// Copyright 2021 The Wuffs Authors. // // Licensed under the Apache License, Version 2.0 or the MIT license // , at your // option. This file may not be copied, modified, or distributed // except according to those terms. // // SPDX-License-Identifier: Apache-2.0 OR MIT pri func ieee_hasher.up_x86_sse42!(x: roslice base.u8), choose cpu_arch >= x86_sse42, { var s : base.u32 var util : base.x86_sse42_utility var kk : base.x86_m128i var x0 : base.x86_m128i var x1 : base.x86_m128i var x2 : base.x86_m128i var x3 : base.x86_m128i var x4 : base.x86_m128i var x5 : base.x86_m128i var x6 : base.x86_m128i var x7 : base.x86_m128i var y0 : base.x86_m128i var y1 : base.x86_m128i var y2 : base.x86_m128i var y3 : base.x86_m128i var y4 : base.x86_m128i var y5 : base.x86_m128i var y6 : base.x86_m128i var y7 : base.x86_m128i s = 0xFFFF_FFFF ^ this.state // Align to a 16-byte boundary. while (args.x.length() > 0) and ((15 & args.x.uintptr_low_12_bits()) <> 0) { s = IEEE_TABLE[0][((s & 0xFF) as base.u8) ^ args.x[0]] ^ (s >> 8) args.x = args.x[1 ..] } if args.x.length() >= 128 { // BEGIN script/print-crc32-x86-sse42-code.go generated code. x0 = util.make_m128i_slice128(a: args.x[0x00 .. 0x10]) x1 = util.make_m128i_slice128(a: args.x[0x10 .. 0x20]) x2 = util.make_m128i_slice128(a: args.x[0x20 .. 0x30]) x3 = util.make_m128i_slice128(a: args.x[0x30 .. 0x40]) x4 = util.make_m128i_slice128(a: args.x[0x40 .. 0x50]) x5 = util.make_m128i_slice128(a: args.x[0x50 .. 0x60]) x6 = util.make_m128i_slice128(a: args.x[0x60 .. 0x70]) x7 = util.make_m128i_slice128(a: args.x[0x70 .. 0x80]) kk = util.make_m128i_multiple_u32(a00: 0x33FF_F533, a01: 0, a02: 0x910E_EEC1, a03: 0) x0 = x0._mm_xor_si128(b: util.make_m128i_single_u32(a: s)) args.x = args.x[128 ..] while args.x.length() >= 128 { y0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x00) x0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x11) y1 = x1._mm_clmulepi64_si128(b: kk, imm8: 0x00) x1 = x1._mm_clmulepi64_si128(b: kk, imm8: 0x11) y2 = x2._mm_clmulepi64_si128(b: kk, imm8: 0x00) x2 = x2._mm_clmulepi64_si128(b: kk, imm8: 0x11) y3 = x3._mm_clmulepi64_si128(b: kk, imm8: 0x00) x3 = x3._mm_clmulepi64_si128(b: kk, imm8: 0x11) y4 = x4._mm_clmulepi64_si128(b: kk, imm8: 0x00) x4 = x4._mm_clmulepi64_si128(b: kk, imm8: 0x11) y5 = x5._mm_clmulepi64_si128(b: kk, imm8: 0x00) x5 = x5._mm_clmulepi64_si128(b: kk, imm8: 0x11) y6 = x6._mm_clmulepi64_si128(b: kk, imm8: 0x00) x6 = x6._mm_clmulepi64_si128(b: kk, imm8: 0x11) y7 = x7._mm_clmulepi64_si128(b: kk, imm8: 0x00) x7 = x7._mm_clmulepi64_si128(b: kk, imm8: 0x11) y0 = y0._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x00 .. 0x10])) x0 = x0._mm_xor_si128(b: y0) y1 = y1._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x10 .. 0x20])) x1 = x1._mm_xor_si128(b: y1) y2 = y2._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x20 .. 0x30])) x2 = x2._mm_xor_si128(b: y2) y3 = y3._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x30 .. 0x40])) x3 = x3._mm_xor_si128(b: y3) y4 = y4._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x40 .. 0x50])) x4 = x4._mm_xor_si128(b: y4) y5 = y5._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x50 .. 0x60])) x5 = x5._mm_xor_si128(b: y5) y6 = y6._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x60 .. 0x70])) x6 = x6._mm_xor_si128(b: y6) y7 = y7._mm_xor_si128(b: util.make_m128i_slice128(a: args.x[0x70 .. 0x80])) x7 = x7._mm_xor_si128(b: y7) args.x = args.x[128 ..] } kk = util.make_m128i_multiple_u32(a00: 0xAE68_9191, a01: 0, a02: 0xCCAA_009E, a03: 0) y0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x00) x0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x11) y2 = x2._mm_clmulepi64_si128(b: kk, imm8: 0x00) x2 = x2._mm_clmulepi64_si128(b: kk, imm8: 0x11) y4 = x4._mm_clmulepi64_si128(b: kk, imm8: 0x00) x4 = x4._mm_clmulepi64_si128(b: kk, imm8: 0x11) y6 = x6._mm_clmulepi64_si128(b: kk, imm8: 0x00) x6 = x6._mm_clmulepi64_si128(b: kk, imm8: 0x11) y0 = y0._mm_xor_si128(b: x1) x0 = x0._mm_xor_si128(b: y0) y2 = y2._mm_xor_si128(b: x3) x2 = x2._mm_xor_si128(b: y2) y4 = y4._mm_xor_si128(b: x5) x4 = x4._mm_xor_si128(b: y4) y6 = y6._mm_xor_si128(b: x7) x6 = x6._mm_xor_si128(b: y6) kk = util.make_m128i_multiple_u32(a00: 0xF1DA_05AA, a01: 0, a02: 0x8125_6527, a03: 0) y0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x00) x0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x11) y4 = x4._mm_clmulepi64_si128(b: kk, imm8: 0x00) x4 = x4._mm_clmulepi64_si128(b: kk, imm8: 0x11) y0 = y0._mm_xor_si128(b: x2) x0 = x0._mm_xor_si128(b: y0) y4 = y4._mm_xor_si128(b: x6) x4 = x4._mm_xor_si128(b: y4) kk = util.make_m128i_multiple_u32(a00: 0x8F35_2D95, a01: 0, a02: 0x1D95_13D7, a03: 0) y0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x00) x0 = x0._mm_clmulepi64_si128(b: kk, imm8: 0x11) y0 = y0._mm_xor_si128(b: x4) x0 = x0._mm_xor_si128(b: y0) kk = util.make_m128i_multiple_u32(a00: 0xF701_1641, a01: 0xB4E5_B025, a02: 0xDB71_0641, a03: 1) s = util.make_m128i_single_u64(a: x0._mm_extract_epi64(imm8: 0)). _mm_clmulepi64_si128(b: kk, imm8: 0x00). _mm_clmulepi64_si128(b: kk, imm8: 0x10). _mm_extract_epi32(imm8: 2) kk = util.make_m128i_multiple_u32(a00: 0xF701_1641, a01: 0xB4E5_B025, a02: 0xDB71_0641, a03: 1) s = util.make_m128i_single_u64(a: x0._mm_extract_epi64(imm8: 1) ^ (s as base.u64)). _mm_clmulepi64_si128(b: kk, imm8: 0x00). _mm_clmulepi64_si128(b: kk, imm8: 0x10). _mm_extract_epi32(imm8: 2) // END script/print-crc32-x86-sse42-code.go generated code. } while args.x.length() >= 8 { kk = util.make_m128i_multiple_u32(a00: 0xF701_1641, a01: 0xB4E5_B025, a02: 0xDB71_0641, a03: 1) s = util.make_m128i_single_u64(a: args.x.peek_u64le() ^ (s as base.u64)). _mm_clmulepi64_si128(b: kk, imm8: 0x00). _mm_clmulepi64_si128(b: kk, imm8: 0x10). _mm_extract_epi32(imm8: 2) args.x = args.x[8 ..] } while args.x.length() > 0 { s = IEEE_TABLE[0][((s & 0xFF) as base.u8) ^ args.x[0]] ^ (s >> 8) args.x = args.x[1 ..] } this.state = 0xFFFF_FFFF ^ s }