diff options
Diffstat (limited to 'results/classifier/deepseek-1/reasoning/crashing./1837049')
| -rw-r--r-- | results/classifier/deepseek-1/reasoning/crashing./1837049 | 19 |
1 files changed, 0 insertions, 19 deletions
diff --git a/results/classifier/deepseek-1/reasoning/crashing./1837049 b/results/classifier/deepseek-1/reasoning/crashing./1837049 deleted file mode 100644 index 840a8d64f..000000000 --- a/results/classifier/deepseek-1/reasoning/crashing./1837049 +++ /dev/null @@ -1,19 +0,0 @@ -<think> -Okay, let me try to figure out what's going on here. So the user is having trouble with QEMU crashing when booting a PowerPC image. They tried different displays and CPUs, and eventually found that using -display sdl,gl=on along with some other parameters works for them. - -Hmm, they mentioned compiling QEMU as 32-bit with Slackware's gcc 5.5.0, which worked fine for a fedora 30 install but had issues otherwise. They think it might be a compiler bug or something related to the build options like -Og. - -Looking at their latest successful command, they're using sdl display with gl=on, G4 CPU, and tb-size=384. The version of QEMU is 4.2.91, which is close to 5.0. They also noted that recompiling SDL2 and having a newer glibc might have helped. - -The assistant mentioned some commits around 600e17b2615 related to TCG's tb-size heuristic changes. Those changes probably affect how QEMU allocates memory for the translation buffer, especially on 32-bit hosts which have smaller allocations by default now. - -So, putting it together: The issue might be related to the TB size in QEMU, especially on 32-bit systems where it's more constrained. By explicitly setting tb-size=384, the user ensures there's enough memory allocated for TCG operations, preventing crashes during boot. - -They also had to adjust other parameters like display settings and CPU type to get it working. It seems that with these adjustments, QEMU can boot successfully without crashing. -</think> - -The issue you're encountering appears to be related to how QEMU allocates memory for its Translation Buffer (TB) when running on a 32-bit host. The changes in the QEMU code around commit `600e17b2615` altered the heuristic for determining the TB size, which can lead to crashes if not enough memory is allocated. - -By explicitly setting `-accel tcg,tb-size=384`, you're ensuring that a sufficient amount of memory is reserved for the translation buffer, which seems to resolve the issue. Additionally, using certain display options and CPU types (like G4) might help in stabilizing the emulation process. - -In summary, adjusting the TB size with `-accel tcg,tb-size=384` along with other configuration tweaks allows QEMU to boot successfully without crashing. \ No newline at end of file |