aboutsummaryrefslogtreecommitdiff
path: root/src
diff options
context:
space:
mode:
authorisharipo <iskander.sharipov@intel.com>2017-11-15 20:56:10 +0300
committerIlya Tocar <ilya.tocar@intel.com>2017-11-17 21:17:58 +0000
commit4e19cfcdf2ff316cc14357516c01f30722f19000 (patch)
tree6d4fd7d0a67732b1f1a02a9b1b2d3d1b6575ea0d /src
parent3a395e22832774ff56d6b82cb8d91bc31167bd8e (diff)
downloadgo-4e19cfcdf2ff316cc14357516c01f30722f19000.tar.xz
cmd/internal/obj/x86: add AVX2 gather and VSIB
Enables AVX2 gather instructions and VSIB support, which makes vm32{x,y} vm64{x,y} operands encodable. AXXX constants placed with respect to sorting order. New VEX optabs inserted near non-VEX entries to simplify potential transition to auto-generated VSIB optabs. Tests go into new AMD64 encoder test file (amd64enc_extra.s) to avoid unnecessary interactions with auto-generated "amd64enc.s". Side note: x86avxgen did not produced these instructions because x86.v0.2.csv misses them. This also explains why x86 test suite have no AVX2 gather instructions tests. List of new instructions: VGATHERPDP VGATHERDPS VGATHERQPD VGATHERQPS VPGATHERDD VPGATHERDQ VPGATHERQD VPGATHERQQ Change-Id: Iac852f3c5016523670bd99de6bec6a48f66fb4f6 Reviewed-on: https://go-review.googlesource.com/77970 Run-TryBot: Iskander Sharipov <iskander.sharipov@intel.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Ilya Tocar <ilya.tocar@intel.com>
Diffstat (limited to 'src')
-rw-r--r--src/cmd/asm/internal/asm/endtoend_test.go1
-rw-r--r--src/cmd/asm/internal/asm/testdata/amd64enc_extra.s237
-rw-r--r--src/cmd/asm/internal/asm/testdata/amd64error.s25
-rw-r--r--src/cmd/internal/obj/x86/aenum.go8
-rw-r--r--src/cmd/internal/obj/x86/anames.go8
-rw-r--r--src/cmd/internal/obj/x86/asm6.go146
6 files changed, 423 insertions, 2 deletions
diff --git a/src/cmd/asm/internal/asm/endtoend_test.go b/src/cmd/asm/internal/asm/endtoend_test.go
index e5bc34edec..092b237efb 100644
--- a/src/cmd/asm/internal/asm/endtoend_test.go
+++ b/src/cmd/asm/internal/asm/endtoend_test.go
@@ -391,6 +391,7 @@ func TestAMD64EndToEnd(t *testing.T) {
func TestAMD64Encoder(t *testing.T) {
testEndToEnd(t, "amd64", "amd64enc")
+ testEndToEnd(t, "amd64", "amd64enc_extra")
}
func TestAMD64Errors(t *testing.T) {
diff --git a/src/cmd/asm/internal/asm/testdata/amd64enc_extra.s b/src/cmd/asm/internal/asm/testdata/amd64enc_extra.s
new file mode 100644
index 0000000000..6b4d7c7356
--- /dev/null
+++ b/src/cmd/asm/internal/asm/testdata/amd64enc_extra.s
@@ -0,0 +1,237 @@
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// This input extends auto-generated amd64enc.s test suite
+// with manually added tests.
+
+#include "../../../../../runtime/textflag.h"
+
+TEXT asmtest(SB),DUPOK|NOSPLIT,$0
+ // AVX2GATHER: basic combinations.
+ VPGATHERDQ Y2, (BP)(X7*2), Y1 // c4e2ed904c7d00
+ VPGATHERDQ X12, (R13)(X14*2), X11 // c40299905c7500
+ VPGATHERDQ Y12, (R13)(X14*2), Y11 // c4029d905c7500
+ VPGATHERDQ Y0, 8(X4*1), Y6 // c4e2fd90342508000000
+ VPGATHERDQ Y0, -8(X4*1), Y6 // c4e2fd903425f8ffffff
+ VPGATHERDQ Y0, 0(X4*1), Y6 // c4e2fd90342500000000
+ VPGATHERDQ Y0, 664(X4*1), Y6 // c4e2fd90342598020000
+ VPGATHERDQ Y0, 8(X4*8), Y6 // c4e2fd9034e508000000
+ VPGATHERDQ Y0, -8(X4*8), Y6 // c4e2fd9034e5f8ffffff
+ VPGATHERDQ Y0, 0(X4*8), Y6 // c4e2fd9034e500000000
+ VPGATHERDQ Y0, 664(X4*8), Y6 // c4e2fd9034e598020000
+ VPGATHERDQ Y0, 8(X14*1), Y6 // c4a2fd90343508000000
+ VPGATHERDQ Y0, -8(X14*1), Y6 // c4a2fd903435f8ffffff
+ VPGATHERDQ Y0, 0(X14*1), Y6 // c4a2fd90343500000000
+ VPGATHERDQ Y0, 664(X14*1), Y6 // c4a2fd90343598020000
+ VPGATHERDQ Y0, 8(X14*8), Y6 // c4a2fd9034f508000000
+ VPGATHERDQ Y0, -8(X14*8), Y6 // c4a2fd9034f5f8ffffff
+ VPGATHERDQ Y0, 0(X14*8), Y6 // c4a2fd9034f500000000
+ VPGATHERDQ Y0, 664(X14*8), Y6 // c4a2fd9034f598020000
+ VPGATHERDQ X2, (BP)(X7*2), X1 // c4e2e9904c7d00
+ VPGATHERDQ Y2, (BP)(X7*2), Y1 // c4e2ed904c7d00
+ VPGATHERDQ X12, (R13)(X14*2), X11 // c40299905c7500
+ VPGATHERDQ Y12, (R13)(X14*2), Y11 // c4029d905c7500
+ VPGATHERDQ Y0, 8(X4*1), Y6 // c4e2fd90342508000000
+ VPGATHERDQ Y0, -8(X4*1), Y6 // c4e2fd903425f8ffffff
+ VPGATHERDQ Y0, 0(X4*1), Y6 // c4e2fd90342500000000
+ VPGATHERDQ Y0, 664(X4*1), Y6 // c4e2fd90342598020000
+ VPGATHERDQ Y0, 8(X4*8), Y6 // c4e2fd9034e508000000
+ VPGATHERDQ Y0, -8(X4*8), Y6 // c4e2fd9034e5f8ffffff
+ VPGATHERDQ Y0, 0(X4*8), Y6 // c4e2fd9034e500000000
+ VPGATHERDQ Y0, 664(X4*8), Y6 // c4e2fd9034e598020000
+ VPGATHERDQ Y0, 8(X14*1), Y6 // c4a2fd90343508000000
+ VPGATHERDQ Y0, -8(X14*1), Y6 // c4a2fd903435f8ffffff
+ VPGATHERDQ Y0, 0(X14*1), Y6 // c4a2fd90343500000000
+ VPGATHERDQ Y0, 664(X14*1), Y6 // c4a2fd90343598020000
+ VPGATHERDQ Y0, 8(X14*8), Y6 // c4a2fd9034f508000000
+ VPGATHERDQ Y0, -8(X14*8), Y6 // c4a2fd9034f5f8ffffff
+ VPGATHERDQ Y0, 0(X14*8), Y6 // c4a2fd9034f500000000
+ VPGATHERDQ Y0, 664(X14*8), Y6 // c4a2fd9034f598020000
+ VPGATHERQQ X2, (BP)(X7*2), X1 // c4e2e9914c7d00
+ VPGATHERQQ Y2, (BP)(Y7*2), Y1 // c4e2ed914c7d00
+ VPGATHERQQ X12, (R13)(X14*2), X11 // c40299915c7500
+ VPGATHERQQ Y12, (R13)(Y14*2), Y11 // c4029d915c7500
+ VPGATHERQQ X2, (BP)(X7*2), X1 // c4e2e9914c7d00
+ VPGATHERQQ Y2, (BP)(Y7*2), Y1 // c4e2ed914c7d00
+ VPGATHERQQ X12, (R13)(X14*2), X11 // c40299915c7500
+ VPGATHERQQ Y12, (R13)(Y14*2), Y11 // c4029d915c7500
+ VGATHERDPD X2, (BP)(X7*2), X1 // c4e2e9924c7d00
+ VGATHERDPD Y2, (BP)(X7*2), Y1 // c4e2ed924c7d00
+ VGATHERDPD X12, (R13)(X14*2), X11 // c40299925c7500
+ VGATHERDPD Y12, (R13)(X14*2), Y11 // c4029d925c7500
+ VGATHERDPD Y0, 8(X4*1), Y6 // c4e2fd92342508000000
+ VGATHERDPD Y0, -8(X4*1), Y6 // c4e2fd923425f8ffffff
+ VGATHERDPD Y0, 0(X4*1), Y6 // c4e2fd92342500000000
+ VGATHERDPD Y0, 664(X4*1), Y6 // c4e2fd92342598020000
+ VGATHERDPD Y0, 8(X4*8), Y6 // c4e2fd9234e508000000
+ VGATHERDPD Y0, -8(X4*8), Y6 // c4e2fd9234e5f8ffffff
+ VGATHERDPD Y0, 0(X4*8), Y6 // c4e2fd9234e500000000
+ VGATHERDPD Y0, 664(X4*8), Y6 // c4e2fd9234e598020000
+ VGATHERDPD Y0, 8(X14*1), Y6 // c4a2fd92343508000000
+ VGATHERDPD Y0, -8(X14*1), Y6 // c4a2fd923435f8ffffff
+ VGATHERDPD Y0, 0(X14*1), Y6 // c4a2fd92343500000000
+ VGATHERDPD Y0, 664(X14*1), Y6 // c4a2fd92343598020000
+ VGATHERDPD Y0, 8(X14*8), Y6 // c4a2fd9234f508000000
+ VGATHERDPD Y0, -8(X14*8), Y6 // c4a2fd9234f5f8ffffff
+ VGATHERDPD Y0, 0(X14*8), Y6 // c4a2fd9234f500000000
+ VGATHERDPD Y0, 664(X14*8), Y6 // c4a2fd9234f598020000
+ VGATHERDPD X2, (BP)(X7*2), X1 // c4e2e9924c7d00
+ VGATHERDPD Y2, (BP)(X7*2), Y1 // c4e2ed924c7d00
+ VGATHERDPD X12, (R13)(X14*2), X11 // c40299925c7500
+ VGATHERDPD Y12, (R13)(X14*2), Y11 // c4029d925c7500
+ VGATHERDPD Y0, 8(X4*1), Y6 // c4e2fd92342508000000
+ VGATHERDPD Y0, -8(X4*1), Y6 // c4e2fd923425f8ffffff
+ VGATHERDPD Y0, 0(X4*1), Y6 // c4e2fd92342500000000
+ VGATHERDPD Y0, 664(X4*1), Y6 // c4e2fd92342598020000
+ VGATHERDPD Y0, 8(X4*8), Y6 // c4e2fd9234e508000000
+ VGATHERDPD Y0, -8(X4*8), Y6 // c4e2fd9234e5f8ffffff
+ VGATHERDPD Y0, 0(X4*8), Y6 // c4e2fd9234e500000000
+ VGATHERDPD Y0, 664(X4*8), Y6 // c4e2fd9234e598020000
+ VGATHERDPD Y0, 8(X14*1), Y6 // c4a2fd92343508000000
+ VGATHERDPD Y0, -8(X14*1), Y6 // c4a2fd923435f8ffffff
+ VGATHERDPD Y0, 0(X14*1), Y6 // c4a2fd92343500000000
+ VGATHERDPD Y0, 664(X14*1), Y6 // c4a2fd92343598020000
+ VGATHERDPD Y0, 8(X14*8), Y6 // c4a2fd9234f508000000
+ VGATHERDPD Y0, -8(X14*8), Y6 // c4a2fd9234f5f8ffffff
+ VGATHERDPD Y0, 0(X14*8), Y6 // c4a2fd9234f500000000
+ VGATHERDPD Y0, 664(X14*8), Y6 // c4a2fd9234f598020000
+ VGATHERQPD X2, (BP)(X7*2), X1 // c4e2e9934c7d00
+ VGATHERQPD Y2, (BP)(Y7*2), Y1 // c4e2ed934c7d00
+ VGATHERQPD X12, (R13)(X14*2), X11 // c40299935c7500
+ VGATHERQPD Y12, (R13)(Y14*2), Y11 // c4029d935c7500
+ VGATHERQPD X2, (BP)(X7*2), X1 // c4e2e9934c7d00
+ VGATHERQPD Y2, (BP)(Y7*2), Y1 // c4e2ed934c7d00
+ VGATHERQPD X12, (R13)(X14*2), X11 // c40299935c7500
+ VGATHERQPD Y12, (R13)(Y14*2), Y11 // c4029d935c7500
+ VGATHERDPS X2, (BP)(X7*2), X1 // c4e269924c7d00
+ VGATHERDPS Y2, (BP)(Y7*2), Y1 // c4e26d924c7d00
+ VGATHERDPS X12, (R13)(X14*2), X11 // c40219925c7500
+ VGATHERDPS Y12, (R13)(Y14*2), Y11 // c4021d925c7500
+ VGATHERDPS X3, 8(X4*1), X6 // c4e26192342508000000
+ VGATHERDPS X3, -8(X4*1), X6 // c4e261923425f8ffffff
+ VGATHERDPS X3, 0(X4*1), X6 // c4e26192342500000000
+ VGATHERDPS X3, 664(X4*1), X6 // c4e26192342598020000
+ VGATHERDPS X3, 8(X4*8), X6 // c4e2619234e508000000
+ VGATHERDPS X3, -8(X4*8), X6 // c4e2619234e5f8ffffff
+ VGATHERDPS X3, 0(X4*8), X6 // c4e2619234e500000000
+ VGATHERDPS X3, 664(X4*8), X6 // c4e2619234e598020000
+ VGATHERDPS X3, 8(X14*1), X6 // c4a26192343508000000
+ VGATHERDPS X3, -8(X14*1), X6 // c4a261923435f8ffffff
+ VGATHERDPS X3, 0(X14*1), X6 // c4a26192343500000000
+ VGATHERDPS X3, 664(X14*1), X6 // c4a26192343598020000
+ VGATHERDPS X3, 8(X14*8), X6 // c4a2619234f508000000
+ VGATHERDPS X3, -8(X14*8), X6 // c4a2619234f5f8ffffff
+ VGATHERDPS X3, 0(X14*8), X6 // c4a2619234f500000000
+ VGATHERDPS X3, 664(X14*8), X6 // c4a2619234f598020000
+ VGATHERDPS X2, (BP)(X7*2), X1 // c4e269924c7d00
+ VGATHERDPS Y2, (BP)(Y7*2), Y1 // c4e26d924c7d00
+ VGATHERDPS X12, (R13)(X14*2), X11 // c40219925c7500
+ VGATHERDPS Y12, (R13)(Y14*2), Y11 // c4021d925c7500
+ VGATHERDPS X5, 8(X4*1), X6 // c4e25192342508000000
+ VGATHERDPS X3, -8(X4*1), X6 // c4e261923425f8ffffff
+ VGATHERDPS X3, 0(X4*1), X6 // c4e26192342500000000
+ VGATHERDPS X3, 664(X4*1), X6 // c4e26192342598020000
+ VGATHERDPS X3, 8(X4*8), X6 // c4e2619234e508000000
+ VGATHERDPS X3, -8(X4*8), X6 // c4e2619234e5f8ffffff
+ VGATHERDPS X3, 0(X4*8), X6 // c4e2619234e500000000
+ VGATHERDPS X3, 664(X4*8), X6 // c4e2619234e598020000
+ VGATHERDPS X3, 8(X14*1), X6 // c4a26192343508000000
+ VGATHERDPS X3, -8(X14*1), X6 // c4a261923435f8ffffff
+ VGATHERDPS X3, 0(X14*1), X6 // c4a26192343500000000
+ VGATHERDPS X3, 664(X14*1), X6 // c4a26192343598020000
+ VGATHERDPS X3, 8(X14*8), X6 // c4a2619234f508000000
+ VGATHERDPS X3, -8(X14*8), X6 // c4a2619234f5f8ffffff
+ VGATHERDPS X3, 0(X14*8), X6 // c4a2619234f500000000
+ VGATHERDPS X3, 664(X14*8), X6 // c4a2619234f598020000
+ VGATHERQPS X2, (BP)(X7*2), X1 // c4e269934c7d00
+ VGATHERQPS X2, (BP)(Y7*2), X1 // c4e26d934c7d00
+ VGATHERQPS X12, (R13)(X14*2), X11 // c40219935c7500
+ VGATHERQPS X12, (R13)(Y14*2), X11 // c4021d935c7500
+ VGATHERQPS X2, (BP)(X7*2), X1 // c4e269934c7d00
+ VGATHERQPS X2, (BP)(Y7*2), X1 // c4e26d934c7d00
+ VGATHERQPS X12, (R13)(X14*2), X11 // c40219935c7500
+ VGATHERQPS X12, (R13)(Y14*2), X11 // c4021d935c7500
+ VPGATHERDD X2, (BP)(X7*2), X1 // c4e269904c7d00
+ VPGATHERDD Y2, (BP)(Y7*2), Y1 // c4e26d904c7d00
+ VPGATHERDD X12, (R13)(X14*2), X11 // c40219905c7500
+ VPGATHERDD Y12, (R13)(Y14*2), Y11 // c4021d905c7500
+ VPGATHERDD X3, 8(X4*1), X6 // c4e26190342508000000
+ VPGATHERDD X3, -8(X4*1), X6 // c4e261903425f8ffffff
+ VPGATHERDD X3, 0(X4*1), X6 // c4e26190342500000000
+ VPGATHERDD X3, 664(X4*1), X6 // c4e26190342598020000
+ VPGATHERDD X3, 8(X4*8), X6 // c4e2619034e508000000
+ VPGATHERDD X3, -8(X4*8), X6 // c4e2619034e5f8ffffff
+ VPGATHERDD X3, 0(X4*8), X6 // c4e2619034e500000000
+ VPGATHERDD X3, 664(X4*8), X6 // c4e2619034e598020000
+ VPGATHERDD X3, 8(X14*1), X6 // c4a26190343508000000
+ VPGATHERDD X3, -8(X14*1), X6 // c4a261903435f8ffffff
+ VPGATHERDD X3, 0(X14*1), X6 // c4a26190343500000000
+ VPGATHERDD X3, 664(X14*1), X6 // c4a26190343598020000
+ VPGATHERDD X3, 8(X14*8), X6 // c4a2619034f508000000
+ VPGATHERDD X3, -8(X14*8), X6 // c4a2619034f5f8ffffff
+ VPGATHERDD X3, 0(X14*8), X6 // c4a2619034f500000000
+ VPGATHERDD X3, 664(X14*8), X6 // c4a2619034f598020000
+ VPGATHERDD X2, (BP)(X7*2), X1 // c4e269904c7d00
+ VPGATHERDD Y2, (BP)(Y7*2), Y1 // c4e26d904c7d00
+ VPGATHERDD X12, (R13)(X14*2), X11 // c40219905c7500
+ VPGATHERDD Y12, (R13)(Y14*2), Y11 // c4021d905c7500
+ VPGATHERDD X3, 8(X4*1), X6 // c4e26190342508000000
+ VPGATHERDD X3, -8(X4*1), X6 // c4e261903425f8ffffff
+ VPGATHERDD X3, 0(X4*1), X6 // c4e26190342500000000
+ VPGATHERDD X3, 664(X4*1), X6 // c4e26190342598020000
+ VPGATHERDD X3, 8(X4*8), X6 // c4e2619034e508000000
+ VPGATHERDD X3, -8(X4*8), X6 // c4e2619034e5f8ffffff
+ VPGATHERDD X3, 0(X4*8), X6 // c4e2619034e500000000
+ VPGATHERDD X3, 664(X4*8), X6 // c4e2619034e598020000
+ VPGATHERDD X3, 8(X14*1), X6 // c4a26190343508000000
+ VPGATHERDD X3, -8(X14*1), X6 // c4a261903435f8ffffff
+ VPGATHERDD X3, 0(X14*1), X6 // c4a26190343500000000
+ VPGATHERDD X3, 664(X14*1), X6 // c4a26190343598020000
+ VPGATHERDD X3, 8(X14*8), X6 // c4a2619034f508000000
+ VPGATHERDD X3, -8(X14*8), X6 // c4a2619034f5f8ffffff
+ VPGATHERDD X3, 0(X14*8), X6 // c4a2619034f500000000
+ VPGATHERDD X3, 664(X14*8), X6 // c4a2619034f598020000
+ VPGATHERQD X2, (BP)(X7*2), X1 // c4e269914c7d00
+ VPGATHERQD X2, (BP)(Y7*2), X1 // c4e26d914c7d00
+ VPGATHERQD X12, (R13)(X14*2), X11 // c40219915c7500
+ VPGATHERQD X12, (R13)(Y14*2), X11 // c4021d915c7500
+ VPGATHERQD X2, (BP)(X7*2), X1 // c4e269914c7d00
+ VPGATHERQD X2, (BP)(Y7*2), X1 // c4e26d914c7d00
+ VPGATHERQD X12, (R13)(X14*2), X11 // c40219915c7500
+ VPGATHERQD X12, (R13)(Y14*2), X11 // c4021d915c7500
+ VPGATHERQQ X0, 0(X1*1), X2 // c4e2f991140d00000000
+ VPGATHERQQ Y0, 0(Y1*1), Y2 // c4e2fd91140d00000000
+ VPGATHERQQ X8, 0(X9*1), X10 // c422b991140d00000000
+ VPGATHERQQ Y8, 0(Y9*1), Y10 // c422bd91140d00000000
+ VPGATHERQQ X0, 0(X1*4), X2 // c4e2f991148d00000000
+ VPGATHERQQ Y0, 0(Y1*4), Y2 // c4e2fd91148d00000000
+ VPGATHERQQ X8, 0(X9*4), X10 // c422b991148d00000000
+ VPGATHERQQ Y8, 0(Y9*4), Y10 // c422bd91148d00000000
+ // AVX2GATHER: test SP/BP base with different displacements.
+ VPGATHERQQ X0, (SP)(X1*1), X2 // c4e2f991140c
+ VPGATHERQQ X0, 16(SP)(X1*1), X2 // c4e2f991540c10
+ VPGATHERQQ X0, 512(SP)(X1*1), X2 // c4e2f991940c00020000
+ VPGATHERQQ X0, (R12)(X1*1), X2 // c4c2f991140c
+ VPGATHERQQ X0, 16(R12)(X1*1), X2 // c4c2f991540c10
+ VPGATHERQQ X0, 512(R12)(X1*1), X2 // c4c2f991940c00020000
+ VPGATHERQQ X0, (BP)(X1*1), X2 // c4e2f991540d00
+ VPGATHERQQ X0, 16(BP)(X1*1), X2 // c4e2f991540d10
+ VPGATHERQQ X0, 512(BP)(X1*1), X2 // c4e2f991940d00020000
+ VPGATHERQQ X0, (R13)(X1*1), X2 // c4c2f991540d00
+ VPGATHERQQ X0, 16(R13)(X1*1), X2 // c4c2f991540d10
+ VPGATHERQQ X0, 512(R13)(X1*1), X2 // c4c2f991940d00020000
+ VPGATHERQQ Y0, (SP)(Y1*1), Y2 // c4e2fd91140c
+ VPGATHERQQ Y0, 16(SP)(Y1*1), Y2 // c4e2fd91540c10
+ VPGATHERQQ Y0, 512(SP)(Y1*1), Y2 // c4e2fd91940c00020000
+ VPGATHERQQ Y0, (R12)(Y1*1), Y2 // c4c2fd91140c
+ VPGATHERQQ Y0, 16(R12)(Y1*1), Y2 // c4c2fd91540c10
+ VPGATHERQQ Y0, 512(R12)(Y1*1), Y2 // c4c2fd91940c00020000
+ VPGATHERQQ Y0, (BP)(Y1*1), Y2 // c4e2fd91540d00
+ VPGATHERQQ Y0, 16(BP)(Y1*1), Y2 // c4e2fd91540d10
+ VPGATHERQQ Y0, 512(BP)(Y1*1), Y2 // c4e2fd91940d00020000
+ VPGATHERQQ Y0, (R13)(Y1*1), Y2 // c4c2fd91540d00
+ VPGATHERQQ Y0, 16(R13)(Y1*1), Y2 // c4c2fd91540d10
+ VPGATHERQQ Y0, 512(R13)(Y1*1), Y2 // c4c2fd91940d00020000
+ // End of tests.
+ RET
diff --git a/src/cmd/asm/internal/asm/testdata/amd64error.s b/src/cmd/asm/internal/asm/testdata/amd64error.s
index 2cb082dacc..7d850f7844 100644
--- a/src/cmd/asm/internal/asm/testdata/amd64error.s
+++ b/src/cmd/asm/internal/asm/testdata/amd64error.s
@@ -7,4 +7,29 @@ TEXT errors(SB),$0
MOVL (AX)(SP*1), AX // ERROR "invalid instruction"
EXTRACTPS $4, X2, (BX) // ERROR "invalid instruction"
EXTRACTPS $-1, X2, (BX) // ERROR "invalid instruction"
+ // VSIB addressing does not permit non-vector (X/Y)
+ // scaled index register.
+ VPGATHERDQ X12,(R13)(AX*2), X11 // ERROR "invalid instruction"
+ VPGATHERDQ X2, 664(BX*1), X1 // ERROR "invalid instruction"
+ VPGATHERDQ Y2, (BP)(AX*2), Y1 // ERROR "invalid instruction"
+ VPGATHERDQ Y5, 664(DX*8), Y6 // ERROR "invalid instruction"
+ VPGATHERDQ Y5, (DX), Y0 // ERROR "invalid instruction"
+ // VM/X rejects Y index register.
+ VPGATHERDQ Y5, 664(Y14*8), Y6 // ERROR "invalid instruction"
+ VPGATHERQQ X2, (BP)(Y7*2), X1 // ERROR "invalid instruction"
+ // VM/Y rejects X index register.
+ VPGATHERQQ Y2, (BP)(X7*2), Y1 // ERROR "invalid instruction"
+ VPGATHERDD Y5, -8(X14*8), Y6 // ERROR "invalid instruction"
+ // No VSIB for legacy instructions.
+ MOVL (AX)(X0*1), AX // ERROR "invalid instruction"
+ MOVL (AX)(Y0*1), AX // ERROR "invalid instruction"
+ // AVX2GATHER mask/index/dest #UD cases.
+ VPGATHERQQ Y2, (BP)(X2*2), Y2 // ERROR "mask, index, and destination registers should be distinct"
+ VPGATHERQQ Y2, (BP)(X2*2), Y7 // ERROR "mask, index, and destination registers should be distinct"
+ VPGATHERQQ Y2, (BP)(X7*2), Y2 // ERROR "mask, index, and destination registers should be distinct"
+ VPGATHERQQ Y7, (BP)(X2*2), Y2 // ERROR "mask, index, and destination registers should be distinct"
+ VPGATHERDQ X2, 664(X2*8), X2 // ERROR "mask, index, and destination registers should be distinct"
+ VPGATHERDQ X2, 664(X2*8), X7 // ERROR "mask, index, and destination registers should be distinct"
+ VPGATHERDQ X2, 664(X7*8), X2 // ERROR "mask, index, and destination registers should be distinct"
+ VPGATHERDQ X7, 664(X2*8), X2 // ERROR "mask, index, and destination registers should be distinct"
RET
diff --git a/src/cmd/internal/obj/x86/aenum.go b/src/cmd/internal/obj/x86/aenum.go
index 0b9cbefe53..013d9e0228 100644
--- a/src/cmd/internal/obj/x86/aenum.go
+++ b/src/cmd/internal/obj/x86/aenum.go
@@ -874,6 +874,10 @@ const (
AVFNMSUB231PS
AVFNMSUB231SD
AVFNMSUB231SS
+ AVGATHERDPD
+ AVGATHERDPS
+ AVGATHERQPD
+ AVGATHERQPS
AVHADDPD
AVHADDPS
AVHSUBPD
@@ -978,6 +982,10 @@ const (
AVPEXTRD
AVPEXTRQ
AVPEXTRW
+ AVPGATHERDD
+ AVPGATHERDQ
+ AVPGATHERQD
+ AVPGATHERQQ
AVPHADDD
AVPHADDSW
AVPHADDW
diff --git a/src/cmd/internal/obj/x86/anames.go b/src/cmd/internal/obj/x86/anames.go
index 5d2a920fe7..ec7bea1255 100644
--- a/src/cmd/internal/obj/x86/anames.go
+++ b/src/cmd/internal/obj/x86/anames.go
@@ -873,6 +873,10 @@ var Anames = []string{
"VFNMSUB231PS",
"VFNMSUB231SD",
"VFNMSUB231SS",
+ "VGATHERDPD",
+ "VGATHERDPS",
+ "VGATHERQPD",
+ "VGATHERQPS",
"VHADDPD",
"VHADDPS",
"VHSUBPD",
@@ -977,6 +981,10 @@ var Anames = []string{
"VPEXTRD",
"VPEXTRQ",
"VPEXTRW",
+ "VPGATHERDD",
+ "VPGATHERDQ",
+ "VPGATHERQD",
+ "VPGATHERQQ",
"VPHADDD",
"VPHADDSW",
"VPHADDW",
diff --git a/src/cmd/internal/obj/x86/asm6.go b/src/cmd/internal/obj/x86/asm6.go
index 4e4cae6b44..6451f2cc98 100644
--- a/src/cmd/internal/obj/x86/asm6.go
+++ b/src/cmd/internal/obj/x86/asm6.go
@@ -148,8 +148,10 @@ const (
Ymm
Yxr
Yxm
+ Yxvm // VSIB vector array; vm32x/vm64x
Yyr
Yym
+ Yyvm // VSIB vector array; vm32y/vm64y
Ytls
Ytextsize
Yindir
@@ -1034,6 +1036,21 @@ var yvex_vmovq = []ytab{
{Zvex_r_v_rm, 2, argList{Yxr, Yxm}},
}
+var yvpgatherdq = []ytab{
+ {Zvex_v_rm_r, 2, argList{Yxr, Yxvm, Yxr}},
+ {Zvex_v_rm_r, 2, argList{Yyr, Yxvm, Yyr}},
+}
+
+var yvpgatherqq = []ytab{
+ {Zvex_v_rm_r, 2, argList{Yxr, Yxvm, Yxr}},
+ {Zvex_v_rm_r, 2, argList{Yyr, Yyvm, Yyr}},
+}
+
+var yvgatherqps = []ytab{
+ {Zvex_v_rm_r, 2, argList{Yxr, Yxvm, Yxr}},
+ {Zvex_v_rm_r, 2, argList{Yxr, Yyvm, Yxr}},
+}
+
var ymmxmm0f38 = []ytab{
{Zlitm_r, 3, argList{Ymm, Ymr}},
{Zlitm_r, 5, argList{Yxm, Yxr}},
@@ -1855,6 +1872,44 @@ var optab =
{obj.APCDATA, ypcdata, Px, [23]uint8{0, 0}},
{obj.ADUFFCOPY, yduff, Px, [23]uint8{0xe8}},
{obj.ADUFFZERO, yduff, Px, [23]uint8{0xe8}},
+
+ // AVX2 gather instructions.
+ // Added as a part of VSIB support implementation,
+ // when x86avxgen will output these, they will be moved to
+ // vex_optabs.go where they belong.
+ {AVGATHERDPD, yvpgatherdq, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW1, 0x92,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW1, 0x92,
+ }},
+ {AVGATHERQPD, yvpgatherqq, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW1, 0x93,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW1, 0x93,
+ }},
+ {AVGATHERDPS, yvpgatherqq, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW0, 0x92,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW0, 0x92,
+ }},
+ {AVGATHERQPS, yvgatherqps, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW0, 0x93,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW0, 0x93,
+ }},
+ {AVPGATHERDD, yvpgatherqq, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW0, 0x90,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW0, 0x90,
+ }},
+ {AVPGATHERQD, yvgatherqps, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW0, 0x91,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW0, 0x91,
+ }},
+ {AVPGATHERDQ, yvpgatherdq, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW1, 0x90,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW1, 0x90,
+ }},
+ {AVPGATHERQQ, yvpgatherqq, Pvex, [23]uint8{
+ vexDDS | vex128 | vex66 | vex0F38 | vexW1, 0x91,
+ vexDDS | vex256 | vex66 | vex0F38 | vexW1, 0x91,
+ }},
+
{obj.AEND, nil, 0, [23]uint8{}},
{0, nil, 0, [23]uint8{}},
}
@@ -2435,6 +2490,18 @@ func oclass(ctxt *obj.Link, p *obj.Prog, a *obj.Addr) int {
// Can't use SP as the index register
return Yxxx
}
+ if a.Index >= REG_X0 && a.Index <= REG_X15 {
+ if ctxt.Arch.Family == sys.I386 && a.Index > REG_X7 {
+ return Yxxx
+ }
+ return Yxvm
+ }
+ if a.Index >= REG_Y0 && a.Index <= REG_Y15 {
+ if ctxt.Arch.Family == sys.I386 && a.Index > REG_Y7 {
+ return Yxxx
+ }
+ return Yyvm
+ }
if ctxt.Arch.Family == sys.AMD64 {
// Offset must fit in a 32-bit signed field (or fit in a 32-bit unsigned field
// where the sign extension doesn't matter).
@@ -2847,9 +2914,11 @@ func (a *AsmBuf) Reset() { a.off = 0 }
// At returns the byte at offset i.
func (a *AsmBuf) At(i int) byte { return a.buf[i] }
+// asmidx emits SIB byte.
func (asmbuf *AsmBuf) asmidx(ctxt *obj.Link, scale int, index int, base int) {
var i int
+ // X/Y index register is used in VSIB.
switch index {
default:
goto bad
@@ -2865,7 +2934,23 @@ func (asmbuf *AsmBuf) asmidx(ctxt *obj.Link, scale int, index int, base int) {
REG_R12,
REG_R13,
REG_R14,
- REG_R15:
+ REG_R15,
+ REG_X8,
+ REG_X9,
+ REG_X10,
+ REG_X11,
+ REG_X12,
+ REG_X13,
+ REG_X14,
+ REG_X15,
+ REG_Y8,
+ REG_Y9,
+ REG_Y10,
+ REG_Y11,
+ REG_Y12,
+ REG_Y13,
+ REG_Y14,
+ REG_Y15:
if ctxt.Arch.Family == sys.I386 {
goto bad
}
@@ -2877,7 +2962,23 @@ func (asmbuf *AsmBuf) asmidx(ctxt *obj.Link, scale int, index int, base int) {
REG_BX,
REG_BP,
REG_SI,
- REG_DI:
+ REG_DI,
+ REG_X0,
+ REG_X1,
+ REG_X2,
+ REG_X3,
+ REG_X4,
+ REG_X5,
+ REG_X6,
+ REG_X7,
+ REG_Y0,
+ REG_Y1,
+ REG_Y2,
+ REG_Y3,
+ REG_Y4,
+ REG_Y5,
+ REG_Y6,
+ REG_Y7:
i = reg[index] << 3
}
@@ -3488,6 +3589,35 @@ func (asmbuf *AsmBuf) asmvex(ctxt *obj.Link, rm, v, r *obj.Addr, vex, opcode uin
asmbuf.Put1(opcode)
}
+// regIndex returns register index that fits in 4 bits.
+//
+// Examples:
+// REG_X15 => 15
+// REG_R9 => 9
+// REG_AX => 0
+//
+func regIndex(r int16) int {
+ lower3bits := reg[r]
+ high4bit := regrex[r] & Rxr << 1
+ return lower3bits | high4bit
+}
+
+// avx2gatherValid returns true if p satisfies AVX2 gather constraints.
+// Reports errors via ctxt.
+func avx2gatherValid(ctxt *obj.Link, p *obj.Prog) bool {
+ // If any pair of the index, mask, or destination registers
+ // are the same, this instruction results a #UD fault.
+ index := regIndex(p.GetFrom3().Index)
+ mask := regIndex(p.From.Reg)
+ dest := regIndex(p.To.Reg)
+ if dest == mask || dest == index || mask == index {
+ ctxt.Diag("mask, index, and destination registers should be distinct: %v", p)
+ return false
+ }
+
+ return true
+}
+
func (asmbuf *AsmBuf) doasm(ctxt *obj.Link, cursym *obj.LSym, p *obj.Prog) {
o := opindex[p.As&obj.AMask]
@@ -3536,6 +3666,18 @@ func (asmbuf *AsmBuf) doasm(ctxt *obj.Link, cursym *obj.LSym, p *obj.Prog) {
p.To.Offset = p.GetFrom3().Offset
p.GetFrom3().Offset = 0
}
+
+ case AVGATHERDPD,
+ AVGATHERQPD,
+ AVGATHERDPS,
+ AVGATHERQPS,
+ AVPGATHERDD,
+ AVPGATHERQD,
+ AVPGATHERDQ,
+ AVPGATHERQQ:
+ if !avx2gatherValid(ctxt, p) {
+ return
+ }
}
if p.Ft == 0 {