[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PATCH v2 4/9] Hexagon (target/hexagon) Add v68 HVX instructions
|
From: |
Taylor Simpson |
|
Subject: |
[PATCH v2 4/9] Hexagon (target/hexagon) Add v68 HVX instructions |
|
Date: |
Thu, 27 Apr 2023 15:40:52 -0700 |
The following instructions are added
V6_v6mpyvubs10_vxx
V6_v6mpyhubs10_vxx
V6_v6mpyvubs10
V6_v6mpyhubs10
Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
Reviewed-by: Anton Johansson <anjo@rev.ng>
---
target/hexagon/mmvec/macros.h | 9 +-
target/hexagon/imported/mmvec/encode_ext.def | 8 +-
target/hexagon/imported/mmvec/ext.idef | 281 ++++++++++++++++++-
3 files changed, 295 insertions(+), 3 deletions(-)
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",
{
--
2.25.1
- [PATCH v2 0/9] Hexagon (target/hexagon) New architecture support, Taylor Simpson, 2023/04/27
- [PATCH v2 2/9] Hexagon (target/hexagon) Add v68 scalar instructions, Taylor Simpson, 2023/04/27
- [PATCH v2 8/9] Hexagon (target/hexagon) Add v73 scalar instructions, Taylor Simpson, 2023/04/27
- [PATCH v2 4/9] Hexagon (target/hexagon) Add v68 HVX instructions,
Taylor Simpson <=
- [PATCH v2 3/9] Hexagon (tests/tcg/hexagon) Add v68 scalar tests, Taylor Simpson, 2023/04/27
- [PATCH v2 6/9] Hexagon (target/hexagon) Add v69 HVX instructions, Taylor Simpson, 2023/04/27
- [PATCH v2 5/9] Hexagon (tests/tcg/hexagon) Add v68 HVX tests, Taylor Simpson, 2023/04/27
- [PATCH v2 9/9] Hexagon (tests/tcg/hexagon) Add v73 scalar tests, Taylor Simpson, 2023/04/27
- [PATCH v2 7/9] Hexagon (tests/tcg/hexagon) Add v69 HVX tests, Taylor Simpson, 2023/04/27
- [PATCH v2 1/9] Hexagon (target/hexagon) Add support for v68/v69/v71/v73, Taylor Simpson, 2023/04/27
- Re: [PATCH v2 0/9] Hexagon (target/hexagon) New architecture support, Richard Henderson, 2023/04/28