Add ceilf16 and ceilf128

Use the generic algorithms to provide implementations for these
routines.

br-ceil.txt

@@ -0,0 +1,44 @@
+	.globl	libm::math::ceil::ceil
+	.p2align	2
+libm::math::ceil::ceil:
+Lfunc_begin17:
+	.cfi_startproc
+	sub sp, sp, #16
+	.cfi_def_cfa_offset 16
+	fcmp d0, #0.0
+	b.eq LBB17_5
+	fmov x8, d0
+	ubfx x9, x8, #52, #11
+	cmp w9, #1074
+	b.hi LBB17_5
+	mov x10, #-4382002437431492608
+	fmov d1, x10
+	fadd d2, d0, d1
+	mov x10, #4841369599423283200
+	fmov d3, x10
+	fadd d2, d2, d3
+	fadd d3, d0, d3
+	fadd d1, d3, d1
+	cmp x8, #0
+	fcsel d1, d2, d1, lt
+	fsub d1, d1, d0
+	str d1, [sp, #8]
+	cmp w9, #1023
+	b.hs LBB17_4
+	ldr d0, [sp, #8]
+	cmp x8, #0
+	mov x8, #-9223372036854775808
+	fmov d0, x8
+	fmov d1, #1.00000000
+	fcsel d0, d0, d1, lt
+	b LBB17_5
+LBB17_4:
+	fadd d0, d1, d0
+	fmov d2, #1.00000000
+	fadd d2, d0, d2
+	fcmp d1, #0.0
+	fcsel d0, d2, d0, mi
+LBB17_5:
+	add sp, sp, #16
+	.cfi_def_cfa_offset 0
+	ret

br-ceil1.txt

@@ -0,0 +1,41 @@
+	.globl	libm::math::ceilf::ceilf
+	.p2align	2
+libm::math::ceilf::ceilf:
+Lfunc_begin18:
+	.cfi_startproc
+	sub sp, sp, #16
+	.cfi_def_cfa_offset 16
+	fcmp s0, #0.0
+	b.eq LBB18_5
+	fmov w8, s0
+	ubfx w9, w8, #23, #8
+	cmp w9, #149
+	b.hi LBB18_5
+	movi.2s v1, #203, lsl #24
+	fadd s2, s0, s1
+	movi.2s v3, #75, lsl #24
+	fadd s2, s2, s3
+	fadd s3, s0, s3
+	fadd s1, s3, s1
+	cmp w8, #0
+	fcsel s1, s2, s1, lt
+	fsub s1, s1, s0
+	str s1, [sp, #12]
+	cmp w9, #127
+	b.hs LBB18_4
+	ldr s0, [sp, #12]
+	cmp w8, #0
+	movi.2s v0, #128, lsl #24
+	fmov s1, #1.00000000
+	fcsel s0, s0, s1, lt
+	b LBB18_5
+LBB18_4:
+	fadd s0, s1, s0
+	fmov s2, #1.00000000
+	fadd s2, s0, s2
+	fcmp s1, #0.0
+	fcsel s0, s2, s0, mi
+LBB18_5:
+	add sp, sp, #16
+	.cfi_def_cfa_offset 0
+	ret

crates/libm-macros/src/shared.rs

@@ -9,7 +9,7 @@ const ALL_OPERATIONS_NESTED: &[(FloatTy, Signature, Option<Signature>, &[&str])]
         FloatTy::F16,
         Signature { args: &[Ty::F16], returns: &[Ty::F16] },
         None,
-        &["fabsf16", "sqrtf16", "truncf16"],
+        &["ceilf16", "fabsf16", "sqrtf16", "truncf16"],
     ),
     (
         // `fn(f32) -> f32`
@@ -40,7 +40,7 @@ const ALL_OPERATIONS_NESTED: &[(FloatTy, Signature, Option<Signature>, &[&str])]
         FloatTy::F128,
         Signature { args: &[Ty::F128], returns: &[Ty::F128] },
         None,
-        &["fabsf128", "sqrtf128", "truncf128"],
+        &["ceilf128", "fabsf128", "sqrtf128", "truncf128"],
     ),
     (
         // `(f16, f16) -> f16`

crates/libm-test/benches/random.rs

@@ -117,7 +117,9 @@ libm_macros::for_each_function! {
         exp10 | exp10f | exp2 | exp2f => (true, Some(musl_math_sys::MACRO_FN_NAME)),
 
         // Musl does not provide `f16` and `f128` functions
-        copysignf128
+        ceilf128
+        | ceilf16
+        | copysignf128
         | copysignf16
         | fabsf128
         | fabsf16

crates/libm-test/src/mpfloat.rs

@@ -137,6 +137,8 @@ libm_macros::for_each_function! {
         // Most of these need a manual implementation
         ceil,
         ceilf,
+        ceilf128,
+        ceilf16,
         copysign,
         copysignf,
         copysignf128,
@@ -237,12 +239,14 @@ impl_no_round! {
 #[cfg(f16_enabled)]
 impl_no_round! {
     fabsf16 => abs_mut;
+    ceilf16 => ceil_mut;
     truncf16 => trunc_mut;
 }
 
 #[cfg(f128_enabled)]
 impl_no_round! {
     fabsf128 => abs_mut;
+    ceilf128 => ceil_mut;
     truncf128 => trunc_mut;
 }
 

crates/libm-test/tests/compare_built_musl.rs

@@ -79,6 +79,8 @@ libm_macros::for_each_function! {
         ynf,
 
         // Not provided by musl
+        ceilf128,
+        ceilf16,
         copysignf128,
         copysignf16,
         fabsf128,

crates/util/src/main.rs

@@ -84,7 +84,9 @@ fn do_eval(basis: &str, op: &str, inputs: &[&str]) {
         emit_types: [CFn, RustFn, RustArgs],
         extra: (basis, op, inputs),
         fn_extra: match MACRO_FN_NAME {
-            copysignf128
+            ceilf128
+            | ceilf16
+            | copysignf128
             | copysignf16
             | fabsf128
             | fabsf16

etc/function-definitions.json

@@ -109,17 +109,33 @@
             "src/libm_helper.rs",
             "src/math/arch/i586.rs",
             "src/math/arch/wasm32.rs",
-            "src/math/ceil.rs"
+            "src/math/ceil.rs",
+            "src/math/generic/ceil.rs"
         ],
         "type": "f64"
     },
     "ceilf": {
         "sources": [
             "src/math/arch/wasm32.rs",
-            "src/math/ceilf.rs"
+            "src/math/ceilf.rs",
+            "src/math/generic/ceil.rs"
         ],
         "type": "f32"
     },
+    "ceilf128": {
+        "sources": [
+            "src/math/ceilf128.rs",
+            "src/math/generic/ceil.rs"
+        ],
+        "type": "f128"
+    },
+    "ceilf16": {
+        "sources": [
+            "src/math/ceilf16.rs",
+            "src/math/generic/ceil.rs"
+        ],
+        "type": "f16"
+    },
     "copysign": {
         "sources": [
             "src/libm_helper.rs",

etc/function-list.txt

@@ -17,6 +17,8 @@ cbrt
 cbrtf
 ceil
 ceilf
+ceilf128
+ceilf16
 copysign
 copysignf
 copysignf128

m-ceil.txt

@@ -0,0 +1,52 @@
+	.globl	libm::math::ceil::ceil
+	.p2align	2
+libm::math::ceil::ceil:
+Lfunc_begin17:
+	.cfi_startproc
+	sub sp, sp, #16
+	.cfi_def_cfa_offset 16
+	fcmp d0, #0.0
+	b.eq LBB17_8
+	fmov x9, d0
+	ubfx x8, x9, #52, #11
+	cmp x8, #1074
+	b.hi LBB17_8
+	tbnz x9, #63, LBB17_5
+	mov x9, #4841369599423283200
+	fmov d1, x9
+	fadd d1, d0, d1
+	mov x9, #-4382002437431492608
+	fmov d2, x9
+	fadd d1, d1, d2
+	fsub d1, d1, d0
+	str d1, [sp, #8]
+	cmp x8, #1023
+	b.hs LBB17_7
+	ldr d0, [sp, #8]
+	fmov d0, #1.00000000
+	b LBB17_8
+LBB17_5:
+	mov x9, #-4382002437431492608
+	fmov d1, x9
+	fadd d1, d0, d1
+	mov x9, #4841369599423283200
+	fmov d2, x9
+	fadd d1, d1, d2
+	fsub d1, d1, d0
+	str d1, [sp, #8]
+	cmp x8, #1023
+	b.hs LBB17_7
+	ldr d0, [sp, #8]
+	mov x8, #-9223372036854775808
+	fmov d0, x8
+	b LBB17_8
+LBB17_7:
+	fadd d0, d1, d0
+	fmov d2, #1.00000000
+	fadd d2, d0, d2
+	fcmp d1, #0.0
+	fcsel d0, d2, d0, mi
+LBB17_8:
+	add sp, sp, #16
+	.cfi_def_cfa_offset 0
+	ret

m-ceil1.txt

@@ -0,0 +1,49 @@
+	.globl	libm::math::ceilf::ceilf
+	.p2align	2
+libm::math::ceilf::ceilf:
+Lfunc_begin18:
+	.cfi_startproc
+	sub sp, sp, #16
+	.cfi_def_cfa_offset 16
+	fmov w8, s0
+	ubfx w9, w8, #23, #8
+	cmp w9, #149
+	b.hi LBB18_5
+	cmp w9, #126
+	b.ls LBB18_4
+	lsr w9, w8, #23
+	add w10, w9, #1
+	mov w9, #8388607
+	lsr w9, w9, w10
+	tst w9, w8
+	b.eq LBB18_5
+	and w10, w10, #0x1f
+	mov w11, #2071986176
+	fmov s1, w11
+	fadd s0, s0, s1
+	str s0, [sp, #8]
+	ldr s0, [sp, #8]
+	cmp w8, #0
+	csel w9, wzr, w9, lt
+	mov w11, #-8388608
+	asr w10, w11, w10
+	add w8, w9, w8
+	and w8, w8, w10
+	fmov s0, w8
+	b LBB18_5
+LBB18_4:
+	mov w9, #2071986176
+	fmov s1, w9
+	fadd s1, s0, s1
+	str s1, [sp, #12]
+	ldr s1, [sp, #12]
+	movi.2s v1, #128, lsl #24
+	cmp w8, #0
+	fmov s2, #1.00000000
+	fcsel s0, s0, s2, eq
+	tst w8, #0x80000000
+	fcsel s0, s0, s1, eq
+LBB18_5:
+	add sp, sp, #16
+	.cfi_def_cfa_offset 0
+	ret

src/math/ceilf128.rs

@@ -0,0 +1,7 @@
+/// Ceil (f128)
+///
+/// Finds the nearest integer greater than or equal to `x`.
+#[cfg_attr(all(test, assert_no_panic), no_panic::no_panic)]
+pub fn ceilf128(x: f128) -> f128 {
+    super::generic::ceil(x)
+}

src/math/ceilf16.rs

@@ -0,0 +1,7 @@
+/// Ceil (f16)
+///
+/// Finds the nearest integer greater than or equal to `x`.
+#[cfg_attr(all(test, assert_no_panic), no_panic::no_panic)]
+pub fn ceilf16(x: f16) -> f16 {
+    super::generic::ceil(x)
+}

src/math/generic/ceil.rs

@@ -63,6 +63,14 @@ mod tests {
         }
     }
 
+    /* Skipping f16 / f128 "sanity_check"s due to rejected literal lexing at MSRV */
+
+    #[test]
+    #[cfg(f16_enabled)]
+    fn spec_tests_f16() {
+        spec_test::<f16>();
+    }
+
     #[test]
     fn sanity_check_f32() {
         assert_eq!(ceil(1.1f32), 2.0);
@@ -84,4 +92,10 @@ mod tests {
     fn spec_tests_f64() {
         spec_test::<f64>();
     }
+
+    #[test]
+    #[cfg(f128_enabled)]
+    fn spec_tests_f128() {
+        spec_test::<f128>();
+    }
 }

src/math/mod.rs

@@ -341,12 +341,14 @@ pub use self::truncf::truncf;
 
 cfg_if! {
     if #[cfg(f16_enabled)] {
+        mod ceilf16;
         mod copysignf16;
         mod fabsf16;
         mod fdimf16;
         mod sqrtf16;
         mod truncf16;
 
+        pub use self::ceilf16::ceilf16;
         pub use self::copysignf16::copysignf16;
         pub use self::fabsf16::fabsf16;
         pub use self::fdimf16::fdimf16;
@@ -357,12 +359,14 @@ cfg_if! {
 
 cfg_if! {
     if #[cfg(f128_enabled)] {
+        mod ceilf128;
         mod copysignf128;
         mod fabsf128;
         mod fdimf128;
         mod sqrtf128;
         mod truncf128;
 
+        pub use self::ceilf128::ceilf128;
         pub use self::copysignf128::copysignf128;
         pub use self::fabsf128::fabsf128;
         pub use self::fdimf128::fdimf128;