https://llvm.org/bugs/show_bug.cgi?id=28963
Bug ID: 28963
Summary: AVX512: Update __vectorcall calling conventions
Product: libraries
Version: trunk
Hardware: PC
OS: All
Status: NEW
Severity: normal
Priority: P
Component: Backend: X86
Assignee: unassignedb...@nondot.org
Reporter: wenzel.ja...@epfl.ch
CC: llvm-bugs@lists.llvm.org
Classification: Unclassified
Clang supports the __vectorcall calling convention which can be used to
directly invoke functions which accept SIMD vectors in registers. This works
great for SSE and AVX vectors, but it looks like the calling convention wasn't
yet updated to the new AVX512 512 bit vectors:
Consider the following simple passthrough functions:
struct Wrapper256 { __m256 x; };
struct Wrapper512 { __m512 x; };
Wrapper256 __vectorcall test1(Wrapper256 x) { return x; }
Wrapper512 __vectorcall test2(Wrapper512 x) { return x; }
Using Clang trunk (clang -O3 -mavx512f -fomit-frame-pointer test.cpp -S), this
assembles to:
__Z4add110Wrapper256: ## @_Z4add110Wrapper256
retq
__Z4add210Wrapper512: ## @_Z4add210Wrapper512
pushq %rbp
movq %rsp, %rbp
andq $-64, %rsp
subq $64, %rsp
vmovaps 16(%rbp), %ymm0
vmovaps 48(%rbp), %ymm1
vmovaps %ymm1, 32(%rdi)
vmovaps %ymm0, (%rdi)
movq %rdi, %rax
movq %rbp, %rsp
popq %rbp
retq
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
