search for: bfloat

I don’t think bfloat should be handled this way. What Amara is suggesting is an optimization, i.e., if we drop the information we are still correct. With bfloat, if we do an operation on float16 instead of bfloat16 this is a correctness problem. So that means that either we need to have new opcodes for bfloat or we ne...

> On May 5, 2020, at 2:45 PM, Ties Stuij <Ties.Stuij at arm.com> wrote: > > Quentin: Thanks for the info. I was under the impression that the LLVM community at large would prefer to extend the IR type to a bfloat MVT type. I've made a number of patches to implement this up to a point for AArch64. I can post those on Phab and start a thread to sample opinions. Sounds good to me! I also think having bfloat (all floating point types actually) in GISel (i.e., LLT) is also the right long term plan. Maybe...

Hi, GlobalISel currently drops all type information relating to the integer/FP distinction during the IR translation pass, as the LLT types only represent whether a value is a scalar/vector/pointer and it’s size/shape. To compensate, later passes use the FP operations on those values to guess what kind of value is being stored within that virtual register. This means that i32/float loads get

...ot have the blocks! Doing a fcsk does not fix the situation as shown below. This is on a RH7.3 machine running a 2.4.20 kernel root at 1[bh-sm5-per-006]# debugfs /dev/hda4 debugfs 1.35-WIP (31-Jan-2004) debugfs: stat users/greve/fbirn-hp-fsfast/dunc-data/dunc-101.1/bold/bh-sm5-per-006/beta_008.bfloat Inode: 5980769 Type: regular Mode: 0777 Flags: 0x0 Generation: 4199683 297 User: 5652 Group: 5652 Size: 4243456 File ACL: 0 Directory ACL: 0 Links: 1 Blockcount: 232 Fragment: Address: 0 Number: 0 Size: 0 ctime: 0x406066ae -- Tue Mar 23 11:32:46 2004 atime: 0x40e5b58c -...