From: Paolo Bonzini <pbonzini@redhat.com>
To: qemu-devel@nongnu.org
Cc: paul@nowt.org, richard.henderson@linaro.org
Subject: [PATCH 30/35] target/i386: implement XSAVE and XRSTOR of AVX registers
Date: Thu, 13 Oct 2022 23:46:46 +0200 [thread overview]
Message-ID: <20221013214651.672114-31-pbonzini@redhat.com> (raw)
In-Reply-To: <20221013214651.672114-1-pbonzini@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
target/i386/tcg/fpu_helper.c | 78 ++++++++++++++++++++++++++++++++++--
1 file changed, 75 insertions(+), 3 deletions(-)
diff --git a/target/i386/tcg/fpu_helper.c b/target/i386/tcg/fpu_helper.c
index 5f3f7a1085..7670739abe 100644
--- a/target/i386/tcg/fpu_helper.c
+++ b/target/i386/tcg/fpu_helper.c
@@ -2559,6 +2559,22 @@ static void do_xsave_sse(CPUX86State *env, target_ulong ptr, uintptr_t ra)
}
}
+static void do_xsave_ymmh(CPUX86State *env, target_ulong ptr, uintptr_t ra)
+{
+ int i, nb_xmm_regs;
+
+ if (env->hflags & HF_CS64_MASK) {
+ nb_xmm_regs = 16;
+ } else {
+ nb_xmm_regs = 8;
+ }
+
+ for (i = 0; i < nb_xmm_regs; i++, ptr += 16) {
+ cpu_stq_data_ra(env, ptr, env->xmm_regs[i].ZMM_Q(2), ra);
+ cpu_stq_data_ra(env, ptr + 8, env->xmm_regs[i].ZMM_Q(3), ra);
+ }
+}
+
static void do_xsave_bndregs(CPUX86State *env, target_ulong ptr, uintptr_t ra)
{
target_ulong addr = ptr + offsetof(XSaveBNDREG, bnd_regs);
@@ -2651,6 +2667,9 @@ static void do_xsave(CPUX86State *env, target_ulong ptr, uint64_t rfbm,
if (opt & XSTATE_SSE_MASK) {
do_xsave_sse(env, ptr, ra);
}
+ if (opt & XSTATE_YMM_MASK) {
+ do_xsave_ymmh(env, ptr + XO(avx_state), ra);
+ }
if (opt & XSTATE_BNDREGS_MASK) {
do_xsave_bndregs(env, ptr + XO(bndreg_state), ra);
}
@@ -2725,6 +2744,54 @@ static void do_xrstor_sse(CPUX86State *env, target_ulong ptr, uintptr_t ra)
}
}
+static void do_clear_sse(CPUX86State *env)
+{
+ int i, nb_xmm_regs;
+
+ if (env->hflags & HF_CS64_MASK) {
+ nb_xmm_regs = 16;
+ } else {
+ nb_xmm_regs = 8;
+ }
+
+ for (i = 0; i < nb_xmm_regs; i++) {
+ env->xmm_regs[i].ZMM_Q(0) = 0;
+ env->xmm_regs[i].ZMM_Q(1) = 0;
+ }
+}
+
+static void do_xrstor_ymmh(CPUX86State *env, target_ulong ptr, uintptr_t ra)
+{
+ int i, nb_xmm_regs;
+
+ if (env->hflags & HF_CS64_MASK) {
+ nb_xmm_regs = 16;
+ } else {
+ nb_xmm_regs = 8;
+ }
+
+ for (i = 0; i < nb_xmm_regs; i++, ptr += 16) {
+ env->xmm_regs[i].ZMM_Q(2) = cpu_ldq_data_ra(env, ptr, ra);
+ env->xmm_regs[i].ZMM_Q(3) = cpu_ldq_data_ra(env, ptr + 8, ra);
+ }
+}
+
+static void do_clear_ymmh(CPUX86State *env)
+{
+ int i, nb_xmm_regs;
+
+ if (env->hflags & HF_CS64_MASK) {
+ nb_xmm_regs = 16;
+ } else {
+ nb_xmm_regs = 8;
+ }
+
+ for (i = 0; i < nb_xmm_regs; i++) {
+ env->xmm_regs[i].ZMM_Q(2) = 0;
+ env->xmm_regs[i].ZMM_Q(3) = 0;
+ }
+}
+
static void do_xrstor_bndregs(CPUX86State *env, target_ulong ptr, uintptr_t ra)
{
target_ulong addr = ptr + offsetof(XSaveBNDREG, bnd_regs);
@@ -2831,9 +2898,14 @@ static void do_xrstor(CPUX86State *env, target_ulong ptr, uint64_t rfbm, uintptr
if (xstate_bv & XSTATE_SSE_MASK) {
do_xrstor_sse(env, ptr, ra);
} else {
- /* ??? When AVX is implemented, we may have to be more
- selective in the clearing. */
- memset(env->xmm_regs, 0, sizeof(env->xmm_regs));
+ do_clear_sse(env);
+ }
+ }
+ if (rfbm & XSTATE_YMM_MASK) {
+ if (xstate_bv & XSTATE_YMM_MASK) {
+ do_xrstor_ymmh(env, ptr + XO(avx_state), ra);
+ } else {
+ do_clear_ymmh(env);
}
}
if (rfbm & XSTATE_BNDREGS_MASK) {
--
2.37.3
next prev parent reply other threads:[~2022-10-13 22:26 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-10-13 21:46 [PATCH v3 00/35] target/i386: new decoder + AVX implementation Paolo Bonzini
2022-10-13 21:46 ` [PATCH 01/35] target/i386: Define XMMReg and access macros, align ZMM registers Paolo Bonzini
2022-10-13 21:46 ` [PATCH 02/35] target/i386: make ldo/sto operations consistent with ldq Paolo Bonzini
2022-10-13 21:46 ` [PATCH 03/35] target/i386: add core of new i386 decoder Paolo Bonzini
2022-10-13 21:46 ` [PATCH 04/35] target/i386: add ALU load/writeback core Paolo Bonzini
2022-10-13 21:46 ` [PATCH 05/35] target/i386: add CPUID[EAX=7, ECX=0].ECX to DisasContext Paolo Bonzini
2022-10-13 21:46 ` [PATCH 06/35] target/i386: add CPUID feature checks to new decoder Paolo Bonzini
2022-10-13 21:46 ` [PATCH 07/35] target/i386: add AVX_EN hflag Paolo Bonzini
2022-10-13 21:46 ` [PATCH 08/35] target/i386: validate VEX prefixes via the instructions' exception classes Paolo Bonzini
2022-10-13 21:46 ` [PATCH 09/35] target/i386: validate SSE prefixes directly in the decoding table Paolo Bonzini
2022-10-13 21:46 ` [PATCH 10/35] target/i386: move scalar 0F 38 and 0F 3A instruction to new decoder Paolo Bonzini
2022-10-13 21:46 ` [PATCH 11/35] target/i386: Prepare ops_sse_header.h for 256 bit AVX Paolo Bonzini
2022-10-13 21:46 ` [PATCH 12/35] target/i386: extend helpers to support VEX.V 3- and 4- operand encodings Paolo Bonzini
2022-10-13 21:46 ` [PATCH 13/35] target/i386: support operand merging in binary scalar helpers Paolo Bonzini
2022-10-13 21:46 ` [PATCH 14/35] target/i386: provide 3-operand versions of unary " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 15/35] target/i386: implement additional AVX comparison operators Paolo Bonzini
2022-10-13 21:46 ` [PATCH 16/35] target/i386: Introduce 256-bit vector helpers Paolo Bonzini
2022-10-13 21:46 ` [PATCH 17/35] target/i386: reimplement 0x0f 0x60-0x6f, add AVX Paolo Bonzini
2022-10-13 21:46 ` [PATCH 18/35] target/i386: reimplement 0x0f 0xd8-0xdf, 0xe8-0xef, 0xf8-0xff, " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 19/35] target/i386: reimplement 0x0f 0x50-0x5f, " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 20/35] target/i386: reimplement 0x0f 0x78-0x7f, " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 21/35] target/i386: reimplement 0x0f 0x70-0x77, " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 22/35] target/i386: reimplement 0x0f 0xd0-0xd7, 0xe0-0xe7, 0xf0-0xf7, " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 23/35] target/i386: clarify (un)signedness of immediates from 0F3Ah opcodes Paolo Bonzini
2022-10-13 21:46 ` [PATCH 24/35] target/i386: reimplement 0x0f 0x3a, add AVX Paolo Bonzini
2022-10-13 21:46 ` [PATCH 25/35] target/i386: Use tcg gvec ops for pmovmskb Paolo Bonzini
2022-10-13 21:46 ` [PATCH 26/35] target/i386: reimplement 0x0f 0x38, add AVX Paolo Bonzini
2022-10-13 21:46 ` [PATCH 27/35] target/i386: reimplement 0x0f 0xc2, 0xc4-0xc6, " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 28/35] target/i386: reimplement 0x0f 0x10-0x17, " Paolo Bonzini
2022-10-13 21:46 ` [PATCH 29/35] target/i386: reimplement 0x0f 0x28-0x2f, " Paolo Bonzini
2022-10-13 21:46 ` Paolo Bonzini [this message]
2022-10-13 21:46 ` [PATCH 31/35] target/i386: implement VLDMXCSR/VSTMXCSR Paolo Bonzini
2022-10-13 21:46 ` [PATCH 32/35] target/i386: Enable AVX cpuid bits when using TCG Paolo Bonzini
2022-10-13 21:46 ` [PATCH 33/35] tests/tcg: extend SSE tests to AVX Paolo Bonzini
2022-10-13 21:46 ` [PATCH 34/35] target/i386: move 3DNow to the new decoder Paolo Bonzini
2022-10-13 21:46 ` [PATCH 35/35] target/i386: remove old SSE decoder Paolo Bonzini
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=20221013214651.672114-31-pbonzini@redhat.com \
--to=pbonzini@redhat.com \
--cc=paul@nowt.org \
--cc=qemu-devel@nongnu.org \
--cc=richard.henderson@linaro.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).