summary refs log tree commit diff stats
path: root/results/classifier/no-thinking-deepseek-r1:32b/reasoning/runtime/1734
diff options
context:
space:
mode:
Diffstat (limited to 'results/classifier/no-thinking-deepseek-r1:32b/reasoning/runtime/1734')
-rw-r--r--results/classifier/no-thinking-deepseek-r1:32b/reasoning/runtime/17347
1 files changed, 7 insertions, 0 deletions
diff --git a/results/classifier/no-thinking-deepseek-r1:32b/reasoning/runtime/1734 b/results/classifier/no-thinking-deepseek-r1:32b/reasoning/runtime/1734
new file mode 100644
index 000000000..7646e6199
--- /dev/null
+++ b/results/classifier/no-thinking-deepseek-r1:32b/reasoning/runtime/1734
@@ -0,0 +1,7 @@
+The problem involves an issue where mmap calls fail when using QEMU version 8.0 for ARM binaries on a host system, specifically when the mapped files exceed around 1GB in size. The user observed that this does not happen with older versions like 6.2. The failure seems to be related to memory addressing limits imposed by QEMU's behavior, such as the ELF binary load address.
+
+**Analysis:**
+The issue likely stems from how QEMU handles memory mapping for ARM binaries. In version 8.0, there might be a stricter enforcement or change in the way virtual memory is allocated and mapped, leading to exhaustion of the available address space when multiple large files are mmaped. The mention of using the `-B` option to adjust the load address suggests that QEMU's memory layout configuration could be a contributing factor.
+
+**Category:**
+runtime
\ No newline at end of file