...we patched Safari to set the DIT bit in the rendering process and observed that such a mitigation results in an overhead of 4.5% on the Speedometer 3.0 benchmark.
Wait does this affect Granite Rapids too?The shadow of the Spectre is long and far reaching:
Even Lunar & Arrow Lake get a patch.
Would seem so.Wait does this affect Granite Rapids too?
No mention of AMD in any of that ??Would seem so.
"This is also the first time seeing microcode updates for Arrow Lake U / S / HX / H, Granite Rapids AP/SP, and Lunar Lake as part of the Linux CPU microcode collection."
Depends on how much performance is lost from further mitigations/patches.Do "we" still care about all this?
These vulnerabilities are made possible by the patches made against Spectre for various architectures, the exploits need specific combos of architecture + Spectre patch to work. For now there are no reported vulnerabilities on the AMD side, though we'll have to wait and see if none were found later than the Intel ones.No mention of AMD in any of that ??
Just asking. ThanksThese vulnerabilities are made possible by the patches made against Spectre for various architectures, the exploits need specific combos of architecture + Spectre patch to work. For now there are no reported vulnerabilities on the AMD side, though we'll have to wait and see if none were found later than the Intel ones.
Yes, "we" care:Do "we" still care about all this?
It was just yesterday that Training Solo was made public as a new speculative execution CPU vulnerability affecting some Intel and Arm CPUs... Today another one is now public for Intel processors: Branch Privilege Injection.
Intel has developed a microcode update for affected processors and provided us with one to evaluate on Alder Lake. We were able to verify that the microcode update stops our primitives that we use in the paper to detect the vulnerabilities. Our performance evaluation shows up to 2.7% overhead for the microcode mitigation on Alder Lake. We have also evaluated several potential alternative mitigation strategies in software with overheads between 1.6% (Coffee Lake Refresh) and 8.3% (Rocket lake).
Branch Privilege Injection affects all Intel CPUs since the 9th Gen Core CPUs (Coffee Lake Refresh) while the Indirect Branch Prediction Barrier vector goes back to 7th Gen Core (Kaby Lake) processors.