Software Tuning, Performance Optimization & Platform Monitoring
Discussion regarding monitoring and software tuning methodologies, Performance Monitoring Unit (PMU) of Intel microprocessors, and platform updating.

SandyBrdige AVX instruction test

GHui
Novice
283 Views

I had test the following instructions on SNB-Server with intel xhpl_intel64.
    FP_COMP_OPS_EXE.SSE_FP_PACKED_DOUBLE
    FP_COMP_OPS_EXE.SSE_FP_SCALAR_SINGLE
    FP_COMP_OPS_EXE.SSE_FP_PACKED_SINGLE
    FP_COMP_OPS_EXE.SSE_FP_SCALAR_DOUBLE
    SIMD_FP_256.PACKED_SINGLE
    SIMD_FP_256.PACKED_DOUBLE
    Inst_retired.any

The result is following.

SIMD_FP_256:0                     INST_RETIRED.ANY:16658554         FP_PACKED:0                    FP_PACKED+FP_SCALAR:614
SIMD_FP_256:0                     INST_RETIRED.ANY:24346539         FP_PACKED:8                    FP_PACKED+FP_SCALAR:770
SIMD_FP_256:0                     INST_RETIRED.ANY:4118767277      FP_PACKED:0                     FP_PACKED+FP_SCALAR:626
SIMD_FP_256:40797767806 INST_RETIRED.ANY:91073484456    FP_PACKED:5357321       FP_PACKED+FP_SCALAR:5357951
SIMD_FP_256:36031781859 INST_RETIRED.ANY:93890133930    FP_PACKED:14608423      FP_PACKED+FP_SCALAR:14609205
SIMD_FP_256:38525006382 INST_RETIRED.ANY:94851761353    FP_PACKED:21183153      FP_PACKED+FP_SCALAR:21183781
SIMD_FP_256:36086436410 INST_RETIRED.ANY:102680984689   FP_PACKED:19713892      FP_PACKED+FP_SCALAR:19714544
SIMD_FP_256:0                     INST_RETIRED.ANY:133528668729   FP_PACKED:132581174     FP_PACKED+FP_SCALAR:132581970
SIMD_FP_256:40433241390 INST_RETIRED.ANY:87005333191    FP_PACKED:10917051      FP_PACKED+FP_SCALAR:10917697
SIMD_FP_256:36830644259 INST_RETIRED.ANY:86226280428    FP_PACKED:12                   FP_PACKED+FP_SCALAR:646
SIMD_FP_256:37879549749 INST_RETIRED.ANY:87789307282    FP_PACKED:10601926      FP_PACKED+FP_SCALAR:10602752
SIMD_FP_256:37249427972 INST_RETIRED.ANY:90346415515    FP_PACKED:12                   FP_PACKED+FP_SCALAR:640
SIMD_FP_256:42385389559 INST_RETIRED.ANY:76898675453    FP_PACKED:10284314      FP_PACKED+FP_SCALAR:10284320
SIMD_FP_256:40925758255 INST_RETIRED.ANY:85786887341    FP_PACKED:3237391       FP_PACKED+FP_SCALAR:3237397
SIMD_FP_256:39869990182 INST_RETIRED.ANY:90620577415    FP_PACKED:10066573      FP_PACKED+FP_SCALAR:10066573
SIMD_FP_256:42517130517 INST_RETIRED.ANY:81060569386    FP_PACKED:9847867       FP_PACKED+FP_SCALAR:9847867
SIMD_FP_256:40583935522 INST_RETIRED.ANY:85878911769    FP_PACKED:5323687       FP_PACKED+FP_SCALAR:5323687
SIMD_FP_256:40692009511 INST_RETIRED.ANY:84406387285    FP_PACKED:9498810       FP_PACKED+FP_SCALAR:9498810
SIMD_FP_256:39164008416 INST_RETIRED.ANY:85252947496    FP_PACKED:9367475       FP_PACKED+FP_SCALAR:9367475
SIMD_FP_256:39499350537 INST_RETIRED.ANY:86089126231    FP_PACKED:9166266       FP_PACKED+FP_SCALAR:9166266
SIMD_FP_256:39825993274 INST_RETIRED.ANY:82245906185    FP_PACKED:8976830       FP_PACKED+FP_SCALAR:8976830
SIMD_FP_256:38406313354 INST_RETIRED.ANY:86213786798    FP_PACKED:4447855       FP_PACKED+FP_SCALAR:4447855
SIMD_FP_256:42803223565 INST_RETIRED.ANY:83831133051    FP_PACKED:8607717       FP_PACKED+FP_SCALAR:8608351
SIMD_FP_256:41494618200 INST_RETIRED.ANY:81369409356    FP_PACKED:8442452       FP_PACKED+FP_SCALAR:8443082
SIMD_FP_256:40296801919 INST_RETIRED.ANY:90421461428    FP_PACKED:8203471       FP_PACKED+FP_SCALAR:8204099
SIMD_FP_256:37736776514 INST_RETIRED.ANY:87027057004    FP_PACKED:11281330      FP_PACKED+FP_SCALAR:11281968
SIMD_FP_256:38854705700 INST_RETIRED.ANY:90937293150    FP_PACKED:7778238       FP_PACKED+FP_SCALAR:7778876
SIMD_FP_256:39018361317 INST_RETIRED.ANY:83614878671    FP_PACKED:11267587      FP_PACKED+FP_SCALAR:11268223
SIMD_FP_256:41477070338 INST_RETIRED.ANY:81017007833    FP_PACKED:14511070      FP_PACKED+FP_SCALAR:14511704
SIMD_FP_256:39959344530 INST_RETIRED.ANY:88172943912    FP_PACKED:13984729      FP_PACKED+FP_SCALAR:13985375
SIMD_FP_256:38816509076 INST_RETIRED.ANY:86050129246    FP_PACKED:6666217       FP_PACKED+FP_SCALAR:6666863
SIMD_FP_256:37107723338 INST_RETIRED.ANY:92202567946    FP_PACKED:12664978      FP_PACKED+FP_SCALAR:12664984
SIMD_FP_256:39983175346 INST_RETIRED.ANY:89325837271    FP_PACKED:11935649      FP_PACKED+FP_SCALAR:11935649
SIMD_FP_256:40790214215 INST_RETIRED.ANY:93712065347    FP_PACKED:15191960      FP_PACKED+FP_SCALAR:15191960
SIMD_FP_256:36852066324 INST_RETIRED.ANY:94228733432    FP_PACKED:15031060      FP_PACKED+FP_SCALAR:15031060
SIMD_FP_256:41837394119 INST_RETIRED.ANY:97649941717    FP_PACKED:17691120      FP_PACKED+FP_SCALAR:17691120
SIMD_FP_256:33866605617 INST_RETIRED.ANY:101718456362   FP_PACKED:21546820      FP_PACKED+FP_SCALAR:21546820
SIMD_FP_256:8037518837  INST_RETIRED.ANY:132773466350   FP_PACKED:27519826      FP_PACKED+FP_SCALAR:27519826
SIMD_FP_256:0                    INST_RETIRED.ANY:110025119          FP_PACKED:1863575050    FP_PACKED+FP_SCALAR:1863575050
SIMD_FP_256:0                    INST_RETIRED.ANY:113987166          FP_PACKED:0                     FP_PACKED+FP_SCALAR:0
SIMD_FP_256:0                    INST_RETIRED.ANY:277501614          FP_PACKED:0                    FP_PACKED+FP_SCALAR:0
SIMD_FP_256:0                    INST_RETIRED.ANY:112435100          FP_PACKED:0                     FP_PACKED+FP_SCALAR:0
SIMD_FP_256:0                    INST_RETIRED.ANY:109972261          FP_PACKED:0                    FP_PACKED+FP_SCALAR:0
SIMD_FP_256:0                    INST_RETIRED.ANY:275336059          FP_PACKED:0                    FP_PACKED+FP_SCALAR:0
SIMD_FP_256:0                    INST_RETIRED.ANY:110758273          FP_PACKED:0                    FP_PACKED+FP_SCALAR:0

Note:

SIMD_FP_256   is short for SIMD_FP_256.PACKED_SINGLE+SIMD_FP_256.PACKED_DOUBLE
FP_PACKED     is short for FP_COMP_OPS_EXE.SSE_FP_PACKED_DOUBLE+FP_COMP_OPS_EXE.SSE_FP_PACKED_SINGLE
FP_PACKED+FP_SCALAR    is short for FP_COMP_OPS_EXE.SSE_FP_PACKED_DOUBLE+FP_COMP_OPS_EXE.SSE_FP_PACKED_SINGLE
                                                             +FP_COMP_OPS_EXE.SSE_FP_SCALAR_DOUBLE+FP_COMP_OPS_EXE.SSE_FP_SCALAR_SINGLE

I confused that total SIMD_FP_256 is so big, but total SSE_FP is so small.

0 Kudos
1 Reply
Patrick_F_Intel1
Employee
283 Views
Hello GHui, Is the test case something you can post? Pat
0 Kudos
Reply