From fa9116947997da72a7b7b99a04e51f2e8eca5ab4 Mon Sep 17 00:00:00 2001 From: Gopesh Bhardwaj Date: Wed, 21 Aug 2024 11:54:30 +0530 Subject: [PATCH] Fixing missing counters for gfx900 (#1028) --- .../counters/xml/derived_counters.xml | 3 + .../counters/yaml/counter_defs.yaml | 70 +++++++++---------- 2 files changed, 38 insertions(+), 35 deletions(-) diff --git a/source/lib/rocprofiler-sdk/counters/xml/derived_counters.xml b/source/lib/rocprofiler-sdk/counters/xml/derived_counters.xml index e998120906..54fd722d74 100755 --- a/source/lib/rocprofiler-sdk/counters/xml/derived_counters.xml +++ b/source/lib/rocprofiler-sdk/counters/xml/derived_counters.xml @@ -132,6 +132,9 @@ + + + # EA1 diff --git a/source/lib/rocprofiler-sdk/counters/yaml/counter_defs.yaml b/source/lib/rocprofiler-sdk/counters/yaml/counter_defs.yaml index 9712eefe5f..1107ff45c2 100644 --- a/source/lib/rocprofiler-sdk/counters/yaml/counter_defs.yaml +++ b/source/lib/rocprofiler-sdk/counters/yaml/counter_defs.yaml @@ -220,7 +220,7 @@ FETCH_SIZE: expression: (TCC_MC_RDREQ_sum*32)/1024 gfx906: expression: (TCC_EA_RDREQ_32B_sum*32+(TCC_EA_RDREQ_sum-TCC_EA_RDREQ_32B_sum)*64+RDATA1_SIZE)/1024 - gfx908/gfx90a/gfx9: + gfx908/gfx90a/gfx9/gfx900: expression: (TCC_EA_RDREQ_32B_sum*32+(TCC_EA_RDREQ_sum-TCC_EA_RDREQ_32B_sum)*64)/1024 gfx942/gfx941/gfx940: expression: (TCC_EA0_RDREQ_32B_sum*32+(TCC_EA0_RDREQ_sum-TCC_EA0_RDREQ_32B_sum)*64)/1024 @@ -228,19 +228,19 @@ FETCH_SIZE: and any cache or memory effects taken into account. FetchSize: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: FETCH_SIZE description: The total kilobytes fetched from the video memory. This is measured with all extra fetches and any cache or memory effects taken into account. FlatLDSInsts: architectures: - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: SQ_INSTS_FLAT_LDS_ONLY/SQ_WAVES description: The average number of FLAT instructions that read or write to LDS executed per work item (affected by flow control). FlatVMemInsts: architectures: - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: (SQ_INSTS_FLAT-SQ_INSTS_FLAT_LDS_ONLY)/SQ_WAVES description: The average number of FLAT instructions that read from or write to the video memory executed per work item (affected by flow control). Includes FLAT instructions that read from or write to scratch. @@ -373,12 +373,12 @@ GL2C_WRREQ_STALL_max: description: Number of cycles a write request was stalled. Max over GL2C instances. GPUBusy: architectures: - gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx908/gfx8/gfx90a/gfx9: + gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: 100*GRBM_GUI_ACTIVE/GRBM_COUNT description: The percentage of time GPU was busy. GPU_UTIL: architectures: - gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx90a/gfx9: + gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx90a/gfx9/gfx900: expression: 100*GRBM_GUI_ACTIVE/GRBM_COUNT description: Percentage of the time that GUI is active # Block GRBM (Graphics Register Bus Manager Block) @@ -473,7 +473,7 @@ L2CacheHit: architectures: gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx1100/gfx1101: expression: 100*reduce(GL2C_HIT,sum)/(reduce(GL2C_HIT,sum)+reduce(GL2C_MISS,sum)) - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: 100*reduce(TCC_HIT,sum)/(reduce(TCC_HIT,sum)+reduce(TCC_MISS,sum)) description: 'The percentage of fetch, write, atomic, and other instructions that hit the data in L2 cache. Value range: 0% (no hit) to 100% (optimal).' @@ -491,13 +491,13 @@ LDSBankConflict: architectures: gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx1100/gfx1101: expression: 100*SQC_LDS_BANK_CONFLICT/SQC_LDS_IDX_ACTIVE - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: 100*SQ_LDS_BANK_CONFLICT/GRBM_GUI_ACTIVE/CU_NUM description: 'The percentage of GPUTime LDS is stalled by bank conflicts. Value range: 0% (optimal) to 100% (bad).' LDSInsts: architectures: - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: (SQ_INSTS_LDS-SQ_INSTS_FLAT_LDS_ONLY)/SQ_WAVES description: The average number of LDS read or LDS write instructions executed per work item (affected by flow control). Excludes FLAT instructions that read from or write to LDS. @@ -553,13 +553,13 @@ MemUnitStalled: architectures: gfx8: expression: 100*reduce(TCP_TCP_TA_DATA_STALL_CYCLES,max)/GRBM_GUI_ACTIVE/SE_NUM - gfx942/gfx941/gfx906/gfx940/gfx908/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx90a/gfx9/gfx900: expression: 100*TCP_TCP_TA_DATA_STALL_CYCLES_max/GRBM_GUI_ACTIVE/SE_NUM description: 'The percentage of GPUTime the memory unit is stalled. Try reducing the number or size of fetches and writes if possible. Value range: 0% (optimal) to 100% (bad).' MemWrites32B: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: WRITE_REQ_32B description: The total number of effective 32B write transactions to the memory MfmaFlops: @@ -599,7 +599,7 @@ RDATA1_SIZE: description: The total kilobytes fetched from the video memory. This is measured on EA1s. SALUBusy: architectures: - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: 100*SQ_INST_CYCLES_SALU*4/SIMD_NUM/GRBM_GUI_ACTIVE gfx942/gfx941/gfx940: expression: 100*reduce(SQ_INST_CYCLES_SALU,sum)*4/SIMD_NUM/reduce(GRBM_GUI_ACTIVE,sum) @@ -2031,7 +2031,7 @@ SQ_WAVES_SAVED: space). Returns one value per-SE (aggregates of SIMD values). SQ_WAVES_sum: architectures: - gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(SQ_WAVES,sum) description: Gives the total number of waves currently enqueued by the application during the collection timeframe (for dispatch profiling this is the timeframe of kernel execution, for agent profiling it @@ -2223,17 +2223,17 @@ TA_BUFFER_WRITE_WAVEFRONTS_sum: description: Number of buffer write wavefronts processed by TA. Sum over TA instances. TA_BUSY_avr: architectures: - gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(TA_TA_BUSY,avr) description: TA block is busy. Average over TA instances. TA_BUSY_max: architectures: - gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(TA_TA_BUSY,max) description: TA block is busy. Max over TA instances. TA_BUSY_min: architectures: - gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx906/gfx1100/gfx1101/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(TA_TA_BUSY,min) description: TA block is busy. Min over TA instances. TA_DATA_STALLED_BY_TC_CYCLES: @@ -2288,7 +2288,7 @@ TA_FLAT_READ_WAVEFRONTS: description: Number of flat opcode reads processed by the TA. TA_FLAT_READ_WAVEFRONTS_sum: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(TA_FLAT_READ_WAVEFRONTS,sum) description: Number of flat opcode reads processed by the TA. Sum over TA instances. TA_FLAT_STORE_WAVEFRONTS: @@ -2328,7 +2328,7 @@ TA_FLAT_WRITE_WAVEFRONTS: description: Number of flat opcode writes processed by the TA. TA_FLAT_WRITE_WAVEFRONTS_sum: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(TA_FLAT_WRITE_WAVEFRONTS,sum) description: Number of flat opcode writes processed by the TA. Sum over TA instances. TA_TA_BUSY: @@ -2808,7 +2808,7 @@ TCC_EA_RDREQ_32B: description: Number of 32-byte TCC/EA read requests TCC_EA_RDREQ_32B_sum: architectures: - gfx906/gfx908/gfx90a/gfx9: + gfx906/gfx908/gfx90a/gfx9/gfx900: expression: reduce(TCC_EA_RDREQ_32B,sum) description: Number of 32-byte TCC/EA read requests. Sum over TCC instances. TCC_EA_RDREQ_DRAM: @@ -2878,7 +2878,7 @@ TCC_EA_RDREQ_LEVEL_sum: Sum over TCC instances. TCC_EA_RDREQ_sum: architectures: - gfx906/gfx908/gfx90a/gfx9: + gfx906/gfx908/gfx90a/gfx9/gfx900: expression: reduce(TCC_EA_RDREQ,sum) description: Number of TCC/EA read requests (either 32-byte or 64-byte). Sum over TCC instances. TCC_EA_RD_UNCACHED_32B: @@ -2916,7 +2916,7 @@ TCC_EA_WRREQ_64B: description: Number of 64-byte transactions going (64-byte write or CMPSWAP) over the TC_EA_wrreq interface. TCC_EA_WRREQ_64B_sum: architectures: - gfx906/gfx908/gfx90a/gfx9: + gfx906/gfx908/gfx90a/gfx9/gfx900: expression: reduce(TCC_EA_WRREQ_64B,sum) description: Number of 64-byte transactions going (64-byte write or CMPSWAP) over the TC_EA_wrreq interface. Sum over TCC instances. @@ -2998,7 +2998,7 @@ TCC_EA_WRREQ_STALL_sum: description: Number of cycles a write request was stalled. Sum over TCC instances. TCC_EA_WRREQ_sum: architectures: - gfx906/gfx908/gfx90a/gfx9: + gfx906/gfx908/gfx90a/gfx9/gfx900: expression: reduce(TCC_EA_WRREQ,sum) description: Number of transactions (either 32-byte or 64-byte) going over the TC_EA_wrreq interface. Sum over TCC instances. @@ -3031,7 +3031,7 @@ TCC_HIT: description: Number of cache hits. TCC_HIT_sum: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(TCC_HIT,sum) description: Number of cache hits. Sum over TCC instances. TCC_INTERNAL_PROBE: @@ -3081,7 +3081,7 @@ TCC_MISS: description: Number of cache misses. UC reads count as misses. TCC_MISS_sum: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: reduce(TCC_MISS,sum) description: Number of cache misses. UC reads count as misses. Sum over TCC instances. TCC_NC_REQ: @@ -3269,7 +3269,7 @@ TCC_WRREQ_STALL_max: architectures: gfx8: expression: reduce(TCC_MC_WRREQ_STALL,max) - gfx906/gfx908/gfx90a/gfx9: + gfx906/gfx908/gfx90a/gfx9/gfx900: expression: reduce(TCC_EA_WRREQ_STALL,max) gfx942/gfx941/gfx940: expression: reduce(TCC_EA0_WRREQ_STALL,max) @@ -3625,12 +3625,12 @@ TCP_TCP_TA_DATA_STALL_CYCLES: description: TCP stalls TA data interface. Now Windowed. TCP_TCP_TA_DATA_STALL_CYCLES_max: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx90a/gfx9/gfx900: expression: reduce(TCP_TCP_TA_DATA_STALL_CYCLES,max) description: Maximum number of TCP stalls TA data interface. TCP_TCP_TA_DATA_STALL_CYCLES_sum: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx90a/gfx9/gfx900: expression: reduce(TCP_TCP_TA_DATA_STALL_CYCLES,sum) description: Total number of TCP stalls TA data interface. TCP_TCR_TCP_STALL_CYCLES: @@ -3965,7 +3965,7 @@ TcUtil: description: 'Unit: percent' VALUBusy: architectures: - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: 100*SQ_ACTIVE_INST_VALU*4/SIMD_NUM/GRBM_GUI_ACTIVE gfx942/gfx941/gfx940: expression: 100*reduce(SQ_ACTIVE_INST_VALU,sum)*4/SIMD_NUM/reduce(GRBM_GUI_ACTIVE,sum) @@ -3979,20 +3979,20 @@ VALUInsts: control). VALUUtilization: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: 100*SQ_THREAD_CYCLES_VALU/(SQ_ACTIVE_INST_VALU*MAX_WAVE_SIZE) description: 'The percentage of active vector ALU threads in a wave. A lower number can mean either more thread divergence in a wave or that the work-group size is not a multiple of 64. Value range: 0% (bad), 100% (ideal - no thread divergence).' VFetchInsts: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: (SQ_INSTS_VMEM_RD-TA_FLAT_READ_WAVEFRONTS_sum)/SQ_WAVES description: The average number of vector fetch instructions from the video memory executed per work-item (affected by flow control). Excludes FLAT instructions that fetch from video memory. VWriteInsts: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: (SQ_INSTS_VMEM_WR-TA_FLAT_WRITE_WAVEFRONTS_sum)/SQ_WAVES description: The average number of vector write instructions to the video memory executed per work-item (affected by flow control). Excludes FLAT instructions that write to video memory. @@ -4037,7 +4037,7 @@ WRITE_REQ_32B: expression: TCC_MC_WRREQ_sum gfx906: expression: (TCC_EA_WRREQ_sum-TCC_EA_WRREQ_64B_sum)+(TCC_EA1_WRREQ_sum-TCC_EA1_WRREQ_64B_sum)+(TCC_EA_WRREQ_64B_sum+TCC_EA1_WRREQ_64B_sum)*2 - gfx908/gfx90a/gfx9: + gfx908/gfx90a/gfx9/gfx900: expression: TCC_EA_WRREQ_64B_sum*2+(TCC_EA_WRREQ_sum-TCC_EA_WRREQ_64B_sum) gfx942/gfx941/gfx940: expression: TCC_EA0_WRREQ_64B_sum*2+(TCC_EA0_WRREQ_sum-TCC_EA0_WRREQ_64B_sum) @@ -4048,7 +4048,7 @@ WRITE_SIZE: expression: (TCC_MC_WRREQ_sum*32)/1024 gfx906: expression: ((TCC_EA_WRREQ_sum-TCC_EA_WRREQ_64B_sum)*32+TCC_EA_WRREQ_64B_sum*64+WDATA1_SIZE)/1024 - gfx908/gfx90a/gfx9: + gfx908/gfx90a/gfx9/gfx900: expression: ((TCC_EA_WRREQ_sum-TCC_EA_WRREQ_64B_sum)*32+TCC_EA_WRREQ_64B_sum*64)/1024 gfx942/gfx941/gfx940: expression: ((TCC_EA0_WRREQ_sum-TCC_EA0_WRREQ_64B_sum)*32+TCC_EA0_WRREQ_64B_sum*64)/1024 @@ -4086,7 +4086,7 @@ Wavefronts: description: Total wavefronts. WriteSize: architectures: - gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9: + gfx942/gfx941/gfx906/gfx940/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: WRITE_SIZE description: The total kilobytes written to the video memory. This is measured with all extra fetches and any cache or memory effects taken into account. @@ -4094,7 +4094,7 @@ WriteUnitStalled: architectures: gfx10/gfx1010/gfx1030/gfx1031/gfx11/gfx1032/gfx1102/gfx1100/gfx1101: expression: 100*GL2C_WRREQ_STALL_max/GRBM_GUI_ACTIVE - gfx906/gfx908/gfx8/gfx90a/gfx9: + gfx906/gfx908/gfx8/gfx90a/gfx9/gfx900: expression: 100*TCC_WRREQ_STALL_max/GRBM_GUI_ACTIVE description: 'The percentage of GPUTime the Write unit is stalled. Value range: 0% to 100% (bad).' sL1dCacheHitRate: