diff options
| author | Evan Cheng <evan.cheng@apple.com> | 2008-11-27 00:49:46 +0000 |
|---|---|---|
| committer | Evan Cheng <evan.cheng@apple.com> | 2008-11-27 00:49:46 +0000 |
| commit | 884c70c912b699a4d14d05dcac1ac4f2d47c0f9a (patch) | |
| tree | eea845867e8c466b91a3bf184a9c9c1df7049e2e /test/CodeGen/X86 | |
| parent | 56eca9103ecbf2293842bfd6067ac26629221eaf (diff) | |
On x86 favors folding short immediate into some arithmetic operations (e.g. add, and, xor, etc.) because materializing an immediate in a register is expensive in turns of code size.
e.g.
movl 4(%esp), %eax
addl $4, %eax
is 2 bytes shorter than
movl $4, %eax
addl 4(%esp), %eax
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@60139 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86')
| -rw-r--r-- | test/CodeGen/X86/fold-imm.ll | 14 | ||||
| -rw-r--r-- | test/CodeGen/X86/isel-sink3.ll | 2 |
2 files changed, 15 insertions, 1 deletions
diff --git a/test/CodeGen/X86/fold-imm.ll b/test/CodeGen/X86/fold-imm.ll new file mode 100644 index 0000000000..1623f31d74 --- /dev/null +++ b/test/CodeGen/X86/fold-imm.ll @@ -0,0 +1,14 @@ +; RUN: llvm-as < %s | llc -march=x86 | grep inc +; RUN: llvm-as < %s | llc -march=x86 | grep add | grep 4 + +define i32 @test(i32 %X) nounwind { +entry: + %0 = add i32 %X, 1 + ret i32 %0 +} + +define i32 @test2(i32 %X) nounwind { +entry: + %0 = add i32 %X, 4 + ret i32 %0 +} diff --git a/test/CodeGen/X86/isel-sink3.ll b/test/CodeGen/X86/isel-sink3.ll index 75c23c3435..4e678c42cf 100644 --- a/test/CodeGen/X86/isel-sink3.ll +++ b/test/CodeGen/X86/isel-sink3.ll @@ -1,4 +1,4 @@ -; RUN: llvm-as < %s | llc | grep {addl.(%eax), %ecx} +; RUN: llvm-as < %s | llc | grep {addl.\$4, %ecx} ; RUN: llvm-as < %s | llc | not grep leal ; this should not sink %1 into bb1, that would increase reg pressure. |