SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 INTrate: PGO+LTO -Ofast native - LLVM, AOCC and GCC

-Ofast -march=native -mtune=native -flto and -fprofile-instr-use vs. -profile-use

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark LLVM AOCC GCC AOCC GCC
500.perlbench_r 100.00% 98.13% 83.16% 100.00% 84.75%
502.gcc_r 100.00% 100.92% 97.47% 100.00% 96.58%
505.mcf_r 100.00% 90.27% 90.77% 100.00% 100.55%
520.omnetpp_r 100.00% 96.81% 88.48% 100.00% 91.39%
523.xalancbmk_r 100.00% 97.15% 88.19% 100.00% 90.78%
525.x264_r 100.00% 116.26% 95.16% 100.00% 81.85%
531.deepsjeng_r 100.00% 80.62% 90.43% 100.00% 112.17%
541.leela_r 100.00% 99.76% 105.58% 100.00% 105.84%
548.exchange2_r CE 100.00% 56.73%
557.xz_r 100.00% 100.42% 96.64% 100.00% 96.23%
LLVM clang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc)
AOCC AMD clang version 17.0.6 (CLANG: AOCC_5.2.0-Build#2035 2026_04_10)
GCC GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM AOCC GCC AOCC GCC
500.perlbench_r 100.00% 127.41% 96.47% 100.00% 75.72%
502.gcc_r 100.00% 130.71% 91.52% 100.00% 70.02%
505.mcf_r 100.00% 137.77% 66.10% 100.00% 47.98%
520.omnetpp_r 100.00% 104.56% 82.51% 100.00% 78.91%
523.xalancbmk_r 100.00% 104.38% 82.70% 100.00% 79.23%
525.x264_r 100.00% 123.68% 91.58% 100.00% 74.05%
531.deepsjeng_r 100.00% 100.14% 102.09% 100.00% 101.95%
541.leela_r 100.00% 99.78% 78.33% 100.00% 78.50%
548.exchange2_r CE 100.00% 57.77%
557.xz_r 100.00% 116.45% 95.26% 100.00% 81.81%
Smaller binary sizes are always better.