diff --git a/crates/core_arch/src/aarch64/neon/generated.rs b/crates/core_arch/src/aarch64/neon/generated.rs index 8e3493dfac..cca3879571 100644 --- a/crates/core_arch/src/aarch64/neon/generated.rs +++ b/crates/core_arch/src/aarch64/neon/generated.rs @@ -268,19 +268,6 @@ pub fn vabdl_high_s8(a: int8x16_t, b: int8x16_t) -> int16x8_t { } } #[doc = "Unsigned Absolute difference Long"] -#[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vabdl_high_u8)"] -#[inline] -#[target_feature(enable = "neon")] -#[stable(feature = "neon_intrinsics", since = "1.59.0")] -#[cfg_attr(all(test, target_endian = "little"), assert_instr(uabdl2))] -pub fn vabdl_high_u8(a: uint8x16_t, b: uint8x16_t) -> uint16x8_t { - unsafe { - let c: uint8x8_t = simd_shuffle!(a, a, [8, 9, 10, 11, 12, 13, 14, 15]); - let d: uint8x8_t = simd_shuffle!(b, b, [8, 9, 10, 11, 12, 13, 14, 15]); - simd_cast(vabd_u8(c, d)) - } -} -#[doc = "Unsigned Absolute difference Long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vabdl_high_u16)"] #[inline] #[target_feature(enable = "neon")] @@ -306,6 +293,19 @@ pub fn vabdl_high_u32(a: uint32x4_t, b: uint32x4_t) -> uint64x2_t { simd_cast(vabd_u32(c, d)) } } +#[doc = "Unsigned Absolute difference Long"] +#[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vabdl_high_u8)"] +#[inline] +#[target_feature(enable = "neon")] +#[stable(feature = "neon_intrinsics", since = "1.59.0")] +#[cfg_attr(all(test, target_endian = "little"), assert_instr(uabdl2))] +pub fn vabdl_high_u8(a: uint8x16_t, b: uint8x16_t) -> uint16x8_t { + unsafe { + let c: uint8x8_t = simd_shuffle!(a, a, [8, 9, 10, 11, 12, 13, 14, 15]); + let d: uint8x8_t = simd_shuffle!(b, b, [8, 9, 10, 11, 12, 13, 14, 15]); + simd_cast(vabd_u8(c, d)) + } +} #[doc = "Floating-point absolute value"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vabs_f64)"] #[inline] @@ -25952,7 +25952,7 @@ pub fn vtbl1_p8(a: poly8x8_t, b: uint8x8_t) -> poly8x8_t { #[cfg_attr(test, assert_instr(tbl))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbl2_s8(a: int8x8x2_t, b: int8x8_t) -> int8x8_t { - unsafe { vqtbl1(transmute(vcombine_s8(a.0, a.1)), transmute(b)) } + vqtbl1_s8(vcombine_s8(a.0, a.1), vreinterpret_u8_s8(b)) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl2_u8)"] @@ -25961,7 +25961,7 @@ pub fn vtbl2_s8(a: int8x8x2_t, b: int8x8_t) -> int8x8_t { #[cfg_attr(test, assert_instr(tbl))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbl2_u8(a: uint8x8x2_t, b: uint8x8_t) -> uint8x8_t { - unsafe { transmute(vqtbl1(transmute(vcombine_u8(a.0, a.1)), b)) } + vqtbl1_u8(vcombine_u8(a.0, a.1), b) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl2_p8)"] @@ -25970,7 +25970,7 @@ pub fn vtbl2_u8(a: uint8x8x2_t, b: uint8x8_t) -> uint8x8_t { #[cfg_attr(test, assert_instr(tbl))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbl2_p8(a: poly8x8x2_t, b: uint8x8_t) -> poly8x8_t { - unsafe { transmute(vqtbl1(transmute(vcombine_p8(a.0, a.1)), b)) } + vqtbl1_p8(vcombine_p8(a.0, a.1), b) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl3_s8)"] @@ -25983,7 +25983,7 @@ pub fn vtbl3_s8(a: int8x8x3_t, b: int8x8_t) -> int8x8_t { vcombine_s8(a.0, a.1), vcombine_s8(a.2, unsafe { crate::mem::zeroed() }), ); - unsafe { transmute(vqtbl2(transmute(x.0), transmute(x.1), transmute(b))) } + vqtbl2_s8(x, vreinterpret_u8_s8(b)) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl3_u8)"] @@ -25996,7 +25996,7 @@ pub fn vtbl3_u8(a: uint8x8x3_t, b: uint8x8_t) -> uint8x8_t { vcombine_u8(a.0, a.1), vcombine_u8(a.2, unsafe { crate::mem::zeroed() }), ); - unsafe { transmute(vqtbl2(transmute(x.0), transmute(x.1), b)) } + vqtbl2_u8(x, b) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl3_p8)"] @@ -26009,7 +26009,7 @@ pub fn vtbl3_p8(a: poly8x8x3_t, b: uint8x8_t) -> poly8x8_t { vcombine_p8(a.0, a.1), vcombine_p8(a.2, unsafe { crate::mem::zeroed() }), ); - unsafe { transmute(vqtbl2(transmute(x.0), transmute(x.1), b)) } + vqtbl2_p8(x, b) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl4_s8)"] @@ -26019,7 +26019,7 @@ pub fn vtbl3_p8(a: poly8x8x3_t, b: uint8x8_t) -> poly8x8_t { #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbl4_s8(a: int8x8x4_t, b: int8x8_t) -> int8x8_t { let x = int8x16x2_t(vcombine_s8(a.0, a.1), vcombine_s8(a.2, a.3)); - unsafe { transmute(vqtbl2(transmute(x.0), transmute(x.1), transmute(b))) } + vqtbl2_s8(x, vreinterpret_u8_s8(b)) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl4_u8)"] @@ -26029,7 +26029,7 @@ pub fn vtbl4_s8(a: int8x8x4_t, b: int8x8_t) -> int8x8_t { #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbl4_u8(a: uint8x8x4_t, b: uint8x8_t) -> uint8x8_t { let x = uint8x16x2_t(vcombine_u8(a.0, a.1), vcombine_u8(a.2, a.3)); - unsafe { transmute(vqtbl2(transmute(x.0), transmute(x.1), b)) } + vqtbl2_u8(x, b) } #[doc = "Table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbl4_p8)"] @@ -26039,7 +26039,7 @@ pub fn vtbl4_u8(a: uint8x8x4_t, b: uint8x8_t) -> uint8x8_t { #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbl4_p8(a: poly8x8x4_t, b: uint8x8_t) -> poly8x8_t { let x = poly8x16x2_t(vcombine_p8(a.0, a.1), vcombine_p8(a.2, a.3)); - unsafe { transmute(vqtbl2(transmute(x.0), transmute(x.1), b)) } + vqtbl2_p8(x, b) } #[doc = "Extended table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbx1_s8)"] @@ -26051,11 +26051,11 @@ pub fn vtbx1_s8(a: int8x8_t, b: int8x8_t, c: int8x8_t) -> int8x8_t { unsafe { simd_select( simd_lt::(c, transmute(i8x8::splat(8))), - transmute(vqtbx1( - transmute(a), - transmute(vcombine_s8(b, crate::mem::zeroed())), - transmute(c), - )), + vqtbx1_s8( + a, + vcombine_s8(b, crate::mem::zeroed()), + vreinterpret_u8_s8(c), + ), a, ) } @@ -26070,11 +26070,7 @@ pub fn vtbx1_u8(a: uint8x8_t, b: uint8x8_t, c: uint8x8_t) -> uint8x8_t { unsafe { simd_select( simd_lt::(c, transmute(u8x8::splat(8))), - transmute(vqtbx1( - transmute(a), - transmute(vcombine_u8(b, crate::mem::zeroed())), - c, - )), + vqtbx1_u8(a, vcombine_u8(b, crate::mem::zeroed()), c), a, ) } @@ -26089,11 +26085,7 @@ pub fn vtbx1_p8(a: poly8x8_t, b: poly8x8_t, c: uint8x8_t) -> poly8x8_t { unsafe { simd_select( simd_lt::(c, transmute(u8x8::splat(8))), - transmute(vqtbx1( - transmute(a), - transmute(vcombine_p8(b, crate::mem::zeroed())), - c, - )), + vqtbx1_p8(a, vcombine_p8(b, crate::mem::zeroed()), c), a, ) } @@ -26105,7 +26097,13 @@ pub fn vtbx1_p8(a: poly8x8_t, b: poly8x8_t, c: uint8x8_t) -> poly8x8_t { #[cfg_attr(test, assert_instr(tbx))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbx2_s8(a: int8x8_t, b: int8x8x2_t, c: int8x8_t) -> int8x8_t { - unsafe { vqtbx1(transmute(a), transmute(vcombine_s8(b.0, b.1)), transmute(c)) } + unsafe { + simd_select( + simd_lt::(c, transmute(i8x8::splat(16))), + vqtbx1_s8(a, vcombine_s8(b.0, b.1), vreinterpret_u8_s8(c)), + a, + ) + } } #[doc = "Extended table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbx2_u8)"] @@ -26114,7 +26112,13 @@ pub fn vtbx2_s8(a: int8x8_t, b: int8x8x2_t, c: int8x8_t) -> int8x8_t { #[cfg_attr(test, assert_instr(tbx))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbx2_u8(a: uint8x8_t, b: uint8x8x2_t, c: uint8x8_t) -> uint8x8_t { - unsafe { transmute(vqtbx1(transmute(a), transmute(vcombine_u8(b.0, b.1)), c)) } + unsafe { + simd_select( + simd_lt::(c, transmute(u8x8::splat(16))), + vqtbx1_u8(a, vcombine_u8(b.0, b.1), c), + a, + ) + } } #[doc = "Extended table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbx2_p8)"] @@ -26123,7 +26127,13 @@ pub fn vtbx2_u8(a: uint8x8_t, b: uint8x8x2_t, c: uint8x8_t) -> uint8x8_t { #[cfg_attr(test, assert_instr(tbx))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbx2_p8(a: poly8x8_t, b: poly8x8x2_t, c: uint8x8_t) -> poly8x8_t { - unsafe { transmute(vqtbx1(transmute(a), transmute(vcombine_p8(b.0, b.1)), c)) } + unsafe { + simd_select( + simd_lt::(c, transmute(u8x8::splat(16))), + vqtbx1_p8(a, vcombine_p8(b.0, b.1), c), + a, + ) + } } #[doc = "Extended table look-up"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vtbx3_s8)"] @@ -26137,16 +26147,11 @@ pub fn vtbx3_s8(a: int8x8_t, b: int8x8x3_t, c: int8x8_t) -> int8x8_t { vcombine_s8(b.2, unsafe { crate::mem::zeroed() }), ); unsafe { - transmute(simd_select( - simd_lt::(transmute(c), transmute(i8x8::splat(24))), - transmute(vqtbx2( - transmute(a), - transmute(x.0), - transmute(x.1), - transmute(c), - )), + simd_select( + simd_lt::(c, transmute(i8x8::splat(24))), + vqtbx2_s8(a, x, vreinterpret_u8_s8(c)), a, - )) + ) } } #[doc = "Extended table look-up"] @@ -26161,11 +26166,11 @@ pub fn vtbx3_u8(a: uint8x8_t, b: uint8x8x3_t, c: uint8x8_t) -> uint8x8_t { vcombine_u8(b.2, unsafe { crate::mem::zeroed() }), ); unsafe { - transmute(simd_select( - simd_lt::(transmute(c), transmute(u8x8::splat(24))), - transmute(vqtbx2(transmute(a), transmute(x.0), transmute(x.1), c)), + simd_select( + simd_lt::(c, transmute(u8x8::splat(24))), + vqtbx2_u8(a, x, c), a, - )) + ) } } #[doc = "Extended table look-up"] @@ -26180,11 +26185,11 @@ pub fn vtbx3_p8(a: poly8x8_t, b: poly8x8x3_t, c: uint8x8_t) -> poly8x8_t { vcombine_p8(b.2, unsafe { crate::mem::zeroed() }), ); unsafe { - transmute(simd_select( - simd_lt::(transmute(c), transmute(u8x8::splat(24))), - transmute(vqtbx2(transmute(a), transmute(x.0), transmute(x.1), c)), + simd_select( + simd_lt::(c, transmute(u8x8::splat(24))), + vqtbx2_p8(a, x, c), a, - )) + ) } } #[doc = "Extended table look-up"] @@ -26194,12 +26199,12 @@ pub fn vtbx3_p8(a: poly8x8_t, b: poly8x8x3_t, c: uint8x8_t) -> poly8x8_t { #[cfg_attr(test, assert_instr(tbx))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbx4_s8(a: int8x8_t, b: int8x8x4_t, c: int8x8_t) -> int8x8_t { + let x = int8x16x2_t(vcombine_s8(b.0, b.1), vcombine_s8(b.2, b.3)); unsafe { - vqtbx2( - transmute(a), - transmute(vcombine_s8(b.0, b.1)), - transmute(vcombine_s8(b.2, b.3)), - transmute(c), + simd_select( + simd_lt::(c, transmute(i8x8::splat(32))), + vqtbx2_s8(a, x, vreinterpret_u8_s8(c)), + a, ) } } @@ -26210,13 +26215,13 @@ pub fn vtbx4_s8(a: int8x8_t, b: int8x8x4_t, c: int8x8_t) -> int8x8_t { #[cfg_attr(test, assert_instr(tbx))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbx4_u8(a: uint8x8_t, b: uint8x8x4_t, c: uint8x8_t) -> uint8x8_t { + let x = uint8x16x2_t(vcombine_u8(b.0, b.1), vcombine_u8(b.2, b.3)); unsafe { - transmute(vqtbx2( - transmute(a), - transmute(vcombine_u8(b.0, b.1)), - transmute(vcombine_u8(b.2, b.3)), - c, - )) + simd_select( + simd_lt::(c, transmute(u8x8::splat(32))), + vqtbx2_u8(a, x, c), + a, + ) } } #[doc = "Extended table look-up"] @@ -26226,13 +26231,13 @@ pub fn vtbx4_u8(a: uint8x8_t, b: uint8x8x4_t, c: uint8x8_t) -> uint8x8_t { #[cfg_attr(test, assert_instr(tbx))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vtbx4_p8(a: poly8x8_t, b: poly8x8x4_t, c: uint8x8_t) -> poly8x8_t { + let x = poly8x16x2_t(vcombine_p8(b.0, b.1), vcombine_p8(b.2, b.3)); unsafe { - transmute(vqtbx2( - transmute(a), - transmute(vcombine_p8(b.0, b.1)), - transmute(vcombine_p8(b.2, b.3)), - c, - )) + simd_select( + simd_lt::(c, transmute(u8x8::splat(32))), + vqtbx2_p8(a, x, c), + a, + ) } } #[doc = "Transpose vectors"] diff --git a/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml b/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml index 163145e7ba..9cceacec9d 100644 --- a/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml +++ b/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml @@ -11470,8 +11470,8 @@ intrinsics: - FnCall: ["vdup_n_{type[0]}", [b]] - '0' - - name: "vabdl_high_{neon_type[0]}" - doc: "Unsigned Absolute difference Long" + - name: "vabdl_high{neon_type[0].noq}" + doc: Unsigned Absolute difference Long arguments: ["a: {neon_type[0]}", "b: {neon_type[0]}"] return_type: "{neon_type[1]}" attr: @@ -11479,13 +11479,99 @@ intrinsics: - FnCall: [cfg_attr, [*all-test-little-endian, {FnCall: [assert_instr, [uabdl2]]}]] safety: safe types: - - [uint8x16_t, uint16x8_t, uint8x8_t, '[8, 9, 10, 11, 12, 13, 14, 15]'] - - [uint16x8_t, uint32x4_t, uint16x4_t, '[4, 5, 6, 7]'] - - [uint32x4_t, uint64x2_t, uint32x2_t, '[2, 3]'] + - [uint8x16_t, uint16x8_t, uint8x8_t] + compose: + - Let: + - c + - "{neon_type[2]}" + - FnCall: + - simd_shuffle! + - - a + - a + - [8, 9, 10, 11, 12, 13, 14, 15] + - Let: + - d + - "{neon_type[2]}" + - FnCall: + - simd_shuffle! + - - b + - b + - [8, 9, 10, 11, 12, 13, 14, 15] + - FnCall: + - simd_cast + - - FnCall: + - "vabd_{neon_type[0]}" + - - c + - d + + - name: "vabdl_high{neon_type[0].noq}" + doc: Unsigned Absolute difference Long + arguments: ["a: {neon_type[0]}", "b: {neon_type[0]}"] + return_type: "{neon_type[1]}" + attr: + - *neon-stable + - FnCall: [cfg_attr, [*all-test-little-endian, {FnCall: [assert_instr, [uabdl2]]}]] + safety: safe + types: + - [uint16x8_t, uint32x4_t, uint16x4_t] + compose: + - Let: + - c + - "{neon_type[2]}" + - FnCall: + - simd_shuffle! + - - a + - a + - [4, 5, 6, 7] + - Let: + - d + - "{neon_type[2]}" + - FnCall: + - simd_shuffle! + - - b + - b + - [4, 5, 6, 7] + + - FnCall: + - simd_cast + - - FnCall: + - "vabd_{neon_type[0]}" + - - c + - d + + - name: "vabdl_high{neon_type[0].noq}" + doc: Unsigned Absolute difference Long + arguments: ["a: {neon_type[0]}", "b: {neon_type[0]}"] + return_type: "{neon_type[1]}" + attr: + - *neon-stable + - FnCall: [cfg_attr, [*all-test-little-endian, {FnCall: [assert_instr, [uabdl2]]}]] + safety: safe + types: + - [uint32x4_t, uint64x2_t, uint32x2_t, int32x2_t] compose: - - Let: [c, "{neon_type[2]}", {FnCall: [simd_shuffle!, [a, a, "{type[3]}"]]}] - - Let: [d, "{neon_type[2]}", {FnCall: [simd_shuffle!, [b, b, "{type[3]}"]]}] - - FnCall: [simd_cast, [{FnCall: ["vabd_{neon_type[0]}", [c, d]]}]] + - Let: + - c + - "{neon_type[2]}" + - FnCall: + - simd_shuffle! + - - a + - a + - [2, 3] + - Let: + - d + - "{neon_type[2]}" + - FnCall: + - simd_shuffle! + - - b + - b + - [2, 3] + - FnCall: + - simd_cast + - - FnCall: + - "vabd_{neon_type[0]}" + - - c + - d - name: "vfms_n_f64" doc: "Floating-point fused Multiply-subtract to accumulator(vector)" @@ -12138,20 +12224,29 @@ intrinsics: attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbx]]}]] - *neon-stable + big_endian_inverse: false safety: safe types: - - [int8x8_t, int8x8x4_t] + - [int8x8_t, 'int8x8x4_t', 'int8x16x2', 'int8x8', 'i8x8::splat(32)'] compose: + - Let: + - x + - FnCall: + - '{type[2]}_t' + - - FnCall: ['vcombine{neon_type[0].no}', ['b.0', 'b.1']] + - FnCall: ['vcombine{neon_type[0].no}', ['b.2', 'b.3']] - FnCall: - - "vqtbx2" - - - FnCall: [transmute, [a]] - - FnCall: - - transmute - - - FnCall: ["vcombine{neon_type[0].noq}", ["b.0", "b.1"]] - - FnCall: - - transmute - - - FnCall: ["vcombine{neon_type[0].noq}", ["b.2", "b.3"]] - - FnCall: [transmute, [c]] + - simd_select + - - FnCall: + - "simd_lt::<{type[3]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[4]}"]] + - FnCall: + - 'vqtbx2{neon_type[0].no}' + - - a + - x + - FnCall: ['vreinterpret_u8{neon_type[0].no}', [c]] + - a - name: "vtbx4{neon_type[0].no}" doc: "Extended table look-up" @@ -12163,21 +12258,27 @@ intrinsics: big_endian_inverse: false safety: safe types: - - [uint8x8_t, uint8x8x4_t, uint8x8_t] - - [poly8x8_t, poly8x8x4_t, uint8x8_t] + - [uint8x8_t, 'uint8x8x4_t', 'uint8x8_t', 'uint8x16x2', 'uint8x8', 'u8x8::splat(32)'] + - [poly8x8_t, 'poly8x8x4_t', 'uint8x8_t', 'poly8x16x2', 'uint8x8', 'u8x8::splat(32)'] compose: + - Let: + - x + - FnCall: + - '{type[3]}_t' + - - FnCall: ['vcombine{neon_type[0].no}', ['b.0', 'b.1']] + - FnCall: ['vcombine{neon_type[0].no}', ['b.2', 'b.3']] - FnCall: - - transmute - - - FnCall: - - "vqtbx2" - - - FnCall: [transmute, [a]] - - FnCall: - - transmute - - - FnCall: ["vcombine{neon_type[0].noq}", ["b.0", "b.1"]] - - FnCall: - - transmute - - - FnCall: ["vcombine{neon_type[0].noq}", ["b.2", "b.3"]] - - c + - simd_select + - - FnCall: + - "simd_lt::<{type[4]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[5]}"]] + - FnCall: + - 'vqtbx2{neon_type[0].no}' + - - a + - x + - c + - a - name: "vtbl1{neon_type[0].no}" doc: "Table look-up" @@ -12187,6 +12288,7 @@ intrinsics: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbl]]}]] - *neon-stable safety: safe + big_endian_inverse: false types: - [int8x8_t, 'int8x8_t', 'unsafe {{ transmute(b) }}'] - [uint8x8_t, 'uint8x8_t', 'b'] @@ -12200,26 +12302,22 @@ intrinsics: - 'unsafe {{ crate::mem::zeroed() }}' - Identifier: ['{type[2]}', Symbol] - - name: "vtbl2{neon_type[1].noq}" + - name: "vtbl2{neon_type[2].no}" doc: "Table look-up" arguments: ["a: {neon_type[0]}", "b: {neon_type[1]}"] - return_type: "{neon_type[1]}" + return_type: "{neon_type[2]}" attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbl]]}]] - *neon-stable + big_endian_inverse: false safety: safe types: - - [int8x8x2_t, 'int8x8_t'] + - ['int8x8x2_t', 'int8x8_t', 'int8x8_t'] compose: - FnCall: - - vqtbl1 - - - FnCall: - - transmute - - - FnCall: - - 'vcombine{neon_type[1].noq}' - - - 'a.0' - - 'a.1' - - FnCall: [transmute, [b]] + - 'vqtbl1{neon_type[2].noq}' + - - FnCall: ['vcombine{neon_type[2].noq}', ['a.0', 'a.1']] + - FnCall: ['vreinterpret_u8{neon_type[2].noq}', [b]] - name: "vtbl2{neon_type[2].no}" doc: "Table look-up" @@ -12231,124 +12329,108 @@ intrinsics: big_endian_inverse: false safety: safe types: - - [uint8x8x2_t, 'uint8x8_t', 'uint8x8_t'] - - [poly8x8x2_t, 'uint8x8_t', 'poly8x8_t'] + - ['uint8x8x2_t', 'uint8x8_t', 'uint8x8_t'] + - ['poly8x8x2_t', 'uint8x8_t', 'poly8x8_t'] compose: - FnCall: - - transmute - - - FnCall: - - vqtbl1 - - - FnCall: - - transmute - - - FnCall: - - 'vcombine{neon_type[2].noq}' - - - 'a.0' - - 'a.1' - - b + - 'vqtbl1{neon_type[2].noq}' + - - FnCall: ['vcombine{neon_type[2].noq}', ['a.0', 'a.1']] + - b - name: "vtbl3{neon_type[1].no}" doc: "Table look-up" arguments: ["a: {neon_type[0]}", "b: {neon_type[1]}"] - return_type: "{neon_type[1]}" + return_type: "{neon_type[2]}" attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbl]]}]] - *neon-stable safety: safe + big_endian_inverse: false types: - - [int8x8x3_t, 'int8x8_t', 'int8x16x2'] + - ['int8x8x3_t', 'int8x8_t', 'int8x8_t', 'int8x16x2'] compose: - Let: - - x - - FnCall: - - '{type[2]}_t' - - - FnCall: ['vcombine{neon_type[1].no}', ['a.0', 'a.1']] - - FnCall: ['vcombine{neon_type[1].no}', ['a.2', 'unsafe {{ crate::mem::zeroed() }}']] + - x + - FnCall: + - '{type[3]}_t' + - - FnCall: ['vcombine{neon_type[2].no}', ['a.0', 'a.1']] + - FnCall: ['vcombine{neon_type[2].no}', ['a.2', 'unsafe {{ crate::mem::zeroed() }}']] - FnCall: - - transmute - - - FnCall: - - vqtbl2 - - - FnCall: [transmute, ['x.0']] - - FnCall: [transmute, ['x.1']] - - FnCall: [transmute, [b]] + - 'vqtbl2{neon_type[2].no}' + - - x + - FnCall: ['vreinterpret_u8{neon_type[2].noq}', [b]] - - name: "vtbl3{neon_type[3].no}" + + - name: "vtbl3{neon_type[2].no}" doc: "Table look-up" arguments: ["a: {neon_type[0]}", "b: {neon_type[1]}"] - return_type: "{neon_type[3]}" + return_type: "{neon_type[2]}" attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbl]]}]] - *neon-stable safety: safe - types: - - [uint8x8x3_t, 'uint8x8_t', 'uint8x16x2', 'uint8x8_t'] - - [poly8x8x3_t, 'uint8x8_t', 'poly8x16x2', 'poly8x8_t'] big_endian_inverse: false + types: + - [uint8x8x3_t, 'uint8x8_t', 'uint8x8_t', 'uint8x16x2'] + - [poly8x8x3_t, 'uint8x8_t', 'poly8x8_t', 'poly8x16x2'] compose: - Let: - - x - - FnCall: - - '{type[2]}_t' - - - FnCall: ['vcombine{neon_type[3].no}', ['a.0', 'a.1']] - - FnCall: ['vcombine{neon_type[3].no}', ['a.2', 'unsafe {{ crate::mem::zeroed() }}']] - - FnCall: - - transmute - - - FnCall: - - vqtbl2 - - - FnCall: [transmute, ['x.0']] - - FnCall: [transmute, ['x.1']] - - b + - x + - FnCall: + - '{type[3]}_t' + - - FnCall: ['vcombine{neon_type[2].no}', ['a.0', 'a.1']] + - FnCall: ['vcombine{neon_type[2].no}', ['a.2', 'unsafe {{ crate::mem::zeroed() }}']] + - FnCall: + - 'vqtbl2{neon_type[2].no}' + - - x + - b - - name: "vtbl4{neon_type[1].no}" + - name: "vtbl4{neon_type[2].no}" doc: "Table look-up" arguments: ["a: {neon_type[0]}", "b: {neon_type[1]}"] - return_type: "{neon_type[1]}" + return_type: "{neon_type[2]}" attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbl]]}]] - *neon-stable safety: safe + big_endian_inverse: false types: - - [int8x8x4_t, 'int8x8_t', 'int8x16x2'] + - ['int8x8x4_t', 'int8x8_t', 'int8x8_t', 'int8x16x2'] compose: - Let: - - x - - FnCall: - - '{type[2]}_t' - - - FnCall: ['vcombine{neon_type[1].no}', ['a.0', 'a.1']] - - FnCall: ['vcombine{neon_type[1].no}', ['a.2', 'a.3']] + - x + - FnCall: + - '{type[3]}_t' + - - FnCall: ['vcombine{neon_type[1].no}', ['a.0', 'a.1']] + - FnCall: ['vcombine{neon_type[1].no}', ['a.2', 'a.3']] - FnCall: - - transmute - - - FnCall: - - 'vqtbl2' - - - FnCall: [transmute, ['x.0']] - - FnCall: [transmute, ['x.1']] - - FnCall: [transmute, [b]] + - 'vqtbl2{neon_type[2].no}' + - - x + - FnCall: ['vreinterpret_u8{neon_type[2].noq}', [b]] - - name: "vtbl4{neon_type[3].no}" + - name: "vtbl4{neon_type[2].no}" doc: "Table look-up" arguments: ["a: {neon_type[0]}", "b: {neon_type[1]}"] - return_type: "{neon_type[3]}" + return_type: "{neon_type[2]}" attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbl]]}]] - *neon-stable safety: safe - types: - - [uint8x8x4_t, 'uint8x8_t', 'uint8x16x2', 'uint8x8_t'] - - [poly8x8x4_t, 'uint8x8_t', 'poly8x16x2', 'poly8x8_t'] big_endian_inverse: false + types: + - [uint8x8x4_t, 'uint8x8_t', 'uint8x8_t', 'uint8x16x2'] + - [poly8x8x4_t, 'uint8x8_t', 'poly8x8_t', 'poly8x16x2'] compose: - Let: - - x - - FnCall: - - '{type[2]}_t' - - - FnCall: ['vcombine{neon_type[3].no}', ['a.0', 'a.1']] - - FnCall: ['vcombine{neon_type[3].no}', ['a.2', 'a.3']] - - FnCall: - - transmute - - - FnCall: - - 'vqtbl2' - - - FnCall: [transmute, ['x.0']] - - FnCall: [transmute, ['x.1']] - - b + - x + - FnCall: + - '{type[3]}_t' + - - FnCall: ['vcombine{neon_type[2].no}', ['a.0', 'a.1']] + - FnCall: ['vcombine{neon_type[2].no}', ['a.2', 'a.3']] + - FnCall: + - 'vqtbl2{neon_type[2].no}' + - - x + - b - name: "vqtbx1{neon_type[0].no}" doc: "Extended table look-up" @@ -12394,28 +12476,49 @@ intrinsics: attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbx]]}]] - *neon-stable + big_endian_inverse: false safety: safe types: - - [int8x8_t, "int8x8_t", "transmute(c)", "i8x8::splat(8)", "int8x8"] - - [uint8x8_t, "uint8x8_t", "c", "u8x8::splat(8)", "uint8x8"] - - [poly8x8_t, "uint8x8_t", "c", "u8x8::splat(8)", "uint8x8"] + - [int8x8_t, "int8x8_t", "int8x8", "i8x8::splat(8)"] compose: - FnCall: - - simd_select - - - FnCall: - - "simd_lt::<{type[4]}_t, int8x8_t>" - - - c - - FnCall: [transmute, ["{type[3]}"]] - - FnCall: - - transmute - - - FnCall: - - "vqtbx1" - - - "transmute(a)" - - FnCall: - - transmute - - - FnCall: ["vcombine{neon_type[0].no}", [b, "crate::mem::zeroed()"]] - - "{type[2]}" - - a + - simd_select + - - FnCall: + - "simd_lt::<{type[2]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[3]}"]] + - FnCall: + - 'vqtbx1{neon_type[0].no}' + - - a + - FnCall: ['vcombine{neon_type[0].no}', [b, 'crate::mem::zeroed()']] + - FnCall: ['vreinterpret_u8{neon_type[0].no}', [c]] + - a + + - name: "vtbx1{neon_type[0].no}" + doc: "Extended table look-up" + arguments: ["a: {neon_type[0]}", "b: {neon_type[0]}", "c: {neon_type[1]}"] + return_type: "{neon_type[0]}" + attr: + - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbx]]}]] + - *neon-stable + big_endian_inverse: false + safety: safe + types: + - [uint8x8_t, "uint8x8_t", "uint8x8", "u8x8::splat(8)"] + - [poly8x8_t, "uint8x8_t", "uint8x8", "u8x8::splat(8)"] + compose: + - FnCall: + - simd_select + - - FnCall: + - "simd_lt::<{type[2]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[3]}"]] + - FnCall: + - 'vqtbx1{neon_type[0].no}' + - - a + - FnCall: ['vcombine{neon_type[0].no}', [b, 'crate::mem::zeroed()']] + - c + - a - name: "vtbx2{neon_type[0].no}" doc: "Extended table look-up" @@ -12424,17 +12527,23 @@ intrinsics: attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbx]]}]] - *neon-stable + big_endian_inverse: false safety: safe types: - - [int8x8_t, 'int8x8x2_t'] + - [int8x8_t, 'int8x8x2_t', 'int8x8', 'i8x8::splat(16)'] compose: - FnCall: - - vqtbx1 - - - FnCall: [transmute, [a]] - - FnCall: - - transmute - - - FnCall: ["vcombine{neon_type[0].no}", ['b.0', 'b.1']] - - FnCall: [transmute, [c]] + - simd_select + - - FnCall: + - "simd_lt::<{type[2]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[3]}"]] + - FnCall: + - 'vqtbx1{neon_type[0].no}' + - - a + - FnCall: ["vcombine{neon_type[0].no}", ['b.0', 'b.1']] + - FnCall: ['vreinterpret_u8{neon_type[0].no}', [c]] + - a - name: "vtbx2{neon_type[0].no}" doc: "Extended table look-up" @@ -12446,18 +12555,21 @@ intrinsics: big_endian_inverse: false safety: safe types: - - [uint8x8_t, 'uint8x8x2_t', uint8x8_t] - - [poly8x8_t, 'poly8x8x2_t', uint8x8_t] + - [uint8x8_t, 'uint8x8x2_t', uint8x8_t, 'uint8x8', 'u8x8::splat(16)'] + - [poly8x8_t, 'poly8x8x2_t', uint8x8_t, 'uint8x8', 'u8x8::splat(16)'] compose: - FnCall: - - transmute - - - FnCall: - - vqtbx1 - - - FnCall: [transmute, [a]] - - FnCall: - - transmute - - - FnCall: ["vcombine{neon_type[0].no}", ['b.0', 'b.1']] - - c + - simd_select + - - FnCall: + - "simd_lt::<{type[3]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[4]}"]] + - FnCall: + - 'vqtbx1{neon_type[0].no}' + - - a + - FnCall: ["vcombine{neon_type[0].no}", ['b.0', 'b.1']] + - c + - a - name: "vtbx3{neon_type[0].no}" doc: "Extended table look-up" @@ -12466,33 +12578,29 @@ intrinsics: attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbx]]}]] - *neon-stable + big_endian_inverse: false safety: safe types: - - [int8x8_t, 'int8x8x3_t', 'int8x16x2', 'i8x8::splat(24)', 'int8x8'] + - [int8x8_t, 'int8x8x3_t', 'int8x16x2', 'int8x8', 'i8x8::splat(24)'] compose: - Let: - - x - - FnCall: - - '{type[2]}_t' - - - FnCall: ['vcombine{neon_type[0].no}', ['b.0', 'b.1']] - - FnCall: ['vcombine{neon_type[0].no}', ['b.2', 'unsafe {{ crate::mem::zeroed() }}']] + - x + - FnCall: + - '{type[2]}_t' + - - FnCall: ['vcombine{neon_type[0].no}', ['b.0', 'b.1']] + - FnCall: ['vcombine{neon_type[0].no}', ['b.2', 'unsafe {{ crate::mem::zeroed() }}']] - FnCall: - - transmute - - - FnCall: - - simd_select - - - FnCall: - - 'simd_lt::<{type[4]}_t, int8x8_t>' - - - FnCall: [transmute, [c]] - - FnCall: [transmute, ['{type[3]}']] - - FnCall: - - transmute - - - FnCall: - - 'vqtbx2' - - - FnCall: [transmute, [a]] - - FnCall: [transmute, ['x.0']] - - FnCall: [transmute, ['x.1']] - - FnCall: [transmute, [c]] - - a + - simd_select + - - FnCall: + - "simd_lt::<{type[3]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[4]}"]] + - FnCall: + - 'vqtbx2{neon_type[0].no}' + - - a + - x + - FnCall: ['vreinterpret_u8{neon_type[0].no}', [c]] + - a - name: "vtbx3{neon_type[0].no}" doc: "Extended table look-up" @@ -12501,11 +12609,11 @@ intrinsics: attr: - FnCall: [cfg_attr, [test, {FnCall: [assert_instr, [tbx]]}]] - *neon-stable + big_endian_inverse: false safety: safe types: - - [uint8x8_t, 'uint8x8x3_t', 'uint8x16x2', 'u8x8::splat(24)', 'uint8x8'] - - [poly8x8_t, 'poly8x8x3_t', 'poly8x16x2', 'u8x8::splat(24)', 'poly8x8'] - big_endian_inverse: false + - [uint8x8_t, 'uint8x8x3_t', 'uint8x16x2', 'uint8x8', 'u8x8::splat(24)'] + - [poly8x8_t, 'poly8x8x3_t', 'poly8x16x2', 'uint8x8', 'u8x8::splat(24)'] compose: - Let: - x @@ -12514,22 +12622,17 @@ intrinsics: - - FnCall: ['vcombine{neon_type[0].no}', ['b.0', 'b.1']] - FnCall: ['vcombine{neon_type[0].no}', ['b.2', 'unsafe {{ crate::mem::zeroed() }}']] - FnCall: - - transmute - - - FnCall: - - simd_select - - - FnCall: - - 'simd_lt::<{type[4]}_t, int8x8_t>' - - - FnCall: [transmute, [c]] - - FnCall: [transmute, ['{type[3]}']] - - FnCall: - - transmute - - - FnCall: - - 'vqtbx2' - - - FnCall: [transmute, [a]] - - FnCall: [transmute, ['x.0']] - - FnCall: [transmute, ['x.1']] - - c - - a + - simd_select + - - FnCall: + - "simd_lt::<{type[3]}_t, int8x8_t>" + - - c + - FnCall: [transmute, ["{type[4]}"]] + - FnCall: + - 'vqtbx2{neon_type[0].no}' + - - a + - x + - c + - a - name: "vqtbl1{neon_type[3].no}" doc: "Table look-up"