Rollup merge of rust-lang#117671 - kjetilkjeka:nvptx_c_abi_avoid_direct, r=davidtwco

NVPTX: Avoid PassMode::Direct for args in C abi

Fixes rust-lang#117480

I must admit that I'm confused about `PassMode` altogether, is there a good sum-up threads for this anywhere? I'm especially confused about how "indirect" and "byval" goes together. To me it seems like "indirect" basically means "use a indirection through a pointer", while "byval" basically means "do not use indirection through a pointer".

The return used to keep `PassMode::Direct` for small aggregates. It turns out that `make_indirect` messes up the tests and one way to fix it is to keep `PassMode::Direct` for all aggregates. I have mostly seen this PassMode mentioned for args. Is it also a problem for returns? When experimenting with `byval` as an alternative i ran into [this assert](https://github.com/rust-lang/rust/blob/61a3eea8043cc1c7a09c2adda884e27ffa8a1172/compiler/rustc_codegen_llvm/src/abi.rs#L463C22-L463C22)

I have added tests for the same kind of types that is already tested for the "ptx-kernel" abi. The tests cannot be enabled until something like rust-lang#117458 is completed and merged.

CC: `@RalfJung` since you seem to be the expert on this and have already helped me out tremendously

CC: `@RDambrosio016` in case this influence your work on `rustc_codegen_nvvm`

`@rustbot` label +O-NVPTX

Author: matthiaskrgr

compiler/rustc_target/src/abi/call/nvptx64.rs

@@ -11,11 +11,10 @@ fn classify_ret<Ty>(ret: &mut ArgAbi<'_, Ty>) {
 }
 
 fn classify_arg<Ty>(arg: &mut ArgAbi<'_, Ty>) {
-    if arg.layout.is_aggregate() && arg.layout.size.bits() > 64 {
-        arg.make_indirect();
-    } else {
-        // FIXME: this is wrong! Need to decide which ABI we really want here.
-        arg.make_direct_deprecated();
+    if arg.layout.is_aggregate() {
+        arg.make_indirect_byval(None);
+    } else if arg.layout.size.bits() < 32 {
+        arg.extend_integer_width_to(32);
     }
 }
 

tests/assembly/nvptx-c-abi-arg-v7.rs

@@ -0,0 +1,206 @@
+//@ assembly-output: ptx-linker
+//@ compile-flags: --crate-type cdylib -C target-cpu=sm_86 -Z unstable-options -Clinker-flavor=llbc
+//@ only-nvptx64
+
+// The PTX ABI stability is tied to major versions of the PTX ISA
+// These tests assume major version 7
+
+// CHECK: .version 7
+
+#![feature(abi_ptx, lang_items, no_core)]
+#![no_core]
+
+#[lang = "sized"]
+trait Sized {}
+#[lang = "copy"]
+trait Copy {}
+
+#[repr(C)]
+pub struct SingleU8 {
+    f: u8,
+}
+
+#[repr(C)]
+pub struct DoubleU8 {
+    f: u8,
+    g: u8,
+}
+
+#[repr(C)]
+pub struct TripleU8 {
+    f: u8,
+    g: u8,
+    h: u8,
+}
+
+#[repr(C)]
+pub struct TripleU16 {
+    f: u16,
+    g: u16,
+    h: u16,
+}
+#[repr(C)]
+pub struct TripleU32 {
+    f: u32,
+    g: u32,
+    h: u32,
+}
+#[repr(C)]
+pub struct TripleU64 {
+    f: u64,
+    g: u64,
+    h: u64,
+}
+
+#[repr(C)]
+pub struct DoubleFloat {
+    f: f32,
+    g: f32,
+}
+
+#[repr(C)]
+pub struct TripleFloat {
+    f: f32,
+    g: f32,
+    h: f32,
+}
+
+#[repr(C)]
+pub struct TripleDouble {
+    f: f64,
+    g: f64,
+    h: f64,
+}
+
+#[repr(C)]
+pub struct ManyIntegers {
+    f: u8,
+    g: u16,
+    h: u32,
+    i: u64,
+}
+
+#[repr(C)]
+pub struct ManyNumerics {
+    f: u8,
+    g: u16,
+    h: u32,
+    i: u64,
+    j: f32,
+    k: f64,
+}
+
+// CHECK: .visible .func f_u8_arg(
+// CHECK: .param .b32 f_u8_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_u8_arg(_a: u8) {}
+
+// CHECK: .visible .func f_u16_arg(
+// CHECK: .param .b32 f_u16_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_u16_arg(_a: u16) {}
+
+// CHECK: .visible .func f_u32_arg(
+// CHECK: .param .b32 f_u32_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_u32_arg(_a: u32) {}
+
+// CHECK: .visible .func f_u64_arg(
+// CHECK: .param .b64 f_u64_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_u64_arg(_a: u64) {}
+
+// CHECK: .visible .func f_u128_arg(
+// CHECK: .param .align 16 .b8 f_u128_arg_param_0[16]
+#[no_mangle]
+pub unsafe extern "C" fn f_u128_arg(_a: u128) {}
+
+// CHECK: .visible .func f_i8_arg(
+// CHECK: .param .b32 f_i8_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_i8_arg(_a: i8) {}
+
+// CHECK: .visible .func f_i16_arg(
+// CHECK: .param .b32 f_i16_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_i16_arg(_a: i16) {}
+
+// CHECK: .visible .func f_i32_arg(
+// CHECK: .param .b32 f_i32_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_i32_arg(_a: i32) {}
+
+// CHECK: .visible .func f_i64_arg(
+// CHECK: .param .b64 f_i64_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_i64_arg(_a: i64) {}
+
+// CHECK: .visible .func f_i128_arg(
+// CHECK: .param .align 16 .b8 f_i128_arg_param_0[16]
+#[no_mangle]
+pub unsafe extern "C" fn f_i128_arg(_a: i128) {}
+
+// CHECK: .visible .func f_f32_arg(
+// CHECK: .param .b32 f_f32_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_f32_arg(_a: f32) {}
+
+// CHECK: .visible .func f_f64_arg(
+// CHECK: .param .b64 f_f64_arg_param_0
+#[no_mangle]
+pub unsafe extern "C" fn f_f64_arg(_a: f64) {}
+
+// CHECK: .visible .func f_single_u8_arg(
+// CHECK: .param .align 1 .b8 f_single_u8_arg_param_0[1]
+#[no_mangle]
+pub unsafe extern "C" fn f_single_u8_arg(_a: SingleU8) {}
+
+// CHECK: .visible .func f_double_u8_arg(
+// CHECK: .param .align 1 .b8 f_double_u8_arg_param_0[2]
+#[no_mangle]
+pub unsafe extern "C" fn f_double_u8_arg(_a: DoubleU8) {}
+
+// CHECK: .visible .func f_triple_u8_arg(
+// CHECK: .param .align 1 .b8 f_triple_u8_arg_param_0[3]
+#[no_mangle]
+pub unsafe extern "C" fn f_triple_u8_arg(_a: TripleU8) {}
+
+// CHECK: .visible .func f_triple_u16_arg(
+// CHECK: .param .align 2 .b8 f_triple_u16_arg_param_0[6]
+#[no_mangle]
+pub unsafe extern "C" fn f_triple_u16_arg(_a: TripleU16) {}
+
+// CHECK: .visible .func f_triple_u32_arg(
+// CHECK: .param .align 4 .b8 f_triple_u32_arg_param_0[12]
+#[no_mangle]
+pub unsafe extern "C" fn f_triple_u32_arg(_a: TripleU32) {}
+
+// CHECK: .visible .func f_triple_u64_arg(
+// CHECK: .param .align 8 .b8 f_triple_u64_arg_param_0[24]
+#[no_mangle]
+pub unsafe extern "C" fn f_triple_u64_arg(_a: TripleU64) {}
+
+// CHECK: .visible .func f_many_integers_arg(
+// CHECK: .param .align 8 .b8 f_many_integers_arg_param_0[16]
+#[no_mangle]
+pub unsafe extern "C" fn f_many_integers_arg(_a: ManyIntegers) {}
+
+// CHECK: .visible .func f_double_float_arg(
+// CHECK: .param .align 4 .b8 f_double_float_arg_param_0[8]
+#[no_mangle]
+pub unsafe extern "C" fn f_double_float_arg(_a: DoubleFloat) {}
+
+// CHECK: .visible .func f_triple_float_arg(
+// CHECK: .param .align 4 .b8 f_triple_float_arg_param_0[12]
+#[no_mangle]
+pub unsafe extern "C" fn f_triple_float_arg(_a: TripleFloat) {}
+
+// CHECK: .visible .func f_triple_double_arg(
+// CHECK: .param .align 8 .b8 f_triple_double_arg_param_0[24]
+#[no_mangle]
+pub unsafe extern "C" fn f_triple_double_arg(_a: TripleDouble) {}
+
+// CHECK: .visible .func f_many_numerics_arg(
+// CHECK: .param .align 8 .b8 f_many_numerics_arg_param_0[32]
+#[no_mangle]
+pub unsafe extern "C" fn f_many_numerics_arg(_a: ManyNumerics) {}