From 1404a1e5b0fe197fe3ab04dafbbef8c2a1d281bc Mon Sep 17 00:00:00 2001 From: sayantn Date: Sat, 9 May 2026 05:03:26 +0530 Subject: [PATCH] Change implementation of `vqdmull{_high}{_lane}` and `vqrdmulh_lane` --- .../core_arch/src/aarch64/neon/generated.rs | 76 ++++++---------- .../src/arm_shared/neon/generated.rs | 90 +++++-------------- .../spec/neon/aarch64.spec.yml | 46 +++++----- .../spec/neon/arm_shared.spec.yml | 24 ++--- 4 files changed, 83 insertions(+), 153 deletions(-) diff --git a/library/stdarch/crates/core_arch/src/aarch64/neon/generated.rs b/library/stdarch/crates/core_arch/src/aarch64/neon/generated.rs index eb590b3af3e2..2c49925431cb 100644 --- a/library/stdarch/crates/core_arch/src/aarch64/neon/generated.rs +++ b/library/stdarch/crates/core_arch/src/aarch64/neon/generated.rs @@ -14837,11 +14837,9 @@ pub fn vqdmulhs_laneq_s32(a: i32, b: int32x4_t) -> i32 { #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_lane_s16(a: int16x8_t, b: int16x4_t) -> int32x4_t { static_assert_uimm_bits!(N, 2); - unsafe { - let a: int16x4_t = simd_shuffle!(a, a, [4, 5, 6, 7]); - let b: int16x4_t = simd_shuffle!(b, b, [N as u32, N as u32, N as u32, N as u32]); - vqdmull_s16(a, b) - } + let a = vget_high_s16(a); + let b = vdup_lane_s16::(b); + vqdmull_s16(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_high_laneq_s32)"] @@ -14852,11 +14850,9 @@ pub fn vqdmull_high_lane_s16(a: int16x8_t, b: int16x4_t) -> int32x #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_laneq_s32(a: int32x4_t, b: int32x4_t) -> int64x2_t { static_assert_uimm_bits!(N, 2); - unsafe { - let a: int32x2_t = simd_shuffle!(a, a, [2, 3]); - let b: int32x2_t = simd_shuffle!(b, b, [N as u32, N as u32]); - vqdmull_s32(a, b) - } + let a = vget_high_s32(a); + let b = vdup_laneq_s32::(b); + vqdmull_s32(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_high_lane_s32)"] @@ -14867,11 +14863,9 @@ pub fn vqdmull_high_laneq_s32(a: int32x4_t, b: int32x4_t) -> int64 #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_lane_s32(a: int32x4_t, b: int32x2_t) -> int64x2_t { static_assert_uimm_bits!(N, 1); - unsafe { - let a: int32x2_t = simd_shuffle!(a, a, [2, 3]); - let b: int32x2_t = simd_shuffle!(b, b, [N as u32, N as u32]); - vqdmull_s32(a, b) - } + let a = vget_high_s32(a); + let b = vdup_lane_s32::(b); + vqdmull_s32(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_high_laneq_s16)"] @@ -14882,11 +14876,9 @@ pub fn vqdmull_high_lane_s32(a: int32x4_t, b: int32x2_t) -> int64x #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_laneq_s16(a: int16x8_t, b: int16x8_t) -> int32x4_t { static_assert_uimm_bits!(N, 3); - unsafe { - let a: int16x4_t = simd_shuffle!(a, a, [4, 5, 6, 7]); - let b: int16x4_t = simd_shuffle!(b, b, [N as u32, N as u32, N as u32, N as u32]); - vqdmull_s16(a, b) - } + let a = vget_high_s16(a); + let b = vdup_laneq_s16::(b); + vqdmull_s16(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_high_n_s16)"] @@ -14895,11 +14887,9 @@ pub fn vqdmull_high_laneq_s16(a: int16x8_t, b: int16x8_t) -> int32 #[cfg_attr(all(test, target_endian = "little"), assert_instr(sqdmull2))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_n_s16(a: int16x8_t, b: i16) -> int32x4_t { - unsafe { - let a: int16x4_t = simd_shuffle!(a, a, [4, 5, 6, 7]); - let b: int16x4_t = vdup_n_s16(b); - vqdmull_s16(a, b) - } + let a = vget_high_s16(a); + let b = vdup_n_s16(b); + vqdmull_s16(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_high_n_s32)"] @@ -14908,11 +14898,9 @@ pub fn vqdmull_high_n_s16(a: int16x8_t, b: i16) -> int32x4_t { #[cfg_attr(all(test, target_endian = "little"), assert_instr(sqdmull2))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_n_s32(a: int32x4_t, b: i32) -> int64x2_t { - unsafe { - let a: int32x2_t = simd_shuffle!(a, a, [2, 3]); - let b: int32x2_t = vdup_n_s32(b); - vqdmull_s32(a, b) - } + let a = vget_high_s32(a); + let b = vdup_n_s32(b); + vqdmull_s32(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_high_s16)"] @@ -14921,11 +14909,9 @@ pub fn vqdmull_high_n_s32(a: int32x4_t, b: i32) -> int64x2_t { #[cfg_attr(all(test, target_endian = "little"), assert_instr(sqdmull2))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_s16(a: int16x8_t, b: int16x8_t) -> int32x4_t { - unsafe { - let a: int16x4_t = simd_shuffle!(a, a, [4, 5, 6, 7]); - let b: int16x4_t = simd_shuffle!(b, b, [4, 5, 6, 7]); - vqdmull_s16(a, b) - } + let a = vget_high_s16(a); + let b = vget_high_s16(b); + vqdmull_s16(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_high_s32)"] @@ -14934,11 +14920,9 @@ pub fn vqdmull_high_s16(a: int16x8_t, b: int16x8_t) -> int32x4_t { #[cfg_attr(all(test, target_endian = "little"), assert_instr(sqdmull2))] #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_high_s32(a: int32x4_t, b: int32x4_t) -> int64x2_t { - unsafe { - let a: int32x2_t = simd_shuffle!(a, a, [2, 3]); - let b: int32x2_t = simd_shuffle!(b, b, [2, 3]); - vqdmull_s32(a, b) - } + let a = vget_high_s32(a); + let b = vget_high_s32(b); + vqdmull_s32(a, b) } #[doc = "Vector saturating doubling long multiply by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_laneq_s16)"] @@ -14949,10 +14933,8 @@ pub fn vqdmull_high_s32(a: int32x4_t, b: int32x4_t) -> int64x2_t { #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_laneq_s16(a: int16x4_t, b: int16x8_t) -> int32x4_t { static_assert_uimm_bits!(N, 3); - unsafe { - let b: int16x4_t = simd_shuffle!(b, b, [N as u32, N as u32, N as u32, N as u32]); - vqdmull_s16(a, b) - } + let b = vdup_laneq_s16::(b); + vqdmull_s16(a, b) } #[doc = "Vector saturating doubling long multiply by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_laneq_s32)"] @@ -14963,10 +14945,8 @@ pub fn vqdmull_laneq_s16(a: int16x4_t, b: int16x8_t) -> int32x4_t #[stable(feature = "neon_intrinsics", since = "1.59.0")] pub fn vqdmull_laneq_s32(a: int32x2_t, b: int32x4_t) -> int64x2_t { static_assert_uimm_bits!(N, 2); - unsafe { - let b: int32x2_t = simd_shuffle!(b, b, [N as u32, N as u32]); - vqdmull_s32(a, b) - } + let b = vdup_laneq_s32::(b); + vqdmull_s32(a, b) } #[doc = "Signed saturating doubling multiply long"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmullh_lane_s16)"] diff --git a/library/stdarch/crates/core_arch/src/arm_shared/neon/generated.rs b/library/stdarch/crates/core_arch/src/arm_shared/neon/generated.rs index 6d54e84f2508..3f5b7073e264 100644 --- a/library/stdarch/crates/core_arch/src/arm_shared/neon/generated.rs +++ b/library/stdarch/crates/core_arch/src/arm_shared/neon/generated.rs @@ -35001,10 +35001,8 @@ pub fn vqdmulhq_s32(a: int32x4_t, b: int32x4_t) -> int32x4_t { )] pub fn vqdmull_lane_s16(a: int16x4_t, b: int16x4_t) -> int32x4_t { static_assert_uimm_bits!(N, 2); - unsafe { - let b: int16x4_t = simd_shuffle!(b, b, [N as u32; 4]); - vqdmull_s16(a, b) - } + let b = vdup_lane_s16::(b); + vqdmull_s16(a, b) } #[doc = "Vector saturating doubling long multiply by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_lane_s32)"] @@ -35027,10 +35025,8 @@ pub fn vqdmull_lane_s16(a: int16x4_t, b: int16x4_t) -> int32x4_t { )] pub fn vqdmull_lane_s32(a: int32x2_t, b: int32x2_t) -> int64x2_t { static_assert_uimm_bits!(N, 1); - unsafe { - let b: int32x2_t = simd_shuffle!(b, b, [N as u32; 2]); - vqdmull_s32(a, b) - } + let b = vdup_lane_s32::(b); + vqdmull_s32(a, b) } #[doc = "Vector saturating doubling long multiply with scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqdmull_n_s16)"] @@ -35588,11 +35584,8 @@ pub fn vqnegq_s32(a: int32x4_t) -> int32x4_t { )] pub fn vqrdmulh_lane_s16(a: int16x4_t, b: int16x4_t) -> int16x4_t { static_assert_uimm_bits!(LANE, 2); - unsafe { - let b: int16x4_t = - simd_shuffle!(b, b, [LANE as u32, LANE as u32, LANE as u32, LANE as u32]); - vqrdmulh_s16(a, b) - } + let b = vdup_lane_s16::(b); + vqrdmulh_s16(a, b) } #[doc = "Vector rounding saturating doubling multiply high by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulh_lane_s32)"] @@ -35615,10 +35608,8 @@ pub fn vqrdmulh_lane_s16(a: int16x4_t, b: int16x4_t) -> int16x4 )] pub fn vqrdmulh_lane_s32(a: int32x2_t, b: int32x2_t) -> int32x2_t { static_assert_uimm_bits!(LANE, 1); - unsafe { - let b: int32x2_t = simd_shuffle!(b, b, [LANE as u32, LANE as u32]); - vqrdmulh_s32(a, b) - } + let b = vdup_lane_s32::(b); + vqrdmulh_s32(a, b) } #[doc = "Vector rounding saturating doubling multiply high by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulh_laneq_s16)"] @@ -35641,11 +35632,8 @@ pub fn vqrdmulh_lane_s32(a: int32x2_t, b: int32x2_t) -> int32x2 )] pub fn vqrdmulh_laneq_s16(a: int16x4_t, b: int16x8_t) -> int16x4_t { static_assert_uimm_bits!(LANE, 3); - unsafe { - let b: int16x4_t = - simd_shuffle!(b, b, [LANE as u32, LANE as u32, LANE as u32, LANE as u32]); - vqrdmulh_s16(a, b) - } + let b = vdup_laneq_s16::(b); + vqrdmulh_s16(a, b) } #[doc = "Vector rounding saturating doubling multiply high by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulh_laneq_s32)"] @@ -35668,10 +35656,8 @@ pub fn vqrdmulh_laneq_s16(a: int16x4_t, b: int16x8_t) -> int16x )] pub fn vqrdmulh_laneq_s32(a: int32x2_t, b: int32x4_t) -> int32x2_t { static_assert_uimm_bits!(LANE, 2); - unsafe { - let b: int32x2_t = simd_shuffle!(b, b, [LANE as u32, LANE as u32]); - vqrdmulh_s32(a, b) - } + let b = vdup_laneq_s32::(b); + vqrdmulh_s32(a, b) } #[doc = "Vector rounding saturating doubling multiply high by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulhq_lane_s16)"] @@ -35694,23 +35680,8 @@ pub fn vqrdmulh_laneq_s32(a: int32x2_t, b: int32x4_t) -> int32x )] pub fn vqrdmulhq_lane_s16(a: int16x8_t, b: int16x4_t) -> int16x8_t { static_assert_uimm_bits!(LANE, 2); - unsafe { - let b: int16x8_t = simd_shuffle!( - b, - b, - [ - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32 - ] - ); - vqrdmulhq_s16(a, b) - } + let b = vdupq_lane_s16::(b); + vqrdmulhq_s16(a, b) } #[doc = "Vector rounding saturating doubling multiply high by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulhq_lane_s32)"] @@ -35733,11 +35704,8 @@ pub fn vqrdmulhq_lane_s16(a: int16x8_t, b: int16x4_t) -> int16x )] pub fn vqrdmulhq_lane_s32(a: int32x4_t, b: int32x2_t) -> int32x4_t { static_assert_uimm_bits!(LANE, 1); - unsafe { - let b: int32x4_t = - simd_shuffle!(b, b, [LANE as u32, LANE as u32, LANE as u32, LANE as u32]); - vqrdmulhq_s32(a, b) - } + let b = vdupq_lane_s32::(b); + vqrdmulhq_s32(a, b) } #[doc = "Vector rounding saturating doubling multiply high by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulhq_laneq_s16)"] @@ -35760,23 +35728,8 @@ pub fn vqrdmulhq_lane_s32(a: int32x4_t, b: int32x2_t) -> int32x )] pub fn vqrdmulhq_laneq_s16(a: int16x8_t, b: int16x8_t) -> int16x8_t { static_assert_uimm_bits!(LANE, 3); - unsafe { - let b: int16x8_t = simd_shuffle!( - b, - b, - [ - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32, - LANE as u32 - ] - ); - vqrdmulhq_s16(a, b) - } + let b = vdupq_laneq_s16::(b); + vqrdmulhq_s16(a, b) } #[doc = "Vector rounding saturating doubling multiply high by scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulhq_laneq_s32)"] @@ -35799,11 +35752,8 @@ pub fn vqrdmulhq_laneq_s16(a: int16x8_t, b: int16x8_t) -> int16 )] pub fn vqrdmulhq_laneq_s32(a: int32x4_t, b: int32x4_t) -> int32x4_t { static_assert_uimm_bits!(LANE, 2); - unsafe { - let b: int32x4_t = - simd_shuffle!(b, b, [LANE as u32, LANE as u32, LANE as u32, LANE as u32]); - vqrdmulhq_s32(a, b) - } + let b = vdupq_laneq_s32::(b); + vqrdmulhq_s32(a, b) } #[doc = "Vector saturating rounding doubling multiply high with scalar"] #[doc = "[Arm's documentation](https://developer.arm.com/architectures/instruction-sets/intrinsics/vqrdmulh_n_s16)"] diff --git a/library/stdarch/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml b/library/stdarch/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml index a6208fac7c35..11ba7c955dd6 100644 --- a/library/stdarch/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml +++ b/library/stdarch/crates/stdarch-gen-arm/spec/neon/aarch64.spec.yml @@ -6963,11 +6963,11 @@ intrinsics: - *neon-stable safety: safe types: - - [int16x8_t, int32x4_t, int16x4_t, '[4, 5, 6, 7]'] - - [int32x4_t, int64x2_t, int32x2_t, '[2, 3]'] + - [int16x8_t, int32x4_t] + - [int32x4_t, int64x2_t] compose: - - Let: [a, "{neon_type[2]}", {FnCall: [simd_shuffle!, [a, a, '{type[3]}']]}] - - Let: [b, "{neon_type[2]}", {FnCall: [simd_shuffle!, [b, b, '{type[3]}']]}] + - Let: [a, {FnCall: ['vget_high_{neon_type[0]}', [a]]}] + - Let: [b, {FnCall: ['vget_high_{neon_type[0]}', [b]]}] - FnCall: ["vqdmull{neon_type[0].noq}", [a, b]] - name: "vqdmull_high_n_{type[1]}" @@ -6979,11 +6979,11 @@ intrinsics: - *neon-stable safety: safe types: - - [int16x8_t, "i16", int32x4_t, int16x4_t, '[4, 5, 6, 7]'] - - [int32x4_t, "i32", int64x2_t, int32x2_t, '[2, 3]'] + - [int16x8_t, "i16", int32x4_t] + - [int32x4_t, "i32", int64x2_t] compose: - - Let: [a, "{neon_type[3]}", {FnCall: [simd_shuffle!, [a, a, "{type[4]}"]]}] - - Let: [b, "{neon_type[3]}", {FnCall: ["vdup_n{neon_type[0].noq}", [b]]}] + - Let: [a, {FnCall: ['vget_high_{neon_type[0]}', [a]]}] + - Let: [b, {FnCall: ["vdup_n{neon_type[0].noq}", [b]]}] - FnCall: ["vqdmull{neon_type[0].noq}", [a, b]] - name: "vqdmull{type[3]}" @@ -7038,7 +7038,7 @@ intrinsics: - Let: [b, "{type[0]}", {FnCall: ['vget{neon_type[1].lane_nox}', [b], [N]]}] - FnCall: ["vqdmulls_s32", [a, b]] - - name: "vqdmull{type[6]}" + - name: "vqdmull{type[3]}" doc: "Signed saturating doubling multiply long" arguments: ["a: {neon_type[0]}", "b: {neon_type[1]}"] return_type: "{neon_type[2]}" @@ -7049,12 +7049,12 @@ intrinsics: static_defs: ['const N: i32'] safety: safe types: - - [int16x8_t, int16x4_t, int32x4_t, int16x4_t, '[4, 5, 6, 7]', '[N as u32, N as u32, N as u32, N as u32]', '_high_lane_s16'] - - [int32x4_t, int32x4_t, int64x2_t, int32x2_t, '[2, 3]', '[N as u32, N as u32]', '_high_laneq_s32'] + - [int16x8_t, int16x4_t, int32x4_t, '_high_lane_s16'] + - [int32x4_t, int32x4_t, int64x2_t, '_high_laneq_s32'] compose: - FnCall: [static_assert_uimm_bits!, [N, '2']] - - Let: [a, "{neon_type[3]}", {FnCall: [simd_shuffle!, [a, a, "{type[4]}"]]}] - - Let: [b, "{neon_type[3]}", {FnCall: [simd_shuffle!, [b, b, "{type[5]}"]]}] + - Let: [a, {FnCall: ['vget_high_{neon_type[0]}', [a]]}] + - Let: [b, {FnCall: ['vdup_lane{neon_type[1].nox}', [b], [N]]}] - FnCall: ["vqdmull{neon_type[0].noq}", [a, b]] - name: "vqdmull_high_lane_s32" @@ -7068,11 +7068,11 @@ intrinsics: static_defs: ['const N: i32'] safety: safe types: - - [int32x4_t, int32x2_t, int64x2_t, int32x2_t, '[2, 3]', '[N as u32, N as u32]'] + - [int32x4_t, int32x2_t, int64x2_t] compose: - FnCall: [static_assert_uimm_bits!, [N, '1']] - - Let: [a, "{neon_type[3]}", {FnCall: [simd_shuffle!, [a, a, "{type[4]}"]]}] - - Let: [b, "{neon_type[3]}", {FnCall: [simd_shuffle!, [b, b, "{type[5]}"]]}] + - Let: [a, {FnCall: ['vget_high_{neon_type[0]}', [a]]}] + - Let: [b, {FnCall: ['vdup_lane{neon_type[1].nox}', [b], [N]]}] - FnCall: ["vqdmull{neon_type[0].noq}", [a, b]] - name: "vqdmull_high_laneq_s16" @@ -7086,11 +7086,11 @@ intrinsics: static_defs: ['const N: i32'] safety: safe types: - - [int16x8_t, int16x8_t, int32x4_t, int16x4_t, '[4, 5, 6, 7]', '[N as u32, N as u32, N as u32, N as u32]'] + - [int16x8_t, int16x8_t, int32x4_t] compose: - FnCall: [static_assert_uimm_bits!, [N, '3']] - - Let: [a, "{neon_type[3]}", {FnCall: [simd_shuffle!, [a, a, "{type[4]}"]]}] - - Let: [b, "{neon_type[3]}", {FnCall: [simd_shuffle!, [b, b, "{type[5]}"]]}] + - Let: [a, {FnCall: ['vget_high_{neon_type[0]}', [a]]}] + - Let: [b, {FnCall: ['vdup_lane{neon_type[1].nox}', [b], [N]]}] - FnCall: ["vqdmull{neon_type[0].noq}", [a, b]] - name: "vqdmull_laneq_s16" @@ -7104,10 +7104,10 @@ intrinsics: static_defs: ['const N: i32'] safety: safe types: - - [int16x4_t, int16x8_t, int32x4_t, '[N as u32, N as u32, N as u32, N as u32]'] + - [int16x4_t, int16x8_t, int32x4_t] compose: - FnCall: [static_assert_uimm_bits!, [N, '3']] - - Let: [b, "{neon_type[0]}", {FnCall: [simd_shuffle!, [b, b, "{type[3]}"]]}] + - Let: [b, {FnCall: ['vdup_lane{neon_type[1].nox}', [b], [N]]}] - FnCall: [vqdmull_s16, [a, b]] - name: "vqdmull_laneq_s32" @@ -7121,10 +7121,10 @@ intrinsics: static_defs: ['const N: i32'] safety: safe types: - - [int32x2_t, int32x4_t, int64x2_t, '[N as u32, N as u32]'] + - [int32x2_t, int32x4_t, int64x2_t] compose: - FnCall: [static_assert_uimm_bits!, [N, '2']] - - Let: [b, "{neon_type[0]}", {FnCall: [simd_shuffle!, [b, b, "{type[3]}"]]}] + - Let: [b, {FnCall: ['vdup_lane{neon_type[1].nox}', [b], [N]]}] - FnCall: [vqdmull_s32, [a, b]] - name: "vqdmlal{type[4]}" diff --git a/library/stdarch/crates/stdarch-gen-arm/spec/neon/arm_shared.spec.yml b/library/stdarch/crates/stdarch-gen-arm/spec/neon/arm_shared.spec.yml index 4024e624ca9b..e646d22369f8 100644 --- a/library/stdarch/crates/stdarch-gen-arm/spec/neon/arm_shared.spec.yml +++ b/library/stdarch/crates/stdarch-gen-arm/spec/neon/arm_shared.spec.yml @@ -7590,7 +7590,7 @@ intrinsics: - [int16x4_t, int16x4_t, int32x4_t, '[N as u32; 4]'] compose: - FnCall: [static_assert_uimm_bits!, [N, '2']] - - Let: [b, "{neon_type[0]}", {FnCall: [simd_shuffle!, [b, b, "{type[3]}"]]}] + - Let: [b, {FnCall: ['vdup{neon_type[0].lane_nox}', [b], [N]]}] - FnCall: [vqdmull_s16, [a, b]] - name: "vqdmull_lane_s32" @@ -7607,10 +7607,10 @@ intrinsics: static_defs: ['const N: i32'] safety: safe types: - - [int32x2_t, int32x2_t, int64x2_t, '[N as u32; 2]'] + - [int32x2_t, int32x2_t, int64x2_t] compose: - FnCall: [static_assert_uimm_bits!, [N, '1']] - - Let: [b, "{neon_type[0]}", {FnCall: [simd_shuffle!, [b, b, "{type[3]}"]]}] + - Let: [b, {FnCall: ['vdup{neon_type[0].lane_nox}', [b], [N]]}] - FnCall: [vqdmull_s32, [a, b]] - name: "vqdmlal{neon_type[1].noq}" @@ -11230,17 +11230,17 @@ intrinsics: static_defs: ['const LANE: i32'] safety: safe types: - - [_lane_s16, int16x4_t, int16x4_t, '2', '[LANE as u32, LANE as u32, LANE as u32, LANE as u32]'] - - [_laneq_s16, int16x4_t, int16x8_t, '3', '[LANE as u32, LANE as u32, LANE as u32, LANE as u32]'] - - [q_lane_s16, int16x8_t, int16x4_t, '2', '[LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32]'] - - [q_laneq_s16, int16x8_t, int16x8_t, '3', '[LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32, LANE as u32]'] - - [_lane_s32, int32x2_t, int32x2_t, '1', '[LANE as u32, LANE as u32]'] - - [_laneq_s32, int32x2_t, int32x4_t, '2', '[LANE as u32, LANE as u32]'] - - [q_lane_s32, int32x4_t, int32x2_t, '1', '[LANE as u32, LANE as u32, LANE as u32, LANE as u32]'] - - [q_laneq_s32, int32x4_t, int32x4_t, '2', '[LANE as u32, LANE as u32, LANE as u32, LANE as u32]'] + - [_lane_s16, int16x4_t, int16x4_t, '2'] + - [_laneq_s16, int16x4_t, int16x8_t, '3'] + - [q_lane_s16, int16x8_t, int16x4_t, '2'] + - [q_laneq_s16, int16x8_t, int16x8_t, '3'] + - [_lane_s32, int32x2_t, int32x2_t, '1'] + - [_laneq_s32, int32x2_t, int32x4_t, '2'] + - [q_lane_s32, int32x4_t, int32x2_t, '1'] + - [q_laneq_s32, int32x4_t, int32x4_t, '2'] compose: - FnCall: [static_assert_uimm_bits!, [LANE, "{type[3]}"]] - - Let: [b, "{neon_type[1]}", {FnCall: [simd_shuffle!, [b, b, '{type[4]}']]}] + - Let: [b, {FnCall: ['vdup{type[0]}', [b], [LANE]]}] - FnCall: ["vqrdmulh{neon_type[1].no}", [a, b]] - name: "vqrdmulh{neon_type[0].N}"