All of lore.kernel.org
 help / color / mirror / Atom feed
From: Huang Tao <eric.huang@linux.alibaba.com>
To: qemu-devel@nongnu.org
Cc: qemu-riscv@nongnu.org, zhiwei_liu@linux.alibaba.com,
	dbarboza@ventanamicro.com, liwei1518@gmail.com,
	bin.meng@windriver.com, alistair.francis@wdc.com,
	palmer@dabbelt.com, Huang Tao <eric.huang@linux.alibaba.com>
Subject: [PATCH 07/65] target/riscv: implement th.vsetvl{i} for XTheadVector
Date: Fri, 12 Apr 2024 15:36:37 +0800	[thread overview]
Message-ID: <20240412073735.76413-8-eric.huang@linux.alibaba.com> (raw)
In-Reply-To: <20240412073735.76413-1-eric.huang@linux.alibaba.com>

In this patch, we implement the th.vetvl{i} instructions. In the th_vsetvl
function, some work has been done according to the difference between RVV1.0
and XTheadVector.

th.vsetvl{i} differs from vsetvl{i} in the following points:
1. th.vsetvl{i} does not have the option to maintain the existing vl.
2. XTheadVector has different vtype encoding from RVV1.0.

Signed-off-by: Huang Tao <eric.huang@linux.alibaba.com>
---
 .../riscv/insn_trans/trans_xtheadvector.c.inc | 93 ++++++++++++++++++-
 1 file changed, 91 insertions(+), 2 deletions(-)

diff --git a/target/riscv/insn_trans/trans_xtheadvector.c.inc b/target/riscv/insn_trans/trans_xtheadvector.c.inc
index 2dd77d74ab..0461b53893 100644
--- a/target/riscv/insn_trans/trans_xtheadvector.c.inc
+++ b/target/riscv/insn_trans/trans_xtheadvector.c.inc
@@ -25,14 +25,103 @@ static bool require_xtheadvector(DisasContext *s)
            s->mstatus_vs != EXT_STATUS_DISABLED;
 }
 
+/*
+ * XTheadVector has different vtype encoding from RVV1.0.
+ * We recode the value in RVV1.0 vtype format to reuse the RVV1.0 functions.
+ * In RVV1.0:
+ *   vtype[7] -> vma
+ *   vtype[6] -> vta
+ *   vtype[5:3] -> vsew
+ *   vtype[2:0] -> vlmul
+ * In XTheadVector:
+ *   vtype[6:5] -> vediv (reserved)
+ *   vtype[4:2] -> vsew
+ *   vtype[1:0] -> vlmul
+ *
+ * Also th_vsetvl does not have feature of keeping existing vl when
+ * (rd == 0 && rs1 == 0)
+ */
+static bool th_vsetvl(DisasContext *s, int rd, int rs1, TCGv s2)
+{
+    TCGv temp = tcg_temp_new();
+    TCGv dst_s2 = tcg_temp_new();
+    /* illegal value check*/
+    TCGLabel *legal = gen_new_label();
+    tcg_gen_shri_tl(temp, s2, 5);
+    tcg_gen_brcondi_tl(TCG_COND_EQ, temp, 0, legal);
+    /*
+     * if illegal, set unsupported value
+     * s2[8:5] == 0b111, which is reserved field in XTheadVector
+     */
+    tcg_gen_movi_tl(s2, 0xff);
+    gen_set_label(legal);
+    /* get vlmul, s2[1:0] -> dst_s2[2:0] */
+    tcg_gen_andi_tl(dst_s2, s2, 0x3);
+    /* get vsew, s2[4:2] -> dst_s2[5:3] */
+    tcg_gen_andi_tl(temp, s2, 0x1c);
+    tcg_gen_shli_tl(temp, temp, 1);
+    tcg_gen_or_tl(dst_s2, dst_s2, temp);
+    /*
+     * get reserved field when illegal, s2[7:5] -> dst_s2[10:8]
+     * avoid dst_s2[7:6], because dst_s2[7:6] are vma and vta.
+     *
+     * Make the dst_s2 an illegal value for RVV1.0, leads to the illegal
+     * operation processing flow.
+     */
+    tcg_gen_andi_tl(temp, s2, 0xe0);
+    tcg_gen_shli_tl(temp, temp, 3);
+    tcg_gen_or_tl(dst_s2, dst_s2, temp);
+
+    /*
+     * We can't reuse do_vsetvl for we don't have ext_zve32f
+     * The code below is almost copied from rvv do_vsetvl
+     * delete zve32f check and the situation when rd = rs1 = 0
+     */
+    TCGv s1, dst;
+    dst = dest_gpr(s, rd);
+    if (rs1 == 0) {
+        /* As the mask is at least one bit, RV_VLEN_MAX is >= VLMAX */
+        s1 = tcg_constant_tl(RV_VLEN_MAX);
+    } else {
+        s1 = get_gpr(s, rs1, EXT_ZERO);
+    }
+
+    gen_helper_vsetvl(dst, tcg_env, s1, dst_s2);
+    gen_set_gpr(s, rd, dst);
+    mark_vs_dirty(s);
+
+    gen_update_pc(s, s->cur_insn_len);
+    lookup_and_goto_ptr(s);
+    s->base.is_jmp = DISAS_NORETURN;
+    return true;
+}
+
+static bool trans_th_vsetvl(DisasContext *s, arg_th_vsetvl *a)
+{
+    if (!require_xtheadvector(s)) {
+        return false;
+    }
+
+    TCGv s2 = get_gpr(s, a->rs2, EXT_ZERO);
+    return th_vsetvl(s, a->rd, a->rs1, s2);
+}
+
+static bool trans_th_vsetvli(DisasContext *s, arg_th_vsetvli *a)
+{
+    if (!require_xtheadvector(s)) {
+        return false;
+    }
+
+    TCGv s2 = tcg_constant_tl(a->zimm);
+    return th_vsetvl(s, a->rd, a->rs1, s2);
+}
+
 #define TH_TRANS_STUB(NAME)                                \
 static bool trans_##NAME(DisasContext *s, arg_##NAME *a)   \
 {                                                          \
     return require_xtheadvector(s);                        \
 }
 
-TH_TRANS_STUB(th_vsetvli)
-TH_TRANS_STUB(th_vsetvl)
 TH_TRANS_STUB(th_vlb_v)
 TH_TRANS_STUB(th_vlh_v)
 TH_TRANS_STUB(th_vlw_v)
-- 
2.44.0



  parent reply	other threads:[~2024-04-12  7:53 UTC|newest]

Thread overview: 66+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-04-12  7:36 [PATCH 00/65]target/riscv: Support XTheadVector extension Huang Tao
2024-04-12  7:36 ` [PATCH 01/65] riscv: thead: Add th.sxstatus CSR emulation Huang Tao
2024-04-12  7:36 ` [PATCH 02/65] target/riscv: Reuse th_csr.c to add user-mode csrs Huang Tao
2024-04-12  7:36 ` [PATCH 03/65] target/riscv: Add properties for XTheadVector extension Huang Tao
2024-04-12  7:36 ` [PATCH 04/65] target/riscv: Override some csr ops for XTheadVector Huang Tao
2024-04-12  7:36 ` [PATCH 05/65] target/riscv: Add mlen in DisasContext Huang Tao
2024-04-12  7:36 ` [PATCH 06/65] target/riscv: Implement insns decode rules for XTheadVector Huang Tao
2024-04-12  7:36 ` Huang Tao [this message]
2024-04-12  7:36 ` [PATCH 08/65] target/riscv: Add strided load instructions " Huang Tao
2024-04-12  7:36 ` [PATCH 09/65] target/riscv: Add strided store " Huang Tao
2024-04-12  7:36 ` [PATCH 10/65] target/riscv: Add unit-stride load " Huang Tao
2024-04-12  7:36 ` [PATCH 11/65] target/riscv: Add unit-stride store " Huang Tao
2024-04-12  7:36 ` [PATCH 12/65] target/riscv: Add indexed load " Huang Tao
2024-04-12  7:36 ` [PATCH 13/65] target/riscv: Add indexed store " Huang Tao
2024-04-12  7:36 ` [PATCH 14/65] target/riscv: Add unit-stride fault-only-first " Huang Tao
2024-04-12  7:36 ` [PATCH 15/65] target/riscv: Add vector amo operations " Huang Tao
2024-04-12  7:36 ` [PATCH 16/65] target/riscv: Add single-width integer add and subtract instructions " Huang Tao
2024-04-12  7:36 ` [PATCH 17/65] target/riscv: Add widening integer add/subtract " Huang Tao
2024-04-12  7:36 ` [PATCH 18/65] target/riscv: Add integer add-with-carry/sub-with-borrow " Huang Tao
2024-04-12  7:36 ` [PATCH 19/65] target/riscv: Add bitwise logical " Huang Tao
2024-04-12  7:36 ` [PATCH 20/65] target/riscv: Add single-width bit shift " Huang Tao
2024-04-12  7:36 ` [PATCH 21/65] target/riscv: Add narrowing integer right " Huang Tao
2024-04-12  7:36 ` [PATCH 22/65] target/riscv: Add integer compare " Huang Tao
2024-04-12  7:36 ` [PATCH 23/65] target/riscv: Add integer min/max " Huang Tao
2024-04-12  7:36 ` [PATCH 24/65] target/riscv: Add single-width integer multiply " Huang Tao
2024-04-12  7:36 ` [PATCH 25/65] target/riscv: Add integer divide " Huang Tao
2024-04-12  7:36 ` [PATCH 26/65] target/riscv: Add widening integer multiply " Huang Tao
2024-04-12  7:36 ` [PATCH 27/65] target/riscv: Add single-width integer multiply-add " Huang Tao
2024-04-12  7:36 ` [PATCH 28/65] target/riscv: Add widening " Huang Tao
2024-04-12  7:36 ` [PATCH 29/65] target/riscv: Add integer merge and move " Huang Tao
2024-04-12  7:37 ` [PATCH 30/65] target/riscv: Add single-width saturating add and sub " Huang Tao
2024-04-12  7:37 ` [PATCH 31/65] target/riscv: Add single-width average " Huang Tao
2024-04-12  7:37 ` [PATCH 32/65] target/riscv: Add single-width fractional mul with rounding and saturation " Huang Tao
2024-04-12  7:37 ` [PATCH 33/65] target/riscv: Add widening saturating scaled multiply-add instructions " Huang Tao
2024-04-12  7:37 ` [PATCH 34/65] target/riscv: Add single-width scaling shift " Huang Tao
2024-04-12  7:37 ` [PATCH 35/65] target/riscv: Add narrowing fixed-point clip " Huang Tao
2024-04-12  7:37 ` [PATCH 36/65] target/riscv: Add single-width floating-point add/sub " Huang Tao
2024-04-12  7:37 ` [PATCH 37/65] target/riscv: Add widening " Huang Tao
2024-04-12  7:37 ` [PATCH 38/65] target/riscv: Add single-width floating-point multiply/divide " Huang Tao
2024-04-12  7:37 ` [PATCH 39/65] target/riscv: Add widening floating-point multiply " Huang Tao
2024-04-12  7:37 ` [PATCH 40/65] target/riscv: Add single-width floating-point fused multiply-add " Huang Tao
2024-04-12  7:37 ` [PATCH 41/65] target/riscv: Add widening floating-point fused mul-add " Huang Tao
2024-04-12  7:37 ` [PATCH 42/65] target/riscv: Add floating-pointing square-root " Huang Tao
2024-04-12  7:37 ` [PATCH 43/65] target/riscv: Add floating-point MIN/MAX " Huang Tao
2024-04-12  7:37 ` [PATCH 44/65] target/riscv: Add floating-point sign-injection " Huang Tao
2024-04-12  7:37 ` [PATCH 45/65] target/riscv: Add floating-point compare " Huang Tao
2024-04-12  7:37 ` [PATCH 46/65] target/riscv: Add floating-point classify and merge " Huang Tao
2024-04-12  7:37 ` [PATCH 47/65] target/riscv: Add single-width floating-point/integer type-convert " Huang Tao
2024-04-12  7:37 ` [PATCH 48/65] target/riscv: Add widening " Huang Tao
2024-04-12  7:37 ` [PATCH 49/65] target/riscv: Add narrowing " Huang Tao
2024-04-12  7:37 ` [PATCH 50/65] target/riscv: Add single-width integer reduction " Huang Tao
2024-04-12  7:37 ` [PATCH 51/65] target/riscv: Add widening " Huang Tao
2024-04-12  7:37 ` [PATCH 52/65] target/riscv: Add single-width floating-point " Huang Tao
2024-04-12  7:37 ` [PATCH 53/65] target/riscv: Add widening " Huang Tao
2024-04-12  7:37 ` [PATCH 54/65] target/riscv: Add mask-register logical " Huang Tao
2024-04-12  7:37 ` [PATCH 55/65] target/riscv: Add vector mask population count vmpopc " Huang Tao
2024-04-12  7:37 ` [PATCH 56/65] target/riscv: Add th.vmfirst.m " Huang Tao
2024-04-12  7:37 ` [PATCH 57/65] target/riscv: Add set-X-first mask bit instructrions " Huang Tao
2024-04-12  7:37 ` [PATCH 58/65] target/riscv: Add vector iota instruction " Huang Tao
2024-04-12  7:37 ` [PATCH 59/65] target/riscv: Add vector element index " Huang Tao
2024-04-12  7:37 ` [PATCH 60/65] target/riscv: Add integer extract and scalar move instructions " Huang Tao
2024-04-12  7:37 ` [PATCH 61/65] target/riscv: Add floating-point " Huang Tao
2024-04-12  7:37 ` [PATCH 62/65] target/riscv: Add vector slide " Huang Tao
2024-04-12  7:37 ` [PATCH 63/65] target/riscv: Add vector register gather " Huang Tao
2024-04-12  7:37 ` [PATCH 64/65] target/riscv: Add vector compress instruction " Huang Tao
2024-04-12  7:37 ` [PATCH 65/65] target/riscv: Enable XTheadVector extension for c906 Huang Tao

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=20240412073735.76413-8-eric.huang@linux.alibaba.com \
    --to=eric.huang@linux.alibaba.com \
    --cc=alistair.francis@wdc.com \
    --cc=bin.meng@windriver.com \
    --cc=dbarboza@ventanamicro.com \
    --cc=liwei1518@gmail.com \
    --cc=palmer@dabbelt.com \
    --cc=qemu-devel@nongnu.org \
    --cc=qemu-riscv@nongnu.org \
    --cc=zhiwei_liu@linux.alibaba.com \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.