Rollup merge of #136457 - calder:master, r=tgross35

Expose algebraic floating point intrinsics

# Problem

A stable Rust implementation of a simple dot product is 8x slower than C++ on modern x86-64 CPUs. The root cause is an inability to let the compiler reorder floating point operations for better vectorization.

See https://github.com/calder/dot-bench for benchmarks. Measurements below were performed on a i7-10875H.

### C++: 10us ✅

With Clang 18.1.3 and `-O2 -march=haswell`:
<table>
<tr>
    <th>C++</th>
    <th>Assembly</th>
</tr>
<tr>
<td>
<pre lang="cc">
float dot(float *a, float *b, size_t len) {
    #pragma clang fp reassociate(on)
    float sum = 0.0;
    for (size_t i = 0; i < len; ++i) {
        sum += a[i] * b[i];
    }
    return sum;
}
</pre>
</td>
<td>
<img src="https://github.com/user-attachments/assets/739573c0-380a-4d84-9fd9-141343ce7e68" />
</td>
</tr>
</table>

### Nightly Rust: 10us ✅

With rustc 1.86.0-nightly (8239a37f9) and `-C opt-level=3 -C target-feature=+avx2,+fma`:
<table>
<tr>
    <th>Rust</th>
    <th>Assembly</th>
</tr>
<tr>
<td>
<pre lang="rust">
fn dot(a: &[f32], b: &[f32]) -> f32 {
    let mut sum = 0.0;
    for i in 0..a.len() {
        sum = fadd_algebraic(sum, fmul_algebraic(a[i], b[i]));
    }
    sum
}
</pre>
</td>
<td>
<img src="https://github.com/user-attachments/assets/9dcf953a-2cd7-42f3-bc34-7117de4c5fb9" />
</td>
</tr>
</table>

### Stable Rust: 84us ❌

With rustc 1.84.1 (e71f9a9a9) and `-C opt-level=3 -C target-feature=+avx2,+fma`:
<table>
<tr>
    <th>Rust</th>
    <th>Assembly</th>
</tr>
<tr>
<td>
<pre lang="rust">
fn dot(a: &[f32], b: &[f32]) -> f32 {
    let mut sum = 0.0;
    for i in 0..a.len() {
        sum += a[i] * b[i];
    }
    sum
}
</pre>
</td>
<td>
<img src="https://github.com/user-attachments/assets/936a1f7e-33e4-4ff8-a732-c3cdfe068dca" />
</td>
</tr>
</table>

# Proposed Change

Add `core::intrinsics::f*_algebraic` wrappers to `f16`, `f32`, `f64`, and `f128` gated on a new `float_algebraic` feature.

# Alternatives Considered

https://github.com/rust-lang/rust/issues/21690 has a lot of good discussion of various options for supporting fast math in Rust, but is still open a decade later because any choice that opts in more than individual operations is ultimately contrary to Rust's design principles.

In the mean time, processors have evolved and we're leaving major performance on the table by not supporting vectorization. We shouldn't make users choose between an unstable compiler and an 8x performance hit.

# References

* https://github.com/rust-lang/rust/issues/21690
* https://github.com/rust-lang/libs-team/issues/532
* https://github.com/rust-lang/rust/issues/136469
* https://github.com/calder/dot-bench
* https://www.felixcloutier.com/x86/vfmadd132ps:vfmadd213ps:vfmadd231ps

try-job: x86_64-gnu-nopt
try-job: x86_64-gnu-aux

1 files changed, 149 insertions, 0 deletions

diff --git a/tests/codegen/float/algebraic.rs b/tests/codegen/float/algebraic.rs
new file mode 100644
index 00000000000..818a4bcdfe3
--- /dev/null
+++ b/tests/codegen/float/algebraic.rs
@@ -0,0 +1,149 @@
+// Verify that algebraic intrinsics generate the correct LLVM calls
+
+// Ensure operations get inlined
+//@ compile-flags: -Copt-level=1
+
+#![crate_type = "lib"]
+#![feature(f16)]
+#![feature(f128)]
+#![feature(float_algebraic)]
+
+// CHECK-LABEL: @f16_algebraic_add
+#[no_mangle]
+pub fn f16_algebraic_add(a: f16, b: f16) -> f16 {
+    // CHECK: fadd reassoc nsz arcp contract half %{{.+}}, %{{.+}}
+    a.algebraic_add(b)
+}
+
+// CHECK-LABEL: @f16_algebraic_sub
+#[no_mangle]
+pub fn f16_algebraic_sub(a: f16, b: f16) -> f16 {
+    // CHECK: fsub reassoc nsz arcp contract half %{{.+}}, %{{.+}}
+    a.algebraic_sub(b)
+}
+
+// CHECK-LABEL: @f16_algebraic_mul
+#[no_mangle]
+pub fn f16_algebraic_mul(a: f16, b: f16) -> f16 {
+    // CHECK: fmul reassoc nsz arcp contract half %{{.+}}, %{{.+}}
+    a.algebraic_mul(b)
+}
+
+// CHECK-LABEL: @f16_algebraic_div
+#[no_mangle]
+pub fn f16_algebraic_div(a: f16, b: f16) -> f16 {
+    // CHECK: fdiv reassoc nsz arcp contract half %{{.+}}, %{{.+}}
+    a.algebraic_div(b)
+}
+
+// CHECK-LABEL: @f16_algebraic_rem
+#[no_mangle]
+pub fn f16_algebraic_rem(a: f16, b: f16) -> f16 {
+    // CHECK: frem reassoc nsz arcp contract half %{{.+}}, %{{.+}}
+    a.algebraic_rem(b)
+}
+
+// CHECK-LABEL: @f32_algebraic_add
+#[no_mangle]
+pub fn f32_algebraic_add(a: f32, b: f32) -> f32 {
+    // CHECK: fadd reassoc nsz arcp contract float %{{.+}}, %{{.+}}
+    a.algebraic_add(b)
+}
+
+// CHECK-LABEL: @f32_algebraic_sub
+#[no_mangle]
+pub fn f32_algebraic_sub(a: f32, b: f32) -> f32 {
+    // CHECK: fsub reassoc nsz arcp contract float %{{.+}}, %{{.+}}
+    a.algebraic_sub(b)
+}
+
+// CHECK-LABEL: @f32_algebraic_mul
+#[no_mangle]
+pub fn f32_algebraic_mul(a: f32, b: f32) -> f32 {
+    // CHECK: fmul reassoc nsz arcp contract float %{{.+}}, %{{.+}}
+    a.algebraic_mul(b)
+}
+
+// CHECK-LABEL: @f32_algebraic_div
+#[no_mangle]
+pub fn f32_algebraic_div(a: f32, b: f32) -> f32 {
+    // CHECK: fdiv reassoc nsz arcp contract float %{{.+}}, %{{.+}}
+    a.algebraic_div(b)
+}
+
+// CHECK-LABEL: @f32_algebraic_rem
+#[no_mangle]
+pub fn f32_algebraic_rem(a: f32, b: f32) -> f32 {
+    // CHECK: frem reassoc nsz arcp contract float %{{.+}}, %{{.+}}
+    a.algebraic_rem(b)
+}
+
+// CHECK-LABEL: @f64_algebraic_add
+#[no_mangle]
+pub fn f64_algebraic_add(a: f64, b: f64) -> f64 {
+    // CHECK: fadd reassoc nsz arcp contract double %{{.+}}, %{{.+}}
+    a.algebraic_add(b)
+}
+
+// CHECK-LABEL: @f64_algebraic_sub
+#[no_mangle]
+pub fn f64_algebraic_sub(a: f64, b: f64) -> f64 {
+    // CHECK: fsub reassoc nsz arcp contract double %{{.+}}, %{{.+}}
+    a.algebraic_sub(b)
+}
+
+// CHECK-LABEL: @f64_algebraic_mul
+#[no_mangle]
+pub fn f64_algebraic_mul(a: f64, b: f64) -> f64 {
+    // CHECK: fmul reassoc nsz arcp contract double %{{.+}}, %{{.+}}
+    a.algebraic_mul(b)
+}
+
+// CHECK-LABEL: @f64_algebraic_div
+#[no_mangle]
+pub fn f64_algebraic_div(a: f64, b: f64) -> f64 {
+    // CHECK: fdiv reassoc nsz arcp contract double %{{.+}}, %{{.+}}
+    a.algebraic_div(b)
+}
+
+// CHECK-LABEL: @f64_algebraic_rem
+#[no_mangle]
+pub fn f64_algebraic_rem(a: f64, b: f64) -> f64 {
+    // CHECK: frem reassoc nsz arcp contract double %{{.+}}, %{{.+}}
+    a.algebraic_rem(b)
+}
+
+// CHECK-LABEL: @f128_algebraic_add
+#[no_mangle]
+pub fn f128_algebraic_add(a: f128, b: f128) -> f128 {
+    // CHECK: fadd reassoc nsz arcp contract fp128 %{{.+}}, %{{.+}}
+    a.algebraic_add(b)
+}
+
+// CHECK-LABEL: @f128_algebraic_sub
+#[no_mangle]
+pub fn f128_algebraic_sub(a: f128, b: f128) -> f128 {
+    // CHECK: fsub reassoc nsz arcp contract fp128 %{{.+}}, %{{.+}}
+    a.algebraic_sub(b)
+}
+
+// CHECK-LABEL: @f128_algebraic_mul
+#[no_mangle]
+pub fn f128_algebraic_mul(a: f128, b: f128) -> f128 {
+    // CHECK: fmul reassoc nsz arcp contract fp128 %{{.+}}, %{{.+}}
+    a.algebraic_mul(b)
+}
+
+// CHECK-LABEL: @f128_algebraic_div
+#[no_mangle]
+pub fn f128_algebraic_div(a: f128, b: f128) -> f128 {
+    // CHECK: fdiv reassoc nsz arcp contract fp128 %{{.+}}, %{{.+}}
+    a.algebraic_div(b)
+}
+
+// CHECK-LABEL: @f128_algebraic_rem
+#[no_mangle]
+pub fn f128_algebraic_rem(a: f128, b: f128) -> f128 {
+    // CHECK: frem reassoc nsz arcp contract fp128 %{{.+}}, %{{.+}}
+    a.algebraic_rem(b)
+}