X86: Try to use a smaller encoding by transforming (X << C1) & C2 into (X & (C2 >> C1)) & C1. (Part of PR5039)

This tends to happen a lot with bitfield code generated by clang. A simple example for x86_64 is
uint64_t foo(uint64_t x) { return (x&1) << 42; }
which used to compile into bloated code:
	shlq	$42, %rdi               ## encoding: [0x48,0xc1,0xe7,0x2a]
	movabsq	$4398046511104, %rax    ## encoding: [0x48,0xb8,0x00,0x00,0x00,0x00,0x00,0x04,0x00,0x00]
	andq	%rdi, %rax              ## encoding: [0x48,0x21,0xf8]
	ret                             ## encoding: [0xc3]

with this patch we can fold the immediate into the and:
	andq	$1, %rdi                ## encoding: [0x48,0x83,0xe7,0x01]
	movq	%rdi, %rax              ## encoding: [0x48,0x89,0xf8]
	shlq	$42, %rax               ## encoding: [0x48,0xc1,0xe0,0x2a]
	ret                             ## encoding: [0xc3]

It's possible to save another byte by using 'andl' instead of 'andq' but I currently see no way of doing
that without making this code even more complicated. See the TODOs in the code.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@129990 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 101 insertions, 0 deletions

diff --git a/test/CodeGen/X86/narrow-shl-cst.ll b/test/CodeGen/X86/narrow-shl-cst.ll
new file mode 100644
index 0000000..a404f34
--- /dev/null
+++ b/test/CodeGen/X86/narrow-shl-cst.ll
@@ -0,0 +1,101 @@
+; RUN: llc < %s -march=x86-64 | FileCheck %s
+; PR5039
+
+define i32 @test1(i32 %x) nounwind {
+  %and = shl i32 %x, 10
+  %shl = and i32 %and, 31744
+  ret i32 %shl
+; CHECK: test1:
+; CHECK: andl $31
+; CHECK: shll $10
+}
+
+define i32 @test2(i32 %x) nounwind {
+  %or = shl i32 %x, 10
+  %shl = or i32 %or, 31744
+  ret i32 %shl
+; CHECK: test2:
+; CHECK: orl $31
+; CHECK: shll $10
+}
+
+define i32 @test3(i32 %x) nounwind {
+  %xor = shl i32 %x, 10
+  %shl = xor i32 %xor, 31744
+  ret i32 %shl
+; CHECK: test3:
+; CHECK: xorl $31
+; CHECK: shll $10
+}
+
+define i64 @test4(i64 %x) nounwind {
+  %and = shl i64 %x, 40
+  %shl = and i64 %and, 264982302294016
+  ret i64 %shl
+; CHECK: test4:
+; CHECK: andq $241
+; CHECK: shlq $40
+}
+
+define i64 @test5(i64 %x) nounwind {
+  %and = shl i64 %x, 40
+  %shl = and i64 %and, 34084860461056
+  ret i64 %shl
+; CHECK: test5:
+; CHECK: andq $31
+; CHECK: shlq $40
+}
+
+define i64 @test6(i64 %x) nounwind {
+  %and = shl i64 %x, 32
+  %shl = and i64 %and, -281474976710656
+  ret i64 %shl
+; CHECK: test6:
+; CHECK: andq $-65536
+; CHECK: shlq $32
+}
+
+define i64 @test7(i64 %x) nounwind {
+  %or = shl i64 %x, 40
+  %shl = or i64 %or, 264982302294016
+  ret i64 %shl
+; CHECK: test7:
+; CHECK: orq $241
+; CHECK: shlq $40
+}
+
+define i64 @test8(i64 %x) nounwind {
+  %or = shl i64 %x, 40
+  %shl = or i64 %or, 34084860461056
+  ret i64 %shl
+; CHECK: test8:
+; CHECK: orq $31
+; CHECK: shlq $40
+}
+
+define i64 @test9(i64 %x) nounwind {
+  %xor = shl i64 %x, 40
+  %shl = xor i64 %xor, 264982302294016
+  ret i64 %shl
+; CHECK: test9:
+; CHECK: orq $241
+; CHECK: shlq $40
+}
+
+define i64 @test10(i64 %x) nounwind {
+  %xor = shl i64 %x, 40
+  %shl = xor i64 %xor, 34084860461056
+  ret i64 %shl
+; CHECK: test10:
+; CHECK: xorq $31
+; CHECK: shlq $40
+}
+
+define i64 @test11(i64 %x) nounwind {
+  %xor = shl i64 %x, 33
+  %shl = xor i64 %xor, -562949953421312
+  ret i64 %shl
+; CHECK: test11:
+; CHECK: xorq $-65536
+; CHECK: shlq $33
+}