -      DO IVEC=1, VECSIZE_USED
@@ -67,7 +67,7 @@ index 4fbb8e6ba..f9e2335de 100644
 +        ENDDO
  !$OMP END DO
  !$OMP END PARALLEL
-+        call counters_smatrix1multi_stop( -1 ) ! fortran=-1
++        call counters_smatrix1multi_stop( -1 ) ! fortranMEs=-1
 +#ifdef MG5AMC_MEEXPORTER_CUDACPP
 +      ENDIF
 +
@@ -77,9 +77,10 @@ index 4fbb8e6ba..f9e2335de 100644
 +          STOP
 +        ENDIF
 +        IF ( FIRST ) THEN ! exclude first pass (helicity filtering) from timers (#461)
++          call counters_smatrix1multi_start( 1, VECSIZE_USED ) ! cudacppHEL=1
 +          CALL FBRIDGESEQUENCE_NOMULTICHANNEL( FBRIDGE_PBRIDGE, ! multi channel disabled for helicity filtering
 +     &      P_MULTI, ALL_G, HEL_RAND, COL_RAND, OUT2,
-+     &      SELECTED_HEL2, SELECTED_COL2 )
++     &      SELECTED_HEL2, SELECTED_COL2, .TRUE.) ! quit after computing helicities
 +          FIRST = .FALSE.
 +c         ! This is a workaround for https://github.com/oliviermattelaer/mg5amc_test/issues/22 (see PR #486)
 +          IF( FBRIDGE_MODE .EQ. 1 ) THEN ! (CppOnly=1 : SMATRIX1 is not called at all)
@@ -93,22 +94,23 @@ index 4fbb8e6ba..f9e2335de 100644
 +          ENDIF
 +          WRITE (6,*) 'NGOODHEL =', NGOODHEL
 +          WRITE (6,*) 'NCOMB =', NCOMB
++          call counters_smatrix1multi_stop( 1 ) ! cudacppHEL=1
 +        ENDIF
-+        call counters_smatrix1multi_start( 0, VECSIZE_USED ) ! cudacpp=0
++        call counters_smatrix1multi_start( 0, VECSIZE_USED ) ! cudacppMEs=0
 +        IF ( .NOT. MULTI_CHANNEL ) THEN
 +          CALL FBRIDGESEQUENCE_NOMULTICHANNEL( FBRIDGE_PBRIDGE, ! multi channel disabled
 +     &      P_MULTI, ALL_G, HEL_RAND, COL_RAND, OUT2,
-+     &      SELECTED_HEL2, SELECTED_COL2 )
++     &      SELECTED_HEL2, SELECTED_COL2, .FALSE.) ! do not quit after computing helicities
 +        ELSE
 +          IF( SDE_STRAT.NE.1 ) THEN
 +            WRITE(6,*) 'ERROR! The cudacpp bridge requires SDE=1' ! multi channel single-diagram enhancement strategy
 +            STOP
 +          ENDIF
-+          CALL FBRIDGESEQUENCE(FBRIDGE_PBRIDGE, P_MULTI, ALL_G,
++          CALL FBRIDGESEQUENCE(FBRIDGE_PBRIDGE, P_MULTI, ALL_G, ! multi channel enabled
 +     &      HEL_RAND, COL_RAND, CHANNEL, OUT2,
-+     &      SELECTED_HEL2, SELECTED_COL2 ) ! 1-N: multi channel enabled
++     &      SELECTED_HEL2, SELECTED_COL2, .FALSE.) ! do not quit after computing helicities
 +        ENDIF
-+        call counters_smatrix1multi_stop( 0 ) ! cudacpp=0
++        call counters_smatrix1multi_stop( 0 ) ! cudacppMEs=0
 +      ENDIF
 +
 +      IF( FBRIDGE_MODE .LT. 0 ) THEN ! (BothQuiet=-1 or BothDebug=-2)
@@ -284,7 +286,7 @@ 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 e73e654d4..27fbe7302 100644
+index e73e654d4..3072054f2 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
 @@ -72,7 +72,10 @@ C
@@ -299,15 +301,7 @@ index e73e654d4..27fbe7302 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
-@@ -140,6 +143,7 @@ C     ----------
- C     BEGIN CODE
- C     ----------
-
-+      call counters_smatrix1_start()
-       NTRY(IMIRROR)=NTRY(IMIRROR)+1
-       THIS_NTRY(IMIRROR) = THIS_NTRY(IMIRROR)+1
-       DO I=1,NEXTERNAL
-@@ -217,6 +221,17 @@ C     ----------
+@@ -217,6 +220,17 @@ C     ----------
            ENDIF
            IF(NTRY(IMIRROR).EQ.MAXTRIES)THEN
              ISHEL(IMIRROR)=MIN(ISUM_HEL,NGOOD(IMIRROR))
@@ -325,22 +319,3 @@ index e73e654d4..27fbe7302 100644
            ENDIF
          ENDIF
        ELSE IF (.NOT.INIT_MODE) THEN  ! random helicity 
-@@ -234,6 +249,7 @@ C       Include the Jacobian from helicity sampling
-         IHEL = HEL_PICKED
-       ELSE
-         ANS = 1D0
-+        call counters_smatrix1_stop()
-         RETURN
-       ENDIF
-       IF (ANS.NE.0D0.AND.(ISUM_HEL .NE. 1.OR.HEL_PICKED.EQ.-1)) THEN
-@@ -278,9 +294,8 @@ C           Set right sign for ANS, based on sign of chosen helicity
-         ENDIF
-       ENDIF
-       ANS=ANS/DBLE(IDEN)
--
-       CALL SELECT_COLOR(RCOL, JAMP2, ICONFIG,1,  ICOL)
--
-+      call counters_smatrix1_stop()
-       END
-
-

epochX/cudacpp/CODEGEN/PLUGIN/CUDACPP_SA_OUTPUT/madgraph/iolibs/template_files/gpu/Bridge.h

@@ -109,9 +109,9 @@ namespace mg5amcCpu
      * @param rndcol the pointer to the input random numbers for color selection
      * @param channelId the Feynman diagram to enhance in multi-channel mode if 1 to n (disable multi-channel if 0)
      * @param mes the pointer to the output matrix elements
-     * @param goodHelOnly quit after computing good helicities?
      * @param selhel the pointer to the output selected helicities
      * @param selcol the pointer to the output selected colors
+     * @param goodHelOnly quit after computing good helicities?
      */
     void gpu_sequence( const FORTRANFPTYPE* momenta,
                        const FORTRANFPTYPE* gs,

epochX/cudacpp/CODEGEN/PLUGIN/CUDACPP_SA_OUTPUT/madgraph/iolibs/template_files/gpu/counters.cc

@@ -21,26 +21,24 @@ extern "C"
 {
   // Now: fortran=-1, cudacpp=0
   // Eventually: fortran=-1, cuda=0, cpp/none=1, cpp/sse4=2, etc...
-  constexpr unsigned int nimplC = 2;
+  constexpr unsigned int nimplC = 3;
   constexpr unsigned int iimplF2C( int iimplF ) { return iimplF + 1; }
   const char* iimplC2TXT( int iimplC )
   {
     const int iimplF = iimplC - 1;
     switch( iimplF )
     {
-      case -1: return "Fortran"; break;
-      case +0: return "CudaCpp"; break;
+      case -1: return "Fortran MEs"; break;
+      case +0: return "CudaCpp MEs"; break;
+      case +1: return "CudaCpp HEL"; break;
       default: assert( false ); break;
     }
   }
 
   static mgOnGpu::Timer<TIMERTYPE> program_timer;
   static float program_totaltime = 0;
-  static mgOnGpu::Timer<TIMERTYPE> smatrix1_timer;
-  static float smatrix1_totaltime = 0;
   static mgOnGpu::Timer<TIMERTYPE> smatrix1multi_timer[nimplC];
   static float smatrix1multi_totaltime[nimplC] = { 0 };
-  static int smatrix1_counter = 0;
   static int smatrix1multi_counter[nimplC] = { 0 };
 
   void counters_initialise_()
@@ -49,19 +47,6 @@ extern "C"
     return;
   }
 
-  void counters_smatrix1_start_()
-  {
-    smatrix1_counter++;
-    smatrix1_timer.Start();
-    return;
-  }
-
-  void counters_smatrix1_stop_()
-  {
-    smatrix1_totaltime += smatrix1_timer.GetDuration();
-    return;
-  }
-
   void counters_smatrix1multi_start_( const int* iimplF, const int* pnevt )
   {
     const unsigned int iimplC = iimplF2C( *iimplF );
@@ -86,13 +71,23 @@ extern "C"
     printf( " [COUNTERS] PROGRAM TOTAL          : %9.4fs\n", program_totaltime );
     printf( " [COUNTERS] Fortran Overhead ( 0 ) : %9.4fs\n", overhead_totaltime );
     for( unsigned int iimplC = 0; iimplC < nimplC; iimplC++ )
+    {
       if( smatrix1multi_counter[iimplC] > 0 )
-        printf( " [COUNTERS] %7s MEs      ( %1d ) : %9.4fs for %8d events => throughput is %8.2E events/s\n",
-                iimplC2TXT( iimplC ),
-                iimplC + 1,
-                smatrix1multi_totaltime[iimplC],
-                smatrix1multi_counter[iimplC],
-                smatrix1multi_counter[iimplC] / smatrix1multi_totaltime[iimplC] );
+      {
+        if( iimplC < nimplC - 1 ) // MEs
+          printf( " [COUNTERS] %11s      ( %1d ) : %9.4fs for %8d events => throughput is %8.2E events/s\n",
+                  iimplC2TXT( iimplC ),
+                  iimplC + 1,
+                  smatrix1multi_totaltime[iimplC],
+                  smatrix1multi_counter[iimplC],
+                  smatrix1multi_counter[iimplC] / smatrix1multi_totaltime[iimplC] );
+        else
+          printf( " [COUNTERS] %11s      ( %1d ) : %9.4fs\n",
+                  iimplC2TXT( iimplC ),
+                  iimplC + 1,
+                  smatrix1multi_totaltime[iimplC] );
+      }
+    }
     return;
   }
 }

epochX/cudacpp/CODEGEN/PLUGIN/CUDACPP_SA_OUTPUT/madgraph/iolibs/template_files/gpu/fbridge.cc

@@ -83,6 +83,7 @@ extern "C"
    * @param mes the pointer to the output matrix elements
    * @param selhel the pointer to the output selected helicities
    * @param selcol the pointer to the output selected colors
+   * @param goodHelOnly quit after computing good helicities?
    */
   void fbridgesequence_( CppObjectInFortran** ppbridge,
                          const FORTRANFPTYPE* momenta,
@@ -92,18 +93,20 @@ extern "C"
                          const unsigned int* pchannelId,
                          FORTRANFPTYPE* mes,
                          int* selhel,
-                         int* selcol )
+                         int* selcol,
+                         const bool* pgoodHelOnly )
   {
     Bridge<FORTRANFPTYPE>* pbridge = dynamic_cast<Bridge<FORTRANFPTYPE>*>( *ppbridge );
+    //printf("fbridgesequence_ goodHelOnly=%d\n", ( *pgoodHelOnly ? 1 : 0 ) );
     if( pbridge == 0 ) throw std::runtime_error( "fbridgesequence_: invalid Bridge address" );
 #ifdef MGONGPUCPP_GPUIMPL
     // Use the device/GPU implementation in the CUDA library
     // (there is also a host implementation in this library)
-    pbridge->gpu_sequence( momenta, gs, rndhel, rndcol, ( pchannelId ? *pchannelId : 0 ), mes, selhel, selcol );
+    pbridge->gpu_sequence( momenta, gs, rndhel, rndcol, ( pchannelId ? *pchannelId : 0 ), mes, selhel, selcol, *pgoodHelOnly );
 #else
     // Use the host/CPU implementation in the C++ library
     // (there is no device implementation in this library)
-    pbridge->cpu_sequence( momenta, gs, rndhel, rndcol, ( pchannelId ? *pchannelId : 0 ), mes, selhel, selcol );
+    pbridge->cpu_sequence( momenta, gs, rndhel, rndcol, ( pchannelId ? *pchannelId : 0 ), mes, selhel, selcol, *pgoodHelOnly );
 #endif
   }
 
@@ -119,6 +122,7 @@ extern "C"
    * @param mes the pointer to the output matrix elements
    * @param selhel the pointer to the output selected helicities
    * @param selcol the pointer to the output selected colors
+   * @param goodHelOnly quit after computing good helicities?
    */
   void fbridgesequence_nomultichannel_( CppObjectInFortran** ppbridge,
                                         const FORTRANFPTYPE* momenta,
@@ -127,9 +131,11 @@ extern "C"
                                         const FORTRANFPTYPE* rndcol,
                                         FORTRANFPTYPE* mes,
                                         int* selhel,
-                                        int* selcol )
+                                        int* selcol,
+                                        const bool* pgoodHelOnly )
   {
-    fbridgesequence_( ppbridge, momenta, gs, rndhel, rndcol, nullptr, mes, selhel, selcol );
+    //printf("fbridgesequence_nomultichannel_ goodHelOnly=%d\n", ( *pgoodHelOnly ? 1 : 0 ) );
+    fbridgesequence_( ppbridge, momenta, gs, rndhel, rndcol, nullptr, mes, selhel, selcol, pgoodHelOnly );
   }
 
   /**

epochX/cudacpp/CODEGEN/PLUGIN/CUDACPP_SA_OUTPUT/madgraph/iolibs/template_files/gpu/fbridge.inc

@@ -40,10 +40,11 @@ C - CHANID:  the input Feynman diagram to enhance in multi-channel mode if 1 to
 C - MES:     the output matrix element Fortran array
 C - SELHEL:  the output selected helicity Fortran array
 C - SELCOL:  the output selected color Fortran array
+C - HELONLY: input flag, quit after computing good helicities?
 C
       INTERFACE
          SUBROUTINE FBRIDGESEQUENCE(PBRIDGE, MOMENTA, GS,
-     &     RNDHEL, RNDCOL, CHANID, MES, SELHEL, SELCOL)
+     &     RNDHEL, RNDCOL, CHANID, MES, SELHEL, SELCOL, HELONLY)
          INTEGER*8 PBRIDGE
          DOUBLE PRECISION MOMENTA(*)
          DOUBLE PRECISION GS(*)
@@ -53,6 +54,7 @@ C
          DOUBLE PRECISION MES(*)
          INTEGER*4 SELHEL(*)
          INTEGER*4 SELCOL(*)
+         LOGICAL HELONLY
          END SUBROUTINE FBRIDGESEQUENCE
       END INTERFACE
 
@@ -66,10 +68,11 @@ C - RNDCOL:  the input random number Fortran array for color selection
 C - MES:     the output matrix element Fortran array
 C - SELHEL:  the output selected helicity Fortran array
 C - SELCOL:  the output selected color Fortran array
+C - HELONLY: input flag, quit after computing good helicities?
 C
       INTERFACE
          SUBROUTINE FBRIDGESEQUENCE_NOMULTICHANNEL(PBRIDGE, MOMENTA, GS,
-     &     RNDHEL, RNDCOL, MES, SELHEL, SELCOL)
+     &     RNDHEL, RNDCOL, MES, SELHEL, SELCOL, HELONLY)
          INTEGER*8 PBRIDGE
          DOUBLE PRECISION MOMENTA(*)
          DOUBLE PRECISION GS(*)
@@ -78,6 +81,7 @@ C
          DOUBLE PRECISION MES(*)
          INTEGER*4 SELHEL(*)
          INTEGER*4 SELCOL(*)
+         LOGICAL HELONLY
          END SUBROUTINE FBRIDGESEQUENCE_NOMULTICHANNEL
       END INTERFACE
 

epochX/cudacpp/CODEGEN/PLUGIN/CUDACPP_SA_OUTPUT/madgraph/iolibs/template_files/gpu/fcheck_sa.f

@@ -63,7 +63,7 @@ PROGRAM FCHECK_SA
           GS(IEVT) = 1.2177157847767195 ! fixed G for aS=0.118 (hardcoded for now in check_sa.cc, fcheck_sa.f, runTest.cc)
         END DO
         CALL FBRIDGESEQUENCE_NOMULTICHANNEL(BRIDGE, MOMENTA, GS, ! TEMPORARY? disable multi-channel in fcheck.exe and fgcheck.exe #466
-     &    RNDHEL, RNDCOL, MES, SELHEL, SELCOL)
+     &    RNDHEL, RNDCOL, MES, SELHEL, SELCOL, .FALSE.) ! do not quit after computing helicities
         DO IEVT = 1, NEVT
 c         DO IEXTERNAL = 1, NEXTERNAL
 c           WRITE(6,*) 'MOMENTA', IEVT, IEXTERNAL,

epochX/cudacpp/CODEGEN/generateAndCompare.sh

@@ -331,6 +331,7 @@ function codeGenAndDiff()
             | awk -vdate="D:20240301000000+01'00'" '{print gensub("(^/ModDate\\().*(\\)>>endobj$)","\\1"date"\\2","g")}' \
             | awk -vdate="D:20240301000000+01'00'" '{print gensub("(^/CreationDate\\().*(\\)$)","\\1"date"\\2","g")}' \
             | awk -vid="0123456789abcdef0123456789abcdef" '{print gensub("(^/ID \\[<).*><.*(>\\]$)","\\1"id"><"id"\\2","g")}' \
+            | awk -vid="0123456789abcdef0123456789abcdef" '{print gensub("(^/ID \\[\\().*\\)\\(.*(\\)\\]$)","\\1"id")("id"\\2","g")}' \
             | awk -vdate="2024-03-01T00:00:00+01:00" '{print gensub("(<xmp:ModifyDate>).*(</xmp:ModifyDate>)","\\1"date"\\2","g")}' \
             | awk -vdate="2024-03-01T00:00:00+01:00" '{print gensub("(<xmp:CreateDate>).*(</xmp:CreateDate>)","\\1"date"\\2","g")}' \
             | awk -vuuid="'uuid=01234567-89ab-cdef-0123-456789abcdef'" '{print gensub("(xapMM:DocumentID=).*(/>$)","\\1"uuid"\\2","g")}' \

epochX/cudacpp/ee_mumu.mad/SubProcesses/Bridge.h

@@ -109,9 +109,9 @@ namespace mg5amcCpu
      * @param rndcol the pointer to the input random numbers for color selection
      * @param channelId the Feynman diagram to enhance in multi-channel mode if 1 to n (disable multi-channel if 0)
      * @param mes the pointer to the output matrix elements
-     * @param goodHelOnly quit after computing good helicities?
      * @param selhel the pointer to the output selected helicities
      * @param selcol the pointer to the output selected colors
+     * @param goodHelOnly quit after computing good helicities?
      */
     void gpu_sequence( const FORTRANFPTYPE* momenta,
                        const FORTRANFPTYPE* gs,