[llvm-dev] wrong offset addition for store 64B

hameeza ahmed via llvm-dev llvm-dev at lists.llvm.org
Wed Aug 22 13:47:52 PDT 2018


Hello,
I have implemented store instruction for 16xi32 64B. it works finely except
that it has wrong offset addition. the offset should be 64 but here i m
getting it 4.

Legalizing: t277: ch = store<ST64[FixedStack7(align=64)+4](align=4)> t0,
t322, t276, undef:i64

Legalizing: t276: i64 = add FrameIndex:i64<7>, Constant:i64<4>


here it should be
Legalizing: t276: i64 = add FrameIndex:i64<7>, Constant:i64<*64*>

Why is that so?

Where am I doing mistake? otherwise store 128B works fine in this code of
matrix multiplication.

Please help.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20180823/7d2a7f5c/attachment.html>


More information about the llvm-dev mailing list