maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[X86] Drop the 64-bit alignment for LOW32_ADDR_ACCESS register class. 

The only 64-bit register in that register class is RIP and it will not
get spilled in the current ABIs.

llvm-svn: 268963
This commit is contained in:
Quentin Colombet 2016-05-09 19:50:30 +00:00
parent 0c0ada01b6
commit 3126db6fd7
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
llvm/lib/Target/X86/X86RegisterInfo.td
View File

@ -421,11 +421,12 @@ def GR64_NOREX_NOSP : RegisterClass<"X86", [i64], 64,
// In such cases, it is fine to use RIP as we are sure the 32 high
// bits are not set. We do not need variants for NOSP as RIP is not
// allowed there.
// Alignment is 64 because we have RIP.
// RIP is not spilled anywhere for now, so stick to 32-bit alignment
// to save on memory space.
// FIXME: We could allow all 64bit registers, but we would need
// something to check that the 32 high bits are not set,
// which we do not have right now.
def LOW32_ADDR_ACCESS : RegisterClass<"X86", [i32], 64, (add GR32, RIP)>;
def LOW32_ADDR_ACCESS : RegisterClass<"X86", [i32], 32, (add GR32, RIP)>;
// When RBP is used as a base pointer in a 32-bit addresses environement,
// this is also safe to use the full register to access addresses.