riscv: Add support for the Zfa extension

This patch introduces the RISC-V Zfa extension, which introduces
additional floating-point instructions:
* fli (load-immediate) with pre-defined immediates
* fminm/fmaxm (like fmin/fmax but with different NaN behaviour)
* fround/froundmx (round to integer)
* fcvtmod.w.d (Modular Convert-to-Integer)
* fmv* to access high bits of float register bigger than XLEN
* Quiet comparison instructions (fleq/fltq)

Zfa defines its instructions in combination with the following extensions:
* single-precision floating-point (F)
* double-precision floating-point (D)
* quad-precision floating-point (Q)
* half-precision floating-point (Zfh)

Since QEMU does not support the RISC-V quad-precision floating-point
ISA extension (Q), this patch does not include the instructions that
depend on this extension. All other instructions are included in this
patch.

The Zfa specification can be found here:
  https://github.com/riscv/riscv-isa-manual/blob/master/src/zfa.tex
The Zfa specifciation is frozen and is in public review since May 3, 2023:
  https://groups.google.com/a/groups.riscv.org/g/isa-dev/c/SED4ntBkabg

The patch also includes a TCG test for the fcvtmod.w.d instruction.
The test cases test for correct results and flag behaviour.
Note, that the Zfa specification requires fcvtmod's flag behaviour
to be identical to a fcvt with the same operands (which is also
tested).

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>
Message-Id: <20230710071243.282464-1-christoph.muellner@vrull.eu>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

2 files changed, 142 insertions, 0 deletions

diff --git a/disas/riscv.c b/disas/riscv.c
index 9f0195b..cd7b6e8 100644
--- a/disas/riscv.c
+++ b/disas/riscv.c
@@ -829,6 +829,39 @@ typedef enum {
     rv_op_fsh = 798,
     rv_op_fmv_h_x = 799,
     rv_op_fmv_x_h = 800,
+    rv_op_fli_s = 801,
+    rv_op_fli_d = 802,
+    rv_op_fli_q = 803,
+    rv_op_fli_h = 804,
+    rv_op_fminm_s = 805,
+    rv_op_fmaxm_s = 806,
+    rv_op_fminm_d = 807,
+    rv_op_fmaxm_d = 808,
+    rv_op_fminm_q = 809,
+    rv_op_fmaxm_q = 810,
+    rv_op_fminm_h = 811,
+    rv_op_fmaxm_h = 812,
+    rv_op_fround_s = 813,
+    rv_op_froundnx_s = 814,
+    rv_op_fround_d = 815,
+    rv_op_froundnx_d = 816,
+    rv_op_fround_q = 817,
+    rv_op_froundnx_q = 818,
+    rv_op_fround_h = 819,
+    rv_op_froundnx_h = 820,
+    rv_op_fcvtmod_w_d = 821,
+    rv_op_fmvh_x_d = 822,
+    rv_op_fmvp_d_x = 823,
+    rv_op_fmvh_x_q = 824,
+    rv_op_fmvp_q_x = 825,
+    rv_op_fleq_s = 826,
+    rv_op_fltq_s = 827,
+    rv_op_fleq_d = 828,
+    rv_op_fltq_d = 829,
+    rv_op_fleq_q = 830,
+    rv_op_fltq_q = 831,
+    rv_op_fleq_h = 832,
+    rv_op_fltq_h = 833,
 } rv_op;
 
 /* register names */
@@ -854,6 +887,23 @@ static const char rv_vreg_name_sym[32][4] = {
     "v24", "v25", "v26", "v27", "v28", "v29", "v30", "v31"
 };
 
+/* The FLI.[HSDQ] numeric constants (0.0 for symbolic constants).
+ * The constants use the hex floating-point literal representation
+ * that is printed when using the printf %a format specifier,
+ * which matches the output that is generated by the disassembler.
+ */
+static const char rv_fli_name_const[32][9] =
+{
+    "0x1p+0", "min", "0x1p-16", "0x1p-15",
+    "0x1p-8", "0x1p-7", "0x1p-4", "0x1p-3",
+    "0x1p-2", "0x1.4p-2", "0x1.8p-2", "0x1.cp-2",
+    "0x1p-1", "0x1.4p-1", "0x1.8p-1", "0x1.cp-1",
+    "0x1p+0", "0x1.4p+0", "0x1.8p+0", "0x1.cp+0",
+    "0x1p+1", "0x1.4p+1", "0x1.8p+1", "0x1p+2",
+    "0x1p+3", "0x1p+4", "0x1p+7", "0x1p+8",
+    "0x1p+15", "0x1p+16", "inf", "nan"
+};
+
 /* pseudo-instruction constraints */
 
 static const rvc_constraint rvcc_jal[] = { rvc_rd_eq_ra, rvc_end };
@@ -1925,6 +1975,39 @@ const rv_opcode_data rvi_opcode_data[] = {
     { "fsh", rv_codec_s, rv_fmt_frs2_offset_rs1, NULL, 0, 0, 0 },
     { "fmv.h.x", rv_codec_r, rv_fmt_frd_rs1, NULL, 0, 0, 0 },
     { "fmv.x.h", rv_codec_r, rv_fmt_rd_frs1, NULL, 0, 0, 0 },
+    { "fli.s", rv_codec_fli, rv_fmt_fli, NULL, 0, 0, 0 },
+    { "fli.d", rv_codec_fli, rv_fmt_fli, NULL, 0, 0, 0 },
+    { "fli.q", rv_codec_fli, rv_fmt_fli, NULL, 0, 0, 0 },
+    { "fli.h", rv_codec_fli, rv_fmt_fli, NULL, 0, 0, 0 },
+    { "fminm.s", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fmaxm.s", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fminm.d", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fmaxm.d", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fminm.q", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fmaxm.q", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fminm.h", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fmaxm.h", rv_codec_r, rv_fmt_frd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fround.s", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "froundnx.s", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "fround.d", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "froundnx.d", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "fround.q", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "froundnx.q", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "fround.h", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "froundnx.h", rv_codec_r_m, rv_fmt_rm_frd_frs1, NULL, 0, 0, 0 },
+    { "fcvtmod.w.d", rv_codec_r_m, rv_fmt_rm_rd_frs1, NULL, 0, 0, 0 },
+    { "fmvh.x.d", rv_codec_r, rv_fmt_rd_frs1, NULL, 0, 0, 0 },
+    { "fmvp.d.x", rv_codec_r, rv_fmt_frd_rs1_rs2, NULL, 0, 0, 0 },
+    { "fmvh.x.q", rv_codec_r, rv_fmt_rd_frs1, NULL, 0, 0, 0 },
+    { "fmvp.q.x", rv_codec_r, rv_fmt_frd_rs1_rs2, NULL, 0, 0, 0 },
+    { "fleq.s", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fltq.s", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fleq.d", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fltq.d", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fleq.q", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fltq.q", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fleq.h", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
+    { "fltq.h", rv_codec_r, rv_fmt_rd_frs1_frs2, NULL, 0, 0, 0 },
 };
 
 /* CSR names */
@@ -2864,24 +2947,38 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                 switch ((inst >> 12) & 0b111) {
                 case 0: op = rv_op_fmin_s; break;
                 case 1: op = rv_op_fmax_s; break;
+                case 2: op = rv_op_fminm_s; break;
+                case 3: op = rv_op_fmaxm_s; break;
                 }
                 break;
             case 21:
                 switch ((inst >> 12) & 0b111) {
                 case 0: op = rv_op_fmin_d; break;
                 case 1: op = rv_op_fmax_d; break;
+                case 2: op = rv_op_fminm_d; break;
+                case 3: op = rv_op_fmaxm_d; break;
+                }
+                break;
+            case 22:
+                switch (((inst >> 12) & 0b111)) {
+                case 2: op = rv_op_fminm_h; break;
+                case 3: op = rv_op_fmaxm_h; break;
                 }
                 break;
             case 23:
                 switch ((inst >> 12) & 0b111) {
                 case 0: op = rv_op_fmin_q; break;
                 case 1: op = rv_op_fmax_q; break;
+                case 2: op = rv_op_fminm_q; break;
+                case 3: op = rv_op_fmaxm_q; break;
                 }
                 break;
             case 32:
                 switch ((inst >> 20) & 0b11111) {
                 case 1: op = rv_op_fcvt_s_d; break;
                 case 3: op = rv_op_fcvt_s_q; break;
+                case 4: op = rv_op_fround_s; break;
+                case 5: op = rv_op_froundnx_s; break;
                 case 6: op = rv_op_fcvt_s_bf16; break;
                 }
                 break;
@@ -2889,10 +2986,14 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                 switch ((inst >> 20) & 0b11111) {
                 case 0: op = rv_op_fcvt_d_s; break;
                 case 3: op = rv_op_fcvt_d_q; break;
+                case 4: op = rv_op_fround_d; break;
+                case 5: op = rv_op_froundnx_d; break;
                 }
                 break;
             case 34:
                 switch (((inst >> 20) & 0b11111)) {
+                case 4: op = rv_op_fround_h; break;
+                case 5: op = rv_op_froundnx_h; break;
                 case 8: op = rv_op_fcvt_bf16_s; break;
                 }
                 break;
@@ -2900,6 +3001,8 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                 switch ((inst >> 20) & 0b11111) {
                 case 0: op = rv_op_fcvt_q_s; break;
                 case 1: op = rv_op_fcvt_q_d; break;
+                case 4: op = rv_op_fround_q; break;
+                case 5: op = rv_op_froundnx_q; break;
                 }
                 break;
             case 44:
@@ -2922,6 +3025,8 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                 case 0: op = rv_op_fle_s; break;
                 case 1: op = rv_op_flt_s; break;
                 case 2: op = rv_op_feq_s; break;
+                case 4: op = rv_op_fleq_s; break;
+                case 5: op = rv_op_fltq_s; break;
                 }
                 break;
             case 81:
@@ -2929,6 +3034,14 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                 case 0: op = rv_op_fle_d; break;
                 case 1: op = rv_op_flt_d; break;
                 case 2: op = rv_op_feq_d; break;
+                case 4: op = rv_op_fleq_d; break;
+                case 5: op = rv_op_fltq_d; break;
+                }
+                break;
+            case 82:
+                switch (((inst >> 12) & 0b111)) {
+                case 4: op = rv_op_fleq_h; break;
+                case 5: op = rv_op_fltq_h; break;
                 }
                 break;
             case 83:
@@ -2936,6 +3049,18 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                 case 0: op = rv_op_fle_q; break;
                 case 1: op = rv_op_flt_q; break;
                 case 2: op = rv_op_feq_q; break;
+                case 4: op = rv_op_fleq_q; break;
+                case 5: op = rv_op_fltq_q; break;
+                }
+                break;
+            case 89:
+		switch (((inst >> 12) & 0b111)) {
+                case 0: op = rv_op_fmvp_d_x; break;
+                }
+                break;
+            case 91:
+		switch (((inst >> 12) & 0b111)) {
+                case 0: op = rv_op_fmvp_q_x; break;
                 }
                 break;
             case 96:
@@ -2952,6 +3077,7 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                 case 1: op = rv_op_fcvt_wu_d; break;
                 case 2: op = rv_op_fcvt_l_d; break;
                 case 3: op = rv_op_fcvt_lu_d; break;
+                case 8: op = rv_op_fcvtmod_w_d; break;
                 }
                 break;
             case 99:
@@ -2998,6 +3124,7 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                         ((inst >> 12) & 0b00000111)) {
                 case 0: op = rv_op_fmv_x_d; break;
                 case 1: op = rv_op_fclass_d; break;
+                case 8: op = rv_op_fmvh_x_d; break;
                 }
                 break;
             case 114:
@@ -3011,30 +3138,35 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa isa)
                         ((inst >> 12) & 0b00000111)) {
                 case 0: op = rv_op_fmv_x_q; break;
                 case 1: op = rv_op_fclass_q; break;
+                case 8: op = rv_op_fmvh_x_q; break;
                 }
                 break;
             case 120:
                 switch (((inst >> 17) & 0b11111000) |
                         ((inst >> 12) & 0b00000111)) {
                 case 0: op = rv_op_fmv_s_x; break;
+                case 8: op = rv_op_fli_s; break;
                 }
                 break;
             case 121:
                 switch (((inst >> 17) & 0b11111000) |
                         ((inst >> 12) & 0b00000111)) {
                 case 0: op = rv_op_fmv_d_x; break;
+                case 8: op = rv_op_fli_d; break;
                 }
                 break;
             case 122:
                 switch (((inst >> 17) & 0b11111000) |
                         ((inst >> 12) & 0b00000111)) {
                 case 0: op = rv_op_fmv_h_x; break;
+                case 8: op = rv_op_fli_h; break;
                 }
                 break;
             case 123:
                 switch (((inst >> 17) & 0b11111000) |
                         ((inst >> 12) & 0b00000111)) {
                 case 0: op = rv_op_fmv_q_x; break;
+                case 8: op = rv_op_fli_q; break;
                 }
                 break;
             }
@@ -4298,6 +4430,10 @@ static void decode_inst_operands(rv_decode *dec, rv_isa isa)
         break;
     case rv_codec_zcmt_jt:
         dec->imm = operand_tbl_index(inst);
+	break;
+    case rv_codec_fli:
+        dec->rd = operand_rd(inst);
+        dec->imm = operand_rs1(inst);
         break;
     case rv_codec_r2_imm5:
         dec->rd = operand_rd(inst);
@@ -4708,6 +4844,9 @@ static void format_inst(char *buf, size_t buflen, size_t tab, rv_decode *dec)
             append(buf, tmp, buflen);
             break;
         }
+        case 'h':
+            append(buf, rv_fli_name_const[dec->imm], buflen);
+            break;
         default:
             break;
         }
diff --git a/disas/riscv.h b/disas/riscv.h
index 052a0c4..9cf901f 100644
--- a/disas/riscv.h
+++ b/disas/riscv.h
@@ -165,6 +165,7 @@ typedef enum {
     rv_codec_r_imm2,
     rv_codec_r2_immhl,
     rv_codec_r2_imm2_imm5,
+    rv_codec_fli,
 } rv_codec;
 
 /* structures */
@@ -229,6 +230,7 @@ enum {
 #define rv_fmt_rd_offset              "O\t0,o"
 #define rv_fmt_rd_rs1_rs2             "O\t0,1,2"
 #define rv_fmt_frd_rs1                "O\t3,1"
+#define rv_fmt_frd_rs1_rs2            "O\t3,1,2"
 #define rv_fmt_frd_frs1               "O\t3,4"
 #define rv_fmt_rd_frs1                "O\t0,4"
 #define rv_fmt_rd_frs1_frs2           "O\t0,4,5"
@@ -295,5 +297,6 @@ enum {
 #define rv_fmt_rd_rs1_immh_imml       "O\t0,1,i,j"
 #define rv_fmt_rd_rs1_immh_imml_addr  "O\t0,(1),i,j"
 #define rv_fmt_rd2_imm                "O\t0,2,(1),i"
+#define rv_fmt_fli                    "O\t3,h"
 
 #endif /* DISAS_RISCV_H */