Add lane construction and composition APIs

[hez2010]

This PR adds lane construction and composition APIs approved in #122557, and the corresponding JIT intrinsics.

The JIT now recognizes the new vector APIs and expands them using existing SIMD nodes. The managed implementation allows decomposition through smaller vector widths when wider hardware support is unavailable.

The xarch lowering uses fixed shuffle forms where profitable:

• vpbroadcast* for sequence and alternating construction
• vshufps and vpunpck* for 128-bit concat/zip/unzip patterns
• full-width unpack plus vperm2i128 for 256-bit zip/unzip
• EVEX shuffle/permute forms for 512-bit operations when AVX512 is available
• fixed immediate shuffles for reverse

The ARM64 lowering avoids table-lookup forms for small fixed concat/reverse operations and uses direct element moves where applicable, such as ins and rev64.

CreateCauchySequence has been removed as it doesn't make sense at all. And SignSequence has been moved to an extension property as we only want to expose it for signed numbers.

Codegen:

Vector128

; Vector128.CreateGeometricSequence<int>(initial, 3); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M55419_IG01: ;; offset=0x0000G_M55419_IG02: ;; offset=0x0000vpbroadcastdxmm0,edx vpmulld xmm0,xmm0, xmmword ptr [reloc @RWD00]vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M55419_IG03: ;; offset=0x0016retRWD00 dq	0000000300000001h,0000001B00000009h; Total bytes of code 23; Vector128.CreateAlternatingSequence<int>(even, odd); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M23632_IG01: ;; offset=0x0000G_M23632_IG02: ;; offset=0x0000vpbroadcastdxmm0,edxvpbroadcastdxmm1,r8d vpunpckldq xmm0,xmm0,xmm1vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M23632_IG03: ;; offset=0x0017ret; Total bytes of code 24; Vector128<int>.SignSequence; Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M37526_IG01: ;; offset=0x0000G_M37526_IG02: ;; offset=0x0000 vmovddup xmm0, qword ptr [reloc @RWD00]vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M37526_IG03: ;; offset=0x000FretRWD00 dq	FFFFFFFF00000001h; Total bytes of code 16; Vector128.CreateHarmonicSequence<float>(start, step); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M24925_IG01: ;; offset=0x0000G_M24925_IG02: ;; offset=0x0000vbroadcastssxmm0,xmm2 vmulps xmm0,xmm0, xmmword ptr [reloc @RWD00]vbroadcastssxmm1,xmm1 vaddps xmm0,xmm1,xmm0vbroadcastssxmm1, dword ptr [reloc @RWD16] vdivps xmm0,xmm1,xmm0vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M24925_IG03: ;; offset=0x002AretRWD00 dq	3F80000000000000h,4040000040000000hRWD16 dd	3F800000h ; 1; Total bytes of code 43; Vector128.CreateGeometricSequence<float>(initial, 3.0f); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M52564_IG01: ;; offset=0x0000G_M52564_IG02: ;; offset=0x0000vbroadcastssxmm0,xmm1 vmulps xmm0,xmm0, xmmword ptr [reloc @RWD00]vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M52564_IG03: ;; offset=0x0014retRWD00 dq	404000003F800000h,41D8000041100000h; Total bytes of code 21; Vector128.ConcatLowerLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M47101_IG01: ;; offset=0x0000G_M47101_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vshufpsxmm0,xmm0, xmmword ptr [r8],68vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M47101_IG03: ;; offset=0x0011ret; Total bytes of code 18; Vector128.ConcatLowerUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M48932_IG01: ;; offset=0x0000G_M48932_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vshufpsxmm0,xmm0, xmmword ptr [r8],-28vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M48932_IG03: ;; offset=0x0011ret; Total bytes of code 18; Vector128.ConcatUpperLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M24196_IG01: ;; offset=0x0000G_M24196_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vshufpsxmm0,xmm0, xmmword ptr [r8],78vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M24196_IG03: ;; offset=0x0011ret; Total bytes of code 18; Vector128.ConcatUpperUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M45405_IG01: ;; offset=0x0000G_M45405_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vshufpsxmm0,xmm0, xmmword ptr [r8],-18vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M45405_IG03: ;; offset=0x0011ret; Total bytes of code 18; Vector128.ZipLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M10057_IG01: ;; offset=0x0000G_M10057_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx] vpunpckldq xmm0,xmm0, xmmword ptr [r8]vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M10057_IG03: ;; offset=0x0010ret; Total bytes of code 17; Vector128.ZipUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M48936_IG01: ;; offset=0x0000G_M48936_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx] vpunpckhdq xmm0,xmm0, xmmword ptr [r8]vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M48936_IG03: ;; offset=0x0010ret; Total bytes of code 17; Vector128.Zip<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M32727_IG01: ;; offset=0x0000G_M32727_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vmovupsxmm1, xmmword ptr [r8] vpunpckldq xmm2,xmm0,xmm1 vpunpckhdq xmm0,xmm0,xmm1vmovups xmmword ptr [rcx],xmm2vmovups xmmword ptr [rcx+0x10],xmm0movrax,rcxG_M32727_IG03: ;; offset=0x001Dret; Total bytes of code 30; Vector128.UnzipEven<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M32745_IG01: ;; offset=0x0000G_M32745_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vshufpsxmm0,xmm0, xmmword ptr [r8],-120vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M32745_IG03: ;; offset=0x0011ret; Total bytes of code 18; Vector128.UnzipOdd<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M32638_IG01: ;; offset=0x0000G_M32638_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vshufpsxmm0,xmm0, xmmword ptr [r8],-35vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M32638_IG03: ;; offset=0x0011ret; Total bytes of code 18; Vector128.Unzip<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M7116_IG01: ;; offset=0x0000G_M7116_IG02: ;; offset=0x0000vmovupsxmm0, xmmword ptr [rdx]vmovupsxmm1, xmmword ptr [r8]vshufpsxmm2,xmm0,xmm1,-120vshufpsxmm0,xmm0,xmm1,-35vmovups xmmword ptr [rcx],xmm2vmovups xmmword ptr [rcx+0x10],xmm0movrax,rcxG_M7116_IG03: ;; offset=0x001Fret; Total bytes of code 32; Vector128.Reverse<int>(vector); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M36184_IG01: ;; offset=0x0000G_M36184_IG02: ;; offset=0x0000 vpshufd xmm0, xmmword ptr [rdx],27vmovups xmmword ptr [rcx],xmm0movrax,rcxG_M36184_IG03: ;; offset=0x000Cret; Total bytes of code 13

Vector256

; Vector256.CreateGeometricSequence<int>(initial, 3); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M59643_IG01: ;; offset=0x0000G_M59643_IG02: ;; offset=0x0000vpbroadcastdymm0,edx vpmulld ymm0,ymm0, ymmword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M59643_IG03: ;; offset=0x0016vzeroupperretRWD00 dq	0000000300000001h,0000001B00000009h,000000F300000051h,0000088B000002D9h; Total bytes of code 26; Vector256.CreateAlternatingSequence<int>(even, odd); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M40720_IG01: ;; offset=0x0000G_M40720_IG02: ;; offset=0x0000vpbroadcastdymm0,edxvmovupsymm1, ymmword ptr [reloc @RWD00]vpbroadcastdymm2,r8dvpermt2dymm2,ymm1,ymm0vmovups ymmword ptr [rcx],ymm2movrax,rcxG_M40720_IG03: ;; offset=0x0021vzeroupperretRWD00 dq	0000000000000008h,0000000100000009h,000000020000000Ah,000000030000000Bh; Total bytes of code 37; Vector256<int>.SignSequence; Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M7766_IG01: ;; offset=0x0000G_M7766_IG02: ;; offset=0x0000vbroadcastsdymm0, qword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M7766_IG03: ;; offset=0x0010vzeroupperretRWD00 dq	FFFFFFFF00000001h; Total bytes of code 20; Vector256.CreateHarmonicSequence<float>(start, step); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M59741_IG01: ;; offset=0x0000G_M59741_IG02: ;; offset=0x0000vbroadcastssymm0,ymm2 vmulps ymm0,ymm0, ymmword ptr [reloc @RWD00]vbroadcastssymm1,ymm1 vaddps ymm0,ymm1,ymm0vbroadcastssymm1, dword ptr [reloc @RWD32] vdivps ymm0,ymm1,ymm0vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M59741_IG03: ;; offset=0x002AvzeroupperretRWD00 dq	3F80000000000000h,4040000040000000h,40A0000040800000h,40E0000040C00000hRWD32 dd	3F800000h ; 1; Total bytes of code 46; Vector256.CreateGeometricSequence<float>(initial, 3.0f); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M42068_IG01: ;; offset=0x0000G_M42068_IG02: ;; offset=0x0000vbroadcastssymm0,ymm1 vmulps ymm0,ymm0, ymmword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M42068_IG03: ;; offset=0x0014vzeroupperretRWD00 dq	404000003F800000h,41D8000041100000h,4373000042A20000h,4508B00044364000h; Total bytes of code 24; Vector256.ConcatLowerLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M46973_IG01: ;; offset=0x0000G_M46973_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx] vmovdqu xmm1, xmmword ptr [r8]vinserti128ymm0,ymm0,xmm1vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M46973_IG03: ;; offset=0x0016vzeroupperret; Total bytes of code 26; Vector256.ConcatLowerUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M9060_IG01: ;; offset=0x0000G_M9060_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vmovupsymm1, ymmword ptr [r8]vextracti128xmm1,ymm1vinserti128ymm0,ymm0,xmm1vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M9060_IG03: ;; offset=0x001Cvzeroupperret; Total bytes of code 32; Vector256.ConcatUpperLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M57796_IG01: ;; offset=0x0000G_M57796_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vextracti128xmm0,ymm0 vmovdqu xmm1, xmmword ptr [r8]vinserti128ymm0,ymm0,xmm1vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M57796_IG03: ;; offset=0x001Cvzeroupperret; Total bytes of code 32; Vector256.ConcatUpperUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M7901_IG01: ;; offset=0x0000G_M7901_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vextracti128xmm0,ymm0vmovupsymm1, ymmword ptr [r8]vinserti128ymm0,ymm1,xmm0vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M7901_IG03: ;; offset=0x001Cvzeroupperret; Total bytes of code 32; Vector256.ZipLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M61705_IG01: ;; offset=0x0000G_M61705_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vmovupsymm1, ymmword ptr [reloc @RWD00]vpermt2dymm0,ymm1, ymmword ptr [r8]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M61705_IG03: ;; offset=0x0019vzeroupperretRWD00 dq	0000000800000000h,0000000900000001h,0000000A00000002h,0000000B00000003h; Total bytes of code 29; Vector256.ZipUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M38056_IG01: ;; offset=0x0000G_M38056_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vmovupsymm1, ymmword ptr [reloc @RWD00]vpermt2dymm0,ymm1, ymmword ptr [r8]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M38056_IG03: ;; offset=0x0019vzeroupperretRWD00 dq	0000000C00000004h,0000000D00000005h,0000000E00000006h,0000000F00000007h; Total bytes of code 29; Vector256.Zip<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M24445_IG01: ;; offset=0x0000G_M24445_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vmovupsymm1, ymmword ptr [r8] vpunpckldq ymm2,ymm0,ymm1 vpunpckhdq ymm0,ymm0,ymm1vperm2i128ymm1,ymm2,ymm0,32vperm2i128ymm0,ymm2,ymm0,49vmovups ymmword ptr [rcx],ymm1vmovups ymmword ptr [rcx+0x20],ymm0movrax,rcxG_M24445_IG03: ;; offset=0x0029vzeroupperret; Total bytes of code 45; Vector256.UnzipEven<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M4905_IG01: ;; offset=0x0000G_M4905_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vmovupsymm1, ymmword ptr [reloc @RWD00]vpermt2dymm0,ymm1, ymmword ptr [r8]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M4905_IG03: ;; offset=0x0019vzeroupperretRWD00 dq	0000000200000000h,0000000600000004h,0000000A00000008h,0000000E0000000Ch; Total bytes of code 29; Vector256.UnzipOdd<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M54526_IG01: ;; offset=0x0000G_M54526_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [rdx]vmovupsymm1, ymmword ptr [reloc @RWD00]vpermt2dymm0,ymm1, ymmword ptr [r8]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M54526_IG03: ;; offset=0x0019vzeroupperretRWD00 dq	0000000300000001h,0000000700000005h,0000000B00000009h,0000000F0000000Dh; Total bytes of code 29; Vector256.Unzip<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M52326_IG01: ;; offset=0x0000G_M52326_IG02: ;; offset=0x0000 vpshufd ymm0, ymmword ptr [rdx],-40vpermqymm0,ymm0,-40 vpshufd ymm1, ymmword ptr [r8],-40vpermqymm1,ymm1,-40vperm2i128ymm2,ymm0,ymm1,32vperm2i128ymm0,ymm0,ymm1,49vmovups ymmword ptr [rcx],ymm2vmovups ymmword ptr [rcx+0x20],ymm0movrax,rcxG_M52326_IG03: ;; offset=0x002Fvzeroupperret; Total bytes of code 51; Vector256.Reverse<int>(vector); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M23794_IG01: ;; offset=0x0000G_M23794_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [reloc @RWD00]vpermdymm0,ymm0, ymmword ptr [rdx]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M23794_IG03: ;; offset=0x0014vzeroupperretRWD00 dq	0000000600000007h,0000000400000005h,0000000200000003h,0000000000000001h; Total bytes of code 24

Vector512

; Vector512.CreateGeometricSequence<int>(initial, 3); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M55451_IG01: ;; offset=0x0000G_M55451_IG02: ;; offset=0x0000vpbroadcastdzmm0,edx vpmulld zmm0,zmm0, zmmword ptr [reloc @RWD00]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M55451_IG03: ;; offset=0x0019vzeroupperretRWD00 dq	0000000300000001h,0000001B00000009h,000000F300000051h,0000088B000002D9h,00004CE3000019A1h,0002B3FB0000E6A9h,001853D300081BF1h,00DAF26B0048FB79h; Total bytes of code 29; Vector512.CreateAlternatingSequence<int>(even, odd); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M560_IG01: ;; offset=0x0000G_M560_IG02: ;; offset=0x0000vpbroadcastdzmm0,edxvmovupszmm1, zmmword ptr [reloc @RWD00]vpbroadcastdzmm2,r8dvpermt2dzmm2,zmm1,zmm0vmovups zmmword ptr [rcx],zmm2movrax,rcxG_M560_IG03: ;; offset=0x0025vzeroupperretRWD00 dq	0000000000000010h,0000000100000011h,0000000200000012h,0000000300000013h,0000000400000014h,0000000500000015h,0000000600000016h,0000000700000017h; Total bytes of code 41; Vector512<int>.SignSequence; Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M63542_IG01: ;; offset=0x0000G_M63542_IG02: ;; offset=0x0000vbroadcastsdzmm0, qword ptr [reloc @RWD00]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M63542_IG03: ;; offset=0x0013vzeroupperretRWD00 dq	FFFFFFFF00000001h; Total bytes of code 23; Vector512.CreateHarmonicSequence<float>(start, step); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M41245_IG01: ;; offset=0x0000G_M41245_IG02: ;; offset=0x0000vbroadcastsszmm0,zmm2 vmulps zmm0,zmm0, zmmword ptr [reloc @RWD00]vbroadcastsszmm1,zmm1 vaddps zmm0,zmm1,zmm0vbroadcastsszmm1, dword ptr [reloc @RWD64] vdivps zmm0,zmm1,zmm0vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M41245_IG03: ;; offset=0x0035vzeroupperretRWD00 dq	3F80000000000000h,4040000040000000h,40A0000040800000h,40E0000040C00000h,4110000041000000h,4130000041200000h,4150000041400000h,4170000041600000hRWD64 dd	3F800000h ; 1; Total bytes of code 57; Vector512.CreateGeometricSequence<float>(initial, 3.0f); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M35892_IG01: ;; offset=0x0000G_M35892_IG02: ;; offset=0x0000vbroadcastsszmm0,zmm1 vmulps zmm0,zmm0, zmmword ptr [reloc @RWD00]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M35892_IG03: ;; offset=0x0019vzeroupperretRWD00 dq	404000003F800000h,41D8000041100000h,4373000042A20000h,4508B00044364000h,4699C60045CD0800h,482CFEC04766A900h,49C29E984901BF10h,4B5AF26B4A91F6F2h; Total bytes of code 29; Vector512.ConcatLowerLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M13789_IG01: ;; offset=0x0000G_M13789_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx] vmovdqu ymm1, ymmword ptr [r8]vinserti32x8zmm0,zmm0,ymm1,1vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M13789_IG03: ;; offset=0x001Bvzeroupperret; Total bytes of code 31; Vector512.ConcatLowerUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M36420_IG01: ;; offset=0x0000G_M36420_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vmovupszmm1, zmmword ptr [r8]vextracti32x8ymm1,zmm1,1vinserti32x8zmm0,zmm0,ymm1,1vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M36420_IG03: ;; offset=0x0023vzeroupperret; Total bytes of code 39; Vector512.ConcatUpperLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M56164_IG01: ;; offset=0x0000G_M56164_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vextracti32x8ymm0,zmm0,1 vmovdqu ymm1, ymmword ptr [r8]vinserti32x8zmm0,zmm0,ymm1,1vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M56164_IG03: ;; offset=0x0022vzeroupperret; Total bytes of code 38; Vector512.ConcatUpperUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M13821_IG01: ;; offset=0x0000G_M13821_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vextracti32x8ymm0,zmm0,1vmovupszmm1, zmmword ptr [r8]vinserti32x8zmm0,zmm1,ymm0,0vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M13821_IG03: ;; offset=0x0023vzeroupperret; Total bytes of code 39; Vector512.ZipLower<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M59881_IG01: ;; offset=0x0000G_M59881_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vmovupszmm1, zmmword ptr [reloc @RWD00]vpermt2dzmm0,zmm1, zmmword ptr [r8]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M59881_IG03: ;; offset=0x001FvzeroupperretRWD00 dq	0000001000000000h,0000001100000001h,0000001200000002h,0000001300000003h,0000001400000004h,0000001500000005h,0000001600000006h,0000001700000007h; Total bytes of code 35; Vector512.ZipUpper<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M10568_IG01: ;; offset=0x0000G_M10568_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vmovupszmm1, zmmword ptr [reloc @RWD00]vpermt2dzmm0,zmm1, zmmword ptr [r8]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M10568_IG03: ;; offset=0x001FvzeroupperretRWD00 dq	0000001800000008h,0000001900000009h,0000001A0000000Ah,0000001B0000000Bh,0000001C0000000Ch,0000001D0000000Dh,0000001E0000000Eh,0000001F0000000Fh; Total bytes of code 35; Vector512.Zip<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M59130_IG01: ;; offset=0x0000G_M59130_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vmovupszmm1, zmmword ptr [r8] vpunpckhdq zmm2,zmm0,zmm1 vpunpckldq zmm0,zmm0,zmm1vshufi32x4zmm1,zmm0,zmm2,68vshufi32x4zmm1,zmm1,zmm1,-40vshufi32x4zmm0,zmm0,zmm2,-18vshufi32x4zmm0,zmm0,zmm0,-40vmovups zmmword ptr [rcx],zmm1vmovups zmmword ptr [rcx+0x40],zmm0movrax,rcxG_M59130_IG03: ;; offset=0x0044vzeroupperret; Total bytes of code 72; Vector512.UnzipEven<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M30057_IG01: ;; offset=0x0000G_M30057_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vmovupszmm1, zmmword ptr [reloc @RWD00]vpermt2dzmm0,zmm1, zmmword ptr [r8]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M30057_IG03: ;; offset=0x001FvzeroupperretRWD00 dq	0000000200000000h,0000000600000004h,0000000A00000008h,0000000E0000000Ch,0000001200000010h,0000001600000014h,0000001A00000018h,0000001E0000001Ch; Total bytes of code 35; Vector512.UnzipOdd<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M65310_IG01: ;; offset=0x0000G_M65310_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vmovupszmm1, zmmword ptr [reloc @RWD00]vpermt2dzmm0,zmm1, zmmword ptr [r8]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M65310_IG03: ;; offset=0x001FvzeroupperretRWD00 dq	0000000300000001h,0000000700000005h,0000000B00000009h,0000000F0000000Dh,0000001300000011h,0000001700000015h,0000001B00000019h,0000001F0000001Dh; Total bytes of code 35; Vector512.Unzip<int>(left, right); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M1377_IG01: ;; offset=0x0000G_M1377_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [rdx]vmovupszmm1, zmmword ptr [reloc @RWD00]vmovupszmm2, zmmword ptr [r8] vmovaps zmm3,zmm2vpermt2dzmm3,zmm1,zmm0vmovupszmm1, zmmword ptr [reloc @RWD64]vpermt2dzmm2,zmm1,zmm0vmovups zmmword ptr [rcx],zmm2vmovups zmmword ptr [rcx+0x40],zmm3movrax,rcxG_M1377_IG03: ;; offset=0x0042vzeroupperretRWD00 dq	0000001300000011h,0000001700000015h,0000001B00000019h,0000001F0000001Dh,0000000300000001h,0000000700000005h,0000000B00000009h,0000000F0000000DhRWD64 dq	0000001200000010h,0000001600000014h,0000001A00000018h,0000001E0000001Ch,0000000200000000h,0000000600000004h,0000000A00000008h,0000000E0000000Ch; Total bytes of code 70; Vector512.Reverse<int>(vector); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M34517_IG01: ;; offset=0x0000G_M34517_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [reloc @RWD00]vpermdzmm0,zmm0, zmmword ptr [rdx]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M34517_IG03: ;; offset=0x0019vzeroupperretRWD00 dq	0000000E0000000Fh,0000000C0000000Dh,0000000A0000000Bh,0000000800000009h,0000000600000007h,0000000400000005h,0000000200000003h,0000000000000001h; Total bytes of code 29

ARM64

; Vector64.CreateGeometricSequence<int>(initial, 3); Emitting BLENDED_CODE for arm64 on WindowsG_M6971_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M6971_IG02: ;; offset=0x0008 dup v0.2s, w0 ldr d16,[@RWD00]mul v0.2s, v0.2s, v16.2sG_M6971_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lrRWD00 dq 0000000300000001h; Total bytes of code 28; Vector64.CreateAlternatingSequence<int>(even, odd); Emitting BLENDED_CODE for arm64 on WindowsG_M10160_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M10160_IG02: ;; offset=0x0008 dup v0.2s, w0 dup v16.2s, w1 zip1 v0.2s, v0.2s, v16.2sG_M10160_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 28; Vector64<int>.SignSequence; Emitting BLENDED_CODE for arm64 on WindowsG_M46198_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M46198_IG02: ;; offset=0x0008 ldr d0,[@RWD00]G_M46198_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lrRWD00 dq FFFFFFFF00000001h; Total bytes of code 20; Vector64.CreateHarmonicSequence<float>(start, step); Emitting BLENDED_CODE for arm64 on WindowsG_M43901_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M43901_IG02: ;; offset=0x0008 ldr d16,[@RWD00]fmul v16.2s, v16.2s, v1.s[0] dup v0.2s, v0.s[0]fadd v0.2s, v16.2s, v0.2s ldr d16,[@RWD08]fdiv v0.2s, v16.2s, v0.2sG_M43901_IG03: ;; offset=0x0020 ldp fp, lr,[sp], #0x10ret lrRWD00 dq	3F80000000000000hRWD08 dq	3F8000003F800000h; Total bytes of code 40; Vector64.ConcatLowerLower<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M48061_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M48061_IG02: ;; offset=0x0008ins v0.s[1], v1.s[0]G_M48061_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector64.ConcatLowerUpper<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M62724_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M62724_IG02: ;; offset=0x0008ins v0.s[1], v1.s[1]G_M62724_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector64.ConcatUpperLower<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M2660_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M2660_IG02: ;; offset=0x0008ins v0.s[0], v0.s[1]ins v0.s[1], v1.s[0]G_M2660_IG03: ;; offset=0x0010 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 24; Vector64.ConcatUpperUpper<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M55837_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M55837_IG02: ;; offset=0x0008ins v0.s[0], v0.s[1]ins v0.s[1], v1.s[1]G_M55837_IG03: ;; offset=0x0010 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 24; Vector64.ZipLower<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M32585_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M32585_IG02: ;; offset=0x0008 zip1 v0.2s, v0.2s, v1.2sG_M32585_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector64.ZipUpper<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M43976_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M43976_IG02: ;; offset=0x0008 zip2 v0.2s, v0.2s, v1.2sG_M43976_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector64.Zip<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M14030_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M14030_IG02: ;; offset=0x0008 zip1 v16.2s, v0.2s, v1.2s zip2 v1.2s, v0.2s, v1.2smov v0.8b, v16.8bG_M14030_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 28; Vector64.UnzipEven<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M26249_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M26249_IG02: ;; offset=0x0008 uzp1 v0.2s, v0.2s, v1.2sG_M26249_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector64.UnzipOdd<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M25438_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M25438_IG02: ;; offset=0x0008 uzp2 v0.2s, v0.2s, v1.2sG_M25438_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector64.Unzip<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M41813_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M41813_IG02: ;; offset=0x0008 uzp1 v16.2s, v0.2s, v1.2s uzp2 v1.2s, v0.2s, v1.2smov v0.8b, v16.8bG_M41813_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 28; Vector64.Reverse<int>(vector); Emitting BLENDED_CODE for arm64 on WindowsG_M63937_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M63937_IG02: ;; offset=0x0008 rev64 v0.2s, v0.2sG_M63937_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector128.CreateGeometricSequence<int>(initial, 3); Emitting BLENDED_CODE for arm64 on WindowsG_M55419_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M55419_IG02: ;; offset=0x0008 dup v0.4s, w0 ldr q16,[@RWD00]mul v0.4s, v0.4s, v16.4sG_M55419_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lrRWD00 dq 0000000300000001h,0000001B00000009h; Total bytes of code 28; Vector128.CreateAlternatingSequence<int>(even, odd); Emitting BLENDED_CODE for arm64 on WindowsG_M23632_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M23632_IG02: ;; offset=0x0008 dup v0.4s, w0 dup v16.4s, w1 zip1 v0.4s, v0.4s, v16.4sG_M23632_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 28; Vector128<int>.SignSequence; Emitting BLENDED_CODE for arm64 on WindowsG_M37526_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M37526_IG02: ;; offset=0x0008 ldr q0,[@RWD00]G_M37526_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lrRWD00 dq FFFFFFFF00000001h,FFFFFFFF00000001h; Total bytes of code 20; Vector128.CreateHarmonicSequence<float>(start, step); Emitting BLENDED_CODE for arm64 on WindowsG_M24925_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M24925_IG02: ;; offset=0x0008 ldr q16,[@RWD00]fmul v16.4s, v16.4s, v1.s[0] dup v0.4s, v0.s[0]fadd v0.4s, v16.4s, v0.4s ldr q16,[@RWD16]fdiv v0.4s, v16.4s, v0.4sG_M24925_IG03: ;; offset=0x0020 ldp fp, lr,[sp], #0x10ret lrRWD00 dq	3F80000000000000h,4040000040000000hRWD16 dq	3F8000003F800000h,3F8000003F800000h; Total bytes of code 40; Vector128.ConcatLowerLower<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M47101_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M47101_IG02: ;; offset=0x0008ins v0.d[1], v1.d[0]G_M47101_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector128.ConcatLowerUpper<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M48932_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M48932_IG02: ;; offset=0x0008 ext v16.16b, v1.16b, v1.16b, #8ins v0.d[1], v16.d[0]G_M48932_IG03: ;; offset=0x0010 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 24; Vector128.ConcatUpperLower<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M24196_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M24196_IG02: ;; offset=0x0008 ext v0.16b, v0.16b, v0.16b, #8ins v0.d[1], v1.d[0]G_M24196_IG03: ;; offset=0x0010 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 24; Vector128.ConcatUpperUpper<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M45405_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M45405_IG02: ;; offset=0x0008 ext v0.16b, v0.16b, v0.16b, #8ins v1.d[0], v0.d[0]mov v0.16b, v1.16bG_M45405_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 28; Vector128.ZipLower<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M10057_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M10057_IG02: ;; offset=0x0008 zip1 v0.4s, v0.4s, v1.4sG_M10057_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector128.ZipUpper<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M48936_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M48936_IG02: ;; offset=0x0008 zip2 v0.4s, v0.4s, v1.4sG_M48936_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector128.Zip<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M32727_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M32727_IG02: ;; offset=0x0008 zip1 v16.4s, v0.4s, v1.4s zip2 v1.4s, v0.4s, v1.4smov v0.16b, v16.16bG_M32727_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 28; Vector128.UnzipEven<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M32745_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M32745_IG02: ;; offset=0x0008 uzp1 v0.4s, v0.4s, v1.4sG_M32745_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector128.UnzipOdd<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M32638_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M32638_IG02: ;; offset=0x0008 uzp2 v0.4s, v0.4s, v1.4sG_M32638_IG03: ;; offset=0x000C ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 20; Vector128.Unzip<int>(left, right); Emitting BLENDED_CODE for arm64 on WindowsG_M7116_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M7116_IG02: ;; offset=0x0008 uzp1 v16.4s, v0.4s, v1.4s uzp2 v1.4s, v0.4s, v1.4smov v0.16b, v16.16bG_M7116_IG03: ;; offset=0x0014 ldp fp, lr,[sp], #0x10ret lr; Total bytes of code 28; Vector128.Reverse<int>(vector); Emitting BLENDED_CODE for arm64 on WindowsG_M36184_IG01: ;; offset=0x0000 stp fp, lr,[sp, #-0x10]!mov fp,spG_M36184_IG02: ;; offset=0x0008 ldr q16,[@RWD00] tbl v0.16b, {v0.16b}, v16.16bG_M36184_IG03: ;; offset=0x0010 ldp fp, lr,[sp], #0x10ret lrRWD00 dq 0B0A09080F0E0D0Ch,0302010007060504h; Total bytes of code 24

Codegen for constant input:

Vector256

; Vector256.CreateGeometricSequence<int>(2, 3); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M53517_IG01: ;; offset=0x0000G_M53517_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M53517_IG03: ;; offset=0x000FvzeroupperretRWD00 dq	0000000600000002h,0000003600000012h,000001E6000000A2h,00001116000005B2h; Total bytes of code 19; Vector256.CreateAlternatingSequence<int>(5, 7); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M39644_IG01: ;; offset=0x0000G_M39644_IG02: ;; offset=0x0000vbroadcastsdymm0, qword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M39644_IG03: ;; offset=0x0010vzeroupperretRWD00 dq	0000000700000005h; Total bytes of code 20; Vector256.CreateHarmonicSequence<float>(1.0f, 2.0f); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M64276_IG01: ;; offset=0x0000G_M64276_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0movrax,rcxG_M64276_IG03: ;; offset=0x000FvzeroupperretRWD00 dq	3EAAAAAB3F800000h,3E1249253E4CCCCDh,3DBA2E8C3DE38E39h,3D8888893D9D89D9h; Total bytes of code 19

Vector512

; Vector512.CreateGeometricSequence<int>(2, 3); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M22445_IG01: ;; offset=0x0000G_M22445_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [reloc @RWD00]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M22445_IG03: ;; offset=0x0013vzeroupperretRWD00 dq	0000000600000002h,0000003600000012h,000001E6000000A2h,00001116000005B2h,000099C600003342h,000567F60001CD52h,0030A7A6001037E2h,01B5E4D60091F6F2h; Total bytes of code 23; Vector512.CreateAlternatingSequence<int>(5, 7); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M9468_IG01: ;; offset=0x0000G_M9468_IG02: ;; offset=0x0000vbroadcastsdzmm0, qword ptr [reloc @RWD00]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M9468_IG03: ;; offset=0x0013vzeroupperretRWD00 dq	0000000700000005h; Total bytes of code 23; Vector512.CreateHarmonicSequence<float>(1.0f, 2.0f); Emitting BLENDED_CODE for x64 + VEX + EVEX on WindowsG_M13716_IG01: ;; offset=0x0000G_M13716_IG02: ;; offset=0x0000vmovupszmm0, zmmword ptr [reloc @RWD00]vmovups zmmword ptr [rcx],zmm0movrax,rcxG_M13716_IG03: ;; offset=0x0013vzeroupperretRWD00 dq	3EAAAAAB3F800000h,3E1249253E4CCCCDh,3DBA2E8C3DE38E39h,3D8888893D9D89D9h,3D5794363D70F0F1h,3D3216433D430C31h,3D17B4263D23D70Ah,3D0421083D0D3DCBh; Total bytes of code 23

Vector512 without AVX512 - Vector256 decomposition path

; Vector512.CreateGeometricSequence<int>(2, 3); Emitting BLENDED_CODE for x64 + VEX on WindowsG_M22445_IG01: ;; offset=0x0000G_M22445_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [reloc @RWD00]vmovupsymm1, ymmword ptr [reloc @RWD32]vmovups ymmword ptr [rcx],ymm1vmovups ymmword ptr [rcx+0x20],ymm0movrax,rcxG_M22445_IG03: ;; offset=0x001CvzeroupperretRWD00 dq	000099C600003342h,000567F60001CD52h,0030A7A6001037E2h,01B5E4D60091F6F2hRWD32 dq	0000000600000002h,0000003600000012h,000001E6000000A2h,00001116000005B2h; Total bytes of code 32; Vector512.CreateAlternatingSequence<int>(5, 7); Emitting BLENDED_CODE for x64 + VEX on WindowsG_M9468_IG01: ;; offset=0x0000G_M9468_IG02: ;; offset=0x0000vbroadcastsdymm0, qword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0vmovups ymmword ptr [rcx+0x20],ymm0movrax,rcxG_M9468_IG03: ;; offset=0x0015vzeroupperretRWD00 dq	0000000700000005h; Total bytes of code 25; Vector512.CreateHarmonicSequence<float>(1.0f, 2.0f); Emitting BLENDED_CODE for x64 + VEX on WindowsG_M13716_IG01: ;; offset=0x0000G_M13716_IG02: ;; offset=0x0000vmovupsymm0, ymmword ptr [reloc @RWD00]vmovups ymmword ptr [rcx],ymm0vmovupsymm0, ymmword ptr [reloc @RWD32]vmovups ymmword ptr [rcx+0x20],ymm0movrax,rcxG_M13716_IG03: ;; offset=0x001CvzeroupperretRWD00 dq	3EAAAAAB3F800000h,3E1249253E4CCCCDh,3DBA2E8C3DE38E39h,3D8888893D9D89D9hRWD32 dq	3D5794363D70F0F1h,3D3216433D430C31h,3D17B4263D23D70Ah,3D0421083D0D3DCBh; Total bytes of code 32

Closes#122557

cc: @tannergooding

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Adds new vector sequence-generation helpers (geometric/alternating/harmonic/cauchy), sign-sequence helpers, and lane-manipulation operations (zip/unzip/concat/reverse) across Vector<T> and Vector{64,128,256,512}<T>, including JIT recognition and test coverage.

Changes:

• Introduces new public APIs in the ref assemblies for sequence creation + lane operations and SignSequence.
• Implements the APIs in CoreLib for Vector<T> and Vector{64,128,256,512}<T>, with some JIT fast-paths.
• Adds unit tests validating the new behaviors across vector widths.

Reviewed changes

Copilot reviewed 23 out of 23 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
src/libraries/System.Runtime.Intrinsics/tests/Vectors/Vector64Tests.cs	Adds tests for new Vector64 sequence + lane APIs
src/libraries/System.Runtime.Intrinsics/tests/Vectors/Vector128Tests.cs	Adds tests for new Vector128 sequence + lane APIs
src/libraries/System.Runtime.Intrinsics/tests/Vectors/Vector256Tests.cs	Adds tests for new Vector256 sequence + lane APIs
src/libraries/System.Runtime.Intrinsics/tests/Vectors/Vector512Tests.cs	Adds tests for new Vector512 sequence + lane APIs
src/libraries/System.Runtime.Intrinsics/ref/System.Runtime.Intrinsics.cs	Exposes new Vector{64,128,256,512} APIs in the reference contract
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector64_1.cs	Adds Vector64<T>.SignSequence
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector64.cs	Implements Vector64 sequence + lane APIs
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector128_1.cs	Adds Vector128<T>.SignSequence
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector128.cs	Implements Vector128 sequence + lane APIs
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector256_1.cs	Adds Vector256<T>.SignSequence
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector256.cs	Implements Vector256 sequence + lane APIs
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector512_1.cs	Adds Vector512<T>.SignSequence
src/libraries/System.Private.CoreLib/src/System/Runtime/Intrinsics/Vector512.cs	Implements Vector512 sequence + lane APIs + AVX-512 special-cases
src/libraries/System.Private.CoreLib/src/System/Numerics/Vector_1.cs	Adds Vector<T>.SignSequence
src/libraries/System.Private.CoreLib/src/System/Numerics/Vector.cs	Implements Vector sequence + lane APIs
src/libraries/System.Numerics.Vectors/tests/GenericVectorTests.cs	Adds tests for new System.Numerics.Vector APIs
src/libraries/System.Numerics.Vectors/ref/System.Numerics.Vectors.cs	Exposes new System.Numerics.Vector APIs in the reference contract
src/coreclr/jit/hwintrinsicxarch.cpp	Adds xarch JIT special-import support for new intrinsics
src/coreclr/jit/hwintrinsiclistxarch.h	Registers new xarch HW intrinsic IDs
src/coreclr/jit/hwintrinsicarm64.cpp	Adds arm64 JIT special-import support for new intrinsics
src/coreclr/jit/hwintrinsiclistarm64.h	Registers new arm64 HW intrinsic IDs
src/coreclr/jit/compiler.h	Declares new SIMD IR node builders used by importer/lowering

[Copilot]

Pull request overview

Copilot reviewed 23 out of 23 changed files in this pull request and generated 8 comments.

Comments suppressed due to low confidence (4)

src/coreclr/jit/gentree.cpp:1

• Several simdCount==1 special-cases use gtWrapWithSideEffects in a way that can reverse the required left-to-right argument evaluation order (op1 then op2) for intrinsics like Concat/Zip/Unzip, and for CreateAlternatingSequence. This is observable for Vector64/Vector64 (and similar) where the vector length is 1 and arguments may have side-effects. Consider materializing op1 into a temp before sequencing op2, or otherwise constructing the tree so op1 is evaluated before op2 while still returning op1 (or the correct constant for UnzipOdd).
  src/coreclr/jit/gentree.cpp:1
• Several simdCount==1 special-cases use gtWrapWithSideEffects in a way that can reverse the required left-to-right argument evaluation order (op1 then op2) for intrinsics like Concat/Zip/Unzip, and for CreateAlternatingSequence. This is observable for Vector64/Vector64 (and similar) where the vector length is 1 and arguments may have side-effects. Consider materializing op1 into a temp before sequencing op2, or otherwise constructing the tree so op1 is evaluated before op2 while still returning op1 (or the correct constant for UnzipOdd).
  src/coreclr/jit/gentree.cpp:1
• Several simdCount==1 special-cases use gtWrapWithSideEffects in a way that can reverse the required left-to-right argument evaluation order (op1 then op2) for intrinsics like Concat/Zip/Unzip, and for CreateAlternatingSequence. This is observable for Vector64/Vector64 (and similar) where the vector length is 1 and arguments may have side-effects. Consider materializing op1 into a temp before sequencing op2, or otherwise constructing the tree so op1 is evaluated before op2 while still returning op1 (or the correct constant for UnzipOdd).
  src/coreclr/jit/gentree.cpp:1
• Several simdCount==1 special-cases use gtWrapWithSideEffects in a way that can reverse the required left-to-right argument evaluation order (op1 then op2) for intrinsics like Concat/Zip/Unzip, and for CreateAlternatingSequence. This is observable for Vector64/Vector64 (and similar) where the vector length is 1 and arguments may have side-effects. Consider materializing op1 into a temp before sequencing op2, or otherwise constructing the tree so op1 is evaluated before op2 while still returning op1 (or the correct constant for UnzipOdd).

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-runtime-intrinsics
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

Copilot reviewed 23 out of 23 changed files in this pull request and generated 4 comments.

[Copilot]

Pull request overview

Copilot reviewed 21 out of 21 changed files in this pull request and generated 3 comments.

[hez2010]

Ping.

[hez2010]

It seems that win-arm64 CI is lagging.

[tannergooding]

/ba-g unrelated timeouts

[MichalStrehovsky]

This change broke main:

/__w/1/s/src/coreclr/jit/gentree.cpp:27454:2: error: Unsupported platform
27454 | #error Unsupported platform
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27492:2: error: Unsupported platform
27492 | #error Unsupported platform
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27495:41: error: use of undeclared identifier 'result'; did you mean 'FindLinkData::result'?
27495 | return gtNewSimdWithUpperNode(type, result, upper, simdBaseType, simdSize);
| ^~~~~~
| FindLinkData::result
/__w/1/s/src/coreclr/jit/compiler.h:4154:19: note: 'FindLinkData::result' declared here
4154 | GenTree** result;
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27495:41: error: invalid use of non-static data member 'result'
27495 | return gtNewSimdWithUpperNode(type, result, upper, simdBaseType, simdSize);
| ^~~~~~
/__w/1/s/src/coreclr/jit/gentree.cpp:27596:2: error: Unsupported platform
27596 | #error Unsupported platform
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27786:2: error: Unsupported platform
27786 | #error Unsupported platform
| ^

@dotnet/jit-contrib unless you have a quick fix, I staged a revert at #129610. I think we can merge without waiting for CI.

[hez2010]

This change broke main:

/__w/1/s/src/coreclr/jit/gentree.cpp:27454:2: error: Unsupported platform
27454 | #error Unsupported platform
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27492:2: error: Unsupported platform
27492 | #error Unsupported platform
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27495:41: error: use of undeclared identifier 'result'; did you mean 'FindLinkData::result'?
27495 | return gtNewSimdWithUpperNode(type, result, upper, simdBaseType, simdSize);
| ^~~~~~
| FindLinkData::result
/__w/1/s/src/coreclr/jit/compiler.h:4154:19: note: 'FindLinkData::result' declared here
4154 | GenTree** result;
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27495:41: error: invalid use of non-static data member 'result'
27495 | return gtNewSimdWithUpperNode(type, result, upper, simdBaseType, simdSize);
| ^~~~~~
/__w/1/s/src/coreclr/jit/gentree.cpp:27596:2: error: Unsupported platform
27596 | #error Unsupported platform
| ^
/__w/1/s/src/coreclr/jit/gentree.cpp:27786:2: error: Unsupported platform
27786 | #error Unsupported platform
| ^

@dotnet/jit-contrib unless you have a quick fix, I staged a revert at #129610. I think we can merge without waiting for CI.

Wondering why this build failure was not captured by the CI.

[pavelsavara]

This is untested fix attempt https://github.com/dotnet/runtime/compare/main...pavelsavara:wasm_fix_jit?expand=1

[MichalStrehovsky]

Reverted. This broke two pull requests and was hit locally in #129607. Not going to wait.

[MichalStrehovsky]

Wondering why this build failure was not captured by the CI.

The results were 3 days stale. The older the results, the more likely this conflicts with something else.

[hez2010]

It seems that recently landed wasm jit simd enablement change breaks the build of this PR. Now we need to #ifdef out the wasm path.

@tannergooding

[tannergooding]

I had missed that it was 3 days stale and that another PR had gone in touching the area (which is rarely touched by anyone else).

@hez2010 could you get up a revert of the revert + one additional commit including the relevant fix so WASM doesn't fail?

Having the paths do NYI_WASM_SIMD is correct for now, as impSpecialIntrinsic under hwintrinsicwasm.cpp is doing the same.