Skip to content

proj_dis_perf_improvements_1

Jump to bottom
Tsukasa OI edited this page Nov 6, 2022 · 33 revisions

Project: Disassembler Performance Improvements (Optimization 1)

Benchmarking System

This benchmark is performed on:

  • Ubuntu 22.04 LTS
  • AMD Ryzen 5 PRO 5650G processor.

In the parallel run, I ran 6 parallel jobs with -j6 (corresponding 6 cores; although the processor has 12 hardware threads, -j12 just slowed the benchmark).

Aggregate Performance Improvements

Type Expected Improvements (in general)
Binary Files 50-130%
RISC-V ELF Programs 30-40%
RISC-V ELF Libraries 30-500%

This is relative to the latest master (commit beaad4923538) and taken on 2022-11-02.

Columns

objdump -d (ELF)

Program Prerequisites HTable/Caching Mapping Syms Notes
Busybox 1.35.1 (RV64GC) 2.9-3.9% 34.8-38.3% 38.6-39.8%
OpenSBI 1.1 (generic fw_*.elf) 5.5-7.8% 64.3-65.8% 70.6-71.6%
Linux kernel 5.19 (vmlinux) 15.0-20.3% 47.6-52.2% 65.3-81.5%
Linux kernel 5.19 (vmlinux.o) 58.6-68.9% 69.5-75.3% 234.0-471.7% Not finally linked
glibc (libc.so.6) 4.9-6.6% 33.3-37.7% 38.6-43.0%

objdump -d (ELF-based archive)

Program Prerequisites HTable/Caching Mapping Syms
glibc (libc.a) 96.8-102.7% 104.0-108.9% 1067.8-1127.6%
newlib (libc.a) 63.8-71.6% 70.3-80.1% 265.8-295.3%

objdump -D (binary)

Program Prerequisites HTable/Caching Mapping Syms
Linux kernel 5.19 (vmlinux) (-2.1)-(-0.9)% 114.7-138.9% 115.7-138.8%
Random files (/dev/urandom) (-2.4)-(-2.2)% 142.5-144.2% 142.8-143.7%
1M (1048576) CSR instructions 2.2% 672.3% 681.2%

gdb: disas of near all code region

Program Prerequisites HTable/Caching Mapping Syms
Linux kernel 5.19 (vmlinux) with debug info 34.8% 38.8% 39.3%
Linux kernel 5.19 (vmlinux) without debug info 92.7% 106.5% 106.4%
OpenSBI 1.1 (generic fw_*.elf) 65.6-66.6% 80.2-80.7% 80.2-80.3%
1M (1048576) CSR instructions (ELF) 64.2% 300.7% 304.3%

Batch: objdump -d on Linux distribution

Serial Run: All ELF Files Under the Directory

System Path N Prerequisites HTable/Caching Mapping Syms
Ubuntu 22.04 LTS
(image for HiFive Unmatched)		 /usr/bin 563 2.3% 29.6% 31.3%
Debian unstable
(as of 2022-07-20)		 /usr/bin 269 1.8% 29.4% 31.0%
Ubuntu 22.04 LTS
(image for HiFive Unmatched)		 /usr/lib 6797 49.8% 67.7% 183.9%
Debian unstable
(as of 2022-07-20)		 /usr/lib 548 82.5% 101.4% 433.3%

Parallel Run: All (including data-only ELFs)

System N Prerequisites HTable/Caching Mapping Syms
Ubuntu 22.04 LTS
(image for HiFive Unmatched)		 7666 38.6% 54.6% 128.5%
Debian unstable
(as of 2022-07-20)		 946 119.8% 123.5% 389.7%

Batch: objdump -D (as binary) on Linux distribution

Serial Run: All ELF Files Under the Directory

System Path N Prerequisites HTable/Caching Mapping Syms
Ubuntu 22.04 LTS
(image for HiFive Unmatched)		 /usr/bin 563 1.4% 79.0% 80.4%
Debian unstable
(as of 2022-07-20)		 /usr/bin 269 1.1% 66.8% 68.0%

Parallel Run: All (including data-only ELFs)

System N Prerequisites HTable/Caching Mapping Syms
Ubuntu 22.04 LTS
(image for HiFive Unmatched)		 7666 (-0.4)% 82.2% 83.2%
Debian unstable
(as of 2022-07-20)		 946 0.7% 66.9% 67.5%

objdump -d (ELF): Extreme Examples

Program N Prerequisites HTable/Caching Mapping Syms (in other words)
OpenSSL 2 119.7-124.5% 124.6-129.6% 2664.3%-2677.8% x27.643-27.778
LLVM 1 129.2% 128.7% 22164.8% x222.648
  • OpenSSL
    1. Ubuntu 22.04 LTS (image for HiFive Unmatched) : /usr/lib/riscv64-linux-gnu/libcrypto.so.3
    2. Debian unstable (as of 2022-07-20) : /usr/lib/riscv64-linux-gnu/libcrypto.so.3
  • LLVM
    1. Ubuntu 22.04 LTS (Package libllvm14 Version 1:14.0.0-1ubuntu1) : /usr/lib/riscv64-linux-gnu/libLLVM-14.so.1

For large library with many symbols, the effect of the mapping symbol optimization is huge. I did expect some improvements but not that huge.

This optimization also benefits Arm architecture (not AArch64, due to different mapping symbol handlings).

Mapping Symbol Related

Program w/ISA Prerequisites HTable/Caching Mapping Syms
Busybox 1.35.1 (RV64GC) with debug info N 4.3% 34.9% 38.8%
Busybox 1.35.1 (RV64GC) with debug info Y 176.3% 256.7% 264.8%
Linux kernel 5.19 (vmlinux) without debug info N 15.0% 47.6% 64.5%
Linux kernel 5.19 (vmlinux) without debug info Y 269.0% 350.6% 501.9%
Linux kernel 5.19 (vmlinux.o) without debug info N 58.4% 75.2% 235.8%
Linux kernel 5.19 (vmlinux.o) without debug info Y 192.4% 224.4% 454.8%

Mapping symbols with ISA benefits from new optimizations.

Clone this wiki locally