SPEC2026 INTrate: PGO+ThinLTO and PGO+LTO native -Ofast - LLVM and GCC
-Ofast -march=native -mtune=native -flto=(thin) and -fprofile-instr-use vs. -profile-use
Please note that even though the header of this comparison
says -Ofast, benchmarks 710.omnetpp_r, 734.vpr_r,
735.gem5_r and 753.ns3_r were additionally compiled
with -fno-finite-math-only.
AMD EPYC 8635P (zen5 based Sorano)
Run-times
Benchmark
LLVM
LLVM FullLTO
GCC
LLVM FullLTO
GCC
706.stockfish_r
100.00%
100.72%
37.28%
100.00%
37.01%
707.ntest_r
100.00%
97.68%
93.04%
100.00%
95.25%
708.sqlite_r
100.00%
100.00%
92.27%
100.00%
92.27%
710.omnetpp_r
100.00%
96.61%
77.34%
100.00%
80.05%
714.cpython_r
100.00%
101.40%
95.79%
100.00%
94.46%
721.gcc_r
100.00%
99.30%
97.22%
100.00%
97.90%
723.llvm_r
100.00%
100.00%
99.25%
100.00%
99.25%
727.cppcheck_r
100.00%
102.43%
92.48%
100.00%
90.28%
729.abc_r
100.00%
100.89%
102.67%
100.00%
101.76%
734.vpr_r
100.00%
95.57%
91.38%
100.00%
95.62%
735.gem5_r
100.00%
97.04%
94.08%
100.00%
96.95%
750.sealcrypto_r
100.00%
100.00%
100.00%
100.00%
100.00%
753.ns3_r
100.00%
93.14%
94.33%
100.00%
101.27%
777.zstd_r
100.00%
100.74%
104.22%
100.00%
103.45%
| Benchmark | LLVM | LLVM FullLTO | GCC | LLVM FullLTO | GCC |
|---|---|---|---|---|---|
| 706.stockfish_r | 100.00% | 100.72% | 37.28% | 100.00% | 37.01% |
| 707.ntest_r | 100.00% | 97.68% | 93.04% | 100.00% | 95.25% |
| 708.sqlite_r | 100.00% | 100.00% | 92.27% | 100.00% | 92.27% |
| 710.omnetpp_r | 100.00% | 96.61% | 77.34% | 100.00% | 80.05% |
| 714.cpython_r | 100.00% | 101.40% | 95.79% | 100.00% | 94.46% |
| 721.gcc_r | 100.00% | 99.30% | 97.22% | 100.00% | 97.90% |
| 723.llvm_r | 100.00% | 100.00% | 99.25% | 100.00% | 99.25% |
| 727.cppcheck_r | 100.00% | 102.43% | 92.48% | 100.00% | 90.28% |
| 729.abc_r | 100.00% | 100.89% | 102.67% | 100.00% | 101.76% |
| 734.vpr_r | 100.00% | 95.57% | 91.38% | 100.00% | 95.62% |
| 735.gem5_r | 100.00% | 97.04% | 94.08% | 100.00% | 96.95% |
| 750.sealcrypto_r | 100.00% | 100.00% | 100.00% | 100.00% | 100.00% |
| 753.ns3_r | 100.00% | 93.14% | 94.33% | 100.00% | 101.27% |
| 777.zstd_r | 100.00% | 100.74% | 104.22% | 100.00% | 103.45% |
| LLVM | clang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc) |
| GCC | GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d) |
Non-debug executable sizes
Benchmark
LLVM
LLVM FullLTO
GCC
LLVM FullLTO
GCC
706.stockfish_r
100.00%
104.87%
94.44%
100.00%
90.06%
707.ntest_r
100.00%
97.55%
86.70%
100.00%
88.87%
708.sqlite_r
100.00%
99.94%
93.88%
100.00%
93.94%
710.omnetpp_r
100.00%
99.82%
79.87%
100.00%
80.02%
714.cpython_r
100.00%
100.76%
82.88%
100.00%
82.26%
721.gcc_r
100.00%
98.55%
86.91%
100.00%
88.19%
723.llvm_r
100.00%
103.72%
77.43%
100.00%
74.65%
727.cppcheck_r
100.00%
99.15%
87.92%
100.00%
88.68%
729.abc_r
100.00%
102.52%
73.93%
100.00%
72.11%
734.vpr_r
100.00%
99.22%
80.26%
100.00%
80.90%
735.gem5_r
100.00%
97.80%
79.26%
100.00%
81.04%
750.sealcrypto_r
100.00%
95.13%
65.48%
100.00%
68.83%
753.ns3_r
100.00%
94.79%
76.07%
100.00%
80.25%
777.zstd_r
100.00%
97.80%
68.70%
100.00%
70.25%
Smaller binary sizes are always better.
| Benchmark | LLVM | LLVM FullLTO | GCC | LLVM FullLTO | GCC |
|---|---|---|---|---|---|
| 706.stockfish_r | 100.00% | 104.87% | 94.44% | 100.00% | 90.06% |
| 707.ntest_r | 100.00% | 97.55% | 86.70% | 100.00% | 88.87% |
| 708.sqlite_r | 100.00% | 99.94% | 93.88% | 100.00% | 93.94% |
| 710.omnetpp_r | 100.00% | 99.82% | 79.87% | 100.00% | 80.02% |
| 714.cpython_r | 100.00% | 100.76% | 82.88% | 100.00% | 82.26% |
| 721.gcc_r | 100.00% | 98.55% | 86.91% | 100.00% | 88.19% |
| 723.llvm_r | 100.00% | 103.72% | 77.43% | 100.00% | 74.65% |
| 727.cppcheck_r | 100.00% | 99.15% | 87.92% | 100.00% | 88.68% |
| 729.abc_r | 100.00% | 102.52% | 73.93% | 100.00% | 72.11% |
| 734.vpr_r | 100.00% | 99.22% | 80.26% | 100.00% | 80.90% |
| 735.gem5_r | 100.00% | 97.80% | 79.26% | 100.00% | 81.04% |
| 750.sealcrypto_r | 100.00% | 95.13% | 65.48% | 100.00% | 68.83% |
| 753.ns3_r | 100.00% | 94.79% | 76.07% | 100.00% | 80.25% |
| 777.zstd_r | 100.00% | 97.80% | 68.70% | 100.00% | 70.25% |
AMD EPYC 75F3 (zen3 based high-frequency Milan)
Run-times
Benchmark
LLVM
LLVM FullLTO
GCC
LLVM FullLTO
GCC
706.stockfish_r
100.00%
100.00%
51.01%
100.00%
51.01%
707.ntest_r
100.00%
98.25%
98.25%
100.00%
100.00%
708.sqlite_r
100.00%
98.85%
89.84%
100.00%
90.89%
710.omnetpp_r
100.00%
94.81%
75.68%
100.00%
79.83%
714.cpython_r
100.00%
101.79%
97.92%
100.00%
96.20%
721.gcc_r
100.00%
99.52%
96.42%
100.00%
96.88%
723.llvm_r
100.00%
100.00%
97.84%
100.00%
97.84%
727.cppcheck_r
100.00%
102.47%
93.27%
100.00%
91.03%
729.abc_r
100.00%
99.35%
99.35%
100.00%
100.00%
734.vpr_r
100.00%
93.82%
89.02%
100.00%
94.88%
735.gem5_r
100.00%
96.10%
95.45%
100.00%
99.32%
750.sealcrypto_r
100.00%
97.86%
95.19%
100.00%
97.27%
753.ns3_r
100.00%
93.56%
86.34%
100.00%
92.29%
777.zstd_r
100.00%
100.49%
103.93%
100.00%
103.42%
| Benchmark | LLVM | LLVM FullLTO | GCC | LLVM FullLTO | GCC |
|---|---|---|---|---|---|
| 706.stockfish_r | 100.00% | 100.00% | 51.01% | 100.00% | 51.01% |
| 707.ntest_r | 100.00% | 98.25% | 98.25% | 100.00% | 100.00% |
| 708.sqlite_r | 100.00% | 98.85% | 89.84% | 100.00% | 90.89% |
| 710.omnetpp_r | 100.00% | 94.81% | 75.68% | 100.00% | 79.83% |
| 714.cpython_r | 100.00% | 101.79% | 97.92% | 100.00% | 96.20% |
| 721.gcc_r | 100.00% | 99.52% | 96.42% | 100.00% | 96.88% |
| 723.llvm_r | 100.00% | 100.00% | 97.84% | 100.00% | 97.84% |
| 727.cppcheck_r | 100.00% | 102.47% | 93.27% | 100.00% | 91.03% |
| 729.abc_r | 100.00% | 99.35% | 99.35% | 100.00% | 100.00% |
| 734.vpr_r | 100.00% | 93.82% | 89.02% | 100.00% | 94.88% |
| 735.gem5_r | 100.00% | 96.10% | 95.45% | 100.00% | 99.32% |
| 750.sealcrypto_r | 100.00% | 97.86% | 95.19% | 100.00% | 97.27% |
| 753.ns3_r | 100.00% | 93.56% | 86.34% | 100.00% | 92.29% |
| 777.zstd_r | 100.00% | 100.49% | 103.93% | 100.00% | 103.42% |
| LLVM | clang version 23.1.3 (https://github.com/llvm/llvm-project.git 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (https://github.com/llvm/llvm-project.git 0d261d1ca552c95a8f007e061c787ac7132fbcbc) |
| GCC | GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d) |
Non-debug executable sizes
Benchmark
LLVM
LLVM FullLTO
GCC
LLVM FullLTO
GCC
706.stockfish_r
100.00%
99.48%
94.39%
100.00%
94.89%
707.ntest_r
100.00%
97.46%
87.78%
100.00%
90.07%
708.sqlite_r
100.00%
100.05%
94.71%
100.00%
94.66%
710.omnetpp_r
100.00%
99.88%
81.07%
100.00%
81.17%
714.cpython_r
100.00%
100.63%
83.31%
100.00%
82.79%
721.gcc_r
100.00%
98.46%
87.16%
100.00%
88.52%
723.llvm_r
100.00%
103.79%
79.38%
100.00%
76.48%
727.cppcheck_r
100.00%
98.89%
88.15%
100.00%
89.13%
729.abc_r
100.00%
102.76%
75.92%
100.00%
73.88%
734.vpr_r
100.00%
99.40%
81.57%
100.00%
82.07%
735.gem5_r
100.00%
97.82%
79.49%
100.00%
81.27%
750.sealcrypto_r
100.00%
98.01%
67.04%
100.00%
68.40%
753.ns3_r
100.00%
94.67%
76.57%
100.00%
80.88%
777.zstd_r
100.00%
104.82%
70.42%
100.00%
67.18%
Smaller binary sizes are always better.
| Benchmark | LLVM | LLVM FullLTO | GCC | LLVM FullLTO | GCC |
|---|---|---|---|---|---|
| 706.stockfish_r | 100.00% | 99.48% | 94.39% | 100.00% | 94.89% |
| 707.ntest_r | 100.00% | 97.46% | 87.78% | 100.00% | 90.07% |
| 708.sqlite_r | 100.00% | 100.05% | 94.71% | 100.00% | 94.66% |
| 710.omnetpp_r | 100.00% | 99.88% | 81.07% | 100.00% | 81.17% |
| 714.cpython_r | 100.00% | 100.63% | 83.31% | 100.00% | 82.79% |
| 721.gcc_r | 100.00% | 98.46% | 87.16% | 100.00% | 88.52% |
| 723.llvm_r | 100.00% | 103.79% | 79.38% | 100.00% | 76.48% |
| 727.cppcheck_r | 100.00% | 98.89% | 88.15% | 100.00% | 89.13% |
| 729.abc_r | 100.00% | 102.76% | 75.92% | 100.00% | 73.88% |
| 734.vpr_r | 100.00% | 99.40% | 81.57% | 100.00% | 82.07% |
| 735.gem5_r | 100.00% | 97.82% | 79.49% | 100.00% | 81.27% |
| 750.sealcrypto_r | 100.00% | 98.01% | 67.04% | 100.00% | 68.40% |
| 753.ns3_r | 100.00% | 94.67% | 76.57% | 100.00% | 80.88% |
| 777.zstd_r | 100.00% | 104.82% | 70.42% | 100.00% | 67.18% |