From: Peter Maydell <peter.maydell@linaro.org>
To: qemu-arm@nongnu.org, qemu-devel@nongnu.org
Cc: Richard Henderson <richard.henderson@linaro.org>
Subject: [PATCH v2 49/57] target/arm: Implement MVE VQDMLADH and VQRDMLADH
Date: Mon, 14 Jun 2021 16:09:59 +0100 [thread overview]
Message-ID: <20210614151007.4545-50-peter.maydell@linaro.org> (raw)
In-Reply-To: <20210614151007.4545-1-peter.maydell@linaro.org>
Implement the MVE VQDMLADH and VQRDMLADH insns. These multiply
elements, and then add pairs of products, double, possibly round,
saturate and return the high half of the result.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
target/arm/helper-mve.h | 16 +++++++
target/arm/mve.decode | 5 +++
target/arm/mve_helper.c | 89 ++++++++++++++++++++++++++++++++++++++
target/arm/translate-mve.c | 4 ++
4 files changed, 114 insertions(+)
diff --git a/target/arm/helper-mve.h b/target/arm/helper-mve.h
index b7e2243a19a..c3cc6a08476 100644
--- a/target/arm/helper-mve.h
+++ b/target/arm/helper-mve.h
@@ -201,6 +201,22 @@ DEF_HELPER_FLAGS_4(mve_vqrshlub, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
DEF_HELPER_FLAGS_4(mve_vqrshluh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
DEF_HELPER_FLAGS_4(mve_vqrshluw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmladhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmladhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmladhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
+DEF_HELPER_FLAGS_4(mve_vqdmladhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmladhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqdmladhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
+DEF_HELPER_FLAGS_4(mve_vqrdmladhb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmladhh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmladhw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
+DEF_HELPER_FLAGS_4(mve_vqrdmladhxb, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmladhxh, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+DEF_HELPER_FLAGS_4(mve_vqrdmladhxw, TCG_CALL_NO_WG, void, env, ptr, ptr, ptr)
+
DEF_HELPER_FLAGS_4(mve_vadd_scalarb, TCG_CALL_NO_WG, void, env, ptr, ptr, i32)
DEF_HELPER_FLAGS_4(mve_vadd_scalarh, TCG_CALL_NO_WG, void, env, ptr, ptr, i32)
DEF_HELPER_FLAGS_4(mve_vadd_scalarw, TCG_CALL_NO_WG, void, env, ptr, ptr, i32)
diff --git a/target/arm/mve.decode b/target/arm/mve.decode
index c30fb2c1536..d267c8838eb 100644
--- a/target/arm/mve.decode
+++ b/target/arm/mve.decode
@@ -142,6 +142,11 @@ VQSHL_U 111 1 1111 0 . .. ... 0 ... 0 0100 . 1 . 1 ... 0 @2op_rev
VQRSHL_S 111 0 1111 0 . .. ... 0 ... 0 0101 . 1 . 1 ... 0 @2op_rev
VQRSHL_U 111 1 1111 0 . .. ... 0 ... 0 0101 . 1 . 1 ... 0 @2op_rev
+VQDMLADH 1110 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 0 @2op
+VQDMLADHX 1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 0 @2op
+VQRDMLADH 1110 1110 0 . .. ... 0 ... 0 1110 . 0 . 0 ... 1 @2op
+VQRDMLADHX 1110 1110 0 . .. ... 0 ... 1 1110 . 0 . 0 ... 1 @2op
+
# Vector miscellaneous
VCLS 1111 1111 1 . 11 .. 00 ... 0 0100 01 . 0 ... 0 @1op
diff --git a/target/arm/mve_helper.c b/target/arm/mve_helper.c
index 7f0eb207155..1310d45f5f5 100644
--- a/target/arm/mve_helper.c
+++ b/target/arm/mve_helper.c
@@ -642,6 +642,95 @@ DO_2OP_SAT_U(vqshlu, DO_UQSHL_OP)
DO_2OP_SAT_S(vqrshls, DO_SQRSHL_OP)
DO_2OP_SAT_U(vqrshlu, DO_UQRSHL_OP)
+/*
+ * Multiply add dual returning high half
+ * The 'FN' here takes four inputs A, B, C, D, a 0/1 indicator of
+ * whether to add the rounding constant, and the pointer to the
+ * saturation flag, and should do "(A * B + C * D) * 2 + rounding constant",
+ * saturate to twice the input size and return the high half; or
+ * (A * B - C * D) etc for VQDMLSDH.
+ */
+#define DO_VQDMLADH_OP(OP, ESIZE, TYPE, XCHG, ROUND, FN) \
+ void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \
+ void *vm) \
+ { \
+ TYPE *d = vd, *n = vn, *m = vm; \
+ uint16_t mask = mve_element_mask(env); \
+ unsigned e; \
+ bool qc = false; \
+ for (e = 0; e < 16 / ESIZE; e++, mask >>= ESIZE) { \
+ bool sat = false; \
+ if ((e & 1) == XCHG) { \
+ TYPE r = FN(n[H##ESIZE(e)], \
+ m[H##ESIZE(e - XCHG)], \
+ n[H##ESIZE(e + (1 - 2 * XCHG))], \
+ m[H##ESIZE(e + (1 - XCHG))], \
+ ROUND, &sat); \
+ mergemask(&d[H##ESIZE(e)], r, mask); \
+ qc |= sat & mask & 1; \
+ } \
+ } \
+ if (qc) { \
+ env->vfp.qc[0] = qc; \
+ } \
+ mve_advance_vpt(env); \
+ }
+
+static int8_t do_vqdmladh_b(int8_t a, int8_t b, int8_t c, int8_t d,
+ int round, bool *sat)
+{
+ int64_t r = ((int64_t)a * b + (int64_t)c * d) * 2 + (round << 7);
+ return do_sat_bhw(r, INT16_MIN, INT16_MAX, sat) >> 8;
+}
+
+static int16_t do_vqdmladh_h(int16_t a, int16_t b, int16_t c, int16_t d,
+ int round, bool *sat)
+{
+ int64_t r = ((int64_t)a * b + (int64_t)c * d) * 2 + (round << 15);
+ return do_sat_bhw(r, INT32_MIN, INT32_MAX, sat) >> 16;
+}
+
+static int32_t do_vqdmladh_w(int32_t a, int32_t b, int32_t c, int32_t d,
+ int round, bool *sat)
+{
+ int64_t m1 = (int64_t)a * b;
+ int64_t m2 = (int64_t)c * d;
+ int64_t r;
+ /*
+ * Architecturally we should do the entire add, double, round
+ * and then check for saturation. We do three saturating adds,
+ * but we need to be careful about the order. If the first
+ * m1 + m2 saturates then it's impossible for the *2+rc to
+ * bring it back into the non-saturated range. However, if
+ * m1 + m2 is negative then it's possible that doing the doubling
+ * would take the intermediate result below INT64_MAX and the
+ * addition of the rounding constant then brings it back in range.
+ * So we add half the rounding constant before doubling rather
+ * than adding the rounding constant after the doubling.
+ */
+ if (sadd64_overflow(m1, m2, &r) ||
+ sadd64_overflow(r, (round << 30), &r) ||
+ sadd64_overflow(r, r, &r)) {
+ *sat = true;
+ return r < 0 ? INT32_MAX : INT32_MIN;
+ }
+ return r >> 32;
+}
+
+DO_VQDMLADH_OP(vqdmladhb, 1, int8_t, 0, 0, do_vqdmladh_b)
+DO_VQDMLADH_OP(vqdmladhh, 2, int16_t, 0, 0, do_vqdmladh_h)
+DO_VQDMLADH_OP(vqdmladhw, 4, int32_t, 0, 0, do_vqdmladh_w)
+DO_VQDMLADH_OP(vqdmladhxb, 1, int8_t, 1, 0, do_vqdmladh_b)
+DO_VQDMLADH_OP(vqdmladhxh, 2, int16_t, 1, 0, do_vqdmladh_h)
+DO_VQDMLADH_OP(vqdmladhxw, 4, int32_t, 1, 0, do_vqdmladh_w)
+
+DO_VQDMLADH_OP(vqrdmladhb, 1, int8_t, 0, 1, do_vqdmladh_b)
+DO_VQDMLADH_OP(vqrdmladhh, 2, int16_t, 0, 1, do_vqdmladh_h)
+DO_VQDMLADH_OP(vqrdmladhw, 4, int32_t, 0, 1, do_vqdmladh_w)
+DO_VQDMLADH_OP(vqrdmladhxb, 1, int8_t, 1, 1, do_vqdmladh_b)
+DO_VQDMLADH_OP(vqrdmladhxh, 2, int16_t, 1, 1, do_vqdmladh_h)
+DO_VQDMLADH_OP(vqrdmladhxw, 4, int32_t, 1, 1, do_vqdmladh_w)
+
#define DO_2OP_SCALAR(OP, ESIZE, TYPE, FN) \
void HELPER(glue(mve_, OP))(CPUARMState *env, void *vd, void *vn, \
uint32_t rm) \
diff --git a/target/arm/translate-mve.c b/target/arm/translate-mve.c
index f6ed15b4aca..78fb61e83ca 100644
--- a/target/arm/translate-mve.c
+++ b/target/arm/translate-mve.c
@@ -410,6 +410,10 @@ DO_2OP(VQSHL_S, vqshls)
DO_2OP(VQSHL_U, vqshlu)
DO_2OP(VQRSHL_S, vqrshls)
DO_2OP(VQRSHL_U, vqrshlu)
+DO_2OP(VQDMLADH, vqdmladh)
+DO_2OP(VQDMLADHX, vqdmladhx)
+DO_2OP(VQRDMLADH, vqrdmladh)
+DO_2OP(VQRDMLADHX, vqrdmladhx)
static bool do_2op_scalar(DisasContext *s, arg_2scalar *a,
MVEGenTwoOpScalarFn fn)
--
2.20.1
next prev parent reply other threads:[~2021-06-14 15:44 UTC|newest]
Thread overview: 77+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-14 15:09 [PATCH v2 00/57] target/arm: First slice of MVE implementation Peter Maydell
2021-06-14 15:09 ` [PATCH v2 01/57] target/arm: Provide and use H8 and H1_8 macros Peter Maydell
2021-06-14 15:09 ` [PATCH v2 02/57] target/arm: Enable FPSCR.QC bit for MVE Peter Maydell
2021-06-14 15:09 ` [PATCH v2 03/57] target/arm: Handle VPR semantics in existing code Peter Maydell
2021-06-14 15:09 ` [PATCH v2 04/57] target/arm: Add handling for PSR.ECI/ICI Peter Maydell
2021-06-14 19:15 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 05/57] target/arm: Let vfp_access_check() handle late NOCP checks Peter Maydell
2021-06-14 15:09 ` [PATCH v2 06/57] target/arm: Implement MVE LCTP Peter Maydell
2021-06-14 15:09 ` [PATCH v2 07/57] target/arm: Implement MVE WLSTP insn Peter Maydell
2021-06-14 19:20 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 08/57] target/arm: Implement MVE DLSTP Peter Maydell
2021-06-14 15:09 ` [PATCH v2 09/57] target/arm: Implement MVE LETP insn Peter Maydell
2021-06-14 15:09 ` [PATCH v2 10/57] target/arm: Add framework for MVE decode Peter Maydell
2021-06-14 15:09 ` [PATCH v2 11/57] target/arm: Implement MVE VLDR/VSTR (non-widening forms) Peter Maydell
2021-06-14 19:29 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 12/57] target/arm: Implement widening/narrowing MVE VLDR/VSTR insns Peter Maydell
2021-06-14 19:39 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 13/57] target/arm: Move expand_pred_b() data to translate.c Peter Maydell
2021-06-14 19:41 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 14/57] target/arm: Implement MVE VCLZ Peter Maydell
2021-06-14 20:00 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 15/57] target/arm: Implement MVE VCLS Peter Maydell
2021-06-14 15:09 ` [PATCH v2 16/57] bitops.h: Provide hswap32(), hswap64(), wswap64() swapping operations Peter Maydell
2021-06-14 15:09 ` [PATCH v2 17/57] target/arm: Implement MVE VREV16, VREV32, VREV64 Peter Maydell
2021-06-14 20:03 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 18/57] target/arm: Implement MVE VMVN (register) Peter Maydell
2021-06-14 15:09 ` [PATCH v2 19/57] target/arm: Implement MVE VABS Peter Maydell
2021-06-14 20:06 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 20/57] target/arm: Implement MVE VNEG Peter Maydell
2021-06-14 15:09 ` [PATCH v2 21/57] tcg: Make gen_dup_i32() public Peter Maydell
2021-06-14 20:12 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 22/57] target/arm: Implement MVE VDUP Peter Maydell
2021-06-14 20:15 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 23/57] target/arm: Implement MVE VAND, VBIC, VORR, VORN, VEOR Peter Maydell
2021-06-14 15:09 ` [PATCH v2 24/57] target/arm: Implement MVE VADD, VSUB, VMUL Peter Maydell
2021-06-14 15:09 ` [PATCH v2 25/57] target/arm: Implement MVE VMULH Peter Maydell
2021-06-14 15:09 ` [PATCH v2 26/57] target/arm: Implement MVE VRMULH Peter Maydell
2021-06-14 15:09 ` [PATCH v2 27/57] target/arm: Implement MVE VMAX, VMIN Peter Maydell
2021-06-14 15:09 ` [PATCH v2 28/57] target/arm: Implement MVE VABD Peter Maydell
2021-06-14 15:09 ` [PATCH v2 29/57] target/arm: Implement MVE VHADD, VHSUB Peter Maydell
2021-06-14 15:09 ` [PATCH v2 30/57] target/arm: Implement MVE VMULL Peter Maydell
2021-06-14 15:09 ` [PATCH v2 31/57] target/arm: Implement MVE VMLALDAV Peter Maydell
2021-06-14 15:09 ` [PATCH v2 32/57] target/arm: Implement MVE VMLSLDAV Peter Maydell
2021-06-14 15:09 ` [PATCH v2 33/57] include/qemu/int128.h: Add function to create Int128 from int64_t Peter Maydell
2021-06-14 15:09 ` [PATCH v2 34/57] target/arm: Implement MVE VRMLALDAVH, VRMLSLDAVH Peter Maydell
2021-06-14 21:27 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 35/57] target/arm: Implement MVE VADD (scalar) Peter Maydell
2021-06-14 21:30 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 36/57] target/arm: Implement MVE VSUB, VMUL (scalar) Peter Maydell
2021-06-14 15:09 ` [PATCH v2 37/57] target/arm: Implement MVE VHADD, VHSUB (scalar) Peter Maydell
2021-06-14 15:09 ` [PATCH v2 38/57] target/arm: Implement MVE VBRSR Peter Maydell
2021-06-14 15:09 ` [PATCH v2 39/57] target/arm: Implement MVE VPST Peter Maydell
2021-06-14 15:09 ` [PATCH v2 40/57] target/arm: Implement MVE VQADD and VQSUB Peter Maydell
2021-06-14 15:09 ` [PATCH v2 41/57] target/arm: Implement MVE VQDMULH and VQRDMULH (scalar) Peter Maydell
2021-06-14 15:09 ` [PATCH v2 42/57] target/arm: Implement MVE VQDMULL scalar Peter Maydell
2021-06-14 15:09 ` [PATCH v2 43/57] target/arm: Implement MVE VQDMULH, VQRDMULH (vector) Peter Maydell
2021-06-14 15:09 ` [PATCH v2 44/57] target/arm: Implement MVE VQADD, VQSUB (vector) Peter Maydell
2021-06-14 15:09 ` [PATCH v2 45/57] target/arm: Implement MVE VQSHL (vector) Peter Maydell
2021-06-14 21:43 ` Richard Henderson
2021-06-14 15:09 ` [PATCH v2 46/57] target/arm: Implement MVE VQRSHL Peter Maydell
2021-06-14 15:09 ` [PATCH v2 47/57] target/arm: Implement MVE VSHL insn Peter Maydell
2021-06-14 15:09 ` [PATCH v2 48/57] target/arm: Implmement MVE VRSHL Peter Maydell
2021-06-14 15:09 ` Peter Maydell [this message]
2021-06-14 15:10 ` [PATCH v2 50/57] target/arm: Implement MVE VQDMLSDH and VQRDMLSDH Peter Maydell
2021-06-14 15:10 ` [PATCH v2 51/57] target/arm: Implement MVE VQDMULL (vector) Peter Maydell
2021-06-14 15:10 ` [PATCH v2 52/57] target/arm: Implement MVE VRHADD Peter Maydell
2021-06-14 15:10 ` [PATCH v2 53/57] target/arm: Implement MVE VADC, VSBC Peter Maydell
2021-06-14 21:58 ` Richard Henderson
2021-06-14 15:10 ` [PATCH v2 54/57] target/arm: Implement MVE VCADD Peter Maydell
2021-06-14 15:10 ` [PATCH v2 55/57] target/arm: Implement MVE VHCADD Peter Maydell
2021-06-14 22:01 ` Richard Henderson
2021-06-14 15:10 ` [PATCH v2 56/57] target/arm: Implement MVE VADDV Peter Maydell
2021-06-14 15:10 ` [PATCH v2 57/57] target/arm: Make VMOV scalar <-> gpreg beatwise for MVE Peter Maydell
2021-06-14 22:20 ` [PATCH v2 00/57] target/arm: First slice of MVE implementation no-reply
2021-06-14 22:22 ` Richard Henderson
2021-06-21 16:37 ` Peter Maydell
2021-06-21 17:13 ` Richard Henderson
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=20210614151007.4545-50-peter.maydell@linaro.org \
--to=peter.maydell@linaro.org \
--cc=qemu-arm@nongnu.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).