trans: Optimize initialization using repeat expressions

This elides initialization for zero-sized arrays:
* for zero-sized elements we previously emitted an empty loop
* for arrays with a length of zero we previously emitted a loop with zero
  iterations

This emits llvm.memset() instead of a loop over each element when:
* all elements are zero integers
* elements are byte sized

diff --git a/src/librustc_trans/common.rs b/src/librustc_trans/common.rs
index 9b0803908b162582e1dc86bbdc152f4b4f3a4dc5..c5f69bd638f3c75af86886d13180a391a0fa0f23 100644 (file)
--- a/src/librustc_trans/common.rs
+++ b/src/librustc_trans/common.rs
@@ -372,7 +372,7 @@ pub fn const_to_uint(v: ValueRef) -> u64 {
     }
 }
 
-fn is_const_integral(v: ValueRef) -> bool {
+pub fn is_const_integral(v: ValueRef) -> bool {
     unsafe {
         !llvm::LLVMIsAConstantInt(v).is_null()
     }

diff --git a/src/librustc_trans/mir/rvalue.rs b/src/librustc_trans/mir/rvalue.rs
index 4bd5091a4f35f8069ea4225e58a17a3a07916468..2cae2150885a604677474156edba1d81370821da 100644 (file)
--- a/src/librustc_trans/mir/rvalue.rs
+++ b/src/librustc_trans/mir/rvalue.rs
@@ -11,7 +11,7 @@
 use llvm::{self, ValueRef};
 use rustc::ty::{self, Ty};
 use rustc::ty::cast::{CastTy, IntTy};
-use rustc::ty::layout::{Layout, LayoutTyper};
+use rustc::ty::layout::{self, Layout, LayoutTyper, Primitive};
 use rustc::mir::tcx::LvalueTy;
 use rustc::mir;
 use rustc::middle::lang_items::ExchangeMallocFnLangItem;
@@ -20,7 +20,7 @@
 use builder::Builder;
 use callee;
 use common::{self, val_ty, C_bool, C_null, C_uint};
-use common::{C_integral};
+use common::{C_integral, C_i32};
 use adt;
 use machine;
 use monomorphize;
@@ -93,12 +93,47 @@ pub fn trans_rvalue(&mut self,
             }
 
             mir::Rvalue::Repeat(ref elem, ref count) => {
+                let dest_ty = dest.ty.to_ty(bcx.tcx());
+
+                // No need to inizialize memory of a zero-sized slice
+                if common::type_is_zero_size(bcx.ccx, dest_ty) {
+                    return bcx;
+                }
+
                 let tr_elem = self.trans_operand(&bcx, elem);
                 let size = count.as_u64(bcx.tcx().sess.target.uint_type);
                 let size = C_uint(bcx.ccx, size);
                 let base = base::get_dataptr(&bcx, dest.llval);
+                let align = dest.alignment.to_align();
+
+                if let OperandValue::Immediate(v) = tr_elem.val {
+                    if common::is_const_integral(v) && common::const_to_uint(v) == 0 {
+                        let align = align.unwrap_or_else(|| bcx.ccx.align_of(tr_elem.ty));
+                        let align = C_i32(bcx.ccx, align as i32);
+                        let ty = type_of::type_of(bcx.ccx, dest_ty);
+                        let size = machine::llsize_of(bcx.ccx, ty);
+                        let fill = C_integral(Type::i8(bcx.ccx), 0, false);
+                        base::call_memset(&bcx, base, fill, size, align, false);
+                        return bcx;
+                    }
+                }
+
+                // Use llvm.memset.p0i8.* to initialize byte arrays
+                let elem_layout = bcx.ccx.layout_of(tr_elem.ty).layout;
+                match *elem_layout {
+                    Layout::Scalar { value: Primitive::Int(layout::I8), .. } |
+                    Layout::CEnum { discr: layout::I8, .. } => {
+                        let align = align.unwrap_or_else(|| bcx.ccx.align_of(tr_elem.ty));
+                        let align = C_i32(bcx.ccx, align as i32);
+                        let fill = tr_elem.immediate();
+                        base::call_memset(&bcx, base, fill, size, align, false);
+                        return bcx;
+                    }
+                    _ => ()
+                }
+
                 tvec::slice_for_each(&bcx, base, tr_elem.ty, size, |bcx, llslot, loop_bb| {
-                    self.store_operand(bcx, llslot, dest.alignment.to_align(), tr_elem);
+                    self.store_operand(bcx, llslot, align, tr_elem);
                     bcx.br(loop_bb);
                 })
             }

diff --git a/src/test/codegen/slice-init.rs b/src/test/codegen/slice-init.rs
new file mode 100644 (file)
index 0000000..cb684af
--- /dev/null
+++ b/src/test/codegen/slice-init.rs
@@ -0,0 +1,74 @@
+// Copyright 2017 The Rust Project Developers. See the COPYRIGHT
+// file at the top-level directory of this distribution and at
+// http://rust-lang.org/COPYRIGHT.
+//
+// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
+// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
+// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
+// option. This file may not be copied, modified, or distributed
+// except according to those terms.
+
+// compile-flags: -C no-prepopulate-passes
+
+#![crate_type = "lib"]
+
+// CHECK-LABEL: @zero_sized_elem
+#[no_mangle]
+pub fn zero_sized_elem() {
+    // CHECK-NOT: br label %slice_loop_header{{.*}}
+    // CHECK-NOT: call void @llvm.memset.p0i8
+    let x = [(); 4];
+    drop(&x);
+}
+
+// CHECK-LABEL: @zero_len_array
+#[no_mangle]
+pub fn zero_len_array() {
+    // CHECK-NOT: br label %slice_loop_header{{.*}}
+    // CHECK-NOT: call void @llvm.memset.p0i8
+    let x = [4; 0];
+    drop(&x);
+}
+
+// CHECK-LABEL: @byte_array
+#[no_mangle]
+pub fn byte_array() {
+    // CHECK: call void @llvm.memset.p0i8.i{{[0-9]+}}(i8* {{.*}}, i8 7, i64 4
+    // CHECK-NOT: br label %slice_loop_header{{.*}}
+    let x = [7u8; 4];
+    drop(&x);
+}
+
+#[allow(dead_code)]
+#[derive(Copy, Clone)]
+enum Init {
+    Loop,
+    Memset,
+}
+
+// CHECK-LABEL: @byte_enum_array
+#[no_mangle]
+pub fn byte_enum_array() {
+    // CHECK: call void @llvm.memset.p0i8.i{{[0-9]+}}(i8* {{.*}}, i8 {{.*}}, i64 4
+    // CHECK-NOT: br label %slice_loop_header{{.*}}
+    let x = [Init::Memset; 4];
+    drop(&x);
+}
+
+// CHECK-LABEL: @zeroed_integer_array
+#[no_mangle]
+pub fn zeroed_integer_array() {
+    // CHECK: call void @llvm.memset.p0i8.i{{[0-9]+}}(i8* {{.*}}, i8 0, i64 16
+    // CHECK-NOT: br label %slice_loop_header{{.*}}
+    let x = [0u32; 4];
+    drop(&x);
+}
+
+// CHECK-LABEL: @nonzero_integer_array
+#[no_mangle]
+pub fn nonzero_integer_array() {
+    // CHECK: br label %slice_loop_header{{.*}}
+    // CHECK-NOT: call void @llvm.memset.p0i8
+    let x = [0x1a_2b_3c_4d_u32; 4];
+    drop(&x);
+}