HP EliteBook 820 G3 Intel(R) Core(TM) i5-6200U CPU Performance governor, running at 2.78 GHz gcc (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609 make CC=gcc CFLAGS="-march=native -mtune=native" tinymembench v0.4.9 (simple benchmark for memory throughput and latency) ========================================================================== == Memory bandwidth tests == == == == Note 1: 1MB = 1000000 bytes == == Note 2: Results for 'copy' tests show how many bytes can be == == copied per second (adding together read and writen == == bytes would have provided twice higher numbers) == == Note 3: 2-pass copy means that we are using a small temporary buffer == == to first fetch data into it, and only then write it to the == == destination (source -> L1 cache, L1 cache -> destination) == == Note 4: If sample standard deviation exceeds 0.1%, it is shown in == == brackets == ========================================================================== C copy backwards : 4485.8 MB/s (2.2%) C copy backwards (32 byte blocks) : 4514.7 MB/s (4.5%) C copy backwards (64 byte blocks) : 4479.1 MB/s (4.7%) C copy : 4481.8 MB/s (3.4%) C copy prefetched (32 bytes step) : 4197.6 MB/s (2.0%) C copy prefetched (64 bytes step) : 4447.7 MB/s (3.5%) C 2-pass copy : 4425.7 MB/s (2.8%) C 2-pass copy prefetched (32 bytes step) : 4394.5 MB/s (6.1%) C 2-pass copy prefetched (64 bytes step) : 4208.3 MB/s (5.4%) C fill : 6128.2 MB/s (8.7%) C fill (shuffle within 16 byte blocks) : 6514.0 MB/s (3.1%) C fill (shuffle within 32 byte blocks) : 6490.0 MB/s (13.7%) C fill (shuffle within 64 byte blocks) : 6477.0 MB/s (9.6%) --- standard memcpy : 6128.5 MB/s (7.0%) standard memset : 14595.1 MB/s (15.3%) --- MOVSB copy : 5683.4 MB/s (11.4%) MOVSD copy : 5708.3 MB/s (9.8%) SSE2 copy : 4465.5 MB/s (9.7%) SSE2 nontemporal copy : 6206.5 MB/s (8.9%) SSE2 copy prefetched (32 bytes step) : 4461.1 MB/s (4.9%) SSE2 copy prefetched (64 bytes step) : 4280.0 MB/s (4.6%) SSE2 nontemporal copy prefetched (32 bytes step) : 6030.0 MB/s (11.6%) SSE2 nontemporal copy prefetched (64 bytes step) : 6049.0 MB/s (8.0%) SSE2 2-pass copy : 4429.8 MB/s (2.1%) SSE2 2-pass copy prefetched (32 bytes step) : 4409.0 MB/s (3.2%) SSE2 2-pass copy prefetched (64 bytes step) : 4322.9 MB/s (12.8%) SSE2 2-pass nontemporal copy : 3770.9 MB/s (11.4%) SSE2 fill : 6000.5 MB/s (5.4%) SSE2 nontemporal fill : 12754.8 MB/s (1.6%) ========================================================================== == Memory latency test == == == == Average time is measured for random memory accesses in the buffers == == of different sizes. The larger is the buffer, the more significant == == are relative contributions of TLB, L1/L2 cache misses and SDRAM == == accesses. For extremely large buffer sizes we are expecting to see == == page table walk with several requests to SDRAM for almost every == == memory access (though 64MiB is not nearly large enough to experience == == this effect to its fullest). == == == == Note 1: All the numbers are representing extra time, which needs to == == be added to L1 cache latency. The cycle timings for L1 cache == == latency can be usually found in the processor documentation. == == Note 2: Dual random read means that we are simultaneously performing == == two independent memory accesses at a time. In the case if == == the memory subsystem can't handle multiple outstanding == == requests, dual random read has the same timings as two == == single reads performed one after another. == ========================================================================== block size : single random read / dual random read, [MADV_NOHUGEPAGE] 1024 : 0.0 ns / 0.0 ns 2048 : 0.0 ns / 0.0 ns 4096 : 0.0 ns / 0.0 ns 8192 : 0.0 ns / 0.1 ns 16384 : 0.0 ns / 0.1 ns 32768 : 0.1 ns / 0.2 ns 65536 : 1.4 ns / 2.2 ns 131072 : 2.2 ns / 2.8 ns 262144 : 3.3 ns / 4.5 ns 524288 : 8.6 ns / 11.3 ns 1048576 : 12.4 ns / 14.9 ns 2097152 : 18.5 ns / 20.5 ns 4194304 : 35.9 ns / 48.3 ns 8388608 : 55.4 ns / 72.4 ns 16777216 : 67.9 ns / 82.3 ns 33554432 : 74.6 ns / 88.4 ns 67108864 : 78.7 ns / 90.7 ns block size : single random read / dual random read, [MADV_HUGEPAGE] 1024 : 0.1 ns / 0.3 ns 2048 : 0.2 ns / 0.3 ns 4096 : 0.2 ns / 0.2 ns 8192 : 0.2 ns / 0.2 ns 16384 : 0.1 ns / 0.2 ns 32768 : 0.1 ns / 0.2 ns 65536 : 1.4 ns / 2.0 ns 131072 : 2.9 ns / 2.8 ns 262144 : 2.4 ns / 2.8 ns 524288 : 6.4 ns / 8.6 ns 1048576 : 8.8 ns / 10.6 ns 2097152 : 10.0 ns / 11.3 ns 4194304 : 27.4 ns / 40.3 ns 8388608 : 48.3 ns / 65.5 ns 16777216 : 58.5 ns / 73.9 ns 33554432 : 63.5 ns / 78.1 ns 67108864 : 66.7 ns / 79.3 ns tinymembench v0.4.9 (simple benchmark for memory throughput and latency) ========================================================================== == Memory bandwidth tests == == == == Note 1: 1MB = 1000000 bytes == == Note 2: Results for 'copy' tests show how many bytes can be == == copied per second (adding together read and writen == == bytes would have provided twice higher numbers) == == Note 3: 2-pass copy means that we are using a small temporary buffer == == to first fetch data into it, and only then write it to the == == destination (source -> L1 cache, L1 cache -> destination) == == Note 4: If sample standard deviation exceeds 0.1%, it is shown in == == brackets == ========================================================================== C copy backwards : 4476.8 MB/s (2.8%) C copy backwards (32 byte blocks) : 4503.4 MB/s (2.0%) C copy backwards (64 byte blocks) : 4455.2 MB/s (2.7%) C copy : 4490.0 MB/s (1.2%) C copy prefetched (32 bytes step) : 4417.5 MB/s (13.8%) C copy prefetched (64 bytes step) : 4417.6 MB/s (11.2%) C 2-pass copy : 4352.9 MB/s (1.8%) C 2-pass copy prefetched (32 bytes step) : 4401.7 MB/s (3.3%) C 2-pass copy prefetched (64 bytes step) : 4350.5 MB/s (2.9%) C fill : 6295.9 MB/s (12.2%) C fill (shuffle within 16 byte blocks) : 6388.9 MB/s (13.7%) C fill (shuffle within 32 byte blocks) : 6475.8 MB/s (8.7%) C fill (shuffle within 64 byte blocks) : 6369.0 MB/s (4.7%) --- standard memcpy : 5940.5 MB/s (10.4%) standard memset : 14587.2 MB/s (19.8%) --- MOVSB copy : 5622.8 MB/s (7.0%) MOVSD copy : 5739.1 MB/s (2.8%) SSE2 copy : 4477.9 MB/s (5.2%) SSE2 nontemporal copy : 6139.2 MB/s (1.2%) SSE2 copy prefetched (32 bytes step) : 4370.1 MB/s (2.3%) SSE2 copy prefetched (64 bytes step) : 4319.7 MB/s (10.1%) SSE2 nontemporal copy prefetched (32 bytes step) : 6068.6 MB/s (2.5%) SSE2 nontemporal copy prefetched (64 bytes step) : 6005.6 MB/s (13.8%) SSE2 2-pass copy : 4203.7 MB/s (2.2%) SSE2 2-pass copy prefetched (32 bytes step) : 4167.9 MB/s (5.8%) SSE2 2-pass copy prefetched (64 bytes step) : 4268.4 MB/s (4.4%) SSE2 2-pass nontemporal copy : 3856.5 MB/s (0.6%) SSE2 fill : 6408.3 MB/s (1.7%) SSE2 nontemporal fill : 14722.0 MB/s (0.5%) ========================================================================== == Memory latency test == == == == Average time is measured for random memory accesses in the buffers == == of different sizes. The larger is the buffer, the more significant == == are relative contributions of TLB, L1/L2 cache misses and SDRAM == == accesses. For extremely large buffer sizes we are expecting to see == == page table walk with several requests to SDRAM for almost every == == memory access (though 64MiB is not nearly large enough to experience == == this effect to its fullest). == == == == Note 1: All the numbers are representing extra time, which needs to == == be added to L1 cache latency. The cycle timings for L1 cache == == latency can be usually found in the processor documentation. == == Note 2: Dual random read means that we are simultaneously performing == == two independent memory accesses at a time. In the case if == == the memory subsystem can't handle multiple outstanding == == requests, dual random read has the same timings as two == == single reads performed one after another. == ========================================================================== block size : single random read / dual random read, [MADV_NOHUGEPAGE] 1024 : 0.0 ns / 0.0 ns 2048 : 0.0 ns / 0.0 ns 4096 : 0.0 ns / 0.0 ns 8192 : 0.0 ns / 0.0 ns 16384 : 0.0 ns / 0.0 ns 32768 : 0.0 ns / 0.0 ns 65536 : 1.3 ns / 1.8 ns 131072 : 1.9 ns / 2.3 ns 262144 : 3.4 ns / 4.6 ns 524288 : 8.1 ns / 10.5 ns 1048576 : 11.1 ns / 13.1 ns 2097152 : 14.4 ns / 16.6 ns 4194304 : 33.2 ns / 45.7 ns 8388608 : 55.0 ns / 72.1 ns 16777216 : 67.0 ns / 82.2 ns 33554432 : 74.3 ns / 87.1 ns 67108864 : 79.8 ns / 91.3 ns block size : single random read / dual random read, [MADV_HUGEPAGE] 1024 : 0.0 ns / 0.0 ns 2048 : 0.0 ns / 0.0 ns 4096 : 0.0 ns / 0.0 ns 8192 : 0.0 ns / 0.0 ns 16384 : 0.0 ns / 0.0 ns 32768 : 0.0 ns / 0.0 ns 65536 : 1.2 ns / 1.7 ns 131072 : 1.9 ns / 2.2 ns 262144 : 2.2 ns / 2.4 ns 524288 : 6.4 ns / 8.5 ns 1048576 : 8.6 ns / 10.3 ns 2097152 : 9.8 ns / 10.9 ns 4194304 : 28.7 ns / 41.6 ns 8388608 : 49.0 ns / 66.1 ns 16777216 : 58.7 ns / 74.1 ns 33554432 : 63.4 ns / 77.1 ns 67108864 : 65.6 ns / 78.3 ns