https://llvm.org/bugs/show_bug.cgi?id=30635
Bug ID: 30635
Summary: Improve *_EXTEND_VECTOR_INREG codegen for illegal
types
Product: libraries
Version: trunk
Hardware: PC
OS: Windows NT
Status: NEW
Severity: normal
Priority: P
Component: Backend: X86
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected], [email protected],
[email protected], [email protected]
Classification: Unclassified
https://reviews.llvm.org/D25268 fixed integer promotion but the codegen is
still pretty nasty - there is no need for us to vectorize in this case, we
should just call movzbl/movsbl load+ext:
define <3 x i16> @zext_i8(<3 x i8>) {
%2 = zext <3 x i8> %0 to <3 x i16>
ret <3 x i16> %2
}
zext_i8:
vpxor %xmm0, %xmm0, %xmm0
vpinsrb $0, 4(%esp), %xmm0, %xmm0
vpinsrb $4, 8(%esp), %xmm0, %xmm0
vpinsrb $8, 12(%esp), %xmm0, %xmm0
vmovd %xmm0, %eax
vpextrw $2, %xmm0, %edx
vpextrw $4, %xmm0, %ecx
retl
define <3 x i16> @sext_i8(<3 x i8>) {
%2 = sext <3 x i8> %0 to <3 x i16>
ret <3 x i16> %2
}
sext_i8:
vpinsrb $0, 4(%esp), %xmm0, %xmm0
vpinsrb $4, 8(%esp), %xmm0, %xmm0
vpinsrb $8, 12(%esp), %xmm0, %xmm0
vpslld $24, %xmm0, %xmm0
vpsrad $24, %xmm0, %xmm0
vmovd %xmm0, %eax
vpextrw $2, %xmm0, %edx
vpextrw $4, %xmm0, %ecx
retl
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
llvm-bugs mailing list
[email protected]
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
