On Sun, Jan 7, 2018 at 2:40 PM, Carlos E. R. <robin.listas@telefonica.net> wrote:
On 2018-01-07 20:22, Per Jessen wrote:
Carlos E. R. wrote:
Now I wonder if single core CPUs are affected.
This issue is related to paralelization optimizations. Thus I wonder whether a machine that doesn't paralelize is affected.
Jump prediction in the pipeline is not related to the number of cores.
Well, it can preload in advance the pipeline. But it can not compute the alternatives; it has to wait till it evaluates the condition, then compute a single one, the correct one, that is hopefully already loaded in the pipeline.
With a cpu that can do parallelization in hardware, both tracks can start to compute before reaching the branch; then one is discarded.
That's from reading <https://www.raspberrypi.org/blog/why-raspberry-pi-isnt-vulnerable-to-spectre-or-meltdown/>
Carlos, You fundamentally misunderstand the raspberrypi article. The parallelism it describes is within a single superscalar core. Thus, a modern Intel "core" has internal parallelism as described in the article. Multiple cores are used to run multiple totally unrelated instruction streams (ie. 2 different programs), and each core has parallelism as described n the article. In fact a modern Intel core has so much internal parallelism, much of the parallelism is wasted. That's where hyperthreading comes in. With hyperthreading two instruction streams are executed simultaneously, but there are parts of the parallelism that are shared between the 2 instruction streams. Greg -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org