Skip to content

Commit dcf3a99

Browse files
valassivalassi
committed
[amd] regenerate all processes with the workaround for HIP FPE #1011 
1 parent 13ebdbe commit dcf3a99

File tree

45 files changed

+258
-202
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

45 files changed

+258
-202
lines changed

‎epochX/cudacpp/ee_mumu.mad/CODEGEN_mad_ee_mumu_log.txt

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -58,7 +58,7 @@ generate e+ e- > mu+ mu-
5858
No model currently active, so we import the Standard Model
5959
INFO: load particles
6060
INFO: load vertices
61-
DEBUG: model prefixing takes 0.006308317184448242 
61+
DEBUG: model prefixing takes 0.006434440612792969 
6262
INFO: Restrict model sm with file models/sm/restrict_default.dat .
6363
DEBUG: Simplifying conditional expressions 
6464
DEBUG: remove interactions: u s w+ at order: QED=1 
@@ -150,7 +150,7 @@ INFO: Checking for minimal orders which gives processes.
150150
INFO: Please specify coupling orders to bypass this step.
151151
INFO: Trying process: e+ e- > mu+ mu- WEIGHTED<=4 @1
152152
INFO: Process has 2 diagrams
153-
1 processes with 2 diagrams generated in 0.005 s
153+
1 processes with 2 diagrams generated in 0.004 s
154154
Total: 1 processes with 2 diagrams
155155
output madevent_simd ../TMPOUT/CODEGEN_mad_ee_mumu --hel_recycling=False --vector_size=32
156156
Output will be done with PLUGIN: CUDACPP_OUTPUT
@@ -180,19 +180,19 @@ INFO: Finding symmetric diagrams for subprocess group epem_mupmum
180180
DEBUG: iconfig_to_diag =  {1: 1, 2: 2} [model_handling.py at line 1551] 
181181
DEBUG: diag_to_iconfig =  {1: 1, 2: 2} [model_handling.py at line 1552] 
182182
Generated helas calls for 1 subprocesses (2 diagrams) in 0.004 s
183-
Wrote files for 8 helas calls in 0.070 s
183+
Wrote files for 8 helas calls in 0.069 s
184184
DEBUG: self.vector_size =  32 [export_v4.py at line 7023] 
185185
ALOHA: aloha starts to compute helicity amplitudes
186186
ALOHA: aloha creates FFV1 routines
187187
ALOHA: aloha creates FFV2 routines
188188
ALOHA: aloha creates FFV4 routines
189-
ALOHA: aloha creates 3 routines in 0.199 s
189+
ALOHA: aloha creates 3 routines in 0.201 s
190190
ALOHA: aloha starts to compute helicity amplitudes
191191
ALOHA: aloha creates FFV1 routines
192192
ALOHA: aloha creates FFV2 routines
193193
ALOHA: aloha creates FFV4 routines
194194
ALOHA: aloha creates FFV2_4 routines
195-
ALOHA: aloha creates 7 routines in 0.252 s
195+
ALOHA: aloha creates 7 routines in 0.255 s
196196
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
197197
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
198198
<class 'aloha.create_aloha.AbstractRoutine'> FFV2
@@ -232,9 +232,9 @@ Type "launch" to generate events from this process, or see
232232
Run "open index.html" to see more information about this process.
233233
quit
234234

235-
real 0m2.042s
236-
user 0m1.792s
237-
sys 0m0.243s
235+
real 0m2.097s
236+
user 0m1.775s
237+
sys 0m0.272s
238238
Code generation completed in 2 seconds
239239
************************************************************
240240
* *

‎epochX/cudacpp/ee_mumu.mad/src/HelAmps_sm.h

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -478,7 +478,10 @@ namespace mg5amcCpu
478478
}
479479
else
480480
{
481-
const fptype emp = pvec0 / ( vmass * pp );
481+
//printf( "DEBUG1011 (before emp): pvec0=%f vmass=%f pp=%f vmass*pp=%f\n", pvec0, vmass, pp, vmass * pp );
482+
//const fptype emp = pvec / ( vmass * pp ); // this may give a FPE #1011 (why?! maybe when vmass=+-epsilon?)
483+
const fptype emp = pvec0 / vmass / pp; // workaround for FPE #1011
484+
//printf( "DEBUG1011 (after emp): emp=%f\n", emp );
482485
vc[2] = cxmake( hel0 * pp / vmass, 0. );
483486
vc[5] = cxmake( hel0 * pvec3 * emp + hel * pt / pp * sqh, 0. );
484487
if( pt != 0. )

‎epochX/cudacpp/ee_mumu.sa/CODEGEN_cudacpp_ee_mumu_log.txt

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -58,7 +58,7 @@ generate e+ e- > mu+ mu-
5858
No model currently active, so we import the Standard Model
5959
INFO: load particles
6060
INFO: load vertices
61-
DEBUG: model prefixing takes 0.006402015686035156 
61+
DEBUG: model prefixing takes 0.0062215328216552734 
6262
INFO: Restrict model sm with file models/sm/restrict_default.dat .
6363
DEBUG: Simplifying conditional expressions 
6464
DEBUG: remove interactions: u s w+ at order: QED=1 
@@ -150,7 +150,7 @@ INFO: Checking for minimal orders which gives processes.
150150
INFO: Please specify coupling orders to bypass this step.
151151
INFO: Trying process: e+ e- > mu+ mu- WEIGHTED<=4 @1
152152
INFO: Process has 2 diagrams
153-
1 processes with 2 diagrams generated in 0.004 s
153+
1 processes with 2 diagrams generated in 0.005 s
154154
Total: 1 processes with 2 diagrams
155155
output standalone_cudacpp ../TMPOUT/CODEGEN_cudacpp_ee_mumu
156156
Output will be done with PLUGIN: CUDACPP_OUTPUT
@@ -169,13 +169,13 @@ INFO: Creating files in directory /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TM
169169
FileWriter <class 'MG5aMC_PLUGIN.CUDACPP_OUTPUT.model_handling.PLUGIN_CPPWriter'> for /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_ee_mumu/SubProcesses/P1_Sigma_sm_epem_mupmum/./CPPProcess.h
170170
FileWriter <class 'MG5aMC_PLUGIN.CUDACPP_OUTPUT.model_handling.PLUGIN_CPPWriter'> for /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_ee_mumu/SubProcesses/P1_Sigma_sm_epem_mupmum/./CPPProcess.cc
171171
INFO: Created files CPPProcess.h and CPPProcess.cc in directory /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_ee_mumu/SubProcesses/P1_Sigma_sm_epem_mupmum/.
172-
Generated helas calls for 1 subprocesses (2 diagrams) in 0.004 s
172+
Generated helas calls for 1 subprocesses (2 diagrams) in 0.003 s
173173
ALOHA: aloha starts to compute helicity amplitudes
174174
ALOHA: aloha creates FFV1 routines
175175
ALOHA: aloha creates FFV2 routines
176176
ALOHA: aloha creates FFV4 routines
177177
ALOHA: aloha creates FFV2_4 routines
178-
ALOHA: aloha creates 4 routines in 0.265 s
178+
ALOHA: aloha creates 4 routines in 0.267 s
179179
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
180180
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
181181
<class 'aloha.create_aloha.AbstractRoutine'> FFV2
@@ -194,7 +194,7 @@ INFO: Created files Parameters_sm.h and Parameters_sm.cc in directory
194194
INFO: /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_ee_mumu/src/. and /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_ee_mumu/src/.
195195
quit
196196

197-
real 0m0.659s
198-
user 0m0.600s
199-
sys 0m0.042s
200-
Code generation completed in 0 seconds
197+
real 0m0.781s
198+
user 0m0.590s
199+
sys 0m0.053s
200+
Code generation completed in 1 seconds

‎epochX/cudacpp/ee_mumu.sa/src/HelAmps_sm.h

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -478,7 +478,10 @@ namespace mg5amcCpu
478478
}
479479
else
480480
{
481-
const fptype emp = pvec0 / ( vmass * pp );
481+
//printf( "DEBUG1011 (before emp): pvec0=%f vmass=%f pp=%f vmass*pp=%f\n", pvec0, vmass, pp, vmass * pp );
482+
//const fptype emp = pvec / ( vmass * pp ); // this may give a FPE #1011 (why?! maybe when vmass=+-epsilon?)
483+
const fptype emp = pvec0 / vmass / pp; // workaround for FPE #1011
484+
//printf( "DEBUG1011 (after emp): emp=%f\n", emp );
482485
vc[2] = cxmake( hel0 * pp / vmass, 0. );
483486
vc[5] = cxmake( hel0 * pvec3 * emp + hel * pt / pp * sqh, 0. );
484487
if( pt != 0. )

‎epochX/cudacpp/gg_tt.mad/CODEGEN_mad_gg_tt_log.txt

Lines changed: 6 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -58,7 +58,7 @@ generate g g > t t~
5858
No model currently active, so we import the Standard Model
5959
INFO: load particles
6060
INFO: load vertices
61-
DEBUG: model prefixing takes 0.006418943405151367 
61+
DEBUG: model prefixing takes 0.0059719085693359375 
6262
INFO: Restrict model sm with file models/sm/restrict_default.dat .
6363
DEBUG: Simplifying conditional expressions 
6464
DEBUG: remove interactions: u s w+ at order: QED=1 
@@ -181,12 +181,12 @@ INFO: Finding symmetric diagrams for subprocess group gg_ttx
181181
DEBUG: iconfig_to_diag =  {1: 1, 2: 2, 3: 3} [model_handling.py at line 1551] 
182182
DEBUG: diag_to_iconfig =  {1: 1, 2: 2, 3: 3} [model_handling.py at line 1552] 
183183
Generated helas calls for 1 subprocesses (3 diagrams) in 0.006 s
184-
Wrote files for 10 helas calls in 0.071 s
184+
Wrote files for 10 helas calls in 0.072 s
185185
DEBUG: self.vector_size =  32 [export_v4.py at line 7023] 
186186
ALOHA: aloha starts to compute helicity amplitudes
187187
ALOHA: aloha creates VVV1 set of routines with options: P0
188188
ALOHA: aloha creates FFV1 routines
189-
ALOHA: aloha creates 2 routines in 0.153 s
189+
ALOHA: aloha creates 2 routines in 0.150 s
190190
ALOHA: aloha starts to compute helicity amplitudes
191191
ALOHA: aloha creates VVV1 set of routines with options: P0
192192
ALOHA: aloha creates FFV1 routines
@@ -226,9 +226,9 @@ Type "launch" to generate events from this process, or see
226226
Run "open index.html" to see more information about this process.
227227
quit
228228

229-
real 0m1.900s
230-
user 0m1.626s
231-
sys 0m0.264s
229+
real 0m1.997s
230+
user 0m1.613s
231+
sys 0m0.278s
232232
Code generation completed in 2 seconds
233233
************************************************************
234234
* *

‎epochX/cudacpp/gg_tt.sa/CODEGEN_cudacpp_gg_tt_log.txt

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -58,7 +58,7 @@ generate g g > t t~
5858
No model currently active, so we import the Standard Model
5959
INFO: load particles
6060
INFO: load vertices
61-
DEBUG: model prefixing takes 0.006226539611816406 
61+
DEBUG: model prefixing takes 0.006254673004150391 
6262
INFO: Restrict model sm with file models/sm/restrict_default.dat .
6363
DEBUG: Simplifying conditional expressions 
6464
DEBUG: remove interactions: u s w+ at order: QED=1 
@@ -174,7 +174,7 @@ Generated helas calls for 1 subprocesses (3 diagrams) in 0.006 s
174174
ALOHA: aloha starts to compute helicity amplitudes
175175
ALOHA: aloha creates VVV1 set of routines with options: P0
176176
ALOHA: aloha creates FFV1 routines
177-
ALOHA: aloha creates 2 routines in 0.143 s
177+
ALOHA: aloha creates 2 routines in 0.144 s
178178
<class 'aloha.create_aloha.AbstractRoutine'> VVV1
179179
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
180180
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
@@ -189,7 +189,7 @@ INFO: Created files Parameters_sm.h and Parameters_sm.cc in directory
189189
INFO: /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_gg_tt/src/. and /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_gg_tt/src/.
190190
quit
191191

192-
real 0m0.559s
193-
user 0m0.480s
194-
sys 0m0.044s
192+
real 0m0.532s
193+
user 0m0.478s
194+
sys 0m0.045s
195195
Code generation completed in 0 seconds

‎epochX/cudacpp/gg_tt.sa/src/HelAmps_sm.h

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -478,7 +478,10 @@ namespace mg5amcCpu
478478
}
479479
else
480480
{
481-
const fptype emp = pvec0 / ( vmass * pp );
481+
//printf( "DEBUG1011 (before emp): pvec0=%f vmass=%f pp=%f vmass*pp=%f\n", pvec0, vmass, pp, vmass * pp );
482+
//const fptype emp = pvec / ( vmass * pp ); // this may give a FPE #1011 (why?! maybe when vmass=+-epsilon?)
483+
const fptype emp = pvec0 / vmass / pp; // workaround for FPE #1011
484+
//printf( "DEBUG1011 (after emp): emp=%f\n", emp );
482485
vc[2] = cxmake( hel0 * pp / vmass, 0. );
483486
vc[5] = cxmake( hel0 * pvec3 * emp + hel * pt / pp * sqh, 0. );
484487
if( pt != 0. )

‎epochX/cudacpp/gg_tt01g.mad/CODEGEN_mad_gg_tt01g_log.txt

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -58,7 +58,7 @@ generate g g > t t~
5858
No model currently active, so we import the Standard Model
5959
INFO: load particles
6060
INFO: load vertices
61-
DEBUG: model prefixing takes 0.006343364715576172 
61+
DEBUG: model prefixing takes 0.006289482116699219 
6262
INFO: Restrict model sm with file models/sm/restrict_default.dat .
6363
DEBUG: Simplifying conditional expressions 
6464
DEBUG: remove interactions: u s w+ at order: QED=1 
@@ -159,7 +159,7 @@ INFO: Please specify coupling orders to bypass this step.
159159
INFO: Trying coupling order WEIGHTED<=3: WEIGTHED IS QCD+2*QED
160160
INFO: Trying process: g g > t t~ g WEIGHTED<=3 @2
161161
INFO: Process has 16 diagrams
162-
1 processes with 16 diagrams generated in 0.020 s
162+
1 processes with 16 diagrams generated in 0.019 s
163163
Total: 2 processes with 19 diagrams
164164
output madevent_simd ../TMPOUT/CODEGEN_mad_gg_tt01g --hel_recycling=False --vector_size=32
165165
Output will be done with PLUGIN: CUDACPP_OUTPUT
@@ -201,23 +201,23 @@ INFO: Finding symmetric diagrams for subprocess group gg_ttx
201201
DEBUG: len(subproc_diagrams_for_config) =  3 [model_handling.py at line 1527] 
202202
DEBUG: iconfig_to_diag =  {1: 1, 2: 2, 3: 3} [model_handling.py at line 1551] 
203203
DEBUG: diag_to_iconfig =  {1: 1, 2: 2, 3: 3} [model_handling.py at line 1552] 
204-
Generated helas calls for 2 subprocesses (19 diagrams) in 0.045 s
205-
Wrote files for 46 helas calls in 0.191 s
204+
Generated helas calls for 2 subprocesses (19 diagrams) in 0.042 s
205+
Wrote files for 46 helas calls in 0.189 s
206206
DEBUG: self.vector_size =  32 [export_v4.py at line 7023] 
207207
ALOHA: aloha starts to compute helicity amplitudes
208208
ALOHA: aloha creates VVV1 routines
209209
ALOHA: aloha creates FFV1 routines
210210
ALOHA: aloha creates VVVV1 set of routines with options: P0
211211
ALOHA: aloha creates VVVV3 set of routines with options: P0
212212
ALOHA: aloha creates VVVV4 set of routines with options: P0
213-
ALOHA: aloha creates 5 routines in 0.326 s
213+
ALOHA: aloha creates 5 routines in 0.338 s
214214
ALOHA: aloha starts to compute helicity amplitudes
215215
ALOHA: aloha creates VVV1 routines
216216
ALOHA: aloha creates FFV1 routines
217217
ALOHA: aloha creates VVVV1 set of routines with options: P0
218218
ALOHA: aloha creates VVVV3 set of routines with options: P0
219219
ALOHA: aloha creates VVVV4 set of routines with options: P0
220-
ALOHA: aloha creates 10 routines in 0.306 s
220+
ALOHA: aloha creates 10 routines in 0.311 s
221221
<class 'aloha.create_aloha.AbstractRoutine'> VVV1
222222
<class 'aloha.create_aloha.AbstractRoutine'> VVV1
223223
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
@@ -265,10 +265,10 @@ Type "launch" to generate events from this process, or see
265265
Run "open index.html" to see more information about this process.
266266
quit
267267

268-
real 0m2.598s
269-
user 0m2.282s
270-
sys 0m0.314s
271-
Code generation completed in 3 seconds
268+
real 0m2.618s
269+
user 0m2.304s
270+
sys 0m0.310s
271+
Code generation completed in 2 seconds
272272
************************************************************
273273
* *
274274
* W E L C O M E to *

‎epochX/cudacpp/gg_tt01g.mad/src/HelAmps_sm.h

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -478,7 +478,10 @@ namespace mg5amcCpu
478478
}
479479
else
480480
{
481-
const fptype emp = pvec0 / ( vmass * pp );
481+
//printf( "DEBUG1011 (before emp): pvec0=%f vmass=%f pp=%f vmass*pp=%f\n", pvec0, vmass, pp, vmass * pp );
482+
//const fptype emp = pvec / ( vmass * pp ); // this may give a FPE #1011 (why?! maybe when vmass=+-epsilon?)
483+
const fptype emp = pvec0 / vmass / pp; // workaround for FPE #1011
484+
//printf( "DEBUG1011 (after emp): emp=%f\n", emp );
482485
vc[2] = cxmake( hel0 * pp / vmass, 0. );
483486
vc[5] = cxmake( hel0 * pvec3 * emp + hel * pt / pp * sqh, 0. );
484487
if( pt != 0. )

‎epochX/cudacpp/gg_ttg.mad/CODEGEN_mad_gg_ttg_log.txt

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -58,7 +58,7 @@ generate g g > t t~ g
5858
No model currently active, so we import the Standard Model
5959
INFO: load particles
6060
INFO: load vertices
61-
DEBUG: model prefixing takes 0.006256103515625 
61+
DEBUG: model prefixing takes 0.0062618255615234375 
6262
INFO: Restrict model sm with file models/sm/restrict_default.dat .
6363
DEBUG: Simplifying conditional expressions 
6464
DEBUG: remove interactions: u s w+ at order: QED=1 
@@ -151,7 +151,7 @@ INFO: Please specify coupling orders to bypass this step.
151151
INFO: Trying coupling order WEIGHTED<=3: WEIGTHED IS QCD+2*QED
152152
INFO: Trying process: g g > t t~ g WEIGHTED<=3 @1
153153
INFO: Process has 16 diagrams
154-
1 processes with 16 diagrams generated in 0.021 s
154+
1 processes with 16 diagrams generated in 0.022 s
155155
Total: 1 processes with 16 diagrams
156156
output madevent_simd ../TMPOUT/CODEGEN_mad_gg_ttg --hel_recycling=False --vector_size=32
157157
Output will be done with PLUGIN: CUDACPP_OUTPUT
@@ -180,23 +180,23 @@ INFO: Finding symmetric diagrams for subprocess group gg_ttxg
180180
DEBUG: len(subproc_diagrams_for_config) =  15 [model_handling.py at line 1527] 
181181
DEBUG: iconfig_to_diag =  {1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9, 10: 10, 11: 11, 12: 12, 13: 13, 14: 14, 15: 15} [model_handling.py at line 1551] 
182182
DEBUG: diag_to_iconfig =  {1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 6, 7: 7, 8: 8, 9: 9, 10: 10, 11: 11, 12: 12, 13: 13, 14: 14, 15: 15} [model_handling.py at line 1552] 
183-
Generated helas calls for 1 subprocesses (16 diagrams) in 0.037 s
184-
Wrote files for 36 helas calls in 0.120 s
183+
Generated helas calls for 1 subprocesses (16 diagrams) in 0.038 s
184+
Wrote files for 36 helas calls in 0.123 s
185185
DEBUG: self.vector_size =  32 [export_v4.py at line 7023] 
186186
ALOHA: aloha starts to compute helicity amplitudes
187187
ALOHA: aloha creates VVV1 routines
188188
ALOHA: aloha creates FFV1 routines
189189
ALOHA: aloha creates VVVV1 set of routines with options: P0
190190
ALOHA: aloha creates VVVV3 set of routines with options: P0
191191
ALOHA: aloha creates VVVV4 set of routines with options: P0
192-
ALOHA: aloha creates 5 routines in 0.362 s
192+
ALOHA: aloha creates 5 routines in 1.397 s
193193
ALOHA: aloha starts to compute helicity amplitudes
194194
ALOHA: aloha creates VVV1 routines
195195
ALOHA: aloha creates FFV1 routines
196196
ALOHA: aloha creates VVVV1 set of routines with options: P0
197197
ALOHA: aloha creates VVVV3 set of routines with options: P0
198198
ALOHA: aloha creates VVVV4 set of routines with options: P0
199-
ALOHA: aloha creates 10 routines in 0.311 s
199+
ALOHA: aloha creates 10 routines in 0.315 s
200200
<class 'aloha.create_aloha.AbstractRoutine'> VVV1
201201
<class 'aloha.create_aloha.AbstractRoutine'> VVV1
202202
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
@@ -237,10 +237,10 @@ Type "launch" to generate events from this process, or see
237237
Run "open index.html" to see more information about this process.
238238
quit
239239

240-
real 0m2.525s
241-
user 0m2.188s
242-
sys 0m0.271s
243-
Code generation completed in 3 seconds
240+
real 0m3.568s
241+
user 0m2.185s
242+
sys 0m0.276s
243+
Code generation completed in 4 seconds
244244
************************************************************
245245
* *
246246
* W E L C O M E to *

‎epochX/cudacpp/gg_ttg.mad/src/HelAmps_sm.h

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -478,7 +478,10 @@ namespace mg5amcCpu
478478
}
479479
else
480480
{
481-
const fptype emp = pvec0 / ( vmass * pp );
481+
//printf( "DEBUG1011 (before emp): pvec0=%f vmass=%f pp=%f vmass*pp=%f\n", pvec0, vmass, pp, vmass * pp );
482+
//const fptype emp = pvec / ( vmass * pp ); // this may give a FPE #1011 (why?! maybe when vmass=+-epsilon?)
483+
const fptype emp = pvec0 / vmass / pp; // workaround for FPE #1011
484+
//printf( "DEBUG1011 (after emp): emp=%f\n", emp );
482485
vc[2] = cxmake( hel0 * pp / vmass, 0. );
483486
vc[5] = cxmake( hel0 * pvec3 * emp + hel * pt / pp * sqh, 0. );
484487
if( pt != 0. )

‎epochX/cudacpp/gg_ttg.sa/CODEGEN_cudacpp_gg_ttg_log.txt

Lines changed: 5 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -58,7 +58,7 @@ generate g g > t t~ g
5858
No model currently active, so we import the Standard Model
5959
INFO: load particles
6060
INFO: load vertices
61-
DEBUG: model prefixing takes 0.006231069564819336 
61+
DEBUG: model prefixing takes 0.006242036819458008 
6262
INFO: Restrict model sm with file models/sm/restrict_default.dat .
6363
DEBUG: Simplifying conditional expressions 
6464
DEBUG: remove interactions: u s w+ at order: QED=1 
@@ -177,7 +177,7 @@ ALOHA: aloha creates FFV1 routines
177177
ALOHA: aloha creates VVVV1 set of routines with options: P0
178178
ALOHA: aloha creates VVVV3 set of routines with options: P0
179179
ALOHA: aloha creates VVVV4 set of routines with options: P0
180-
ALOHA: aloha creates 5 routines in 0.330 s
180+
ALOHA: aloha creates 5 routines in 0.326 s
181181
<class 'aloha.create_aloha.AbstractRoutine'> VVV1
182182
<class 'aloha.create_aloha.AbstractRoutine'> VVV1
183183
<class 'aloha.create_aloha.AbstractRoutine'> FFV1
@@ -197,7 +197,7 @@ INFO: Created files Parameters_sm.h and Parameters_sm.cc in directory
197197
INFO: /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_gg_ttg/src/. and /data/avalassi/GPU2023/madgraph4gpuX/MG5aMC/TMPOUT/CODEGEN_cudacpp_gg_ttg/src/.
198198
quit
199199

200-
real 0m0.806s
201-
user 0m0.706s
202-
sys 0m0.070s
200+
real 0m0.777s
201+
user 0m0.714s
202+
sys 0m0.058s
203203
Code generation completed in 1 seconds

‎epochX/cudacpp/gg_ttg.sa/src/HelAmps_sm.h

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -478,7 +478,10 @@ namespace mg5amcCpu
478478
}
479479
else
480480
{
481-
const fptype emp = pvec0 / ( vmass * pp );
481+
//printf( "DEBUG1011 (before emp): pvec0=%f vmass=%f pp=%f vmass*pp=%f\n", pvec0, vmass, pp, vmass * pp );
482+
//const fptype emp = pvec / ( vmass * pp ); // this may give a FPE #1011 (why?! maybe when vmass=+-epsilon?)
483+
const fptype emp = pvec0 / vmass / pp; // workaround for FPE #1011
484+
//printf( "DEBUG1011 (after emp): emp=%f\n", emp );
482485
vc[2] = cxmake( hel0 * pp / vmass, 0. );
483486
vc[5] = cxmake( hel0 * pvec3 * emp + hel * pt / pp * sqh, 0. );
484487
if( pt != 0. )

0 commit comments

Comments
 (0)