summaryrefslogtreecommitdiffstats
path: root/gitlab/issues/target_arm/host_missing/accel_TCG/1612.toml
diff options
context:
space:
mode:
authorChristian Krinitsin <mail@krinitsin.com>2025-05-30 16:52:07 +0200
committerChristian Krinitsin <mail@krinitsin.com>2025-05-30 16:52:17 +0200
commit9260319e7411ff8281700a532caa436f40120ec4 (patch)
tree2f6bfe5f3458dd49d328d3a9eb508595450adec0 /gitlab/issues/target_arm/host_missing/accel_TCG/1612.toml
parent225caa38269323af1bfc2daadff5ec8bd930747f (diff)
downloademulator-bug-study-9260319e7411ff8281700a532caa436f40120ec4.tar.gz
emulator-bug-study-9260319e7411ff8281700a532caa436f40120ec4.zip
gitlab scraper: download in toml and text format
Diffstat (limited to 'gitlab/issues/target_arm/host_missing/accel_TCG/1612.toml')
-rw-r--r--gitlab/issues/target_arm/host_missing/accel_TCG/1612.toml59
1 files changed, 0 insertions, 59 deletions
diff --git a/gitlab/issues/target_arm/host_missing/accel_TCG/1612.toml b/gitlab/issues/target_arm/host_missing/accel_TCG/1612.toml
deleted file mode 100644
index cb03a7d0..00000000
--- a/gitlab/issues/target_arm/host_missing/accel_TCG/1612.toml
+++ /dev/null
@@ -1,59 +0,0 @@
-id = 1612
-title = "SVE first-faulting gather loads return incorrect data"
-state = "closed"
-created_at = "2023-04-21T16:12:04.887Z"
-closed_at = "2023-05-18T14:51:26.696Z"
-labels = ["Closed::Fixed", "accel: TCG", "target: arm"]
-url = "https://gitlab.com/qemu-project/qemu/-/issues/1612"
-host-os = "Ubuntu 20.04 LTS"
-host-arch = "x86_64"
-qemu-version = "qemu-aarch64 version 5.0.50 (v5.0.0-403-g50de9b78ce-dirty) (and latest master)"
-guest-os = "n/a"
-guest-arch = "n/a"
-description = """The results of `ldff1(b|h|w|d)` seem to be incorrect when `<Zt> == <Zm>`. The first element is duplicated throughout the vector while the FFR indicates that all elements were successfully loaded. This happens since https://gitlab.com/qemu-project/qemu/-/commit/50de9b78cec06e6d16e92a114a505779359ca532, and still happens on the latest master."""
-reproduce = """1. This assembly sequence loads data with an `ldff1d` instruction (and also loads the ffr). Note that with `ldff1d`, `<Zt> == <Zm>`.
-
-asmtest.s
-```
- .type asmtest, @function
- .balign 16
- .global asmtest
-asmtest:
- setffr
- ptrue p0.d
- index z1.d, #0, #1
- ldff1d z1.d, p0/z, [x0, z1.d, LSL #3]
- rdffr p1.b
- st1d {z1.d}, p0, [x1]
- str p1, [x2]
- ret
-```
-
-This harness for convenience intialises some data and checks the element at index 1, which should be 1.
-
-test.c
-```
-#include <arm_sve.h>
-#include <stdio.h>
-
-void asmtest(int64_t const * data, svint64_t * loaded, svbool_t * ffr);
-
-int main() {
- const int64_t data[] = {42, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
- 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21,
- 22, 23, 24, 25, 26, 27, 28, 29, 30, 31};
- svint64_t loaded;
- svbool_t ffr;
-
- asmtest(data, &loaded, &ffr);
-
- // Check value of element at index 1
- svuint64_t lanes = svindex_u64(0, 1);
- svbool_t lane = svcmpeq_n_u64(svptrue_b64(), lanes, 1);
- printf("%ld\\n", svaddv_s64(lane, loaded));
-}
-```
-
-2. ```clang-15 -fuse-ld=lld -march=armv8-a+sve2 -target aarch64-unknown-linux-gnu -static *.c *.s -o svldffgathertest```
-3. ```qemu-aarch64 svldffgathertest``` - the value printed should be 1, but it can be seen that all values in the loaded vector are 42."""
-additional = """The above code was successfully tested on real SVE hardware. Normal gathers work fine in QEMU, as does a non-gather first-fault load."""