riscv_dis_opts_cache_support

Disassembler: Cache instruction class support

• Status: WITHDRAWN
  Combined with another fixes and superseded by riscv-dis-opts-batch-1
• Branch: riscv-dis-opts-cache-support
• Tracking PR: #22 (view Pull Request and Diff)

Requires

• Disassembler: Minor optimizations (batch 1)
  • Requires: Disassembler: Use faster hash table

Project: Disassembler Performance Improvements

Aggregate performance benchmark should be available here.

1. Disassembler: Use faster hash table
2. Disassembler: Minor optimizations (batch 1)
3. (You are here) Disassembler: Cache instruction class support

Feature Description

As I mentioned in the previous patchset, calling riscv_subset_supports repeatedly harms the performance in a measurable way (about 3-13% in total).

As a simple solution, this patch now caches instruction class support as a signed char array.

Performance Improvements

On disassembling linked RISC-V ELF programs using objdump, performance improvements achieved by this patchset is about 5-7%. Not bad for a simple change.

This is relative to the previous optimization.

objdump -d (ELF)

Program	Improvements	Notes
Busybox 1.35.1 (RV64GC)	6.3-7.1%
OpenSBI 1.1 (generic fw_*.elf)	7.4-8.0%
Linux kernel 5.18 (vmlinux)	6.1-6.4%
Linux kernel 5.18 (vmlinux.o)	0.3-2.5%	Not finally linked
glibc (libc.so.6)	6.4-7.0%

objdump -d (ELF-based archive)

Program	Improvements
glibc (libc.a)	0.6-1.0%
newlib (libc.a)	1.2-1.9%

objdump -D (binary)

Program	Improvements
Linux kernel 5.18 (vmlinux)	8.6-8.8%
Random files (/dev/urandom)	8.5-9.0%
1M (1048576) CSR instructions	37.4%

gdb: disas of near all code region

Program	Improvements
Linux kernel 5.18 (vmlinux) with debug info	1.0%
Linux kernel 5.18 (vmlinux) without debug info	1.7%
OpenSBI 1.1 (generic fw_*.elf)	1.5-1.6%
1M (1048576) CSR instructions (ELF)	8.1%

Batch: objdump -d on Linux distribution

Serial Run: All /usr/bin Programs

System	N	Improvements
Ubuntu 22.04 LTS (image for HiFive Unmatched)	563	6.3%
Debian unstable (as of 2022-07-20)	269	5.9%

Parallel Run: Top 100 in Size (including data-only ELFs)

System	N	Improvements
Ubuntu 22.04 LTS (image for HiFive Unmatched)	100	3.0%
Debian unstable (as of 2022-07-20)	100	1.0%

Parallel Run: All (including data-only ELFs)

System	N	Improvements
Ubuntu 22.04 LTS (image for HiFive Unmatched)	7666	2.9%
Debian unstable (as of 2022-07-20)	946	0.9%

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

Serial Run: All /usr/bin Programs

System	N	Improvements
Ubuntu 22.04 LTS (image for HiFive Unmatched)	563	8.0%
Debian unstable (as of 2022-07-20)	269	8.6%

Parallel Run: Top 100 in Size (including data-only ELFs)

System	N	Improvements
Ubuntu 22.04 LTS (image for HiFive Unmatched)	100	8.9%
Debian unstable (as of 2022-07-20)	100	8.3%

Parallel Run: All (including data-only ELFs)

System	N	Improvements
Ubuntu 22.04 LTS (image for HiFive Unmatched)	7666	8.2%
Debian unstable (as of 2022-07-20)	946	7.2%