summary refs log tree commit diff stats
path: root/results/scraper/fex/668
diff options
context:
space:
mode:
authorChristian Krinitsin <mail@krinitsin.com>2025-07-17 09:10:43 +0200
committerChristian Krinitsin <mail@krinitsin.com>2025-07-17 09:10:43 +0200
commitf2ec263023649e596c5076df32c2d328bc9393d2 (patch)
tree5dd86caab46e552bd2e62bf9c4fb1a7504a44db4 /results/scraper/fex/668
parent63d2e9d409831aa8582787234cae4741847504b7 (diff)
downloadqemu-analysis-main.tar.gz
qemu-analysis-main.zip
add downloaded fex bug-reports HEAD main
Diffstat (limited to 'results/scraper/fex/668')
-rw-r--r--results/scraper/fex/66812
1 files changed, 12 insertions, 0 deletions
diff --git a/results/scraper/fex/668 b/results/scraper/fex/668
new file mode 100644
index 000000000..e6e17a928
--- /dev/null
+++ b/results/scraper/fex/668
@@ -0,0 +1,12 @@
+clang -mcpu/tune=native doesn't pick up Snapdragon 865 correctly
+This is causing atomics to be terrible loadstore acquire/release pairs for everything outside of the JIT.

+Even clang 11 is effected by this.

+

+First step would have our cmake script check for this and explicit place cortex-a77 there.

+Also -march=native isn't supported on ARM with clang, so the current cmake check does nothing. It needs -mcpu or -mtune.

+

+Second step would be to upstream changes to clang/llvm so it is picked up correctly. This way it can be supported on at least clang 12.

+

+Might also effect Snapdragon 850, so c630 is also slower?

+

+Causing pain with interpreter cmpxchg fallback unit tests.
\ No newline at end of file