This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: A few SPEC95 results for the new register allocator
- To: Diego Novillo <dnovillo at redhat dot com>
- Subject: Re: A few SPEC95 results for the new register allocator
- From: Daniel Berlin <dberlin at redhat dot com>
- Date: 20 Feb 2001 13:08:42 -0500
- Cc: gcc at gcc dot gnu dot org
- References: <20010220101701.A5198@tornado.cygnus.com>
Diego Novillo <dnovillo@redhat.com> writes:
> Today, the new register allocator branch successfully passed a
> few SPEC95 tests for the first time. The new allocator produces
> slightly worse SPEC marks except for perl (+4% speedup).
This is actually very good, considering we
1. don't yet split live ranges.
2. have no optimized spill code at all. (IE it stores/reloads
everywhere, doesn't do reuse of spill slots)
3. don't do rematerialization.
4. don't do interference region spilling.
5. have a very simple spill heuristic.
I would have actually expected to be doing much worse at this point,
around ~15% slowdown on x86.
I'd guess we'd be doing at least as well as the old allocator already on RISC
processors, however.
I'll run spec on my powerbook.
--Dan
>
> These are the results compared to today's mainline.
>
> -------------------------------------------------------------------------------
> HARDWARE
> --------
> Model Name: IA-32 Dell Dimension 4100
> CPU: Pentium III (Coppermine) at 996.785 MHz
> FPU: Integrated
> Number of CPU(s): 1
> Secondary Cache: 256 KB
> Memory: 257408 kB
>
>
> SOFTWARE
> --------
> Operating System: Linux 2.2.16-22
> Compiler: GCC as of Feb 20, 2001
> -------------------------------------------------------------------------------
>
>
> -------------------------------------------------------------------------------
> GCC mainline (20010220)
> -------------------------------------------------------------------------------
> SPEC Benchmark CINT95 Summary
>
>
> Base Base Base Peak Peak Peak
> Benchmarks Ref Time Run Time Ratio Ref Time Run Time Ratio
> ------------ -------- -------- -------- -------- -------- --------
> 099.go 4600 118 38.9 4600 119 38.6
> 124.m88ksim 1900 65.2 29.2 1900 61.7 30.8
> 126.gcc 1700 51.7 32.9 1700 51.4 33.1
> 129.compress 1800 89.0 20.2 1800 88.2 20.4
> 130.li 1900 57.5 33.0 1900 55.3 34.4
> 132.ijpeg 2400 74.4 32.3 2400 75.1 31.9
> 134.perl 1900 47.5 40.0 1900 47.6 39.9
> 147.vortex 2700 94.2 28.6 2700 94.0 28.7
> SPECint_base95 (Geom. Mean) 31.3
> SPECint95 (Geom. Mean) 31.7
> -------------------------------------------------------------------------------
>
>
> -------------------------------------------------------------------------------
> new-regalloc-branch (20010220)
> -------------------------------------------------------------------------------
> Estimated Estimated
> Base Base Base Peak Peak Peak
> Benchmarks Ref Time Run Time Ratio Ref Time Run Time Ratio
> ------------ -------- -------- -------- -------- -------- --------
> 099.go 4600 128 35.9 4600 127 36.1 (-6%)
> 124.m88ksim 1900 -- -- 1900 -- --
> 126.gcc 1700 -- -- 1700 -- --
> 129.compress 1800 92.8 19.4 1800 91.6 19.7 (-3%)
> 130.li 1900 56.9 33.4 1900 55.8 34.0 (-1%)
> 132.ijpeg 2400 -- -- 2400 -- --
> 134.perl 1900 47.4 40.1 1900 45.6 41.6 (+4%)
> 147.vortex 2700 99.7 27.1 2700 99.7 27.1 (-5%)
> -------------------------------------------------------------------------------