From: Adrian Hunter <adrian.hunter@intel.com>
To: linux-kernel@vger.kernel.org
Cc: "Chang S. Bae" <chang.seok.bae@intel.com>,
Masami Hiramatsu <mhiramat@kernel.org>,
Nikolay Borisov <nik.borisov@suse.com>,
Borislav Petkov <bp@alien8.de>, Ingo Molnar <mingo@redhat.com>,
"H. Peter Anvin" <hpa@zytor.com>,
Dave Hansen <dave.hansen@linux.intel.com>,
Thomas Gleixner <tglx@linutronix.de>,
x86@kernel.org, Arnaldo Carvalho de Melo <acme@kernel.org>,
Jiri Olsa <jolsa@kernel.org>, Namhyung Kim <namhyung@kernel.org>,
Ian Rogers <irogers@google.com>,
linux-perf-users@vger.kernel.org
Subject: [PATCH 07/10] x86/insn: Add support for APX EVEX to the instruction decoder logic
Date: Thu, 2 May 2024 13:58:50 +0300 [thread overview]
Message-ID: <20240502105853.5338-8-adrian.hunter@intel.com> (raw)
In-Reply-To: <20240502105853.5338-1-adrian.hunter@intel.com>
Intel Advanced Performance Extensions (APX) extends the EVEX prefix to
support:
- extended general purpose registers (EGPRs) i.e. r16 to r31
- Push-Pop Acceleration (PPX) hints
- new data destination (NDD) register
- suppress status flags writes (NF) of common instructions
- new instructions
Refer to the Intel Advanced Performance Extensions (Intel APX) Architecture
Specification for details.
The extended EVEX prefix does not need amended instruction decoder logic,
except in one area. Some instructions are defined as SCALABLE which means
the EVEX.W bit and EVEX.pp bits are used to determine operand size.
Specifically, if an instruction is SCALABLE and EVEX.W is zero, then
EVEX.pp value 0 (representing no prefix NP) means default operand size,
whereas EVEX.pp value 1 (representing 66 prefix) means operand size
override i.e. 16 bits
Add an attribute (INAT_EVEX_SCALABLE) to identify such instructions, and
amend the logic appropriately.
Amend the awk script that generates the attribute tables from the opcode
map, to recognise "(es)" as attribute INAT_EVEX_SCALABLE.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
arch/x86/include/asm/inat.h | 6 ++++++
arch/x86/include/asm/insn.h | 7 +++++++
arch/x86/lib/insn.c | 4 ++++
arch/x86/tools/gen-insn-attr-x86.awk | 4 ++++
tools/arch/x86/include/asm/inat.h | 6 ++++++
tools/arch/x86/include/asm/insn.h | 7 +++++++
tools/arch/x86/lib/insn.c | 4 ++++
tools/arch/x86/tools/gen-insn-attr-x86.awk | 4 ++++
8 files changed, 42 insertions(+)
diff --git a/arch/x86/include/asm/inat.h b/arch/x86/include/asm/inat.h
index 1331bdd39a23..53e4015242b4 100644
--- a/arch/x86/include/asm/inat.h
+++ b/arch/x86/include/asm/inat.h
@@ -81,6 +81,7 @@
#define INAT_EVEXONLY (1 << (INAT_FLAG_OFFS + 7))
#define INAT_NO_REX2 (1 << (INAT_FLAG_OFFS + 8))
#define INAT_REX2_VARIANT (1 << (INAT_FLAG_OFFS + 9))
+#define INAT_EVEX_SCALABLE (1 << (INAT_FLAG_OFFS + 10))
/* Attribute making macros for attribute tables */
#define INAT_MAKE_PREFIX(pfx) (pfx << INAT_PFX_OFFS)
#define INAT_MAKE_ESCAPE(esc) (esc << INAT_ESC_OFFS)
@@ -236,4 +237,9 @@ static inline int inat_must_evex(insn_attr_t attr)
{
return attr & INAT_EVEXONLY;
}
+
+static inline int inat_evex_scalable(insn_attr_t attr)
+{
+ return attr & INAT_EVEX_SCALABLE;
+}
#endif
diff --git a/arch/x86/include/asm/insn.h b/arch/x86/include/asm/insn.h
index 95249ec1f24e..7152ea809e6a 100644
--- a/arch/x86/include/asm/insn.h
+++ b/arch/x86/include/asm/insn.h
@@ -215,6 +215,13 @@ static inline insn_byte_t insn_vex_p_bits(struct insn *insn)
return X86_VEX_P(insn->vex_prefix.bytes[2]);
}
+static inline insn_byte_t insn_vex_w_bit(struct insn *insn)
+{
+ if (insn->vex_prefix.nbytes < 3)
+ return 0;
+ return X86_VEX_W(insn->vex_prefix.bytes[2]);
+}
+
/* Get the last prefix id from last prefix or VEX prefix */
static inline int insn_last_prefix_id(struct insn *insn)
{
diff --git a/arch/x86/lib/insn.c b/arch/x86/lib/insn.c
index 6126ddc6e5f5..5952ab41c60f 100644
--- a/arch/x86/lib/insn.c
+++ b/arch/x86/lib/insn.c
@@ -294,6 +294,10 @@ int insn_get_opcode(struct insn *insn)
m = insn_vex_m_bits(insn);
p = insn_vex_p_bits(insn);
insn->attr = inat_get_avx_attribute(op, m, p);
+ /* SCALABLE EVEX uses p bits to encode operand size */
+ if (inat_evex_scalable(insn->attr) && !insn_vex_w_bit(insn) &&
+ p == INAT_PFX_OPNDSZ)
+ insn->opnd_bytes = 2;
if ((inat_must_evex(insn->attr) && !insn_is_evex(insn)) ||
(!inat_accept_vex(insn->attr) &&
!inat_is_group(insn->attr))) {
diff --git a/arch/x86/tools/gen-insn-attr-x86.awk b/arch/x86/tools/gen-insn-attr-x86.awk
index 3f43aa7d8fef..5770c8097f32 100644
--- a/arch/x86/tools/gen-insn-attr-x86.awk
+++ b/arch/x86/tools/gen-insn-attr-x86.awk
@@ -83,6 +83,8 @@ BEGIN {
vexonly_expr = "\\(v\\)"
# All opcodes with (ev) superscript supports *only* EVEX prefix
evexonly_expr = "\\(ev\\)"
+ # (es) is the same as (ev) but also "SCALABLE" i.e. W and pp determine operand size
+ evex_scalable_expr = "\\(es\\)"
prefix_expr = "\\(Prefix\\)"
prefix_num["Operand-Size"] = "INAT_PFX_OPNDSZ"
@@ -332,6 +334,8 @@ function convert_operands(count,opnd, i,j,imm,mod)
# check VEX codes
if (match(ext, evexonly_expr))
flags = add_flags(flags, "INAT_VEXOK | INAT_EVEXONLY")
+ else if (match(ext, evex_scalable_expr))
+ flags = add_flags(flags, "INAT_VEXOK | INAT_EVEXONLY | INAT_EVEX_SCALABLE")
else if (match(ext, vexonly_expr))
flags = add_flags(flags, "INAT_VEXOK | INAT_VEXONLY")
else if (match(ext, vexok_expr) || match(opcode, vexok_opcode_expr))
diff --git a/tools/arch/x86/include/asm/inat.h b/tools/arch/x86/include/asm/inat.h
index 2e65312cae52..253690eb3c26 100644
--- a/tools/arch/x86/include/asm/inat.h
+++ b/tools/arch/x86/include/asm/inat.h
@@ -81,6 +81,7 @@
#define INAT_EVEXONLY (1 << (INAT_FLAG_OFFS + 7))
#define INAT_NO_REX2 (1 << (INAT_FLAG_OFFS + 8))
#define INAT_REX2_VARIANT (1 << (INAT_FLAG_OFFS + 9))
+#define INAT_EVEX_SCALABLE (1 << (INAT_FLAG_OFFS + 10))
/* Attribute making macros for attribute tables */
#define INAT_MAKE_PREFIX(pfx) (pfx << INAT_PFX_OFFS)
#define INAT_MAKE_ESCAPE(esc) (esc << INAT_ESC_OFFS)
@@ -236,4 +237,9 @@ static inline int inat_must_evex(insn_attr_t attr)
{
return attr & INAT_EVEXONLY;
}
+
+static inline int inat_evex_scalable(insn_attr_t attr)
+{
+ return attr & INAT_EVEX_SCALABLE;
+}
#endif
diff --git a/tools/arch/x86/include/asm/insn.h b/tools/arch/x86/include/asm/insn.h
index 1a7e8fc4d75a..0e5abd896ad4 100644
--- a/tools/arch/x86/include/asm/insn.h
+++ b/tools/arch/x86/include/asm/insn.h
@@ -215,6 +215,13 @@ static inline insn_byte_t insn_vex_p_bits(struct insn *insn)
return X86_VEX_P(insn->vex_prefix.bytes[2]);
}
+static inline insn_byte_t insn_vex_w_bit(struct insn *insn)
+{
+ if (insn->vex_prefix.nbytes < 3)
+ return 0;
+ return X86_VEX_W(insn->vex_prefix.bytes[2]);
+}
+
/* Get the last prefix id from last prefix or VEX prefix */
static inline int insn_last_prefix_id(struct insn *insn)
{
diff --git a/tools/arch/x86/lib/insn.c b/tools/arch/x86/lib/insn.c
index f761adeb8e8c..a43b37346a22 100644
--- a/tools/arch/x86/lib/insn.c
+++ b/tools/arch/x86/lib/insn.c
@@ -294,6 +294,10 @@ int insn_get_opcode(struct insn *insn)
m = insn_vex_m_bits(insn);
p = insn_vex_p_bits(insn);
insn->attr = inat_get_avx_attribute(op, m, p);
+ /* SCALABLE EVEX uses p bits to encode operand size */
+ if (inat_evex_scalable(insn->attr) && !insn_vex_w_bit(insn) &&
+ p == INAT_PFX_OPNDSZ)
+ insn->opnd_bytes = 2;
if ((inat_must_evex(insn->attr) && !insn_is_evex(insn)) ||
(!inat_accept_vex(insn->attr) &&
!inat_is_group(insn->attr))) {
diff --git a/tools/arch/x86/tools/gen-insn-attr-x86.awk b/tools/arch/x86/tools/gen-insn-attr-x86.awk
index 3f43aa7d8fef..5770c8097f32 100644
--- a/tools/arch/x86/tools/gen-insn-attr-x86.awk
+++ b/tools/arch/x86/tools/gen-insn-attr-x86.awk
@@ -83,6 +83,8 @@ BEGIN {
vexonly_expr = "\\(v\\)"
# All opcodes with (ev) superscript supports *only* EVEX prefix
evexonly_expr = "\\(ev\\)"
+ # (es) is the same as (ev) but also "SCALABLE" i.e. W and pp determine operand size
+ evex_scalable_expr = "\\(es\\)"
prefix_expr = "\\(Prefix\\)"
prefix_num["Operand-Size"] = "INAT_PFX_OPNDSZ"
@@ -332,6 +334,8 @@ function convert_operands(count,opnd, i,j,imm,mod)
# check VEX codes
if (match(ext, evexonly_expr))
flags = add_flags(flags, "INAT_VEXOK | INAT_EVEXONLY")
+ else if (match(ext, evex_scalable_expr))
+ flags = add_flags(flags, "INAT_VEXOK | INAT_EVEXONLY | INAT_EVEX_SCALABLE")
else if (match(ext, vexonly_expr))
flags = add_flags(flags, "INAT_VEXOK | INAT_VEXONLY")
else if (match(ext, vexok_expr) || match(opcode, vexok_opcode_expr))
--
2.34.1
next prev parent reply other threads:[~2024-05-02 10:59 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-05-02 10:58 [PATCH 00/10] perf intel pt: Update instruction decoder for APX and other new instructions Adrian Hunter
2024-05-02 10:58 ` [PATCH 01/10] x86/insn: Add Key Locker instructions to the opcode map Adrian Hunter
2024-05-02 10:58 ` [PATCH 02/10] x86/insn: Fix PUSH instruction in x86 instruction decoder " Adrian Hunter
2024-05-02 13:41 ` Arnaldo Carvalho de Melo
2024-05-02 19:11 ` Adrian Hunter
2024-05-02 10:58 ` [PATCH 03/10] x86/insn: Add VEX versions of VPDPBUSD, VPDPBUSDS, VPDPWSSD and VPDPWSSDS Adrian Hunter
2024-05-02 10:58 ` [PATCH 04/10] x86/insn: Add misc new Intel instructions Adrian Hunter
2024-05-02 10:58 ` [PATCH 05/10] x86/insn: Add support for REX2 prefix to the instruction decoder logic Adrian Hunter
2024-05-02 18:10 ` Ian Rogers
2024-05-03 5:09 ` Adrian Hunter
2024-05-02 10:58 ` [PATCH 06/10] x86/insn: x86/insn: Add support for REX2 prefix to the instruction decoder opcode map Adrian Hunter
2024-05-02 10:58 ` Adrian Hunter [this message]
2024-05-02 10:58 ` [PATCH 08/10] x86/insn: Add support for APX EVEX instructions to the " Adrian Hunter
2024-05-02 10:58 ` [PATCH 09/10] perf intel pt: Add new JMPABS instruction to the Intel PT instruction decoder Adrian Hunter
2024-05-28 18:14 ` Adrian Hunter
2024-06-18 8:05 ` Adrian Hunter
2024-05-02 10:58 ` [PATCH 10/10] perf tests: Add APX and other new instructions to x86 instruction decoder test Adrian Hunter
2024-06-26 3:59 ` (subset) [PATCH 00/10] perf intel pt: Update instruction decoder for APX and other new instructions Namhyung Kim
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240502105853.5338-8-adrian.hunter@intel.com \
--to=adrian.hunter@intel.com \
--cc=acme@kernel.org \
--cc=bp@alien8.de \
--cc=chang.seok.bae@intel.com \
--cc=dave.hansen@linux.intel.com \
--cc=hpa@zytor.com \
--cc=irogers@google.com \
--cc=jolsa@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-perf-users@vger.kernel.org \
--cc=mhiramat@kernel.org \
--cc=mingo@redhat.com \
--cc=namhyung@kernel.org \
--cc=nik.borisov@suse.com \
--cc=tglx@linutronix.de \
--cc=x86@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).