{"id":182307,"date":"2026-06-24T20:48:35","date_gmt":"2026-06-24T18:48:35","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/odnostrochnoe-izmenenie-v-gcc-privelo-k-uskoreniyu-na-12-v-odnom-teste-i-zamedleniyu-na-14-v-drugom"},"modified":"2026-06-24T20:48:35","modified_gmt":"2026-06-24T18:48:35","slug":"odnostrochnoe-izmenenie-v-gcc-privelo-k-uskoreniyu-na-12-v-odnom-teste-i-zamedleniyu-na-14-v-drugom","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/odnostrochnoe-izmenenie-v-gcc-privelo-k-uskoreniyu-na-12-v-odnom-teste-i-zamedleniyu-na-14-v-drugom","title":{"rendered":"A one-line change in GCC resulted in a 12% speedup in one test and a 14% slowdown in another.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
An engineer from Intel made a single-line change to the GCC compiler set that increases the weight of a mispredicted branch on x86 processors. This modification improved the performance of the generated code by 12.7% in test 544.nab_r when optimizations '-O2 -mtune=graniterapids' were enabled for Intel Granite Rapids\/Xeon 6 CPUs and by 12.1% with optimizations '-O2 -mtune=znver5' on AMD Zen5 CPUs.<\/p>\n
Changing the weight from 'COSTS_N_INSNS (2)' to 'COSTS_N_INSNS (2) + 3' increases the significance of a misprediction error from 2 to 5 conditional instructions, better reflecting the characteristics of instruction pipeline processing in modern CPUs, where branch mispredictions are more costly. This weight change forces the compiler to convert 'if' expressions into conditional commands without branches, such as CMOV, thus eliminating stalls caused by incorrect branch predictions due to the need to flush the pipeline state. Previously, the weight 'COSTS_N_INSNS (2) + 3' was specified in GCC only for Intel Ice Lake and Alder Lake processors, but is now set for the general (generic) profile of x86 processors.<\/p>\n
Interestingly, after the patch was adopted, a regression was revealed, causing the 'Hint' test to run 30% slower for one of the GCC developers when building with the options '-march=generic -mtune=znver5' and '-march=generic -mtune=graniterapids'. The performance of the SPEC2017 and SPEC2026 test passes remained at the previous level after the change. The regression has been confirmed by the author of the original commit, who reports that the slowdown of the Hint test run is 14% when built with the options '-O2 -mtune=generic -march=x86-64-v3'. <\/p>\n
The slowdown is attributed to the fact that the Hint test code contains only one conditional structure, which GCC converts into a CMOV-based representation. This structure occurs infrequently (in 3% of cases), and its optimization does not impact performance. Meanwhile, the change in code generation mode resulted in a side effect that slowed down the execution of more frequently executed code. \n Source: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"