diff options
Diffstat (limited to 'tests/codegen')
| -rw-r--r-- | tests/codegen/array-map.rs | 49 | ||||
| -rw-r--r-- | tests/codegen/autovectorize-f32x4.rs | 13 | ||||
| -rw-r--r-- | tests/codegen/consts.rs | 1 | ||||
| -rw-r--r-- | tests/codegen/merge-functions.rs | 1 | ||||
| -rw-r--r-- | tests/codegen/sse42-implies-crc32.rs | 1 | ||||
| -rw-r--r-- | tests/codegen/uninit-consts.rs | 1 | ||||
| -rw-r--r-- | tests/codegen/vec-in-place.rs | 1 |
7 files changed, 61 insertions, 6 deletions
diff --git a/tests/codegen/array-map.rs b/tests/codegen/array-map.rs new file mode 100644 index 00000000000..9298e89e397 --- /dev/null +++ b/tests/codegen/array-map.rs @@ -0,0 +1,49 @@ +// compile-flags: -C opt-level=3 -C target-cpu=x86-64-v3 +// no-system-llvm +// only-x86_64 +// ignore-debug (the extra assertions get in the way) + +#![crate_type = "lib"] +#![feature(array_zip)] + +// CHECK-LABEL: @short_integer_map +#[no_mangle] +pub fn short_integer_map(x: [u32; 8]) -> [u32; 8] { + // CHECK: load <8 x i32> + // CHECK: shl <8 x i32> + // CHECK: or <8 x i32> + // CHECK: store <8 x i32> + x.map(|x| 2 * x + 1) +} + +// CHECK-LABEL: @short_integer_zip_map +#[no_mangle] +pub fn short_integer_zip_map(x: [u32; 8], y: [u32; 8]) -> [u32; 8] { + // CHECK: %[[A:.+]] = load <8 x i32> + // CHECK: %[[B:.+]] = load <8 x i32> + // CHECK: sub <8 x i32> %[[A]], %[[B]] + // CHECK: store <8 x i32> + x.zip(y).map(|(x, y)| x - y) +} + +// This test is checking that LLVM can SRoA away a bunch of the overhead, +// like fully moving the iterators to registers. Notably, previous implementations +// of `map` ended up `alloca`ing the whole `array::IntoIterator`, meaning both a +// hard-to-eliminate `memcpy` and that the iteration counts needed to be written +// out to stack every iteration, even for infallible operations on `Copy` types. +// +// This is still imperfect, as there's more copies than would be ideal, +// but hopefully work like #103830 will improve that in future, +// and update this test to be stricter. +// +// CHECK-LABEL: @long_integer_map +#[no_mangle] +pub fn long_integer_map(x: [u32; 64]) -> [u32; 64] { + // CHECK: start: + // CHECK-NEXT: alloca [64 x i32] + // CHECK-NEXT: alloca %"core::mem::manually_drop::ManuallyDrop<[u32; 64]>" + // CHECK-NOT: alloca + // CHECK: mul <{{[0-9]+}} x i32> + // CHECK: add <{{[0-9]+}} x i32> + x.map(|x| 13 * x + 7) +} diff --git a/tests/codegen/autovectorize-f32x4.rs b/tests/codegen/autovectorize-f32x4.rs index 6b09c8fc998..9ecea53f1c0 100644 --- a/tests/codegen/autovectorize-f32x4.rs +++ b/tests/codegen/autovectorize-f32x4.rs @@ -1,6 +1,7 @@ -// compile-flags: -C opt-level=3 +// compile-flags: -C opt-level=3 -Z merge-functions=disabled // only-x86_64 #![crate_type = "lib"] +#![feature(array_zip)] // CHECK-LABEL: @auto_vectorize_direct #[no_mangle] @@ -30,3 +31,13 @@ pub fn auto_vectorize_loop(a: [f32; 4], b: [f32; 4]) -> [f32; 4] { } c } + +// CHECK-LABEL: @auto_vectorize_array_zip_map +#[no_mangle] +pub fn auto_vectorize_array_zip_map(a: [f32; 4], b: [f32; 4]) -> [f32; 4] { +// CHECK: load <4 x float> +// CHECK: load <4 x float> +// CHECK: fadd <4 x float> +// CHECK: store <4 x float> + a.zip(b).map(|(a, b)| a + b) +} diff --git a/tests/codegen/consts.rs b/tests/codegen/consts.rs index 260d9de8670..d0418d11142 100644 --- a/tests/codegen/consts.rs +++ b/tests/codegen/consts.rs @@ -1,5 +1,4 @@ // compile-flags: -C no-prepopulate-passes -// min-llvm-version: 14.0 #![crate_type = "lib"] diff --git a/tests/codegen/merge-functions.rs b/tests/codegen/merge-functions.rs index 8e8fe5c964d..d6caeeee896 100644 --- a/tests/codegen/merge-functions.rs +++ b/tests/codegen/merge-functions.rs @@ -1,4 +1,3 @@ -// min-llvm-version: 14.0 // revisions: O Os //[Os] compile-flags: -Copt-level=s //[O] compile-flags: -O diff --git a/tests/codegen/sse42-implies-crc32.rs b/tests/codegen/sse42-implies-crc32.rs index 47b1a899340..56079d32a8d 100644 --- a/tests/codegen/sse42-implies-crc32.rs +++ b/tests/codegen/sse42-implies-crc32.rs @@ -1,5 +1,4 @@ // only-x86_64 -// min-llvm-version: 14.0 // compile-flags: -Copt-level=3 #![crate_type = "lib"] diff --git a/tests/codegen/uninit-consts.rs b/tests/codegen/uninit-consts.rs index 98a6761f8ab..54e9a9e9bb8 100644 --- a/tests/codegen/uninit-consts.rs +++ b/tests/codegen/uninit-consts.rs @@ -1,5 +1,4 @@ // compile-flags: -C no-prepopulate-passes -// min-llvm-version: 14.0 // Check that we use undef (and not zero) for uninitialized bytes in constants. diff --git a/tests/codegen/vec-in-place.rs b/tests/codegen/vec-in-place.rs index 5df3669056d..9992604221b 100644 --- a/tests/codegen/vec-in-place.rs +++ b/tests/codegen/vec-in-place.rs @@ -1,4 +1,3 @@ -// min-llvm-version: 14.0 // ignore-debug: the debug assertions get in the way // compile-flags: -O -Z merge-functions=disabled #![crate_type = "lib"] |
