[v2,4/9] Hexagon (target/hexagon) Add v68 HVX instructions

Message ID	20230427224057.3766963-5-tsimpson@quicinc.com (mailing list archive)
State	New, archived
Headers	show Return-Path: <qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org> From: Taylor Simpson <tsimpson@quicinc.com> To: qemu-devel@nongnu.org Cc: tsimpson@quicinc.com, richard.henderson@linaro.org, philmd@linaro.org, ale@rev.ng, anjo@rev.ng, bcain@quicinc.com, quic_mathbern@quicinc.com Subject: [PATCH v2 4/9] Hexagon (target/hexagon) Add v68 HVX instructions Date: Thu, 27 Apr 2023 15:40:52 -0700 Message-Id: <20230427224057.3766963-5-tsimpson@quicinc.com> In-Reply-To: <20230427224057.3766963-1-tsimpson@quicinc.com> References: <20230427224057.3766963-1-tsimpson@quicinc.com> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: base64 Received-SPF: pass client-ip=205.220.168.131; envelope-from=tsimpson@qualcomm.com; helo=mx0a-0031df01.pphosted.com X-Spam_score_int: -24 X-Spam_score: -2.5 X-Spam_bar: -- X-Spam_report: (-2.5 / 5.0 requ) BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, HEADER_FROM_DIFFERENT_DOMAINS=0.25, RCVD_IN_DNSWL_LOW=-0.7, SPF_HELO_NONE=0.001, SPF_PASS=-0.001, T_SCC_BODY_TEXT_LINE=-0.01 autolearn=ham autolearn_force=no X-Spam_action: no action Precedence: list Errors-To: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org Sender: qemu-devel-bounces+qemu-devel=archiver.kernel.org@nongnu.org
Series	Hexagon (target/hexagon) New architecture support \| expand [v2,0/9] Hexagon (target/hexagon) New architecture support [v2,1/9] Hexagon (target/hexagon) Add support for v68/v69/v71/v73 [v2,2/9] Hexagon (target/hexagon) Add v68 scalar instructions [v2,3/9] Hexagon (tests/tcg/hexagon) Add v68 scalar tests [v2,4/9] Hexagon (target/hexagon) Add v68 HVX instructions [v2,5/9] Hexagon (tests/tcg/hexagon) Add v68 HVX tests [v2,6/9] Hexagon (target/hexagon) Add v69 HVX instructions [v2,7/9] Hexagon (tests/tcg/hexagon) Add v69 HVX tests [v2,8/9] Hexagon (target/hexagon) Add v73 scalar instructions [v2,9/9] Hexagon (tests/tcg/hexagon) Add v73 scalar tests

diff --git a/target/hexagon/mmvec/macros.h b/target/hexagon/mmvec/macros.h index 1201d778d0..a655634fd1 100644 --- a/target/hexagon/mmvec/macros.h +++ b/target/hexagon/mmvec/macros.h @@ -1,5 +1,5 @@ /* - * Copyright(c) 2019-2022 Qualcomm Innovation Center, Inc. All Rights Reserved. + * Copyright(c) 2019-2023 Qualcomm Innovation Center, Inc. All Rights Reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -346,4 +346,11 @@ #define fUARCH_NOTE_PUMP_2X() #define IV1DEAD() + +#define fGET10BIT(COE, VAL, POS) \ + do { \ + COE = (sextract32(VAL, 24 + 2 * POS, 2) << 8) | \ + extract32(VAL, POS * 8, 8); \ + } while (0); + #endif diff --git a/target/hexagon/imported/mmvec/encode_ext.def b/target/hexagon/imported/mmvec/encode_ext.def index 6fbbe2c422..b9b62fef8d 100644 --- a/target/hexagon/imported/mmvec/encode_ext.def +++ b/target/hexagon/imported/mmvec/encode_ext.def @@ -1,5 +1,5 @@ /* - * Copyright(c) 2019-2021 Qualcomm Innovation Center, Inc. All Rights Reserved. + * Copyright(c) 2019-2023 Qualcomm Innovation Center, Inc. All Rights Reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -730,6 +730,8 @@ DEF_ENC(V6_vmaxb, ICLASS_CJ" 1 111 001 vvvvv PP 0 uuuuu 101 ddddd") // DEF_ENC(V6_vsatuwuh, ICLASS_CJ" 1 111 001 vvvvv PP 0 uuuuu 110 ddddd") // DEF_ENC(V6_vdealb4w, ICLASS_CJ" 1 111 001 vvvvv PP 0 uuuuu 111 ddddd") // +DEF_ENC(V6_v6mpyvubs10_vxx, ICLASS_CJ" 1 111 001 vvvvv PP 1 uuuuu 0ii xxxxx") +DEF_ENC(V6_v6mpyhubs10_vxx, ICLASS_CJ" 1 111 001 vvvvv PP 1 uuuuu 1ii xxxxx") DEF_ENC(V6_vmpyowh_rnd, ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 000 ddddd") // DEF_ENC(V6_vshuffeb, ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 001 ddddd") // @@ -740,6 +742,10 @@ DEF_ENC(V6_vshufoeh, ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 101 ddddd") // DEF_ENC(V6_vshufoeb, ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 110 ddddd") // DEF_ENC(V6_vcombine, ICLASS_CJ" 1 111 010 vvvvv PP 0 uuuuu 111 ddddd") // +DEF_ENC(V6_v6mpyvubs10, ICLASS_CJ" 1 111 010 vvvvv PP 1 uuuuu 0ii ddddd") +DEF_ENC(V6_v6mpyhubs10, ICLASS_CJ" 1 111 010 vvvvv PP 1 uuuuu 1ii ddddd") + + DEF_ENC(V6_vmpyieoh, ICLASS_CJ" 1 111 011 vvvvv PP 0 uuuuu 000 ddddd") // DEF_ENC(V6_vadduwsat, ICLASS_CJ" 1 111 011 vvvvv PP 0 uuuuu 001 ddddd") // DEF_ENC(V6_vsathub, ICLASS_CJ" 1 111 011 vvvvv PP 0 uuuuu 010 ddddd") // diff --git a/target/hexagon/imported/mmvec/ext.idef b/target/hexagon/imported/mmvec/ext.idef index 8ca5a606e1..c0d169fd4f 100644 --- a/target/hexagon/imported/mmvec/ext.idef +++ b/target/hexagon/imported/mmvec/ext.idef @@ -1,5 +1,5 @@ /* - * Copyright(c) 2019-2021 Qualcomm Innovation Center, Inc. All Rights Reserved. + * Copyright(c) 2019-2023 Qualcomm Innovation Center, Inc. All Rights Reserved. * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by @@ -116,6 +116,10 @@ ITERATOR_INSN_MPY_SLOT_LATE(WIDTH,TAG, SYNTAX2,DESCR,CODE) EXTINSN(V6_##TAG, SYNTAX, ATTRIBS(A_EXTENSION,A_CVI,A_CVI_VX_DV), \ DESCR, DO_FOR_EACH_CODE(WIDTH, CODE)) +#define ITERATOR_INSN_MPY_SLOT_DOUBLE_VEC_VX_FWD(WIDTH,TAG,SYNTAX,DESCR,CODE) \ +EXTINSN(V6_##TAG, SYNTAX, ATTRIBS(A_EXTENSION,A_CVI,A_CVI_VX_DV), \ +DESCR, DO_FOR_EACH_CODE(WIDTH, CODE)) + #define ITERATOR_INSN2_MPY_SLOT_DOUBLE_VEC(WIDTH,TAG,SYNTAX,SYNTAX2,DESCR,CODE) \ ITERATOR_INSN_MPY_SLOT_DOUBLE_VEC(WIDTH,TAG,SYNTAX2,DESCR,CODE) @@ -2507,6 +2511,281 @@ EXTINSN(V6_vscattermhw , "vscatter(Rt32,Mu2,Vvv32.w).h=Vw32", ATTRIBS(A_EXTENSIO }) +ITERATOR_INSN_MPY_SLOT_DOUBLE_VEC_VX_FWD(32, v6mpyvubs10_vxx, "Vxx32.w+=v6mpy(Vuu32.ub,Vvv32.b,#u2):v", "", + fHIDE(size2s_t c00;) + fGET10BIT(c00, VvvV.v[0].uw[i], 0) + fHIDE(size2s_t c01;) + fGET10BIT(c01, VvvV.v[0].uw[i], 1) + fHIDE(size2s_t c02;) + fGET10BIT(c02, VvvV.v[0].uw[i], 2) + + fHIDE(size2s_t c10;) + fGET10BIT(c10, VvvV.v[1].uw[i], 0) + fHIDE(size2s_t c11;) + fGET10BIT(c11, VvvV.v[1].uw[i], 1) + fHIDE(size2s_t c12;) + fGET10BIT(c12, VvvV.v[1].uw[i], 2) + + if (uiV == 0) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c10); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c11); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c12); + + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c12); + + } else if (uiV == 1) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c12); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c00); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c02); + + } else if (uiV == 2) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c10); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c12); + + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c12); + + } else if (uiV == 3) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c12); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c00); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c01); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c02); + } +) +ITERATOR_INSN_MPY_SLOT_DOUBLE_VEC_VX_FWD(32, v6mpyhubs10_vxx, "Vxx32.w+=v6mpy(Vuu32.ub,Vvv32.b,#u2):h", "", + fHIDE(size2s_t c00;) + fGET10BIT(c00, VvvV.v[0].uw[i], 0) + fHIDE(size2s_t c01;) + fGET10BIT(c01, VvvV.v[0].uw[i], 1) + fHIDE(size2s_t c02;) + fGET10BIT(c02, VvvV.v[0].uw[i], 2) + fHIDE(size2s_t c10;) + fGET10BIT(c10, VvvV.v[1].uw[i], 0) + fHIDE(size2s_t c11;) + fGET10BIT(c11, VvvV.v[1].uw[i], 1) + fHIDE(size2s_t c12;) + fGET10BIT(c12, VvvV.v[1].uw[i], 2) + + if (uiV == 0) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c10); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c11); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c12); + + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c12); + + } else if (uiV == 1) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c12); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c00); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c02); + + } else if (uiV == 2) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c10); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c12); + + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c12); + + } else if (uiV == 3) { + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c00); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c01); + VxxV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c02); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c10); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c12); + + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c00); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c01); + VxxV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c02); + } +) + + +ITERATOR_INSN_MPY_SLOT_DOUBLE_VEC(32, v6mpyvubs10, "Vdd32.w=v6mpy(Vuu32.ub,Vvv32.b,#u2):v", "", + fHIDE(short c00;) + fGET10BIT(c00, VvvV.v[0].uw[i], 0) + fHIDE(short c01;) + fGET10BIT(c01, VvvV.v[0].uw[i], 1) + fHIDE(short c02;) + fGET10BIT(c02, VvvV.v[0].uw[i], 2) + fHIDE(short c10;) + fGET10BIT(c10, VvvV.v[1].uw[i], 0) + fHIDE(short c11;) + fGET10BIT(c11, VvvV.v[1].uw[i], 1) + fHIDE(short c12;) + fGET10BIT(c12, VvvV.v[1].uw[i], 2) + + + + if (uiV == 0) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c10); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c11); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c12); + + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c12); + + } else if (uiV == 1) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c12); + + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c00); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c02); + + } else if (uiV == 2) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c10); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c12); + + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c12); + + } else if (uiV == 3) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c12); + + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c00); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c01); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c02); + } +) + +ITERATOR_INSN_MPY_SLOT_DOUBLE_VEC(32, v6mpyhubs10, "Vdd32.w=v6mpy(Vuu32.ub,Vvv32.b,#u2):h", "", + fHIDE(short c00;) + fGET10BIT(c00, VvvV.v[0].uw[i], 0) + fHIDE(short c01;) + fGET10BIT(c01, VvvV.v[0].uw[i], 1) + fHIDE(short c02;) + fGET10BIT(c02, VvvV.v[0].uw[i], 2) + fHIDE(short c10;) + fGET10BIT(c10, VvvV.v[1].uw[i], 0) + fHIDE(short c11;) + fGET10BIT(c11, VvvV.v[1].uw[i], 1) + fHIDE(short c12;) + fGET10BIT(c12, VvvV.v[1].uw[i], 2) + + if (uiV == 0) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c10); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c11); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c12); + + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c12); + + } else if (uiV == 1) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(3,VuuV.v[1].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c12); + + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[1].uw[i]), c00); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c01); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c02); + + } else if (uiV == 2) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c10); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c12); + + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c12); + + } else if (uiV == 3) { + VddV.v[1].w[i] = fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c00); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c01); + VddV.v[1].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c02); + + VddV.v[0].w[i] = fMPY16US(fGETUBYTE(1,VuuV.v[1].uw[i]), c10); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(3,VuuV.v[0].uw[i]), c11); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(1,VuuV.v[0].uw[i]), c12); + + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[1].uw[i]), c00); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(2,VuuV.v[0].uw[i]), c01); + VddV.v[0].w[i] += fMPY16US(fGETUBYTE(0,VuuV.v[0].uw[i]), c02); + } +) + EXTINSN(V6_vscattermhwq, "if (Qs4) vscatter(Rt32,Mu2,Vvv32.w).h=Vw32", ATTRIBS(A_EXTENSION,A_CVI,A_CVI_SCATTER,A_CVI_VA_DV,A_CVI_VM,A_MEMLIKE), "Scatter halfwords conditional", {

[v2,4/9] Hexagon (target/hexagon) Add v68 HVX instructions

Commit Message

Patch