[llvm] [X86] Truncate i64 add to i32 when upper 33 bits are zeros (PR #144066)

[Phoebe Wang via llvm-commits]

================
@@ -0,0 +1,97 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-unknown-unknown | FileCheck %s --check-prefix=X86
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=X64
+
+define i64 @test1(i16 %a) {
+; X86-LABEL: test1:
+; X86:       # %bb.0:
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $42, %eax
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    retl
+;
+; X64-LABEL: test1:
+; X64:       # %bb.0:
+; X64-NEXT:    movzwl %di, %eax
+; X64-NEXT:    addl $42, %eax
+; X64-NEXT:    retq
+  %zext_a = zext i16 %a to i64
+  %sum = add nuw nsw i64 %zext_a, 42
+  ret i64 %sum
+}
+
+define i64 @test2(i16 %a, i16 %b) {
+; X86-LABEL: test2:
+; X86:       # %bb.0:
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl %ecx, %eax
+; X86-NEXT:    xorl %edx, %edx
+; X86-NEXT:    retl
+;
+; X64-LABEL: test2:
+; X64:       # %bb.0:
+; X64-NEXT:    movzwl %si, %ecx
+; X64-NEXT:    movzwl %di, %eax
+; X64-NEXT:    addl %ecx, %eax
+; X64-NEXT:    retq
+  %zext_a = zext i16 %a to i64
+  %zext_b = zext i16 %b to i64
+; First 48 bits are all zeros so we can safely truncate to 32 bit additon
+  %sum = add nuw nsw i64 %zext_a, %zext_b
+  ret i64 %sum
+}
+
+define i64 @test3(i16 %a) {
+; X86-LABEL: test3:
+; X86:       # %bb.0:
+; X86-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $42, %eax
+; X86-NEXT:    movl $1, %edx
+; X86-NEXT:    retl
+;
+; X64-LABEL: test3:
+; X64:       # %bb.0:
+; X64-NEXT:    movzwl %di, %ecx
+; X64-NEXT:    movabsq $4294967338, %rax # imm = 0x10000002A
+; X64-NEXT:    addq %rcx, %rax
+; X64-NEXT:    retq
+  %zext_a = zext i16 %a to i64
+; Set the 32nd bit of a to force 64 bit addition, we do not truncate to 32 bit addition in this case
+  %or_a = or i64 %zext_a, 4294967296
+  %sum = add nuw nsw i64 %or_a, 42
+  ret i64 %sum
+}
+
+define i64 @test4(i16 %a, i16 %b) {
----------------
phoebewang wrote:

Add `nounwind` to suppress CFI directives.

https://github.com/llvm/llvm-project/pull/144066