# Copyright (C) 2022-2023 Exaloop Inc. <https://exaloop.io>


@tuple(container=False)  # disallow default __getitem__
class Vec[T, N: Static[int]]:
    ZERO_16x8i = Vec[u8,16](u8(0))
    FF_16x8i = Vec[u8,16](u8(0xff))
    ZERO_32x8i = Vec[u8,32](u8(0))
    FF_32x8i = Vec[u8,32](u8(0xff))

    ZERO_4x64f = Vec[f64,4](0.0)

    @llvm
    def _mm_set1_epi8(val: u8) -> Vec[u8, 16]:
        %0 = insertelement <16 x i8> undef, i8 %val, i32 0
        %1 = shufflevector <16 x i8> %0, <16 x i8> undef, <16 x i32> zeroinitializer
        ret <16 x i8> %1

    @llvm
    def _mm_set1_epi32(val: u32) -> Vec[u32, 4]:
        %0 = insertelement <4 x i32> undef, i32 %val, i32 0
        %1 = shufflevector <4 x i32> %0, <4 x i32> undef, <4 x i32> zeroinitializer
        ret <4 x i32> %1

    @llvm
    def _mm256_set1_epi8(val: u8) -> Vec[u8, 32]:
        %0 = insertelement <32 x i8> undef, i8 %val, i32 0
        %1 = shufflevector <32 x i8> %0, <32 x i8> undef, <32 x i32> zeroinitializer
        ret <32 x i8> %1

    @llvm
    def _mm256_set1_epi32(val: u32) -> Vec[u32, 8]:
        %0 = insertelement <8 x i32> undef, i32 %val, i32 0
        %1 = shufflevector <8 x i32> %0, <8 x i32> undef, <8 x i32> zeroinitializer
        ret <8 x i32> %1

    @llvm
    def _mm256_set1_epi64x(val: u64) -> Vec[u64, 4]:
        %0 = insertelement <4 x i64> undef, i64 %val, i32 0
        %1 = shufflevector <4 x i64> %0, <4 x i64> undef, <4 x i32> zeroinitializer
        ret <4 x i64> %1

    @llvm
    def _mm512_set1_epi64(val: u64) -> Vec[u64, 8]:
        %0 = insertelement <8 x i64> undef, i64 %val, i32 0
        %1 = shufflevector <8 x i64> %0, <8 x i64> undef, <8 x i32> zeroinitializer
        ret <8 x i64> %1

    @llvm
    def _mm_load_epi32(data) -> Vec[u32, 4]:
        %0 = bitcast i32* %data to <4 x i32>*
        %1 = load <4 x i32>, <4 x i32>* %0, align 1
        ret <4 x i32> %1

    @llvm
    def _mm_loadu_si128(data) -> Vec[u8, 16]:
        %0 = bitcast i8* %data to <16 x i8>*
        %1 = load <16 x i8>, <16 x i8>* %0, align 1
        ret <16 x i8> %1

    @llvm
    def _mm256_loadu_si256(data) -> Vec[u8, 32]:
        %0 = bitcast i8* %data to <32 x i8>*
        %1 = load <32 x i8>, <32 x i8>* %0, align 1
        ret <32 x i8> %1

    @llvm
    def _mm256_load_epi32(data) -> Vec[u32, 8]:
        %0 = bitcast i32* %data to <8 x i32>*
        %1 = load <8 x i32>, <8 x i32>* %0
        ret <8 x i32> %1

    @llvm
    def _mm256_load_epi64(data) -> Vec[u64, 4]:
        %0 = bitcast i64* %data to <4 x i64>*
        %1 = load <4 x i64>, <4 x i64>* %0
        ret <4 x i64> %1

    @llvm
    def _mm512_load_epi64(data) -> Vec[u64, 8]:
        %0 = bitcast i64* %data to <8 x i64>*
        %1 = load <8 x i64>, <8 x i64>* %0
        ret <8 x i64> %1

    @llvm
    def _mm256_set1_ps(val: f32) -> Vec[f32, 8]:
        %0 = insertelement <8 x float> undef, float %val, i32 0
        %1 = shufflevector <8 x float> %0, <8 x float> undef, <8 x i32> zeroinitializer
        ret <8 x float> %1

    @llvm
    def _mm256_set1_pd(val: f64) -> Vec[f64, 4]:
        %0 = insertelement <4 x double> undef, double %val, i64 0
        %1 = shufflevector <4 x double> %0, <4 x double> undef, <4 x i32> zeroinitializer
        ret <4 x double> %1

    @llvm
    def _mm512_set1_pd(val: f64) -> Vec[f64, 8]:
        %0 = insertelement <8 x double> undef, double %val, i64 0
        %1 = shufflevector <8 x double> %0, <8 x double> undef, <8 x i32> zeroinitializer
        ret <8 x double> %1

    @llvm
    def _mm512_set1_ps(val: f32) -> Vec[f32, 16]:
        %0 = insertelement <16 x float> undef, float %val, i32 0
        %1 = shufflevector <16 x float> %0, <16 x float> undef, <16 x i32> zeroinitializer
        ret <16 x float> %1

    @llvm
    def _mm256_loadu_ps(data: Ptr[f32]) -> Vec[f32, 8]:
        %0 = bitcast float* %data to <8 x float>*
        %1 = load <8 x float>, <8 x float>* %0
        ret <8 x float> %1

    @llvm
    def _mm256_loadu_pd(data: Ptr[f64]) -> Vec[f64, 4]:
        %0 = bitcast double* %data to <4 x double>*
        %1 = load <4 x double>, <4 x double>* %0
        ret <4 x double> %1

    @llvm
    def _mm512_loadu_pd(data: Ptr[f64]) -> Vec[f64, 8]:
        %0 = bitcast double* %data to <8 x double>*
        %1 = load <8 x double>, <8 x double>* %0
        ret <8 x double> %1

    @llvm
    def _mm512_loadu_ps(data: Ptr[f32]) -> Vec[f32, 16]:
        %0 = bitcast float* %data to <16 x float>*
        %1 = load <16 x float>, <16 x float>* %0
        ret <16 x float> %1

    @llvm
    def _mm256_cvtepi8_epi32(vec: Vec[u8, 16]) -> Vec[u32, 8]:
        %0 = shufflevector <16 x i8> %vec, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
        %1 = sext <8 x i8> %0 to <8 x i32>
        ret <8 x i32> %1

    @llvm
    def _mm512_cvtepi8_epi64(vec: Vec[u8, 32]) -> Vec[u32, 16]:
        %0 = shufflevector <32 x i8> %vec, <32 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
        %1 = sext <16 x i8> %0 to <16 x i32>
        ret <16 x i32> %1

    @llvm
    def _mm256_castsi256_ps(vec: Vec[u32, 8]) -> Vec[f32, 8]:
        %0 = bitcast <8 x i32> %vec to <8 x float>
        ret <8 x float> %0

    @llvm
    def _mm256_castsi256_pd(vec: Vec[u64, 4]) -> Vec[f64, 4]:
        %0 = bitcast <4 x i64> %vec to <4 x double>
        ret <4 x double> %0

    @llvm
    def _mm512_castsi512_ps(vec: Vec[u32, 16]) -> Vec[f32, 16]:
        %0 = bitcast <16 x i32> %vec to <16 x float>
        ret <16 x float> %0

    def __new__(x, T: type, N: Static[int]) -> Vec[T, N]:
        if isinstance(T, u8) and N == 16:
            if isinstance(x, u8) or isinstance(x, byte): # TODO: u8<->byte
                return Vec._mm_set1_epi8(x)
            if isinstance(x, Ptr[u8]) or isinstance(x, Ptr[byte]):
                return Vec._mm_loadu_si128(x)
            if isinstance(x, str):
                return Vec._mm_loadu_si128(x.ptr)
        if isinstance(T, u8) and N == 32:
            if isinstance(x, u8) or isinstance(x, byte): # TODO: u8<->byte
                return Vec._mm256_set1_epi8(x)
            if isinstance(x, Ptr[u8]) or isinstance(x, Ptr[byte]):
                return Vec._mm256_loadu_si256(x)
            if isinstance(x, str):
                return Vec._mm256_loadu_si256(x.ptr)
        if isinstance(T, u32) and N == 4:
            if isinstance(x, int):
                assert x >= 0, "SIMD: No support for negative int vectors added yet."
                return Vec._mm_set1_epi32(u32(x))
            if isinstance(x, u32):
                return Vec._mm_set1_epi32(x)
            if isinstance(x, Ptr[u32]):
                return Vec._mm_load_epi32(x)
            if isinstance(x, List[u32]):
                return Vec._mm_load_epi32(x.arr.ptr)
            if isinstance(x, str):
                return Vec._mm_load_epi32(x.ptr)
        if isinstance(T, u32) and N == 8:
            if isinstance(x, int):
                assert x >= 0, "SIMD: No support for negative int vectors added yet."
                return Vec._mm256_set1_epi32(u32(x))
            if isinstance(x, u64):
                return Vec._mm256_set1_epi32(x)
            if isinstance(x, Ptr[u64]):
                return Vec._mm256_load_epi32(x)
            if isinstance(x, List[u64]):
                return Vec._mm256_load_epi32(x.arr.ptr)
            if isinstance(x, str):
                return Vec._mm256_load_epi32(x.ptr)
        if isinstance(T, u64) and N == 4:
            if isinstance(x, int):
                assert x >= 0, "SIMD: No support for negative int vectors added yet."
                return Vec._mm256_set1_epi64x(u64(x))
            if isinstance(x, u64):
                return Vec._mm256_set1_epi64x(x)
            if isinstance(x, Ptr[u64]):
                return Vec._mm256_load_epi64(x)
            if isinstance(x, List[u64]):
                return Vec._mm256_load_epi64(x.arr.ptr)
            if isinstance(x, str):
                return Vec._mm256_load_epi64(x.ptr)
        if isinstance(T, u64) and N == 8:
            if isinstance(x, int):
                assert x >= 0, "SIMD: No support for negative int vectors added yet."
                return Vec._mm512_set1_epi64(u64(x))
            if isinstance(x, u64):
                return Vec._mm512_set1_epi64(x)
            if isinstance(x, Ptr[u64]):
                return Vec._mm512_load_epi64(x)
            if isinstance(x, List[u64]):
                return Vec._mm512_load_epi64(x.arr.ptr)
            if isinstance(x, str):
                return Vec._mm512_load_epi64(x.ptr)
        if isinstance(T, f32) and N == 8:
            if isinstance(x, f32):
                return Vec._mm256_set1_ps(x)
            if isinstance(x, Ptr[f32]):
                return Vec._mm256_loadu_ps(x)
            if isinstance(x, List[f32]):
                return Vec._mm256_loadu_ps(x.arr.ptr)
            if isinstance(x, Vec[u8, 16]):
                return Vec._mm256_castsi256_ps(Vec._mm256_cvtepi8_epi32(x))
        if isinstance(T, f32) and N == 16:
            if isinstance(x, f32):
                return Vec._mm512_set1_ps(x)
            if isinstance(x, Ptr[f32]):
                return Vec._mm512_loadu_ps(x)
            if isinstance(x, List[f32]):
                return Vec._mm512_loadu_ps(x.arr.ptr)
            if isinstance(x, Vec[u8, 32]):
                return Vec._mm512_castsi512_ps(Vec._mm512_cvtepi8_epi64(x))
        if isinstance(T, f64) and N == 4:
            if isinstance(x, f64):
                return Vec._mm256_set1_pd(x)
            if isinstance(x, Ptr[f64]):
                return Vec._mm256_loadu_pd(x)
            if isinstance(x, List[f64]):
                return Vec._mm256_loadu_pd(x.arr.ptr)
        if isinstance(T, f64) and N == 8:
            if isinstance(x, f64):
                return Vec._mm512_set1_pd(x)
            if isinstance(x, Ptr[f64]):
                return Vec._mm512_loadu_pd(x)
            if isinstance(x, List[f64]):
                return Vec._mm512_loadu_pd(x.arr.ptr)
        compile_error("invalid SIMD vector constructor")

    def __new__(x: str, offset: int = 0) -> Vec[u8, N]:
        return Vec(x.ptr + offset, u8, N)

    def __new__(x: List[T], offset: int = 0) -> Vec[T, N]:
        return Vec(x.arr.ptr + offset, T, N)

    def __new__(x) -> Vec[T, N]:
        return Vec(x, T, N)

    @llvm
    def _mm_cmpeq_epi8(x: Vec[u8, 16], y: Vec[u8, 16]) -> Vec[u8, 16]:
        %0 = icmp eq <16 x i8> %x, %y
        %1 = sext <16 x i1> %0 to <16 x i8>
        ret <16 x i8> %1

    def __eq__(self: Vec[u8, 16], other: Vec[u8, 16]) -> Vec[u8, 16]:
        return Vec._mm_cmpeq_epi8(self, other)

    @llvm
    def _mm256_cmpeq_epi8(x: Vec[u8, 32], y: Vec[u8, 32]) -> Vec[u8, 32]:
        %0 = icmp eq <32 x i8> %x, %y
        %1 = sext <32 x i1> %0 to <32 x i8>
        ret <32 x i8> %1

    def __eq__(self: Vec[u8, 32], other: Vec[u8, 32]) -> Vec[u8, 32]:
        return Vec._mm256_cmpeq_epi8(self, other)

    @llvm
    def _mm_andnot_si128(x: Vec[u8, 16], y: Vec[u8, 16]) -> Vec[u8, 16]:
        %0 = xor <16 x i8> %x, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
        %1 = and <16 x i8> %y, %0
        ret <16 x i8> %1

    def __ne__(self: Vec[u8, 16], other: Vec[u8, 16]) -> Vec[u8, 16]:
        return Vec._mm_andnot_si128((self == other), Vec.FF_16x8i)

    @llvm
    def _mm256_andnot_si256(x: Vec[u8, 32], y: Vec[u8, 32]) -> Vec[u8, 32]:
        %0 = xor <32 x i8> %x, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>
        %1 = and <32 x i8> %y, %0
        ret <32 x i8> %1

    def __ne__(self: Vec[u8, 32], other: Vec[u8, 32]) -> Vec[u8, 32]:
        return Vec._mm256_andnot_si256((self == other), Vec.FF_32x8i)

    def __eq__(self: Vec[u8, 16], other: bool) -> Vec[u8, 16]:
        if not other:
            return Vec._mm_andnot_si128(self, Vec.FF_16x8i)
        else:
            return Vec._mm_andnot_si128(self, Vec.ZERO_16x8i)

    def __eq__(self: Vec[u8, 32], other: bool) -> Vec[u8, 32]:
        if not other:
            return Vec._mm256_andnot_si256(self, Vec.FF_32x8i)
        else:
            return Vec._mm256_andnot_si256(self, Vec.ZERO_32x8i)

    @llvm
    def _mm_and_si128(x: Vec[u8, 16], y: Vec[u8, 16]) -> Vec[u8, 16]:
        %0 = and <16 x i8> %x, %y
        ret <16 x i8> %0

    def __and__(self: Vec[u8, 16], other: Vec[u8, 16]) -> Vec[u8, 16]:
        return Vec._mm_and_si128(self, other)

    @llvm
    def _mm_and_si256(x: Vec[u8, 32], y: Vec[u8, 32]) -> Vec[u8, 32]:
        %0 = and <32 x i8> %x, %y
        ret <32 x i8> %0

    def __and__(self: Vec[u8, 32], other: Vec[u8, 32]) -> Vec[u8, 32]:
        return Vec._mm_and_si256(self, other)

    @llvm
    def _mm256_and_si256(x: Vec[u64, 4], y: Vec[u64, 4]) -> Vec[u64, 4]:
        %0 = and <4 x i64> %x, %y
        ret <4 x i64> %0

    def __and__(self: Vec[u64, 4], other: Vec[u64, 4]) -> Vec[u64, 4]:
        return Vec._mm256_and_si256(self, other)

    @llvm
    def _mm256_and_ps(x: Vec[f32, 8], y: Vec[f32, 8]) -> Vec[f32, 8]:
        %0 = bitcast <8 x float> %x to <8 x i32>
        %1 = bitcast <8 x float> %y to <8 x i32>
        %2 = and <8 x i32> %0, %1
        %3 = bitcast <8 x i32> %2 to <8 x float>
        ret <8 x float> %3

    def __and__(self: Vec[f32, 8], other: Vec[f32, 8]) -> Vec[f32, 8]:
        return Vec._mm256_and_ps(self, other)

    @llvm
    def _mm512_and_ps(x: Vec[f32, 16], y: Vec[f32, 16]) -> Vec[f32, 16]:
        %0 = bitcast <16 x float> %x to <16 x i32>
        %1 = bitcast <16 x float> %y to <16 x i32>
        %2 = and <16 x i32> %0, %1
        %3 = bitcast <16 x i32> %2 to <16 x float>
        ret <16 x float> %3

    def __and__(self: Vec[f32, 16], other: Vec[f32, 16]) -> Vec[f32, 16]:
        return Vec._mm512_and_ps(self, other)

    @llvm
    def _mm_or_si128(x: Vec[u8, 16], y: Vec[u8, 16]) -> Vec[u8, 16]:
        %0 = or <16 x i8> %x, %y
        ret <16 x i8> %0

    def __or__(self: Vec[u8, 16], other: Vec[u8, 16]) -> Vec[u8, 16]:
        return Vec._mm_or_si128(self, other)

    @llvm
    def _mm_or_si256(x: Vec[u8, 32], y: Vec[u8, 32]) -> Vec[u8, 32]:
        %0 = or <32 x i8> %x, %y
        ret <32 x i8> %0

    def __or__(self: Vec[u8, 32], other: Vec[u8, 32]) -> Vec[u8, 32]:
        return Vec._mm_or_si256(self, other)

    @llvm
    def _mm256_or_ps(x: Vec[f32, 8], y: Vec[f32, 8]) -> Vec[f32, 8]:
        %0 = bitcast <8 x float> %x to <8 x i32>
        %1 = bitcast <8 x float> %y to <8 x i32>
        %2 = or <8 x i32> %0, %1
        %3 = bitcast <8 x i32> %2 to <8 x float>
        ret <8 x float> %3

    def __or__(self: Vec[f32, 8], other: Vec[f32, 8]) -> Vec[f32, 8]:
        return Vec._mm256_or_ps(self, other)

    @llvm
    def _mm512_or_ps(x: Vec[f32, 16], y: Vec[f32, 16]) -> Vec[f32, 16]:
        %0 = bitcast <16 x float> %x to <16 x i32>
        %1 = bitcast <16 x float> %y to <16 x i32>
        %2 = or <16 x i32> %0, %1
        %3 = bitcast <16 x i32> %2 to <16 x float>
        ret <16 x float> %3

    def __or__(self: Vec[f32, 16], other: Vec[f32, 16]) -> Vec[f32, 16]:
        return Vec._mm512_or_ps(self, other)

    @llvm
    def _mm_bsrli_si128_8(vec: Vec[u8, 16]) -> Vec[u8, 16]:
        %0 = shufflevector <16 x i8> %vec, <16 x i8> zeroinitializer, <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
        ret <16 x i8> %0

    @llvm
    def _mm256_add_ps(x: Vec[f32, 8], y: Vec[f32, 8]) -> Vec[f32, 8]:
        %0 = fadd <8 x float> %x, %y
        ret <8 x float> %0

    def __rshift__(self: Vec[u8, 16], shift: Static[int]) -> Vec[u8, 16]:
        if shift == 0:
            return self
        elif shift == 8:
            return Vec._mm_bsrli_si128_8(self)
        else:
            compile_error("invalid bitshift")

    @llvm
    def _mm_bsrli_256(vec: Vec[u8, 32]) -> Vec[u8, 32]:
        %0 = shufflevector <32 x i8> %vec, <32 x i8> zeroinitializer, <32 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47>
        ret <32 x i8> %0

    def __rshift__(self: Vec[u8, 32], shift: Static[int]) -> Vec[u8, 32]:
        if shift == 0:
            return self
        elif shift == 16:
            return Vec._mm_bsrli_256(self)
        else:
            compile_error("invalid bitshift")

    # @llvm  # https://stackoverflow.com/questions/6996764/fastest-way-to-do-horizontal-sse-vector-sum-or-other-reduction
    # def sum(self: Vec[f32, 8]) -> f32:
    #     %0 = shufflevector <8 x float> %self, <8 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
    #     %1 = shufflevector <8 x float> %self, <8 x float> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
    #     %2 = fadd <4 x float> %0, %1
    #     %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> <i32 1, i32 undef, i32 3, i32 undef>
    #     %4 = fadd <4 x float> %2, %3
    #     %5 = shufflevector <4 x float> %4, <4 x float> poison, <4 x i32> <i32 2, i32 undef, i32 undef, i32 undef>
    #     %6 = fadd <4 x float> %4, %5
    #     %7 = extractelement <4 x float> %6, i32 0
    #     ret float %7

    def sum(self: Vec[f32, 8], x: f32 = f32(0.0)) -> f32:
        return x + self[0] + self[1] + self[2] + self[3] + self[4] + self[5] + self[6] + self[7]

    def __repr__(self):
        return f"<{','.join(self.scatter())}>"

    # Methods below added ad-hoc. TODO: Add Intel intrinsics equivalents for them and integrate them neatly above.

    # Constructors
    @llvm
    def generic_init[SLS: Static[int]](val: T) -> Vec[T, SLS]:
        %0 = insertelement <{=SLS} x {=T}> undef, {=T} %val, i32 0
        %1 = shufflevector <{=SLS} x {=T}> %0, <{=SLS} x {=T}> undef, <{=SLS} x i32> zeroinitializer
        ret <{=SLS} x {=T}> %1

    @llvm
    def generic_load[SLS: Static[int]](data: Ptr[T]) -> Vec[T, SLS]:
        %0 = bitcast T* %data to <{=SLS} x {=T}>*
        %1 = load <{=SLS} x {=T}>, <{=SLS} x {=T}>* %0
        ret <{=SLS} x {=T}> %1

    # Bitwise intrinsics
    @llvm
    def __and__(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = and <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def __and__(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = and <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def __or__(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = or <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def __or__(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = or <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def __xor__(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = xor <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def __xor__(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = xor <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def __lshift__(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = shl <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def __lshift__(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = shl <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def __rshift__(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = lshr <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def __rshift__(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = lshr <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def bit_flip(self: Vec[u1, N]) -> Vec[u1, N]:
        %0 = insertelement <{=N} x i1> undef, i1 1, i32 0
        %1 = shufflevector <{=N} x i1> %0, <{=N} x i1> undef, <{=N} x i32> zeroinitializer
        %2 = xor <{=N} x i1> %self, %1
        ret <{=N} x i1> %2

    @llvm
    def shift_half(self: Vec[u128, N]) -> Vec[u128, N]:
        %0 = insertelement <{=N} x i128> undef, i128 64, i32 0
        %1 = shufflevector <{=N} x i128> %0, <{=N} x i128> undef, <{=N} x i32> zeroinitializer
        %2 = lshr <{=N} x i128> %self, %1
        ret <{=N} x i128> %2

    # Comparisons
    @llvm
    def __ge__(self: Vec[u64, N], other: Vec[u64, N]) -> Vec[u1, N]:
        %0 = icmp uge <{=N} x i64> %self, %other
        ret <{=N} x i1> %0

    @llvm
    def __ge__(self: Vec[u64, N], other: u64) -> Vec[u1, N]:
        %0 = insertelement <{=N} x i64> undef, i64 %other, i32 0
        %1 = shufflevector <{=N} x i64> %0, <{=N} x i64> undef, <{=N} x i32> zeroinitializer
        %2 = icmp uge <{=N} x i64> %self, %1
        ret <{=N} x i1> %2

    @llvm
    def __ge__(self: Vec[f64, N], other: Vec[f64, N]) -> Vec[u1, N]:
        %0 = fcmp oge <{=N} x double> %self, %other
        ret <{=N} x i1> %0

    @llvm
    def __ge__(self: Vec[f64, N], other: f64) -> Vec[u1, N]:
        %0 = insertelement <{=N} x double> undef, double %other, i32 0
        %1 = shufflevector <{=N} x double> %0, <{=N} x double> undef, <{=N} x i32> zeroinitializer
        %2 = fcmp oge <{=N} x double> %self, %1
        ret <{=N} x i1> %2

    @llvm
    def __le__(self: Vec[u64, N], other: Vec[u64, N]) -> Vec[u1, N]:
        %0 = icmp ule <{=N} x i64> %self, %other
        ret <{=N} x i1> %0

    @llvm
    def __le__(self: Vec[u64, N], other: u64) -> Vec[u1, N]:
        %0 = insertelement <{=N} x i64> undef, i64 %other, i32 0
        %1 = shufflevector <{=N} x i64> %0, <{=N} x i64> undef, <{=N} x i32> zeroinitializer
        %2 = icmp ule <{=N} x i64> %self, %1
        ret <{=N} x i1> %2

    # Arithmetic intrinsics
    @llvm
    def __neg__(self: Vec[T, N]) -> Vec[T, N]:
        %0 = sub <{=N} x {=T}> zeroinitializer, %self
        ret <{=N} x {=T}> %0

    @llvm
    def __mod__(self: Vec[u64, N], other: Vec[u64, N]) -> Vec[u64, N]:
        %0 = urem <{=N} x i64> %self, %other
        ret <{=N} x i64> %0

    @llvm
    def __mod__(self: Vec[u64, N], other: u64) -> Vec[u64, N]:
        %0 = insertelement <{=N} x i64> undef, i64 %other, i32 0
        %1 = shufflevector <{=N} x i64> %0, <{=N} x i64> undef, <{=N} x i32> zeroinitializer
        %2 = urem <{=N} x i64> %self, %1
        ret <{=N} x i64> %2

    @llvm
    def add(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = add <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def add(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = add <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def fadd(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = fadd <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def fadd(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = fadd <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    def __add__(self: Vec[T, N], other) -> Vec[T, N]:
        if isinstance(T, u8) or isinstance(T, u64) or isinstance(T, u128):
            return self.add(other)
        if isinstance(T, f32) or isinstance(T, f64):
            return self.fadd(other)
        compile_error("invalid SIMD vector addition")

    @llvm
    def sub(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = sub <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def sub(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = sub <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def fsub(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = fsub <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def fsub(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = fsub <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    def __sub__(self: Vec[T, N], other) -> Vec[T, N]:
        if isinstance(T, u8) or isinstance(T, u64):
            return self.sub(other)
        if isinstance(T, f32) or isinstance(T, f64):
            return self.fsub(other)
        compile_error("invalid SIMD vector subtraction")

    @llvm
    def mul(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = mul <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def fmul(self: Vec[T, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = fmul <{=N} x {=T}> %self, %other
        ret <{=N} x {=T}> %0

    @llvm
    def mul(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = mul <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    @llvm
    def fmul(self: Vec[T, N], other: T) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = fmul <{=N} x {=T}> %self, %1
        ret <{=N} x {=T}> %2

    def __mul__(self: Vec[T, N], other) -> Vec[T, N]:
        if isinstance(T, u8) or isinstance(T, u64) or isinstance(T, u128):
            return self.mul(other)
        if isinstance(T, f32) or isinstance(T, f64):
            return self.fmul(other)
        compile_error("invalid SIMD vector multiplication")

    @llvm
    def __truediv__(self: Vec[u64, N], other: Vec[u64, N]) -> Vec[f64, N]:
        %0 = uitofp <{=N} x i64> %self to <{=N} x double>
        %1 = uitofp <{=N} x i64> %other to <{=N} x double>
        %2 = fdiv <{=N} x double> %0, %1
        ret <{=N} x double> %2

    @llvm
    def __truediv__(self: Vec[u64, 4], other: Vec[u64, 4]) -> Vec[f64, 4]:
        %0 = uitofp <4 x i64> %self to <4 x double>
        %1 = uitofp <4 x i64> %other to <4 x double>
        %2 = fdiv <4 x double> %0, %1
        ret <4 x double> %2

    @llvm
    def __truediv__(self: Vec[u64, 8], other: Vec[u64, 8]) -> Vec[f64, 8]:
        %0 = uitofp <8 x i64> %self to <8 x double>
        %1 = uitofp <8 x i64> %other to <8 x double>
        %2 = fdiv <8 x double> %0, %1
        ret <8 x double> %2

    @llvm
    def __truediv__(self: Vec[u64, 8], other: u64) -> Vec[f64, 8]:
        %0 = uitofp <8 x i64> %self to <8 x double>
        %1 = uitofp i64 %other to double
        %2 = insertelement <8 x double> undef, double %1, i32 0
        %3 = shufflevector <8 x double> %0, <8 x double> undef, <8 x i32> zeroinitializer
        %4 = fdiv <8 x double> %0, %3
        ret <8 x double> %4

    @llvm
    def add_overflow(self: Vec[u64, N], other: Vec[u64, N]) -> Tuple[Vec[u64, N], Vec[u1, N]]:
        declare {<{=N} x i64>, <{=N} x i1>} @llvm.uadd.with.overflow.v{=N}i64(<{=N} x i64>, <{=N} x i64>)
        %0 = call {<{=N} x i64>, <{=N} x i1>} @llvm.uadd.with.overflow.v{=N}i64(<{=N} x i64> %self, <{=N} x i64> %other)
        ret {<{=N} x i64>, <{=N} x i1>} %0

    @llvm
    def add_overflow(self: Vec[u64, N], other: u64) -> Tuple[Vec[u64, N], Vec[u1, N]]:
        declare {<{=N} x i64>, <{=N} x i1>} @llvm.uadd.with.overflow.v{=N}i64(<{=N} x i64>, <{=N} x i64>)
        %0 = insertelement <{=N} x i64> undef, i64 %other, i32 0
        %1 = shufflevector <{=N} x i64> %0, <{=N} x i64> undef, <{=N} x i32> zeroinitializer
        %2 = call {<{=N} x i64>, <{=N} x i1>} @llvm.uadd.with.overflow.v{=N}i64(<{=N} x i64> %self, <{=N} x i64> %1)
        ret {<{=N} x i64>, <{=N} x i1>} %2

    @llvm
    def sub_overflow(self: Vec[u64, N], other: Vec[u64, N]) -> Tuple[Vec[u64, N], Vec[u1, N]]:
        declare {<{=N} x i64>, <{=N} x i1>} @llvm.usub.with.overflow.v{=N}i64(<{=N} x i64>, <{=N} x i64>)
        %0 = call {<{=N} x i64>, <{=N} x i1>} @llvm.usub.with.overflow.v{=N}i64(<{=N} x i64> %self, <{=N} x i64> %other)
        ret {<{=N} x i64>, <{=N} x i1>} %0

    def sub_overflow_commutative(self: Vec[u64, N], other: u64) -> Tuple[Vec[u64, N], Vec[u1, N]]:
        return Vec[u64, N](other).sub_overflow(self)

    @llvm
    def sub_overflow(self: Vec[u64, N], other: u64) -> Tuple[Vec[u64, N], Vec[u1, N]]:
        declare {<{=N} x i64>, <{=N} x i1>} @llvm.usub.with.overflow.v{=N}i64(<{=N} x i64>, <{=N} x i64>)
        %0 = insertelement <{=N} x i64> undef, i64 %other, i32 0
        %1 = shufflevector <{=N} x i64> %0, <{=N} x i64> undef, <{=N} x i32> zeroinitializer
        %2 = call {<{=N} x i64>, <{=N} x i1>} @llvm.usub.with.overflow.v{=N}i64(<{=N} x i64> %self, <{=N} x i64> %1)
        ret {<{=N} x i64>, <{=N} x i1>} %2

    @llvm
    def zext_mul(self: Vec[u64, N], other: Vec[u64, N]) -> Vec[u128, N]:
        %0 = zext <{=N} x i64> %self to <{=N} x i128>
        %1 = zext <{=N} x i64> %other to <{=N} x i128>
        %2 = mul nuw <{=N} x i128> %0, %1
        ret <{=N} x i128> %2

    @llvm
    def zext_mul(self: Vec[u64, N], other: u64) -> Vec[u128, N]:
        %0 = zext <{=N} x i64> %self to <{=N} x i128>
        %1 = insertelement <{=N} x i64> undef, i64 %other, i32 0
        %2 = shufflevector <{=N} x i64> %1, <{=N} x i64> undef, <{=N} x i32> zeroinitializer
        %3 = zext <{=N} x i64> %2 to <{=N} x i128>
        %4 = mul nuw <{=N} x i128> %0, %3
        ret <{=N} x i128> %4

    @nocapture
    @llvm
    def mulx(self: Vec[u64, N], other: Vec[u64, N], hi: Ptr[Vec[u64, N]]) -> Vec[u64, N]:
        %0 = zext <{=N} x i64> %self to <{=N} x i128>
        %1 = zext <{=N} x i64> %other to <{=N} x i128>
        %2 = mul nuw <{=N} x i128> %0, %1
        %3 = lshr <{=N} x i128> %2, <i128 64, i128 64, i128 64, i128 64, i128 64, i128 64, i128 64, i128 64>
        %4 = trunc <{=N} x i128> %3 to <{=N} x i64>
        store <{=N} x i64> %4, <{=N} x i64>* %hi, align 8
        %5 = trunc <{=N} x i128> %2 to <{=N} x i64>
        ret <{=N} x i64> %5

    def mulhi(self: Vec[u64, N], other: Vec[u64, N]) -> Vec[u64, N]:
        hi = Ptr[Vec[u64, N]](1)
        self.mulx(other, hi)
        return hi[0]

    @llvm
    def sqrt(self: Vec[f64, N]) -> Vec[f64, N]:
        declare <{=N} x double> @llvm.sqrt.v{=N}f64(<{=N} x double>)
        %0 = call <{=N} x double> @llvm.sqrt.v{=N}f64(<{=N} x double> %self)
        ret <{=N} x double> %0

    @llvm
    def log(self: Vec[f64, N]) -> Vec[f64, N]:
        declare <{=N} x double> @llvm.log.v{=N}f64(<{=N} x double>)
        %0 = call <{=N} x double> @llvm.log.v{=N}f64(<{=N} x double> %self)
        ret <{=N} x double> %0

    @llvm
    def cos(self: Vec[f64, N]) -> Vec[f64, N]:
        declare <{=N} x double> @llvm.cos.v{=N}f64(<{=N} x double>)
        %0 = call <{=N} x double> @llvm.cos.v{=N}f64(<{=N} x double> %self)
        ret <{=N} x double> %0

    @llvm
    def fabs(self: Vec[f64, N]) -> Vec[f64, N]:
        declare <{=N} x double> @llvm.fabs.v{=N}f64(<{=N} x double>)
        %0 = call <{=N} x double> @llvm.fabs.v{=N}f64(<{=N} x double> %self)
        ret <{=N} x double> %0

    # Conversion intrinsics
    @llvm
    def zext_double(self: Vec[u64, N]) -> Vec[u128, N]:
        %0 = zext <{=N} x i64> %self to <{=N} x i128>
        ret <{=N} x i128> %0

    @llvm
    def trunc_half(self: Vec[u128, N]) -> Vec[u64, N]:
        %0 = trunc <{=N} x i128> %self to <{=N} x i64>
        ret <{=N} x i64> %0

    @llvm
    def shift_trunc_half(self: Vec[u128, N]) -> Vec[u64, N]:
        %0 = insertelement <{=N} x i128> undef, i128 64, i32 0
        %1 = shufflevector <{=N} x i128> %0, <{=N} x i128> undef, <{=N} x i32> zeroinitializer
        %2 = lshr <{=N} x i128> %self, %1
        %3 = trunc <{=N} x i128> %2 to <{=N} x i64>
        ret <{=N} x i64> %3

    @llvm
    def to_u64(self: Vec[f64, N]) -> Vec[u64, N]:
        %0 = fptoui <{=N} x double> %self to <{=N} x i64>
        ret <{=N} x i64> %0

    @llvm
    def to_u64(self: Vec[u1, N]) -> Vec[u64, N]:
        %0 = zext <{=N} x i1> %self to <{=N} x i64>
        ret <{=N} x i64> %0

    @llvm
    def to_float(self: Vec[T, N]) -> Vec[f64, N]:
        %0 = uitofp <{=N} x {=T}> %self to <{=N} x double>
        ret <{=N} x double> %0

    # Predication intrinsics
    @llvm
    def sub_if(self: Vec[T, N], other: Vec[T, N], mask: Vec[u1, N]) -> Vec[T, N]:
        %0 = sub <{=N} x {=T}> %self, %other
        %1 = select <{=N} x i1> %mask, <{=N} x {=T}> %0, <{=N} x {=T}> %self
        ret <{=N} x {=T}> %1

    @llvm
    def sub_if(self: Vec[T, N], other: T, mask: Vec[u1, N]) -> Vec[T, N]:
        %0 = insertelement <{=N} x {=T}> undef, {=T} %other, i32 0
        %1 = shufflevector <{=N} x {=T}> %0, <{=N} x {=T}> undef, <{=N} x i32> zeroinitializer
        %2 = sub <{=N} x {=T}> %self, %1
        %3 = select <{=N} x i1> %mask, <{=N} x {=T}> %2, <{=N} x {=T}> %self
        ret <{=N} x {=T}> %3

    # Gather-scatter
    @llvm
    def __getitem__(self: Vec[T, N], idx) -> T:
        %0 = extractelement <{=N} x {=T}> %self, i64 %idx
        ret {=T} %0

    # Misc
    def copy(self: Vec[T, N]) -> Vec[T, N]:
        return self

    @llvm
    def mask(self: Vec[T, N], mask: Vec[u1, N], other: Vec[T, N]) -> Vec[T, N]:
        %0 = select <{=N} x i1> %mask, <{=N} x {=T}> %self, <{=N} x {=T}> %other
        ret <{=N} x {=T}> %0

    def scatter(self: Vec[T, N]) -> List[T]:
        return [self[i] for i in staticrange(N)]


u8x16 = Vec[u8, 16]
u8x32 = Vec[u8, 32]
f32x8 = Vec[f32, 8]


@llvm
def bitcast_scatter[N: Static[int]](ptr_in: Ptr[Vec[u64, N]]) -> Ptr[u64]:
    %0 = bitcast <{=N} x i64>* %ptr_in to i64*
    ret i64* %0


@llvm
def bitcast_vectorize[N: Static[int]](ptr_in: Ptr[u64]) -> Ptr[Vec[u64, N]]:
    %0 = bitcast i64* %ptr_in to <{=N} x i64>*
    ret <{=N} x i64>* %0