mikros/rust
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Auto merge of #115236 - scottmcm:less-vector, r=compiler-errors

Browse Source 
Stop emitting non-power-of-two vectors in (non-portable-SIMD) codegen

Fixes #115212

It's unclear what makes this not work sometimes, since it often *does* work, so for now just disable the unusual cases.  A future PR can consider doing something smarter, but this is an easy and safe tweak that we can do to resolve the regressions for now.
This commit is contained in:
bors 2023-08-26 06:02:13 +00:00
commit 766c0c0b83
3 changed files with 43 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
compiler/rustc_codegen_llvm/src/type_of.rs
View File

@ -405,7 +405,11 @@ impl<'tcx> LayoutLlvmExt<'tcx> for TyAndLayout<'tcx> {
// Vectors, even for non-power-of-two sizes, have the same layout as
// arrays but don't count as aggregate types
// While LLVM theoretically supports non-power-of-two sizes, and they
// often work fine, sometimes x86-isel deals with them horribly
// (see #115212) so for now only use power-of-two ones.
if let FieldsShape::Array { count, .. } = self.layout.fields()
&& count.is_power_of_two()
&& let element = self.field(cx, 0)
&& element.ty.is_integral()
{

21
tests/codegen/mem-replace-simple-type.rs
View File

@ -33,12 +33,21 @@ pub fn replace_ref_str<'a>(r: &mut &'a str, v: &'a str) -> &'a str {
}
#[no_mangle]
// CHECK-LABEL: @replace_short_array(
pub fn replace_short_array(r: &mut [u32; 3], v: [u32; 3]) -> [u32; 3] {
// CHECK-LABEL: @replace_short_array_3(
pub fn replace_short_array_3(r: &mut [u32; 3], v: [u32; 3]) -> [u32; 3] {
// CHECK-NOT: alloca
// CHECK: %[[R:.+]] = load <3 x i32>, ptr %r, align 4
// CHECK: store <3 x i32> %[[R]], ptr %result
// CHECK: %[[V:.+]] = load <3 x i32>, ptr %v, align 4
// CHECK: store <3 x i32> %[[V]], ptr %r
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 4 %result, ptr align 4 %r, i64 12, i1 false)
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 4 %r, ptr align 4 %v, i64 12, i1 false)
std::mem::replace(r, v)
}
#[no_mangle]
// CHECK-LABEL: @replace_short_array_4(
pub fn replace_short_array_4(r: &mut [u32; 4], v: [u32; 4]) -> [u32; 4] {
// CHECK-NOT: alloca
// CHECK: %[[R:.+]] = load <4 x i32>, ptr %r, align 4
// CHECK: store <4 x i32> %[[R]], ptr %result
// CHECK: %[[V:.+]] = load <4 x i32>, ptr %v, align 4
// CHECK: store <4 x i32> %[[V]], ptr %r
std::mem::replace(r, v)
}

33
tests/codegen/swap-small-types.rs
View File

@ -11,11 +11,12 @@ type RGB48 = [u16; 3];
// CHECK-LABEL: @swap_rgb48_manually(
#[no_mangle]
pub fn swap_rgb48_manually(x: &mut RGB48, y: &mut RGB48) {
// CHECK-NOT: alloca
// CHECK: %[[TEMP0:.+]] = load <3 x i16>, ptr %x, align 2
// CHECK: %[[TEMP1:.+]] = load <3 x i16>, ptr %y, align 2
// CHECK: store <3 x i16> %[[TEMP1]], ptr %x, align 2
// CHECK: store <3 x i16> %[[TEMP0]], ptr %y, align 2
// FIXME: See #115212 for why this has an alloca again
// CHECK: alloca [3 x i16], align 2
// CHECK: call void @llvm.memcpy.p0.p0.i64({{.+}}, i64 6, i1 false)
// CHECK: call void @llvm.memcpy.p0.p0.i64({{.+}}, i64 6, i1 false)
// CHECK: call void @llvm.memcpy.p0.p0.i64({{.+}}, i64 6, i1 false)
let temp = *x;
*x = *y;
@ -25,11 +26,25 @@ pub fn swap_rgb48_manually(x: &mut RGB48, y: &mut RGB48) {
// CHECK-LABEL: @swap_rgb48
#[no_mangle]
pub fn swap_rgb48(x: &mut RGB48, y: &mut RGB48) {
// FIXME: See #115212 for why this has an alloca again
// CHECK: alloca [3 x i16], align 2
// CHECK: call void @llvm.memcpy.p0.p0.i64({{.+}}, i64 6, i1 false)
// CHECK: call void @llvm.memcpy.p0.p0.i64({{.+}}, i64 6, i1 false)
// CHECK: call void @llvm.memcpy.p0.p0.i64({{.+}}, i64 6, i1 false)
swap(x, y)
}
type RGBA64 = [u16; 4];
// CHECK-LABEL: @swap_rgba64
#[no_mangle]
pub fn swap_rgba64(x: &mut RGBA64, y: &mut RGBA64) {
// CHECK-NOT: alloca
// CHECK: load <3 x i16>
// CHECK: load <3 x i16>
// CHECK: store <3 x i16>
// CHECK: store <3 x i16>
// CHECK-DAG: %[[XVAL:.+]] = load <4 x i16>, ptr %x, align 2
// CHECK-DAG: %[[YVAL:.+]] = load <4 x i16>, ptr %y, align 2
// CHECK-DAG: store <4 x i16> %[[YVAL]], ptr %x, align 2
// CHECK-DAG: store <4 x i16> %[[XVAL]], ptr %y, align 2
swap(x, y)
}