Increase pool size of the "leak" base allocator in proxy lib #1072

lplewa · 2025-02-04T12:45:18Z

This improves search time if ptr belongs to leak pool, during free

Description

Checklist

Code compiles without errors locally
All tests pass locally
CI workflows execute properly

CI workflows, not executed per PR (e.g. Nightly), execute properly
New tests added, especially if they will fail without my changes
Added/extended example(s) to cover this functionality
Extended the README/documentation
All newly added source files have a license
All newly added source files are referenced in CMake files
Logger (with debug/info/... messages) is used

github-actions · 2025-02-04T12:46:13Z

Compute Benchmarks run (with params: ):
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13135770149

github-actions · 2025-02-04T13:05:08Z

Compute Benchmarks run ():
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13135770149
Job status: success. Test status: success.

Summary

No diffs to calculate performance change

(result is better)

Performance change in benchmark groups

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:4 glibc	3389.920000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 os_provider	15424.700000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 proxy_pool<os_provider>	16799.700000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	19295.500000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1820.640000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 scalable_pool<os_provider>	266.953000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 umfProxy	309.542000 ns

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:1 glibc	1604.830000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 os_provider	1772.280000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 proxy_pool<os_provider>	2080.350000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2421.930000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	139.721000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 scalable_pool<os_provider>	383.964000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 umfProxy	195.268000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:4 glibc	2325.380000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 os_provider	16778.900000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 proxy_pool<os_provider>	18633.900000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	15183.900000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	2072.560000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 scalable_pool<os_provider>	197.450000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 umfProxy	227.053000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:1 glibc	1527.680000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 os_provider	1843.560000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 proxy_pool<os_provider>	2080.320000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2414.970000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	143.938000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 scalable_pool<os_provider>	372.519000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 umfProxy	351.568000 ns

Relative perf in group alloc/min (8): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 glibc	1058.490000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 glibc	211.060000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 jemalloc_pool<os_provider>	4252.960000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 jemalloc_pool<os_provider>	475.322000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 scalable_pool<os_provider>	2130.960000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 scalable_pool<os_provider>	6196.470000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 umfProxy	1854.730000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 umfProxy	1672.730000 ns

Relative perf in group multiple (22): cannot calculate

Benchmark	This PR	Relative perf	Change	-
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 glibc	37643.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 glibc	7672.170000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 glibc	297516.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 glibc	42071.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 proxy_pool<os_provider>	5029180.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 proxy_pool<os_provider>	1202700.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 os_provider	4903430.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 os_provider	1107540.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 disjoint_pool<os_provider>	8182400.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 disjoint_pool<os_provider>	1374090.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 jemalloc_pool<os_provider>	264275.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 jemalloc_pool<os_provider>	32841.700000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 jemalloc_pool<os_provider>	556949.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 jemalloc_pool<os_provider>	86416.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 scalable_pool<os_provider>	22014.200000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 scalable_pool<os_provider>	6956.080000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 scalable_pool<os_provider>	67516.900000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 scalable_pool<os_provider>	29087.300000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 umfProxy	49565.800000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 umfProxy	12141.800000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 umfProxy	80361.300000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 umfProxy	39279.200000 ns

LD_PRELOAD=/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/lib/libumf_proxy.so

Command:

/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/benchmark/umf-benchmark --benchmark_format=csv --benchmark_filter=glibc

github-actions · 2025-02-04T13:13:31Z

Compute Benchmarks run (with params: ):
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13136267326

github-actions · 2025-02-04T13:31:30Z

Compute Benchmarks run ():
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13136267326
Job status: success. Test status: success.

Summary

No diffs to calculate performance change

(result is better)

Performance change in benchmark groups

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:4 glibc	3360.610000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 os_provider	14509.700000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 proxy_pool<os_provider>	16840.900000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	19114.600000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1939.620000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 scalable_pool<os_provider>	236.910000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 umfProxy	240.517000 ns

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:1 glibc	1572.180000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 os_provider	1763.510000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 proxy_pool<os_provider>	1963.350000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2361.250000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	137.891000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 scalable_pool<os_provider>	359.950000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 umfProxy	159.579000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:4 glibc	2304.560000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 os_provider	16099.900000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 proxy_pool<os_provider>	17432.700000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	15049.300000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1907.210000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 scalable_pool<os_provider>	216.245000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 umfProxy	195.207000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:1 glibc	1528.900000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 os_provider	1772.380000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 proxy_pool<os_provider>	2056.550000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2436.270000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	143.196000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 scalable_pool<os_provider>	335.006000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 umfProxy	323.089000 ns

Relative perf in group alloc/min (8): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 glibc	1032.550000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 glibc	192.345000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 jemalloc_pool<os_provider>	3309.510000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 jemalloc_pool<os_provider>	401.405000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 scalable_pool<os_provider>	1750.050000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 scalable_pool<os_provider>	5727.320000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 umfProxy	1885.580000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 umfProxy	1609.190000 ns

Relative perf in group multiple (22): cannot calculate

Benchmark	This PR	Relative perf	Change	-
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 glibc	31044.500000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 glibc	7447.110000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 glibc	267261.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 glibc	40647.300000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 proxy_pool<os_provider>	4775130.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 proxy_pool<os_provider>	1090480.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 os_provider	4713090.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 os_provider	1031260.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 disjoint_pool<os_provider>	7420490.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 disjoint_pool<os_provider>	1160420.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 jemalloc_pool<os_provider>	272905.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 jemalloc_pool<os_provider>	32436.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 jemalloc_pool<os_provider>	526918.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 jemalloc_pool<os_provider>	73461.800000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 scalable_pool<os_provider>	19961.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 scalable_pool<os_provider>	7054.140000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 scalable_pool<os_provider>	69808.200000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 scalable_pool<os_provider>	25395.600000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 umfProxy	44031.200000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 umfProxy	11237.900000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 umfProxy	68327.900000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 umfProxy	29040.800000 ns

LD_PRELOAD=/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/lib/libumf_proxy.so

Command:

/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/benchmark/umf-benchmark --benchmark_format=csv --benchmark_filter=glibc

github-actions · 2025-02-04T17:59:33Z

Compute Benchmarks run (with params: ):
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13141814453

github-actions · 2025-02-04T18:37:43Z

Compute Benchmarks run ():
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13141814453
Job status: success. Test status: success.

Summary

No diffs to calculate performance change

(result is better)

Performance change in benchmark groups

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:4 glibc	3415.400000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 os_provider	13512.700000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 proxy_pool<os_provider>	16749.700000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	18540.400000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1915.890000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 scalable_pool<os_provider>	233.611000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 umfProxy	259.442000 ns

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:1 glibc	1572.310000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 os_provider	1740.660000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 proxy_pool<os_provider>	1984.160000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2325.140000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	137.510000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 scalable_pool<os_provider>	365.765000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 umfProxy	149.834000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:4 glibc	2320.920000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 os_provider	14848.700000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 proxy_pool<os_provider>	17998.400000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	15182.600000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1826.690000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 scalable_pool<os_provider>	211.096000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 umfProxy	190.532000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:1 glibc	1518.150000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 os_provider	1715.950000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 proxy_pool<os_provider>	2006.950000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2403.190000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	143.176000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 scalable_pool<os_provider>	333.199000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 umfProxy	323.467000 ns

Relative perf in group alloc/min (8): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 glibc	1040.470000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 glibc	186.562000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 jemalloc_pool<os_provider>	3462.690000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 jemalloc_pool<os_provider>	392.779000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 scalable_pool<os_provider>	1892.430000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 scalable_pool<os_provider>	5817.050000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 umfProxy	1944.220000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 umfProxy	1387.530000 ns

Relative perf in group multiple (22): cannot calculate

Benchmark	This PR	Relative perf	Change	-
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 glibc	31894.700000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 glibc	7477.160000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 glibc	264801.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 glibc	41033.200000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 proxy_pool<os_provider>	4809650.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 proxy_pool<os_provider>	1114730.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 os_provider	4718750.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 os_provider	1068180.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 disjoint_pool<os_provider>	7241490.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 disjoint_pool<os_provider>	1162360.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 jemalloc_pool<os_provider>	271685.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 jemalloc_pool<os_provider>	32373.900000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 jemalloc_pool<os_provider>	489028.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 jemalloc_pool<os_provider>	68649.900000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 scalable_pool<os_provider>	19957.400000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 scalable_pool<os_provider>	6836.670000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 scalable_pool<os_provider>	60931.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 scalable_pool<os_provider>	25522.600000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 umfProxy	43314.100000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 umfProxy	11291.300000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 umfProxy	66748.200000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 umfProxy	28720.300000 ns

LD_PRELOAD=/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/lib/libumf_proxy.so

Command:

/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/benchmark/umf-benchmark --benchmark_format=csv --benchmark_filter=glibc

github-actions · 2025-02-04T18:50:28Z

Compute Benchmarks run (with params: --save baseline):
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13142699560

github-actions · 2025-02-04T19:29:00Z

Compute Benchmarks run (--save baseline):
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13142699560
Job status: success. Test status: success.

Summary

No diffs to calculate performance change

(result is better)

Performance change in benchmark groups

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:4 glibc	3417.530000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 os_provider	14325.700000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 proxy_pool<os_provider>	16902.700000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	18907.500000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1848.060000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 scalable_pool<os_provider>	232.829000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 umfProxy	261.538000 ns

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:1 glibc	1577.800000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 os_provider	1745.930000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 proxy_pool<os_provider>	1996.650000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2385.000000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	138.235000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 scalable_pool<os_provider>	380.272000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 umfProxy	172.280000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:4 glibc	2307.000000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 os_provider	14699.000000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 proxy_pool<os_provider>	16955.300000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	15031.400000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1901.660000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 scalable_pool<os_provider>	233.105000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 umfProxy	204.080000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:1 glibc	1514.310000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 os_provider	1741.970000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 proxy_pool<os_provider>	2039.880000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2403.440000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	142.305000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 scalable_pool<os_provider>	341.130000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 umfProxy	340.153000 ns

Relative perf in group alloc/min (8): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 glibc	1049.890000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 glibc	195.003000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 jemalloc_pool<os_provider>	3494.730000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 jemalloc_pool<os_provider>	399.067000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 scalable_pool<os_provider>	2032.310000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 scalable_pool<os_provider>	5911.460000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 umfProxy	1859.530000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 umfProxy	1581.000000 ns

Relative perf in group multiple (22): cannot calculate

Benchmark	This PR	Relative perf	Change	-
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 glibc	32893.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 glibc	7490.530000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 glibc	269263.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 glibc	41097.100000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 proxy_pool<os_provider>	4822860.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 proxy_pool<os_provider>	1127250.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 os_provider	4745080.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 os_provider	1071210.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 disjoint_pool<os_provider>	7443620.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 disjoint_pool<os_provider>	1201270.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 jemalloc_pool<os_provider>	272759.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 jemalloc_pool<os_provider>	32097.700000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 jemalloc_pool<os_provider>	512669.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 jemalloc_pool<os_provider>	69462.400000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 scalable_pool<os_provider>	20108.900000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 scalable_pool<os_provider>	7034.520000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 scalable_pool<os_provider>	61849.500000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 scalable_pool<os_provider>	26390.100000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 umfProxy	44646.700000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 umfProxy	11637.500000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 umfProxy	72323.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 umfProxy	31281.400000 ns

LD_PRELOAD=/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/lib/libumf_proxy.so

Command:

/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/benchmark/umf-benchmark --benchmark_format=csv --benchmark_filter=glibc

github-actions · 2025-02-04T19:42:09Z

Compute Benchmarks run (with params: ):
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13143531847

github-actions · 2025-02-04T19:52:21Z

Compute Benchmarks run ():
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13143531847
Job status: success. Test status: success.

Summary

No diffs to calculate performance change

(result is better)

Performance change in benchmark groups

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:4 glibc	3342.760000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 os_provider	14613.400000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 proxy_pool<os_provider>	16723.100000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	18407.300000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1640.520000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 scalable_pool<os_provider>	219.742000 ns
alloc/size:10000/0/4096/iterations:200000/threads:4 umfProxy	245.111000 ns

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:1 glibc	1547.230000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 os_provider	1695.610000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 proxy_pool<os_provider>	1945.790000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2244.100000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	137.634000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 scalable_pool<os_provider>	357.313000 ns
alloc/size:10000/0/4096/iterations:200000/threads:1 umfProxy	150.143000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:4 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:4 glibc	2351.560000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 os_provider	15189.000000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 proxy_pool<os_provider>	17168.700000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	17721.700000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	1734.930000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 scalable_pool<os_provider>	191.313000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:4 umfProxy	190.998000 ns

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:1 (7): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:1 glibc	1507.040000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 os_provider	1693.860000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 proxy_pool<os_provider>	2013.750000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	2350.290000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	138.273000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 scalable_pool<os_provider>	314.076000 ns
alloc/size:10000/100000/4096/iterations:200000/threads:1 umfProxy	316.439000 ns

Relative perf in group alloc/min (8): cannot calculate

Benchmark	This PR	Relative perf	Change	-
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 glibc	1028.410000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 glibc	183.951000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 jemalloc_pool<os_provider>	3182.160000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 jemalloc_pool<os_provider>	387.302000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 scalable_pool<os_provider>	1877.420000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 scalable_pool<os_provider>	5699.870000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 umfProxy	1729.630000 ns
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 umfProxy	1901.530000 ns

Relative perf in group multiple (22): cannot calculate

Benchmark	This PR	Relative perf	Change	-
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 glibc	31225.100000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 glibc	7391.820000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 glibc	259024.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 glibc	41203.500000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 proxy_pool<os_provider>	4756000.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 proxy_pool<os_provider>	1089820.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 os_provider	4666710.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 os_provider	1036650.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 disjoint_pool<os_provider>	6995720.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 disjoint_pool<os_provider>	1133850.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 jemalloc_pool<os_provider>	276450.000000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 jemalloc_pool<os_provider>	32294.900000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 jemalloc_pool<os_provider>	479051.000000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 jemalloc_pool<os_provider>	65914.700000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 scalable_pool<os_provider>	19678.300000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 scalable_pool<os_provider>	7129.440000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 scalable_pool<os_provider>	57400.900000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 scalable_pool<os_provider>	24286.700000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 umfProxy	43506.400000 ns
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 umfProxy	11470.400000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 umfProxy	67908.500000 ns
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 umfProxy	27550.700000 ns

LD_PRELOAD=/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/lib/libumf_proxy.so

Command:

/home/runner/work/unified-memory-framework/unified-memory-framework/umf-repo/build/benchmark/umf-benchmark --benchmark_format=csv --benchmark_filter=glibc

bratpiorka · 2025-02-06T13:03:16Z

changed to draft: we will keep this PR as a draft for a while to see how our perf workflow (compute bench) works

This improves search time if ptr belongs to leak pool, during free

github-actions · 2025-02-10T16:31:37Z

Compute Benchmarks run (with params: ):
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13245673775

github-actions · 2025-02-10T16:36:24Z

Compute Benchmarks run ():
https://github.com/oneapi-src/unified-memory-framework/actions/runs/13245673775
Job status: success. Test status: success.

Summary

Total 58 benchmarks in mean.
Geomean 121.186%.
Improved 22 Regressed 11 (threshold 2.00%)

(result is better)

Performance change in benchmark groups

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:4 (7): 101.333%

Benchmark	This PR	baseline	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:4 os_provider	2051.760000 ns	2363.680 ns	115.20%	15.20%	.
alloc/size:10000/0/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	3403.530000 ns	3650.920 ns	107.27%	7.27%	.
alloc/size:10000/0/4096/iterations:200000/threads:4 scalable_pool<os_provider>	287.241000 ns	304.282 ns	105.93%	5.93%	.
alloc/size:10000/0/4096/iterations:200000/threads:4 glibc	2727.530 ns	2667.110000 ns	97.78%	-2.22%	.
alloc/size:10000/0/4096/iterations:200000/threads:4 umfProxy	134.035 ns	129.460000 ns	96.59%	-3.41%	.
alloc/size:10000/0/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	4653.740 ns	4471.720000 ns	96.09%	-3.91%	.
alloc/size:10000/0/4096/iterations:200000/threads:4 proxy_pool<os_provider>	3205.760 ns	2960.570000 ns	92.35%	-7.65%	.

Relative perf in group alloc/size:10000/0/4096/iterations:200000/threads:1 (7): 100.964%

Benchmark	This PR	baseline	Relative perf	Change	-
alloc/size:10000/0/4096/iterations:200000/threads:1 scalable_pool<os_provider>	210.923000 ns	221.392 ns	104.96%	4.96%	.
alloc/size:10000/0/4096/iterations:200000/threads:1 umfProxy	98.223200 ns	102.586 ns	104.44%	4.44%	.
alloc/size:10000/0/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	118.293000 ns	119.234 ns	100.80%	0.80%	.
alloc/size:10000/0/4096/iterations:200000/threads:1 proxy_pool<os_provider>	287.488000 ns	287.875 ns	100.13%	0.13%	.
alloc/size:10000/0/4096/iterations:200000/threads:1 glibc	713.650 ns	711.473000 ns	99.69%	-0.31%	.
alloc/size:10000/0/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	504.666 ns	497.277000 ns	98.54%	-1.46%	.
alloc/size:10000/0/4096/iterations:200000/threads:1 os_provider	191.964 ns	188.882000 ns	98.39%	-1.61%	.

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:4 (7): 99.656%

Benchmark	This PR	baseline	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:4 proxy_pool<os_provider>	3242.450000 ns	3407.210 ns	105.08%	5.08%	.
alloc/size:10000/100000/4096/iterations:200000/threads:4 jemalloc_pool<os_provider>	3363.900000 ns	3478.800 ns	103.42%	3.42%	.
alloc/size:10000/100000/4096/iterations:200000/threads:4 scalable_pool<os_provider>	257.309000 ns	263.513 ns	102.41%	2.41%	.
alloc/size:10000/100000/4096/iterations:200000/threads:4 os_provider	1932.780000 ns	1970.950 ns	101.97%	1.97%	.
alloc/size:10000/100000/4096/iterations:200000/threads:4 glibc	1229.450000 ns	1249.440 ns	101.63%	1.63%	.
alloc/size:10000/100000/4096/iterations:200000/threads:4 disjoint_pool<os_provider>	4628.400 ns	4553.370000 ns	98.38%	-1.62%	.
alloc/size:10000/100000/4096/iterations:200000/threads:4 umfProxy	134.734 ns	115.919000 ns	86.04%	-13.96%	.

Relative perf in group alloc/size:10000/100000/4096/iterations:200000/threads:1 (7): 100.874%

Benchmark	This PR	baseline	Relative perf	Change	-
alloc/size:10000/100000/4096/iterations:200000/threads:1 glibc	740.773000 ns	759.433 ns	102.52%	2.52%	.
alloc/size:10000/100000/4096/iterations:200000/threads:1 proxy_pool<os_provider>	307.607000 ns	310.745 ns	101.02%	1.02%	.
alloc/size:10000/100000/4096/iterations:200000/threads:1 jemalloc_pool<os_provider>	119.094000 ns	120.099 ns	100.84%	0.84%	.
alloc/size:10000/100000/4096/iterations:200000/threads:1 umfProxy	202.358000 ns	203.803 ns	100.71%	0.71%	.
alloc/size:10000/100000/4096/iterations:200000/threads:1 disjoint_pool<os_provider>	502.442000 ns	505.405 ns	100.59%	0.59%	.
alloc/size:10000/100000/4096/iterations:200000/threads:1 scalable_pool<os_provider>	208.617000 ns	209.745 ns	100.54%	0.54%	.
alloc/size:10000/100000/4096/iterations:200000/threads:1 os_provider	191.642 ns	191.474000 ns	99.91%	-0.09%	.

Relative perf in group alloc/min (8): 95.007%

Benchmark	This PR	baseline	Relative perf	Change	-
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 umfProxy	635.381000 ns	699.582 ns	110.10%	10.10%	.
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 glibc	176.579000 ns	181.747 ns	102.93%	2.93%	.
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 glibc	810.645000 ns	831.334 ns	102.55%	2.55%	.
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 jemalloc_pool<os_provider>	4104.410000 ns	4191.580 ns	102.12%	2.12%	.
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 scalable_pool<os_provider>	960.592000 ns	973.583 ns	101.35%	1.35%	.
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:1 jemalloc_pool<os_provider>	352.115 ns	351.107000 ns	99.71%	-0.29%	.
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 scalable_pool<os_provider>	992.345 ns	972.995000 ns	98.05%	-1.95%	.
alloc/min size:10000/max size:0/granularity:8/65536/8/iterations:200000/threads:4 umfProxy	946.110 ns	534.020000 ns	56.44%	-43.56%	.

Relative perf in group multiple (22): 167.579%

Benchmark	This PR	baseline	Relative perf	Change	-
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 umfProxy	118033.000000 ns	2755370.000 ns	2334.41%	2234.41%	++++++++++
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 umfProxy	132544.000000 ns	2706190.000 ns	2041.73%	1941.73%	+++++++++
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 umfProxy	672664.000000 ns	10867000.000 ns	1615.52%	1515.52%	+++++++
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 umfProxy	721505.000000 ns	8105340.000 ns	1123.39%	1023.39%	+++++
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 scalable_pool<os_provider>	15067.800000 ns	15738.100 ns	104.45%	4.45%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 glibc	4232.690000 ns	4333.260 ns	102.38%	2.38%	.
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 scalable_pool<os_provider>	74573.400000 ns	76235.600 ns	102.23%	2.23%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 os_provider	1193720.000000 ns	1218820.000 ns	102.10%	2.10%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 glibc	32318.700000 ns	32983.200 ns	102.06%	2.06%	.
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 jemalloc_pool<os_provider>	59435.700000 ns	60403.700 ns	101.63%	1.63%	.
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 jemalloc_pool<os_provider>	623351.000000 ns	630411.000 ns	101.13%	1.13%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 jemalloc_pool<os_provider>	24193.800000 ns	24394.200 ns	100.83%	0.83%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 os_provider	141016.000000 ns	141575.000 ns	100.40%	0.40%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 disjoint_pool<os_provider>	209432.000 ns	209175.000000 ns	99.88%	-0.12%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:1 proxy_pool<os_provider>	162317.000 ns	161788.000000 ns	99.67%	-0.33%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 disjoint_pool<os_provider>	1756840.000 ns	1739990.000000 ns	99.04%	-0.96%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 scalable_pool<os_provider>	46594.700 ns	46092.500000 ns	98.92%	-1.08%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 proxy_pool<os_provider>	1194040.000 ns	1168820.000000 ns	97.89%	-2.11%	.
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:4 glibc	140097.000 ns	137103.000000 ns	97.86%	-2.14%	.
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 glibc	31785.200 ns	31104.500000 ns	97.86%	-2.14%	.
multiple_malloc_free/size:10000/4096/iterations:2000/threads:4 jemalloc_pool<os_provider>	516974.000 ns	497415.000000 ns	96.22%	-3.78%	.
multiple_malloc_free/min size:10000/max size:8/granularity:65536/8/iterations:2000/threads:1 scalable_pool<os_provider>	26745.400 ns	25406.200000 ns	94.99%	-5.01%	.

ldorau approved these changes Feb 4, 2025

View reviewed changes

lplewa force-pushed the leak branch from 15c61b6 to 12f32c2 Compare February 4, 2025 13:12

lplewa mentioned this pull request Feb 4, 2025

proxylib Free Is Noticeably Slower Than Direct UMF Pool Under Multithreaded Workloads #1077

Open

KFilipek approved these changes Feb 6, 2025

View reviewed changes

bratpiorka marked this pull request as draft February 6, 2025 13:01

increase leak pool size in proxy lib

5b01c85

This improves search time if ptr belongs to leak pool, during free

lplewa force-pushed the leak branch from 12f32c2 to 5b01c85 Compare February 10, 2025 16:31

lplewa marked this pull request as ready for review February 10, 2025 16:41

ldorau merged commit f570e7e into oneapi-src:main Feb 13, 2025
76 checks passed

Increase pool size of the "leak" base allocator in proxy lib #1072

Increase pool size of the "leak" base allocator in proxy lib #1072

Uh oh!

Conversation

lplewa commented Feb 4, 2025

Description

Checklist

Uh oh!

github-actions bot commented Feb 4, 2025

Uh oh!

github-actions bot commented Feb 4, 2025

Summary

Performance change in benchmark groups

Details

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command:

Environment Variables:

Command: