similar to: [LLVMdev] Seg faulting on vector ops

On Fri, 20 Jul 2007, Chuck Rose III wrote: > I'm looking to make use of the vectorization primitives in the Intel > chip with the code we generate from LLVM and so I've started > experimenting with it. What is the state of the machine code generated > for vectors? In my tinkering, I seem to be getting some wonky machine > instructions, but I'm most likely just doing

Hrm. This problem shouldn't be target specific. I am pretty sure prologue / epilogue inserter aligns stack correctly if there are stack objects with greater than default stack alignment requirement. Seems to be the initial alloca() instruction should specify 16 byte alignment? Evan On Jul 21, 2007, at 2:51 PM, Chris Lattner wrote: > On Fri, 20 Jul 2007, Chuck Rose III wrote:

Hi Chuck! On Jul 20, 2007, at 11:36 AM, Chuck Rose III wrote: > Hola LLVMers, > > > > I’m looking to make use of the vectorization primitives in the > Intel chip with the code we generate from LLVM and so I’ve started > experimenting with it. What is the state of the machine code > generated for vectors? In my tinkering, I seem to be getting some > wonky

I am fairly certain this is right. Chuck, can you do a quick experiment for me? Go back to your original code but make sure the alloca instruction specify 16-byte alignment. The code should work. If not, please file a bug. Thanks, Evan On Jul 24, 2007, at 1:58 PM, Evan Cheng wrote: > Hrm. This problem shouldn't be target specific. I am pretty sure > prologue / epilogue inserter

I'd like to start a discussion about how clang supports _Float16 for target architectures that don't have direct support for 16-bit floating point arithmetic. The current clang language extensions documentation says, "If half-precision instructions are unavailable, values will be promoted to single-precision, similar to the semantics of __fp16 except that the results will be stored

It seems that there are several issues here: 1. Should the front end be concerned with whether or not the IR that it is emitting can be translated into a well-defined IR? 2. How should the selection DAG handle data types whose representation isn't defined by the ABI we're targeting? 3. What should the ABI do with half-precision floats? Working backward... The third question here is

On 24 Jan 2019, at 4:46, Sjoerd Meijer wrote: > Hello, > > I added _Float16 support to Clang and codegen support in the AArch64 > and ARM backends, but have not looked into x86. Ahmed is right: > AArch64 is fine, only a few ACLE intrinsics are missing. ARM has rough > edges: scalar codegen should be mostly fine, vector codegen needs some > more work. > >

Thanks for the clarification. That behaviour is quite weird as arbiter bricks should hold?only metadata. What does the following show on host?uk3-prod-gfs-arb-01: du -h -x -d 1?/data/glusterfs/gv1/brick1/brickdu -h -x -d 1?/data/glusterfs/gv1/brick3/brickdu -h -x -d 1 /data/glusterfs/gv1/brick2/brick If indeed the shards are taking space -?that is a really strange situation.From which version

Hi Strahil, This is the output from the commands: root at uk3-prod-gfs-arb-01:~# du -h -x -d 1 /data/glusterfs/gv1/brick1/brick 2.2G /data/glusterfs/gv1/brick1/brick/.glusterfs 24M /data/glusterfs/gv1/brick1/brick/scalelite-recordings 16K /data/glusterfs/gv1/brick1/brick/mytute 18M /data/glusterfs/gv1/brick1/brick/.shard 0

Hi, you mentioned that the arbiter bricks run out of inodes.Are you using XFS ?Can you provide the xfs_info of each brick ? Best Regards,Strahil Nikolov? On Sat, Jul 1, 2023 at 19:41, Liam Smith<liam.smith at ek.co> wrote: Hi, We're running a cluster with two data nodes and one arbiter, and have sharding enabled. We had an issue a while back where one of the server's

Hello, we are trying to configure a CTDB-Cluster with Glusterfs. We are using Samba 4.5 together with gluster 3.9. We set up a lvm2 thin-provisioned volume to use gluster-snapshots. Then we configured the first share without using shadow_copy2 and everything was working fine. Then we added the shadow_copy2 parameters, when we did a "smbclient" we got the following message: root at

Hi, We're running a cluster with two data nodes and one arbiter, and have sharding enabled. We had an issue a while back where one of the server's crashed, we got the server back up and running and ensured that all healing entries cleared, and also increased the server spec (CPU/Mem) as this seemed to be the potential cause. Since then however, we've seen some strange behaviour,

Hi Liam, I saw that your XFS uses ?imaxpct=25? which for an arbiter brick is a little bit low. If you have free space on the bricks, increase the maxpct to a bigger value, like:xfs_growfs -m 80 /path/to/brickThat will set 80% of the Filesystem for inodes, which you can verify with df -i /brick/path (compare before and after).?This way?you won?t run out of inodes in the future. Of course, always

Hi Strahil, We're using gluster to act as a share for an application to temporarily process and store files, before they're then archived off over night. The issue we're seeing isn't with the inodes running out of space, but the actual disk space on the arb server running low. This is the df -h? output for the bricks on the arb server: /dev/sdd1 15G 12G 3.3G 79%

Hi, Thanks for your response, please find the xfs_info for each brick on the arbiter below: root at uk3-prod-gfs-arb-01:~# xfs_info /data/glusterfs/gv1/brick1 meta-data=/dev/sdc1 isize=512 agcount=31, agsize=131007 blks = sectsz=512 attr=2, projid32bit=1 = crc=1 finobt=1, sparse=1, rmapbt=0 =

Hello Nithya! Thank you so much, I think we are close to build a stable storage solution according to your recommendations. Here's our rebalance log - please don't pay attention to error messages after 9AM - this is when we manually destroyed volume to recreate it for further testing. Also all remove-brick operations you could see in the log were executed manually when recreating volume.

Hi all, I think I might have found a potential bug when using SSE intrinsic and unaligned memory. Here's the code to reproduce it: #include "llvm/Module.h" #include "llvm/Intrinsics.h" #include "llvm/Instructions.h" #include "llvm/ModuleProvider.h" #include "llvm/ExecutionEngine/JIT.h" #include

Hello Nithya! Thank you for your help on figuring this out! We changed our configuration and after having a successful test yesterday we have run into new issue today. The test including moderate read/write (~20-30 Mb/s) and scaling the storage was running about 3 hours and at some moment system got stuck: On the user level there are such errors when trying to work with filesystem: OSError:

Hi folks, I'm troubled moving an arbiter brick to another server because of I/O load issues. My setup is as follows: # gluster volume info Volume Name: myvol Type: Distributed-Replicate Volume ID: 43ba517a-ac09-461e-99da-a197759a7dc8 Status: Started Snapshot Count: 0 Number of Bricks: 3 x (2 + 1) = 9 Transport-type: tcp Bricks: Brick1: gv0:/data/glusterfs Brick2: gv1:/data/glusterfs Brick3:

Hi Karthik, Thank you for your reply. The heal is still undergoing, as the /var/log/glusterfs/glustershd.log keeps growing, and there's a lot of pending entries in the heal info. The gluster version is 3.10.9 and 3.10.10 (the version update in progress). It doesn't have info summary [yet?], and the heal info is way too long to attach here. (It takes more than 20 minutes just to collect