diff options
Diffstat (limited to 'results/classifier/deepseek-r1:14b/reasoning/performance/2183')
| -rw-r--r-- | results/classifier/deepseek-r1:14b/reasoning/performance/2183 | 15 |
1 files changed, 15 insertions, 0 deletions
diff --git a/results/classifier/deepseek-r1:14b/reasoning/performance/2183 b/results/classifier/deepseek-r1:14b/reasoning/performance/2183 new file mode 100644 index 00000000..4235c6b5 --- /dev/null +++ b/results/classifier/deepseek-r1:14b/reasoning/performance/2183 @@ -0,0 +1,15 @@ +<think> +Okay, I need to classify this bug report from QEMU into one of the given categories. Let's see what the issue is about. + +The user mentioned that aarch64 emulation has become much slower since release 8.1.5 and it's also present in 8.2.1. They noticed longer boot times and memory accesses taking three times longer. They provided steps to reproduce, involving booting an ARM kernel on virt machine and running a script. + +They also tried the master branch and didn't see the issue there, so it's specific to these releases. They traced the problem to a commit ef74024b... by @bonzini on the 8.2 branch, which reverted mapping of PCREL translation blocks. + +So, what category does this fall into? The options are mistranslation, assembly lowering, and others like device, network, etc. Since the issue is about performance degradation in emulation speed, especially related to the TCG (Translation Code Generator) part of QEMU, it's likely a performance problem. + +Mistranslation might involve incorrect mapping from source to target, but here it's more about performance after a specific commit that dealt with translation blocks. The user linked it directly to the commit affecting PCREL handling in TCG, which impacts how code is translated and executed, leading to slower performance. + +Therefore, this bug report should be classified under "performance" because it's causing significant slowdowns. +</think> + +performance \ No newline at end of file |