Ah, it probably segfaulted because we don't actually support bulldozer yet.
I imagine it did not detect as a bulldozer.
I have a few tickets to close for recent cpus before I can fix that. Thanks
for trying though. And I have access to that machine, so I can sort it out
later.
Bill.
On 17 February 2014 22:03, Jean-Pierre Flori <jpfl...@gmail.com> wrote:
> Here are result for a bulldozer cpu (combinat on Sage cluster).
> I removed x86_64/sqr_basecase.asm which was the assembly file used.
> Note that without it, the speed exe segfault...
>
>
> ################################################################################
> # With assembly file
>
> ################################################################################
> jpflori@combinat:~/build/mpir/tune$ ./speed -s 1-100 -t 1 mpn_sqr_basecase
> overhead 0.000000002 secs, precision 1000000 units of 4.35e-10 secs, CPU
> freq 2300.19 MHz
> mpn_sqr_basecase
> 1 0.000000005
> 2 0.000000009
> 3 0.000000014
> 4 0.000000028
> 5 0.000000038
> 6 0.000000052
> 7 0.000000065
> 8 0.000000082
> 9 0.000000098
> 10 0.000000118
> 11 0.000000140
> 12 0.000000160
> 13 0.000000185
> 14 0.000000216
> 15 0.000000233
> 16 0.000000285
> 17 0.000000311
> 18 0.000000339
> 19 0.000000376
> 20 0.000000427
> 21 0.000000457
> 22 0.000000507
> 23 0.000000550
> 24 0.000000594
> 25 0.000000620
> 26 0.000000704
> 27 0.000000764
> 28 0.000000816
> 29 0.000000890
> 30 0.000000962
> 31 0.000000993
> 32 0.000001096
> 33 0.000001119
> 34 0.000001199
> 35 0.000001262
> 36 0.000001330
> 37 0.000001413
> 38 0.000001521
> 39 0.000001572
> 40 0.000001620
> 41 0.000001731
> 42 0.000001827
> 43 0.000001891
> 44 0.000001972
> 45 0.000002099
> 46 0.000002192
> 47 0.000002264
> 48 0.000002393
> 49 0.000002494
> 50 0.000002572
> 51 0.000002682
> 52 0.000002784
> 53 0.000002873
> 54 0.000002950
> 55 0.000003086
> 56 0.000003208
> 57 0.000003314
> 58 0.000003470
> 59 0.000003597
> 60 0.000003734
> 61 0.000003824
> 62 0.000003925
> 63 0.000004042
> 64 0.000004244
> 65 0.000004308
> 66 0.000004404
> 67 0.000004577
> 68 0.000004737
> 69 0.000004907
> 70 0.000004991
> 71 0.000005139
> 72 0.000005274
> 73 0.000005442
> 74 0.000005549
> 75 0.000005704
> 76 0.000005921
> 77 0.000006020
> 78 0.000006226
> 79 0.000006342
> 80 0.000006742
> 81 0.000006855
> 82 0.000007019
> 83 0.000007181
> 84 0.000007380
> 85 0.000007529
> 86 0.000007674
> 87 0.000007733
> 88 0.000008123
> 89 0.000008130
> 90 0.000008421
> 91 0.000008468
> 92 0.000008748
> 93 0.000008864
> 94 0.000009011
> 95 0.000009163
> 96 0.000009462
> 97 0.000009550
> 98 0.000009791
> 99 0.000009950
> 100 0.000010140
>
> ./tuneup
> Parameters for ./mpn/x86_64/k8/k10/k102/gmp-mparam.h
> Using: CPU cycle counter, supplemented by microsecond getrusage()
> speed_precision 1000000, speed_unittime 4.35e-10 secs, CPU freq 2300.19 MHz
> DEFAULT_MAX_SIZE 1000, fft_max_size 50000
>
> /* Generated by tuneup.c, 2014-02-17, gcc 4.6 */
>
> #define MUL_KARATSUBA_THRESHOLD 14
> #define MUL_TOOM3_THRESHOLD 89
> #define MUL_TOOM4_THRESHOLD 238
> #define MUL_TOOM8H_THRESHOLD 351
>
> #define SQR_BASECASE_THRESHOLD 0 /* always (native) */
> #define SQR_KARATSUBA_THRESHOLD 20
> #define SQR_TOOM3_THRESHOLD 122
> #define SQR_TOOM4_THRESHOLD 680
> #define SQR_TOOM8_THRESHOLD 680
>
> #define POWM_THRESHOLD 128
>
> #define HGCD_THRESHOLD 91
> #define GCD_DC_THRESHOLD 418
> #define GCDEXT_DC_THRESHOLD 351
> #define JACOBI_BASE_METHOD 1
>
> #define DIVREM_1_NORM_THRESHOLD MP_SIZE_T_MAX /* never */
> #define DIVREM_1_UNNORM_THRESHOLD MP_SIZE_T_MAX /* never */
> #define MOD_1_NORM_THRESHOLD 0 /* always */
> #define MOD_1_UNNORM_THRESHOLD 0 /* always */
> #define USE_PREINV_DIVREM_1 1 /* native */
> #define USE_PREINV_MOD_1 1
> #define DIVEXACT_1_THRESHOLD 0 /* always */
> #define MODEXACT_1_ODD_THRESHOLD 0 /* always (native) */
> #define MOD_1_1_THRESHOLD 7
> #define MOD_1_2_THRESHOLD 7
> #define MOD_1_3_THRESHOLD 20
> #define DIVREM_HENSEL_QR_1_THRESHOLD 13
> #define RSH_DIVREM_HENSEL_QR_1_THRESHOLD 13
> #define DIVREM_EUCLID_HENSEL_THRESHOLD 24
>
> #define ROOTREM_THRESHOLD 6
>
> #define GET_STR_DC_THRESHOLD 12
> #define GET_STR_PRECOMPUTE_THRESHOLD 19
> #define SET_STR_DC_THRESHOLD 454
> #define SET_STR_PRECOMPUTE_THRESHOLD 538
>
> #define MUL_FFT_FULL_THRESHOLD 3904
>
> #define SQR_FFT_FULL_THRESHOLD 3392
>
> #define MULLOW_BASECASE_THRESHOLD 0 /* always */
> #define MULLOW_DC_THRESHOLD 33
> #define MULLOW_MUL_THRESHOLD 5565
>
> #define MULHIGH_BASECASE_THRESHOLD 7
> #define MULHIGH_DC_THRESHOLD 29
> #define MULHIGH_MUL_THRESHOLD 3336
>
> #define MULMOD_2EXPM1_THRESHOLD 14
>
> #define FAC_UI_THRESHOLD 8671
> #define DC_DIV_QR_THRESHOLD 52
> #define DC_DIVAPPR_Q_N_THRESHOLD 12
> #define INV_DIV_QR_THRESHOLD 2130
> #define INV_DIVAPPR_Q_N_THRESHOLD 12
> #define DC_DIV_Q_THRESHOLD 122
> #define INV_DIV_Q_THRESHOLD 6039
> #define DC_DIVAPPR_Q_THRESHOLD 41
> #define INV_DIVAPPR_Q_THRESHOLD 16039
> #define DC_BDIV_QR_THRESHOLD 54
> #define DC_BDIV_Q_THRESHOLD 35
>
> /* fft_tuning -- autogenerated by tune-fft */
>
> #define FFT_TAB \
> { { 4, 3 }, { 3, 3 }, { 3, 2 }, { 2, 1 }, { 1, 0 } }
>
> #define MULMOD_TAB \
> { 4, 4, 4, 4, 4, 3, 3, 3, 3, 2, 2, 3, 2, 2, 2, 2, 2, 1, 1 }
>
> #define FFT_N_NUM 19
>
> #define FFT_MULMOD_2EXPP1_CUTOFF 256
>
>
> /* Tuneup completed successfully, took 166 seconds */
>
>
>
> ################################################################################
> # Without assembly file
>
> ################################################################################
>
> jpflori@combinat:~/build/mpir/tune$ ./speed -s 1-100 -t 1 mpn_sqr_basecase
> overhead 0.000000002 secs, precision 1000000 units of 4.35e-10 secs, CPU
> freq 2300.19 MHz
> mpn_sqr_basecase
> 1 0.000000008
> 2 0.000000016
> 3 0.000000046
> 4 0.000000061
> 5 0.000000079
> 6 0.000000097
> 7 0.000000125
> 8 0.000000144
> 9 0.000000171
> 10 0.000000189
> 11 0.000000237
> 12 0.000000262
> 13 0.000000293
> 14 0.000000335
> 15 0.000000368
> 16 0.000000395
> 17 0.000000468
> 18 0.000000487
> 19 0.000000545
> 20 0.000000580
> 21 0.000000663
> 22 0.000000684
> 23 0.000000741
> 24 0.000000805
> 25 0.000000893
> 26 0.000000888
> 27 0.000001022
> 28 0.000001021
> 29 0.000001138
> 30 0.000001152
> 31 0.000001260
> Segmentation fault (core dumped)
>
>
--
You received this message because you are subscribed to the Google Groups
"mpir-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to mpir-devel+unsubscr...@googlegroups.com.
To post to this group, send email to mpir-devel@googlegroups.com.
Visit this group at http://groups.google.com/group/mpir-devel.
For more options, visit https://groups.google.com/groups/opt_out.
