From: frank.chang@sifive.com
To: qemu-devel@nongnu.org, qemu-riscv@nongnu.org
Cc: Sagar Karandikar <sagark@eecs.berkeley.edu>,
Frank Chang <frank.chang@sifive.com>,
Bastian Koppelmann <kbastian@mail.uni-paderborn.de>,
Richard Henderson <richard.henderson@linaro.org>,
Alistair Francis <Alistair.Francis@wdc.com>,
Palmer Dabbelt <palmer@dabbelt.com>,
LIU Zhiwei <zhiwei_liu@c-sky.com>
Subject: [PATCH v7 69/75] target/riscv: rvv-1.0: floating-point reciprocal square-root estimate instruction
Date: Fri, 26 Feb 2021 11:18:53 +0800 [thread overview]
Message-ID: <20210226031902.23656-70-frank.chang@sifive.com> (raw)
In-Reply-To: <20210226031902.23656-1-frank.chang@sifive.com>
From: Frank Chang <frank.chang@sifive.com>
Implement the floating-point reciprocal square-root estimate to 7 bits
instruction.
Signed-off-by: Frank Chang <frank.chang@sifive.com>
---
target/riscv/helper.h | 4 +
target/riscv/insn32.decode | 1 +
target/riscv/insn_trans/trans_rvv.c.inc | 1 +
target/riscv/vector_helper.c | 183 ++++++++++++++++++++++++
4 files changed, 189 insertions(+)
diff --git a/target/riscv/helper.h b/target/riscv/helper.h
index 78f330b949e..1be95ebd0aa 100644
--- a/target/riscv/helper.h
+++ b/target/riscv/helper.h
@@ -882,6 +882,10 @@ DEF_HELPER_5(vfsqrt_v_h, void, ptr, ptr, ptr, env, i32)
DEF_HELPER_5(vfsqrt_v_w, void, ptr, ptr, ptr, env, i32)
DEF_HELPER_5(vfsqrt_v_d, void, ptr, ptr, ptr, env, i32)
+DEF_HELPER_5(vfrsqrt7_v_h, void, ptr, ptr, ptr, env, i32)
+DEF_HELPER_5(vfrsqrt7_v_w, void, ptr, ptr, ptr, env, i32)
+DEF_HELPER_5(vfrsqrt7_v_d, void, ptr, ptr, ptr, env, i32)
+
DEF_HELPER_6(vfmin_vv_h, void, ptr, ptr, ptr, ptr, env, i32)
DEF_HELPER_6(vfmin_vv_w, void, ptr, ptr, ptr, ptr, env, i32)
DEF_HELPER_6(vfmin_vv_d, void, ptr, ptr, ptr, ptr, env, i32)
diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
index f670c1282d8..c1e896f6b84 100644
--- a/target/riscv/insn32.decode
+++ b/target/riscv/insn32.decode
@@ -535,6 +535,7 @@ vfwmsac_vf 111110 . ..... ..... 101 ..... 1010111 @r_vm
vfwnmsac_vv 111111 . ..... ..... 001 ..... 1010111 @r_vm
vfwnmsac_vf 111111 . ..... ..... 101 ..... 1010111 @r_vm
vfsqrt_v 010011 . ..... 00000 001 ..... 1010111 @r2_vm
+vfrsqrt7_v 010011 . ..... 00100 001 ..... 1010111 @r2_vm
vfmin_vv 000100 . ..... ..... 001 ..... 1010111 @r_vm
vfmin_vf 000100 . ..... ..... 101 ..... 1010111 @r_vm
vfmax_vv 000110 . ..... ..... 001 ..... 1010111 @r_vm
diff --git a/target/riscv/insn_trans/trans_rvv.c.inc b/target/riscv/insn_trans/trans_rvv.c.inc
index 2a3a437168f..ca290469e23 100644
--- a/target/riscv/insn_trans/trans_rvv.c.inc
+++ b/target/riscv/insn_trans/trans_rvv.c.inc
@@ -2638,6 +2638,7 @@ static bool trans_##NAME(DisasContext *s, arg_rmr *a) \
}
GEN_OPFV_TRANS(vfsqrt_v, opfv_check, RISCV_FRM_DYN)
+GEN_OPFV_TRANS(vfrsqrt7_v, opfv_check, RISCV_FRM_DYN)
/* Vector Floating-Point MIN/MAX Instructions */
GEN_OPFVV_TRANS(vfmin_vv, opfvv_check)
diff --git a/target/riscv/vector_helper.c b/target/riscv/vector_helper.c
index efa32565d9a..471b07f4388 100644
--- a/target/riscv/vector_helper.c
+++ b/target/riscv/vector_helper.c
@@ -18,6 +18,7 @@
#include "qemu/osdep.h"
#include "qemu/host-utils.h"
+#include "qemu/bitops.h"
#include "cpu.h"
#include "exec/memop.h"
#include "exec/exec-all.h"
@@ -3638,6 +3639,188 @@ GEN_VEXT_V_ENV(vfsqrt_v_h, 2, 2)
GEN_VEXT_V_ENV(vfsqrt_v_w, 4, 4)
GEN_VEXT_V_ENV(vfsqrt_v_d, 8, 8)
+/*
+ * Vector Floating-Point Reciprocal Square-Root Estimate Instruction
+ *
+ * Adapted from riscv-v-spec recip.cc:
+ * https://github.com/riscv/riscv-v-spec/blob/master/recip.cc
+ */
+static uint64_t frsqrt7(uint64_t f, int exp_size, int frac_size)
+{
+ bool sign = extract64(f, frac_size + exp_size, 1);
+ uint64_t exp = extract64(f, frac_size, exp_size);
+ uint64_t frac = extract64(f, 0, frac_size);
+
+ const uint8_t lookup_table[] = {
+ 52, 51, 50, 48, 47, 46, 44, 43,
+ 42, 41, 40, 39, 38, 36, 35, 34,
+ 33, 32, 31, 30, 30, 29, 28, 27,
+ 26, 25, 24, 23, 23, 22, 21, 20,
+ 19, 19, 18, 17, 16, 16, 15, 14,
+ 14, 13, 12, 12, 11, 10, 10, 9,
+ 9, 8, 7, 7, 6, 6, 5, 4,
+ 4, 3, 3, 2, 2, 1, 1, 0,
+ 127, 125, 123, 121, 119, 118, 116, 114,
+ 113, 111, 109, 108, 106, 105, 103, 102,
+ 100, 99, 97, 96, 95, 93, 92, 91,
+ 90, 88, 87, 86, 85, 84, 83, 82,
+ 80, 79, 78, 77, 76, 75, 74, 73,
+ 72, 71, 70, 70, 69, 68, 67, 66,
+ 65, 64, 63, 63, 62, 61, 60, 59,
+ 59, 58, 57, 56, 56, 55, 54, 53
+ };
+ const int precision = 7;
+
+ if (exp == 0 && frac != 0) { /* subnormal */
+ /* Normalize the subnormal. */
+ while (extract64(frac, frac_size - 1, 1) == 0) {
+ exp--;
+ frac <<= 1;
+ }
+
+ frac = (frac << 1) & MAKE_64BIT_MASK(0, frac_size);
+ }
+
+ int idx = ((exp & 1) << (precision - 1)) |
+ (frac >> (frac_size - precision + 1));
+ uint64_t out_frac = (uint64_t)(lookup_table[idx]) <<
+ (frac_size - precision);
+ uint64_t out_exp = (3 * MAKE_64BIT_MASK(0, exp_size - 1) + ~exp) / 2;
+
+ uint64_t val = 0;
+ val = deposit64(val, 0, frac_size, out_frac);
+ val = deposit64(val, frac_size, exp_size, out_exp);
+ val = deposit64(val, frac_size + exp_size, 1, sign);
+ return val;
+}
+
+static float16 frsqrt7_h(float16 f, float_status *s)
+{
+ int exp_size = 5, frac_size = 10;
+ bool sign = float16_is_neg(f);
+
+ /*
+ * frsqrt7(sNaN) = canonical NaN
+ * frsqrt7(-inf) = canonical NaN
+ * frsqrt7(-normal) = canonical NaN
+ * frsqrt7(-subnormal) = canonical NaN
+ */
+ if (float16_is_signaling_nan(f, s) ||
+ (float16_is_infinity(f) && sign) ||
+ (float16_is_normal(f) && sign) ||
+ (float16_is_zero_or_denormal(f) && !float16_is_zero(f) && sign)) {
+ s->float_exception_flags |= float_flag_invalid;
+ return float16_default_nan(s);
+ }
+
+ /* frsqrt7(qNaN) = canonical NaN */
+ if (float16_is_quiet_nan(f, s)) {
+ return float16_default_nan(s);
+ }
+
+ /* frsqrt7(+-0) = +-inf */
+ if (float16_is_zero(f)) {
+ s->float_exception_flags |= float_flag_divbyzero;
+ return float16_set_sign(float16_infinity, sign);
+ }
+
+ /* frsqrt7(+inf) = +0 */
+ if (float16_is_infinity(f) && !sign) {
+ return float16_set_sign(float16_zero, sign);
+ }
+
+ /* +normal, +subnormal */
+ uint64_t val = frsqrt7(f, exp_size, frac_size);
+ return make_float16(val);
+}
+
+static float32 frsqrt7_s(float32 f, float_status *s)
+{
+ int exp_size = 8, frac_size = 23;
+ bool sign = float32_is_neg(f);
+
+ /*
+ * frsqrt7(sNaN) = canonical NaN
+ * frsqrt7(-inf) = canonical NaN
+ * frsqrt7(-normal) = canonical NaN
+ * frsqrt7(-subnormal) = canonical NaN
+ */
+ if (float32_is_signaling_nan(f, s) ||
+ (float32_is_infinity(f) && sign) ||
+ (float32_is_normal(f) && sign) ||
+ (float32_is_zero_or_denormal(f) && !float32_is_zero(f) && sign)) {
+ s->float_exception_flags |= float_flag_invalid;
+ return float32_default_nan(s);
+ }
+
+ /* frsqrt7(qNaN) = canonical NaN */
+ if (float32_is_quiet_nan(f, s)) {
+ return float32_default_nan(s);
+ }
+
+ /* frsqrt7(+-0) = +-inf */
+ if (float32_is_zero(f)) {
+ s->float_exception_flags |= float_flag_divbyzero;
+ return float32_set_sign(float32_infinity, sign);
+ }
+
+ /* frsqrt7(+inf) = +0 */
+ if (float32_is_infinity(f) && !sign) {
+ return float32_set_sign(float32_zero, sign);
+ }
+
+ /* +normal, +subnormal */
+ uint64_t val = frsqrt7(f, exp_size, frac_size);
+ return make_float32(val);
+}
+
+static float64 frsqrt7_d(float64 f, float_status *s)
+{
+ int exp_size = 11, frac_size = 52;
+ bool sign = float64_is_neg(f);
+
+ /*
+ * frsqrt7(sNaN) = canonical NaN
+ * frsqrt7(-inf) = canonical NaN
+ * frsqrt7(-normal) = canonical NaN
+ * frsqrt7(-subnormal) = canonical NaN
+ */
+ if (float64_is_signaling_nan(f, s) ||
+ (float64_is_infinity(f) && sign) ||
+ (float64_is_normal(f) && sign) ||
+ (float64_is_zero_or_denormal(f) && !float64_is_zero(f) && sign)) {
+ s->float_exception_flags |= float_flag_invalid;
+ return float64_default_nan(s);
+ }
+
+ /* frsqrt7(qNaN) = canonical NaN */
+ if (float64_is_quiet_nan(f, s)) {
+ return float64_default_nan(s);
+ }
+
+ /* frsqrt7(+-0) = +-inf */
+ if (float64_is_zero(f)) {
+ s->float_exception_flags |= float_flag_divbyzero;
+ return float64_set_sign(float64_infinity, sign);
+ }
+
+ /* frsqrt7(+inf) = +0 */
+ if (float64_is_infinity(f) && !sign) {
+ return float64_set_sign(float64_zero, sign);
+ }
+
+ /* +normal, +subnormal */
+ uint64_t val = frsqrt7(f, exp_size, frac_size);
+ return make_float64(val);
+}
+
+RVVCALL(OPFVV1, vfrsqrt7_v_h, OP_UU_H, H2, H2, frsqrt7_h)
+RVVCALL(OPFVV1, vfrsqrt7_v_w, OP_UU_W, H4, H4, frsqrt7_s)
+RVVCALL(OPFVV1, vfrsqrt7_v_d, OP_UU_D, H8, H8, frsqrt7_d)
+GEN_VEXT_V_ENV(vfrsqrt7_v_h, 2, 2)
+GEN_VEXT_V_ENV(vfrsqrt7_v_w, 4, 4)
+GEN_VEXT_V_ENV(vfrsqrt7_v_d, 8, 8)
+
/* Vector Floating-Point MIN/MAX Instructions */
RVVCALL(OPFVV2, vfmin_vv_h, OP_UUU_H, H2, H2, H2, float16_minnum_noprop)
RVVCALL(OPFVV2, vfmin_vv_w, OP_UUU_W, H4, H4, H4, float32_minnum_noprop)
--
2.17.1
next prev parent reply other threads:[~2021-02-26 4:03 UTC|newest]
Thread overview: 76+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-02-26 3:17 [PATCH v7 00/75] support vector extension v1.0 frank.chang
2021-02-26 3:17 ` [PATCH v7 01/75] target/riscv: drop vector 0.7.1 and add 1.0 support frank.chang
2021-02-26 3:17 ` [PATCH v7 02/75] target/riscv: Use FIELD_EX32() to extract wd field frank.chang
2021-02-26 3:17 ` [PATCH v7 03/75] target/riscv: rvv-1.0: add mstatus VS field frank.chang
2021-02-26 3:17 ` [PATCH v7 04/75] target/riscv: rvv-1.0: add sstatus " frank.chang
2021-02-26 3:17 ` [PATCH v7 05/75] target/riscv: rvv-1.0: introduce writable misa.v field frank.chang
2021-02-26 3:17 ` [PATCH v7 06/75] target/riscv: rvv-1.0: add translation-time vector context status frank.chang
2021-02-26 3:17 ` [PATCH v7 07/75] target/riscv: rvv-1.0: remove rvv related codes from fcsr registers frank.chang
2021-02-26 3:17 ` [PATCH v7 08/75] target/riscv: rvv-1.0: add vcsr register frank.chang
2021-02-26 3:17 ` [PATCH v7 09/75] target/riscv: rvv-1.0: add vlenb register frank.chang
2021-02-26 3:17 ` [PATCH v7 10/75] target/riscv: rvv-1.0: check MSTATUS_VS when accessing vector csr registers frank.chang
2021-02-26 3:17 ` [PATCH v7 11/75] target/riscv: rvv-1.0: remove MLEN calculations frank.chang
2021-02-26 3:17 ` [PATCH v7 12/75] target/riscv: rvv-1.0: add fractional LMUL frank.chang
2021-02-26 3:17 ` [PATCH v7 13/75] target/riscv: rvv-1.0: add VMA and VTA frank.chang
2021-02-26 3:17 ` [PATCH v7 14/75] target/riscv: rvv-1.0: update check functions frank.chang
2021-02-26 3:17 ` [PATCH v7 15/75] target/riscv: introduce more imm value modes in translator functions frank.chang
2021-02-26 3:18 ` [PATCH v7 16/75] target/riscv: rvv:1.0: add translation-time nan-box helper function frank.chang
2021-02-26 3:18 ` [PATCH v7 17/75] target/riscv: rvv-1.0: configure instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 18/75] target/riscv: rvv-1.0: stride load and store instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 19/75] target/riscv: rvv-1.0: index " frank.chang
2021-02-26 3:18 ` [PATCH v7 20/75] target/riscv: rvv-1.0: fix address index overflow bug of indexed load/store insns frank.chang
2021-02-26 3:18 ` [PATCH v7 21/75] target/riscv: rvv-1.0: fault-only-first unit stride load frank.chang
2021-02-26 3:18 ` [PATCH v7 22/75] target/riscv: rvv-1.0: amo operations frank.chang
2021-02-26 3:18 ` [PATCH v7 23/75] target/riscv: rvv-1.0: load/store whole register instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 24/75] target/riscv: rvv-1.0: update vext_max_elems() for load/store insns frank.chang
2021-02-26 3:18 ` [PATCH v7 25/75] target/riscv: rvv-1.0: take fractional LMUL into vector max elements calculation frank.chang
2021-02-26 3:18 ` [PATCH v7 26/75] target/riscv: rvv-1.0: floating-point square-root instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 27/75] target/riscv: rvv-1.0: floating-point classify instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 28/75] target/riscv: rvv-1.0: mask population count instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 29/75] target/riscv: rvv-1.0: find-first-set mask bit instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 30/75] target/riscv: rvv-1.0: set-X-first mask bit instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 31/75] target/riscv: rvv-1.0: iota instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 32/75] target/riscv: rvv-1.0: element index instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 33/75] target/riscv: rvv-1.0: allow load element with sign-extended frank.chang
2021-02-26 3:18 ` [PATCH v7 34/75] target/riscv: rvv-1.0: register gather instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 35/75] target/riscv: rvv-1.0: integer scalar move instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 36/75] target/riscv: rvv-1.0: floating-point move instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 37/75] target/riscv: rvv-1.0: floating-point scalar move instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 38/75] target/riscv: rvv-1.0: whole register " frank.chang
2021-02-26 3:18 ` [PATCH v7 39/75] target/riscv: rvv-1.0: integer extension instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 40/75] target/riscv: rvv-1.0: single-width averaging add and subtract instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 41/75] target/riscv: rvv-1.0: single-width bit shift instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 42/75] target/riscv: rvv-1.0: integer add-with-carry/subtract-with-borrow frank.chang
2021-02-26 3:18 ` [PATCH v7 43/75] target/riscv: rvv-1.0: narrowing integer right shift instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 44/75] target/riscv: rvv-1.0: widening integer multiply-add instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 45/75] target/riscv: rvv-1.0: single-width saturating add and subtract instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 46/75] target/riscv: rvv-1.0: integer comparison instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 47/75] target/riscv: rvv-1.0: floating-point compare instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 48/75] target/riscv: rvv-1.0: mask-register logical instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 49/75] target/riscv: rvv-1.0: slide instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 50/75] target/riscv: rvv-1.0: floating-point " frank.chang
2021-02-26 3:18 ` [PATCH v7 51/75] target/riscv: rvv-1.0: narrowing fixed-point clip instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 52/75] target/riscv: rvv-1.0: single-width floating-point reduction frank.chang
2021-02-26 3:18 ` [PATCH v7 53/75] target/riscv: rvv-1.0: widening floating-point reduction instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 54/75] target/riscv: rvv-1.0: single-width scaling shift instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 55/75] target/riscv: rvv-1.0: remove widening saturating scaled multiply-add frank.chang
2021-02-26 3:18 ` [PATCH v7 56/75] target/riscv: rvv-1.0: remove vmford.vv and vmford.vf frank.chang
2021-02-26 3:18 ` [PATCH v7 57/75] target/riscv: rvv-1.0: remove integer extract instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 58/75] target/riscv: rvv-1.0: floating-point min/max instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 59/75] target/riscv: introduce floating-point rounding mode enum frank.chang
2021-02-26 3:18 ` [PATCH v7 60/75] target/riscv: rvv-1.0: floating-point/integer type-convert instructions frank.chang
2021-02-26 3:18 ` [PATCH v7 61/75] target/riscv: rvv-1.0: widening floating-point/integer type-convert frank.chang
2021-02-26 3:18 ` [PATCH v7 62/75] target/riscv: add "set round to odd" rounding mode helper function frank.chang
2021-02-26 3:18 ` [PATCH v7 63/75] target/riscv: rvv-1.0: narrowing floating-point/integer type-convert frank.chang
2021-02-26 3:18 ` [PATCH v7 64/75] target/riscv: rvv-1.0: relax RV_VLEN_MAX to 1024-bits frank.chang
2021-02-26 3:18 ` [PATCH v7 65/75] target/riscv: rvv-1.0: implement vstart CSR frank.chang
2021-02-26 3:18 ` [PATCH v7 66/75] target/riscv: rvv-1.0: trigger illegal instruction exception if frm is not valid frank.chang
2021-02-26 3:18 ` [PATCH v7 67/75] target/riscv: rvv-1.0: set mstatus.SD bit when writing vector CSRs frank.chang
2021-02-26 3:18 ` [PATCH v7 68/75] target/riscv: gdb: support vector registers for rv64 & rv32 frank.chang
2021-02-26 3:18 ` frank.chang [this message]
2021-02-26 3:18 ` [PATCH v7 70/75] target/riscv: rvv-1.0: floating-point reciprocal estimate instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 71/75] target/riscv: set mstatus.SD bit when writing fp CSRs frank.chang
2021-02-26 3:18 ` [PATCH v7 72/75] target/riscv: rvv-1.0: rename r2_zimm to r2_zimm11 frank.chang
2021-02-26 3:18 ` [PATCH v7 73/75] target/riscv: rvv-1.0: add vsetivli instruction frank.chang
2021-02-26 3:18 ` [PATCH v7 74/75] target/riscv: rvv-1.0: add evl parameter to vext_ldst_us() frank.chang
2021-02-26 3:18 ` [PATCH v7 75/75] target/riscv: rvv-1.0: add vector unit-stride mask load/store insns frank.chang
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=20210226031902.23656-70-frank.chang@sifive.com \
--to=frank.chang@sifive.com \
--cc=Alistair.Francis@wdc.com \
--cc=kbastian@mail.uni-paderborn.de \
--cc=palmer@dabbelt.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-riscv@nongnu.org \
--cc=richard.henderson@linaro.org \
--cc=sagark@eecs.berkeley.edu \
--cc=zhiwei_liu@c-sky.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 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).