Skip to content

SIMD: replace raw AVX512 of floor/trunc/rint with universal intrinsics#20397

Merged
mattip merged 1 commit into
numpy:mainfrom
seiko2plus:replace_rest_unary
Jan 19, 2022
Merged

SIMD: replace raw AVX512 of floor/trunc/rint with universal intrinsics#20397
mattip merged 1 commit into
numpy:mainfrom
seiko2plus:replace_rest_unary

Conversation

@seiko2plus

@seiko2plus seiko2plus commented Nov 17, 2021 •
edited
Loading

Copy link
Copy Markdown
Member

merge after:

Benchmark

X86

CPU 
Architecture:                    x86_64
CPU op-mode(s):                  32-bit, 64-bit
Byte Order:                      Little Endian
Address sizes:                   46 bits physical, 48 bits virtual
CPU(s):                          4
On-line CPU(s) list:             0-3
Thread(s) per core:              2
Core(s) per socket:              2
Socket(s):                       1
NUMA node(s):                    1
Vendor ID:                       GenuineIntel
CPU family:                      6
Model:                           85
Model name:                      Intel(R) Xeon(R) Platinum 8124M CPU @ 3.00GHz
Stepping:                        4
CPU MHz:                         3410.808
BogoMIPS:                        5999.99
Hypervisor vendor:               KVM
Virtualization type:             full
L1d cache:                       64 KiB
L1i cache:                       64 KiB
L2 cache:                        2 MiB
L3 cache:                        24.8 MiB
NUMA node0 CPU(s):               0-3
Vulnerability Itlb multihit:     KVM: Mitigation: VMX unsupported
Vulnerability L1tf:              Mitigation; PTE Inversion
Vulnerability Mds:               Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Vulnerability Meltdown:          Mitigation; PTI
Vulnerability Spec store bypass: Vulnerable
Vulnerability Spectre v1:        Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:        Mitigation; Full generic retpoline, STIBP disabled, RSB filling
Vulnerability Srbds:             Not affected
Vulnerability Tsx async abort:   Vulnerable: Clear CPU buffers attempted, no microcode; SMT Host state unknown
Flags:                           fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant
                                 _tsc rep_good nopl xtopology nonstop_tsc cpuid aperfmperf tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt
                                 tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch invpcid_single pti fsgsbase tsc_adjust bmi1 hle avx2 smep b
                                 mi2 erms invpcid rtm mpx avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves ida arat
                                 pku ospke
OS 
Linux ip-172-31-32-40 5.11.0-1020-aws #21~20.04.2-Ubuntu SMP Fri Oct 1 13:03:59 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
Python 3.8.10
gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0

Benchmark

Before(AVX512f) vs After(SSE41) 
unset NPY_DISABLE_CPU_FEATURES
python runtests.py -n --bench-compare parent/main "rint|floor|trunc" -- --sort ratio
       before           after         ratio
     [8e1c0c8f]       [ca6f7772]
                      <replace_rest_unary>
-         165±1μs        151±0.3μs     0.92  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 2, 'd')
-       513±0.7μs        451±0.4μs     0.88  bench_ufunc.UFunc.time_ufunc_types('rint')
-         131±4μs          115±3μs     0.88  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'd')
-       131±0.1μs        115±0.5μs     0.88  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 4, 'f')
-       255±0.4μs        222±0.4μs     0.87  bench_ufunc.UFunc.time_ufunc_types('floor')
-       252±0.4μs        217±0.3μs     0.86  bench_ufunc.UFunc.time_ufunc_types('trunc')
-         135±2μs        114±0.3μs     0.84  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 2, 'f')
-         193±8μs          152±4μs     0.78  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 4, 'f')
-         149±5μs          115±2μs     0.77  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'd')
-         196±1μs        151±0.5μs     0.77  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 2, 'd')
-       149±0.4μs        114±0.6μs     0.76  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 4, 'f')
-         132±3μs         99.1±2μs     0.75  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 4, 'f')
-       151±0.9μs        113±0.7μs     0.75  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 2, 'f')
-         154±4μs         97.9±2μs     0.64  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 4, 'f')
-         124±2μs         75.6±1μs     0.61  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 2, 'f')
-         192±5μs          115±3μs     0.60  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'd')
-       191±0.2μs        114±0.5μs     0.60  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 4, 'f')
-       192±0.8μs        114±0.5μs     0.59  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 2, 'f')
-         146±2μs         76.9±2μs     0.53  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 2, 'f')
-         200±6μs         98.1±2μs     0.49  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 4, 'f')
-       123±0.1μs       49.2±0.4μs     0.40  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'f')
-         191±3μs         75.6±1μs     0.40  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 2, 'f')
-       145±0.3μs       49.4±0.4μs     0.34  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'f')
-       189±0.3μs       49.1±0.3μs     0.26  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'f')
SSE41 
export NPY_DISABLE_CPU_FEATURES="AVX512F AVX512_SKX"
python runtests.py -n --bench-compare parent/main "rint|floor|trunc" -- --sort ratio
       before           after         ratio
     [8e1c0c8f]       [ca6f7772]
                      <replace_rest_unary>
-       225±0.9μs        214±0.5μs     0.95  bench_ufunc.UFunc.time_ufunc_types('trunc')
-         242±5μs        218±0.6μs     0.90  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 2, 'd')
-         131±4μs          110±2μs     0.84  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'd')
-         130±1μs        109±0.2μs     0.84  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 4, 'f')
-         133±3μs        109±0.1μs     0.82  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 2, 'f')
-       185±0.7μs        145±0.3μs     0.78  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 2, 'd')
-         193±7μs          145±4μs     0.75  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 4, 'f')
-       147±0.5μs        109±0.6μs     0.74  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 4, 'f')
-         197±2μs        145±0.3μs     0.73  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 2, 'd')
-       148±0.9μs        108±0.4μs     0.73  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 2, 'f')
-         131±4μs         92.0±2μs     0.70  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 4, 'f')
-      21.6±0.2μs       14.6±0.8μs     0.68  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 1, 'f')
-         154±5μs         93.7±2μs     0.61  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 4, 'f')
-         183±5μs          110±2μs     0.60  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'd')
-         124±2μs       73.4±0.5μs     0.59  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 2, 'f')
-         191±5μs          110±3μs     0.57  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'd')
-       191±0.2μs        109±0.2μs     0.57  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 4, 'f')
-       192±0.3μs        109±0.2μs     0.57  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 2, 'f')
-         146±2μs       71.6±0.6μs     0.49  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 2, 'f')
-         201±6μs         93.7±2μs     0.47  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 4, 'f')
-       123±0.3μs       47.7±0.6μs     0.39  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'f')
-         190±2μs         73.2±2μs     0.38  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 2, 'f')
-       144±0.1μs      47.5±0.09μs     0.33  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'f')
-       189±0.1μs       47.8±0.6μs     0.25  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'f')

Power little-endian

CPU 
Architecture:                    ppc64le
Byte Order:                      Little Endian
CPU(s):                          8
On-line CPU(s) list:             0-7
Thread(s) per core:              1
Core(s) per socket:              1
Socket(s):                       8
NUMA node(s):                    1
Model:                           2.2 (pvr 004e 1202)
Model name:                      POWER9 (architected), altivec supported
L1d cache:                       256 KiB
L1i cache:                       256 KiB
NUMA node0 CPU(s):               0-7
Vulnerability L1tf:              Not affected
Vulnerability Meltdown:          Mitigation; RFI Flush
Vulnerability Spec store bypass: Mitigation; Kernel entry/exit barrier (eieio)
Vulnerability Spectre v1:        Mitigation; __user pointer sanitization
Vulnerability Spectre v2:        Vulnerable

processor   : 7
cpu     : POWER9 (architected), altivec supported
clock       : 2200.000000MHz
revision    : 2.2 (pvr 004e 1202)

timebase    : 512000000
platform    : pSeries
model       : IBM pSeries (emulated by qemu)
machine     : CHRP IBM pSeries (emulated by qemu)
MMU     : Radix
OS 
Linux e517009a912a 4.19.0-2-powerpc64le #1 SMP Debian 4.19.16-1 (2019-01-17) ppc64le ppc64le ppc64le GNU/Linux
gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0

Benchmark

baseline(VSX2) 
python runtests.py -n --bench-compare parent/main "rint|floor|trunc" -- --sort ratio
       before           after         ratio
     [8e1c0c8f]       [ca6f7772]
                      <replace_rest_unary>
+         142±1μs          153±1μs     1.08  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 4, 'd')
-      125±0.09μs        118±0.4μs     0.94  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 2, 'd')
-       147±0.3μs          119±1μs     0.81  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 2, 'f')
-       147±0.3μs          119±1μs     0.81  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 2, 'f')
-       147±0.6μs        119±0.6μs     0.81  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 4, 'f')
-       130±0.9μs          105±2μs     0.80  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 1, 'd')
-         149±2μs        119±0.5μs     0.80  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 4, 'f')
-         131±1μs          104±1μs     0.80  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 1, 'd')
-       145±0.9μs       116±0.07μs     0.80  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 4, 'f')
-     4.35±0.01ms      3.47±0.01ms     0.80  bench_ufunc.UFunc.time_ufunc_types('rint')
-       145±0.3μs        116±0.3μs     0.80  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 2, 'f')
-       145±0.3μs        116±0.4μs     0.80  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 2, 'f')
-       145±0.3μs        116±0.5μs     0.80  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 4, 'f')
-       125±0.3μs       93.8±0.7μs     0.75  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'd')
-       125±0.7μs         92.6±2μs     0.74  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'd')
-       126±0.4μs       77.6±0.3μs     0.62  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 1, 'd')
-       125±0.2μs       77.3±0.2μs     0.62  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 1, 'd')
-       147±0.7μs         78.6±1μs     0.53  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 4, 'f')
-       147±0.7μs       76.3±0.3μs     0.52  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 4, 'f')
-       145±0.7μs       71.4±0.4μs     0.49  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'f')
-       145±0.4μs       70.7±0.2μs     0.49  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'f')
-       123±0.2μs       58.6±0.3μs     0.47  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 1, 'd')
-       124±0.5μs       58.6±0.2μs     0.47  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 1, 'd')
-       146±0.3μs       67.1±0.3μs     0.46  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 1, 'f')
-       147±0.3μs       66.8±0.2μs     0.46  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 1, 'f')
-       145±0.3μs       65.8±0.5μs     0.45  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 1, 'f')
-       146±0.6μs       65.8±0.4μs     0.45  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 1, 'f')
-      145±0.09μs       33.6±0.2μs     0.23  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 1, 'f')
-       146±0.4μs      33.7±0.04μs     0.23  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 1, 'f')
-     1.76±0.05ms          154±2μs     0.09  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 4, 'd')
-     1.78±0.04ms          147±4μs     0.08  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 4, 'd')
-     1.76±0.05ms          144±3μs     0.08  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 4, 'd')
-     1.77±0.04ms          130±3μs     0.07  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 2, 'd')
-     1.73±0.01ms        119±0.9μs     0.07  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 2, 'f')
-     1.74±0.01ms        119±0.5μs     0.07  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 4, 'f')
-     1.73±0.02ms        116±0.5μs     0.07  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 4, 'f')
-     1.74±0.01ms        116±0.4μs     0.07  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 2, 'f')
-     1.77±0.04ms        117±0.6μs     0.07  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 2, 'd')
-     1.75±0.04ms          104±1μs     0.06  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 1, 'd')
-     1.75±0.04ms         91.3±2μs     0.05  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'd')
-     1.74±0.02ms         77.5±1μs     0.04  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 4, 'f')
-     1.74±0.03ms       77.6±0.2μs     0.04  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 1, 'd')
-     1.73±0.02ms       70.7±0.4μs     0.04  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'f')
-     1.73±0.01ms       66.7±0.3μs     0.04  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 1, 'f')
-     1.72±0.02ms       65.8±0.2μs     0.04  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 1, 'f')
-     1.75±0.03ms       58.5±0.2μs     0.03  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 1, 'd')
-     1.72±0.02ms      33.7±0.03μs     0.02  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 1, 'f')

AArch64

CPU 
Architecture:                    aarch64
CPU op-mode(s):                  32-bit, 64-bit
Byte Order:                      Little Endian
CPU(s):                          2
On-line CPU(s) list:             0,1
Thread(s) per core:              1
Core(s) per socket:              2
Socket(s):                       1
NUMA node(s):                    1
Vendor ID:                       ARM
Model:                           1
Model name:                      Neoverse-N1
Stepping:                        r3p1
BogoMIPS:                        243.75
L1d cache:                       128 KiB
L1i cache:                       128 KiB
L2 cache:                        2 MiB
L3 cache:                        32 MiB
NUMA node0 CPU(s):               0,1
Vulnerability Itlb multihit:     Not affected
Vulnerability L1tf:              Not affected
Vulnerability Mds:               Not affected
Vulnerability Meltdown:          Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:        Mitigation; __user pointer sanitization
Vulnerability Spectre v2:        Not affected
Vulnerability Srbds:             Not affected
Vulnerability Tsx async abort:   Not affected
Flags:                           fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm lrcpc dcpop asimddp ssbs
OS 
Linux ip-172-31-44-172 5.11.0-1020-aws #21~20.04.2-Ubuntu SMP Fri Oct 1 13:01:34 UTC 2021 aarch64 aarch64 aarch64 GNU/Linux
gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0

Benchmark

baseline(ASIMD) 
python runtests.py --bench-compare parent/main "rint|floor|trunc" -- --sort ratio
       before           after         ratio
     [8e1c0c8f]       [ca6f7772]
                      <replace_rest_unary>
+      97.6±0.5μs        108±0.7μs     1.11  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 4, 'f')
+      97.6±0.8μs        108±0.7μs     1.11  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 4, 'f')
+      97.6±0.6μs        107±0.9μs     1.10  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 4, 'f')
-      89.5±0.4μs       83.7±0.2μs     0.93  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 1, 'f')
-      89.1±0.4μs       82.0±0.7μs     0.92  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 4, 'f')
-      89.1±0.4μs       81.9±0.5μs     0.92  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 4, 'f')
-      89.0±0.4μs       81.7±0.6μs     0.92  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 4, 'f')
-        88.8±1μs       80.3±0.7μs     0.90  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'd')
-        88.7±1μs       80.0±0.8μs     0.90  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'd')
-        88.7±1μs       79.9±0.8μs     0.90  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'd')
-      90.8±0.6μs       81.3±0.8μs     0.90  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 1, 'd')
-      91.2±0.5μs         81.4±1μs     0.89  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 1, 'd')
-      90.7±0.7μs       81.0±0.9μs     0.89  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 1, 'd')
-       161±0.7μs          142±2μs     0.88  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 4, 1, 'd')
-         253±5μs        219±0.1μs     0.87  bench_ufunc.UFunc.time_ufunc_types('floor')
-       162±0.6μs          140±2μs     0.87  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 4, 1, 'd')
-         163±1μs          141±3μs     0.86  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 4, 1, 'd')
-         243±7μs        209±0.4μs     0.86  bench_ufunc.UFunc.time_ufunc_types('trunc')
-        89.1±1μs         75.1±1μs     0.84  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 4, 'f')
-        89.2±1μs       75.0±0.8μs     0.84  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 4, 'f')
-        89.3±1μs         74.8±1μs     0.84  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 4, 'f')
-      83.5±0.5μs       68.8±0.2μs     0.82  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 2, 'f')
-      83.4±0.5μs       68.6±0.2μs     0.82  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 2, 'f')
-      83.8±0.3μs       68.5±0.2μs     0.82  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 2, 'f')
-      82.7±0.3μs       55.3±0.3μs     0.67  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 2, 'f')
-     82.7±0.09μs       55.2±0.2μs     0.67  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 2, 'f')
-      82.6±0.3μs       55.0±0.3μs     0.67  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 2, 'f')
-      85.1±0.7μs       40.4±0.7μs     0.47  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 1, 'd')
-      85.2±0.4μs       40.3±0.6μs     0.47  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 1, 'd')
-      85.2±0.4μs       40.0±0.5μs     0.47  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 2, 1, 'f')
-      85.0±0.4μs       39.8±0.6μs     0.47  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 2, 1, 'f')
-      85.3±0.3μs       39.6±0.3μs     0.46  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 2, 1, 'f')
-      85.4±0.7μs       39.4±0.6μs     0.46  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 1, 'd')
-      82.5±0.2μs       24.1±0.8μs     0.29  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'rint'>, 1, 1, 'f')
-      82.5±0.3μs       24.0±0.4μs     0.29  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'trunc'>, 1, 1, 'f')
-      82.4±0.3μs       23.9±0.4μs     0.29  bench_ufunc_strides.Unary.time_ufunc(<ufunc 'floor'>, 1, 1, 'f')

Binary size(striped)

LIB Before(KB) After(KB) Diff(KB)
_multiarray_umath.cpython-38-x86_64-linux-gnu.so 4428 4444 16

@rgommers rgommers added the component: SIMD Issues in SIMD (fast instruction sets) code or machinery label Dec 14, 2021
@seiko2plus seiko2plus mentioned this pull request Jan 18, 2022
Merged
@seiko2plus seiko2plus force-pushed the replace_rest_unary branch 2 times, most recently from 3253810 to 81e8175 Compare January 19, 2022 16:59
@seiko2plus

seiko2plus commented Jan 19, 2022

Copy link
Copy Markdown
Member Author

ping @mattip

@mattip mattip merged commit 44ffb08 into numpy:main Jan 19, 2022
@mattip

mattip commented Jan 19, 2022

Copy link
Copy Markdown
Member

Thanks @seiko2plus

@seiko2plus seiko2plus deleted the replace_rest_unary branch January 20, 2022 00:01
@seiko2plus seiko2plus added the 56 - Needs Release Note. Needs an entry in doc/release/upcoming_changes label Feb 13, 2022
@seiko2plus seiko2plus mentioned this pull request Apr 28, 2022
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

56 - Needs Release Note. Needs an entry in doc/release/upcoming_changes component: SIMD Issues in SIMD (fast instruction sets) code or machinery

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@seiko2plus @mattip @rgommers