İşleme Grafiği

334 İşleme

Yazar SHA1 Mesaj Tarih
jatang e7b96b1e71 SWDEV-458392 - Add gfx12 support.
Change-Id: I91bb6a3329bf77f26005a345c18b63b86922028a
2024-06-17 13:24:48 -04:00
Giovanni LB 4d0fee6637 SWDEV-468021: Improving profiler performance
Change-Id: I91921160a19aaea1fedfa40d11b83114d6a6fc65
2024-06-14 15:29:27 -03:00
Ammar ELWazir a50d5ce4ae Fixing the force of using cmake python3 interpreter
Change-Id: I5d87dfa78052611856652fc887cdfa6764c51028
2024-06-10 16:00:09 -04:00
Saurabh Verma 056156135d SWDEV-457922: Temporary fix to prevent multiple instances of either rocprofiler v1 or v2 to be running, whether it's interception or standalone mode.
The fix works by using a pid file and a lock file, similar to how firefox prevents multiple instances.

Change-Id: I4974a80bf920eaf0376a53410573c1cc74ec17ac
2024-06-10 12:49:29 -04:00
AravindanC 59230206a3 SWDEV-460492 Update permissions, install target types of executables
Change-Id: I23828d4938ce35f2c27d86654f4e122f8886ec1d
2024-06-07 17:42:32 -04:00
Ammar ELWazir c312e9b0f1 Fixing rocsys includes
Change-Id: Ia19030a5ad4caebe7b6703721702a1776e97455b
2024-06-06 15:00:48 -05:00
gobhardw 1e69b3e2f6 SWDEV-465520: RHEL9 toolchain doesnt seem to support experimental filesystem
Change-Id: I6b1e7f42c49b1c7af412c0b68851724861c9970a
2024-06-05 12:50:48 -04:00
Giovanni LB 140f658820 SWDEV-464906: Swap read and stop packets in queue.cpp
Change-Id: I2dd33a71c0cbe69fe6c2cfaf523b9f8c6500093f
2024-06-05 09:11:59 -04:00
Ammar ELWazir 80dce3d672 Fixing Activities Trace Period
Change-Id: If37d8518392112cedc66ce616adaaad22c517d4a
2024-05-31 18:09:53 -05:00
Ammar ELWazir b8726a6225 Adding JSON & Major fixes
Adding Chrome Tracing JSON for V2
Some fixes for the Kernel Dispatch tracing
Changes related to the HIP Runtime Destruction fix

Change-Id: I2fe1f1ef66b415df17c503a4a9fadcfa107d8c49
2024-05-29 01:21:52 -04:00
Giovanni LB b354c54ed7 Setting output buffer to always local
Change-Id: I82bf9f4cb922221d4a2b7ae15adb0f208a0ba70e
2024-05-27 14:53:32 -04:00
Saurabh Verma 34606db8c3 Fix for crash Error: HsaRsrcFactory::SignalWait: signal_value(1), ret_value(-1)
Context::GetData() would crash when it tries to wait on the completion signal of the stop packet. The expected signal value after waiting in HsaRsrcFactory::SignalWait() is supposed to be 0 but sometimes it comes up as -1. The signal being checked has already decremented before. Profile::Finalize() was assigning the same completion signal to the read and stop packet. So those two packets have been assigned different completion signals. This fix only affects standalone profiling mode.

Change-Id: I4b16825019e58d95d70188a72b2cc5871e09dd81
2024-05-26 13:31:39 -04:00
Ammar ELWazir b6576393a2 Temporary fix for SWDEV-459354
Change-Id: I13d7d600ab72d873d1dbc5d5e9b72c2ca5941488
2024-05-24 10:39:32 -04:00
Ammar ELWazir 5d1320928c SWDEV-419012: Adding New Perfetto SDK & Fixing early destruction of perfetto plugin
Change-Id: Ie1d8e753f10a1ba0686b021c14b788b444499f14
2024-05-14 11:24:40 -04:00
Giovanni LB 0fc74490cd Reducing rocprofilerv2 memory usage
Change-Id: I13c2dec763818df38868593bba0223f01132bb7a
2024-05-09 18:57:22 -04:00
Giovanni LB 604738fd8d Fixing occupancy metrics for MI300
Change-Id: I25a150dbcd3582cd397a658aa9a991956577cfa8
2024-04-30 15:23:42 -04:00
Giovanni LB 4ef523c84a SWDEV-459693: Allowing runtime kernels for codeobjects in featureTests
Change-Id: I809abcefaa320a3dee6a9d9fc8843162e3e5930e
2024-04-30 13:11:25 -03:00
Giovanni LB 3ae1b6af74 SWDEV-452226: Changing ATT dispatch IDs to match previous kernel-trace change
Change-Id: Ifd00160f2e8705e9ef11d41d284f6625b70de389
2024-04-29 11:56:32 -04:00
gobhardw dc09052d82 SWDEV-452226-dispatch ids to start from 0 for v1 parity
Change-Id: Ib60a6c108131f67395ab1f9d35ad605859d39987
2024-04-23 00:41:55 -04:00
Giovanni LB 97c6b0b867 Re-enabling ATT tests
Change-Id: Ied23410d40eecbbfb8972d36ed065289f74aaa0c
2024-04-17 18:49:55 -03:00
Giovanni LB 97288fedc9 Disabling ATT Tests
Change-Id: I0e03337de967aedaa15600c8b848ea618376d8a2
2024-03-28 15:22:37 -04:00
Giovanni LB bf749ba8d6 Fixing XCC-based metrics
Change-Id: I7451355b5b5eeefbd1c78b996d4ca5c4b770246a
2024-03-27 16:17:24 -04:00
Giovanni LB 4dd21807c0 Updating to load_delta. Fixing perfetto plugin.
Change-Id: If893f84b0ff108cfa0ccdcf717ee8592aa621032
2024-03-07 15:21:37 -03:00
gobhardw 82905606d5 SWDEV-446917 ub24 compilation fix
Change-Id: I2740d8551c7cdee070eaa0e1a320d1a7a3e4c7e9
2024-03-07 16:44:52 +05:30
Giovanni LB 4711a083de SWDEV-445711: Workaround for race condition on exit
Change-Id: I7391637005ccc7ea89611f28691c02e569197dff
2024-02-27 14:45:47 -05:00
Giovanni LB fe00badc6e Adding late-replace envvar filenames to ATT
Change-Id: I5df51934ffa25358503013e1a230adb894d6ff94
2024-02-22 10:02:32 -05:00
Manjunath-Jakaraddi 8d705f8b96 SWDEV-445654: Adding metrics to MI300
Change-Id: I7c3d78d8739ad219fc05bd99d7fc11d1f023aadc
2024-02-14 13:17:15 -05:00
Giovanni LB 95b49ca593 Adding bank conflict and other MI300 missing metrics to rocprofv2
Change-Id: Ibb7fb6163aa3bd7742f199a03600c737720e3c61
2024-02-14 13:15:54 -05:00
gobhardw 28871e661a node id to be reported as gpu id
Change-Id: I237281b33cfb5bdad67b7d6e8b69e9a86d62ca5d
2024-02-12 12:26:41 +05:30
Giovanni LB 363abb238b Fixing SE numbering and cache invalidade parsing
Change-Id: Ie63c68a0b2b07427586faaf69b68a19c315387e3
2024-01-31 13:11:19 -03:00
Giovanni LB 067071f669 SWDEV-439757: Remove codeobj marker on last kernel. Fixed codeobj size limit. ATT version bump.
Change-Id: Ie0b9c4de6c37acaaad4ae7d8d774d234a3847b7c
2024-01-17 05:01:50 -03:00
Saurabh Verma 5676c27c02 Supporting UMC counters in rocprofv2
Change-Id: Ic96c33244d78ac5031cfd99ac2be0c12c774d1b0
2024-01-04 12:33:08 -05:00
Ammar ELWazir a59d56f9bc Fixing ROCProfilerV2 to match versioning changes in HIP Runtime
Change-Id: Ifb0f98356acebcb7cf5372340b36178de000405d
2024-01-03 12:28:08 -06:00
Giovanni LB 25a02fe03e SWDEV-437776: Fixing typo for in stitch.py. Fixing singleDispatch kernel name.
Change-Id: I8d65598fffa4f15f78a64f741fdcb9a4a77af2a0
2023-12-27 18:08:39 -05:00
Saurabh Verma b18bc9930b Fixed format of presenting MI300 counters for individual xccs
1. Modified rocprofiler_record_profiler_t to include xcc_index
2. Modified file plugin to add additional column XCC_Index to the csv

Change-Id: Ic526c739e16b50dca5704070ccabbcbecd994d6a
2023-12-20 14:04:10 -05:00
Giovanni LB 490ea4c37d SWDEV-432445: ATT - Adding fixes for symbolic stitch and source reference numbering
Change-Id: I31d63bd2500155697129c29d7e0bc857e48cad8e
2023-12-18 08:53:51 -03:00
Jatin Chaudhary 8dd9d9df51 SWDEV-412457 - include assert header in files that use it
Change-Id: Ie1a947e78c4badc5b07b15fefc8aaedacfc7a02c
2023-12-13 13:37:41 -05:00
Giovanni LB e85c6911aa SWDEV-432445: Reorganizing ATT packet start/stop. Bump ATT parser version.
Change-Id: I979648f8a6a25aa7b499b3eea49978578f6c3f5a
2023-12-12 17:43:34 -05:00
Giovanni LB edf93d48ab SWDEV-432445: ATT continuous mode update part2. Added codeobj tracking.
Change-Id: I1b58af70d221bbeb9b4cab960d26357a504045dd
2023-12-12 17:40:59 -05:00
Giovanni LB aba6a1d986 SWDEV-432445: ATT continuous mode
Change-Id: I52732fc1dba41df3373ef6a19a428b00d729bf04
2023-12-12 10:55:23 -03:00
Giovanni LB f13f221780 Att param as unsigned integer
Change-Id: I4cd611c5e60e7c22d17c657a8ea5663dfb57e781
2023-12-04 15:23:48 -05:00
Giovanni LB 99b14fc9f8 Added option to control how codeobj is dumped from ATT
Change-Id: Ie76aeea1193c7ba8fe7f51be159516f8a9eab55f
2023-11-08 17:48:28 -03:00
gobhardw 9954f937d0 Fixing build for opensuse
Change-Id: I7b68382224221e899f64c06cec67b969a212f21a
2023-11-03 17:04:08 +00:00
gobhardw 6ee20035dd rocporf binary update
Change-Id: I2ba07a213acc7db403b804d8136edce28df9dd14
2023-11-02 07:14:59 -04:00
Mythreya 46e02a9866 Add MI200/MI300 counters
Revision - Addition [Impact SoC: MI200, MI300]
    Note: this set of counters are important help understand the
          bottleneck.
    1.	TCC_TAG_STALL
        a.	Metric:  TCC_TAG_STALL/TCC_CYCLE: percentage of time TCC
            tag lookup pipeline is stalled
    2.	TCP_TCR_TCP_STALL_CYCLES
        a.	Metric: TCP_TCR_TCP_STALL_CYCLES/TCP_GATE_EN1: percentage
            of time TCP is stalled by TCR

Revision - Addition [Impact SoC: MI300]
    3.	TCC_BUBBLE:
        a.	Definition: Number of 128-byte read requests sent to EA
        b.	Revised Metric #1, TCC-EA Read BW:
        ReadBW =   128 * TCC_BUBBLE
                + 64 * (TCC_EA0_RDREQ - TCC_BUBBLE - TCC_EA0_RDREQ_32B)
                + 32 * TCC_EA0_RDREQ_32B
        c.	Revised Metric #2: TCC_EA Read Latency
        ReadLatency = TCC_EA0_RDREQ_LEVEL / (TCC_BUBBLE + TCC_EA0_RDREQ)

    /* [Fineprint] More detailed arithmetic:
    * ReadLatency = TCC_EA0_RDREQ_LEVEL / (#32B_req + #64B_req + #128B_req * 2)
    */

Change-Id: I0a2dfc1b64ca97023b1e8ba0f9830330b3034946
2023-10-30 15:38:46 -04:00
Ammar ELWazir 32907ff14c Fixing Tool/Plugin Versioning and adding --version and adding proper documentation for those features
Change-Id: Icdb0d64e02943e638f183ab31759a37736d2dd61
2023-10-27 14:52:13 +00:00
Lancelot Six 372fc3910b src/core/isa_capture/code_object_track.cpp: remove unused dbgapi import
The src/core/isa_capture/code_object_track.cpp file includes the
amd-dbgapi header but does not use the library (and is not configured to
use the library).

Remove the unused import.

Change-Id: Icc6c46a1f52355e9e5631395dfad1b44365bdca4
2023-10-27 00:09:21 -04:00
Mythreya 1fae494b12 Remove non-functional counters for MI200 and MI300
Counters removed for MI300 (gfx940)
    TCP_TCC_WRITE_REQ_HOLE_LATENCY
    TCP_TCC_WRITE_REQ_LATENCY
    TCP_TCC_READ_REQ_LATENCY
    TCP_TCP_LATENCY

Counters removed for MI200 and MI300 (gfx90a and gfx940 respectively)
    TA_BUFFER_COALESCABLE_WAVEFRONT
    TA_FLAT_COALESCABLE_WAVEFRONT
    TCC_EA0_WRREQ_IO_CREDIT_STALL
    TCC_EA0_WRREQ_GMI_CREDIT_STALL
    TCC_EA0_WRREQ_DRAM_CREDIT_STALL
    TCC_EA0_RDREQ_IO_CREDIT_STALL
    TCC_EA0_RDREQ_GMI_CREDIT_STALL
    TCC_EA0_RDREQ_DRAM_CREDIT_STALL

Change-Id: Ic3d1e7bf35495f35b1239f03ca6420e949421386
2023-10-26 12:50:57 -04:00
Ammar ELWazir c9b5d73ca8 Fixing Usage of session after free
If hsa_shut_down() is not called from the application then we may still have async calls back
to the rocprofiler to use session parameters, thats why we need to leak the session up till
this is fixed in the ROCR-Runtime

Change-Id: Ia8fcdaa859815d88a272296770d702968132e3e6
2023-10-19 23:29:39 -04:00
Benjamin Welton cb190b72f1 Migrate tracer performance fixes from roctracer
Change-Id: If9e1349537cce23b7be1d8530d795744a8cd07b1
2023-10-19 20:34:12 +00:00