target-arm: Implement VFPv4 fused multiply-accumulate insns

Implement the fused multiply-accumulate instructions (VFMA, VFMS,
VFNMA, VFNMS) which are new in VFPv4.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

4 files changed, 90 insertions, 0 deletions

diff --git a/target-arm/cpu.h b/target-arm/cpu.h
index af3904d..c4d742f 100644
--- a/target-arm/cpu.h
+++ b/target-arm/cpu.h
@@ -376,6 +376,7 @@ enum arm_features {
     ARM_FEATURE_STRONGARM,
     ARM_FEATURE_VAPA, /* cp15 VA to PA lookups */
     ARM_FEATURE_ARM_DIV, /* divide supported in ARM encoding */
+    ARM_FEATURE_VFP4, /* VFPv4 (implies that NEON is v2) */
 };
 
 static inline int arm_feature(CPUARMState *env, int feature)
diff --git a/target-arm/helper.c b/target-arm/helper.c
index 3a51fd7..eddb923 100644
--- a/target-arm/helper.c
+++ b/target-arm/helper.c
@@ -204,6 +204,7 @@ static void cpu_reset_model_id(CPUARMState *env, uint32_t id)
         set_feature(env, ARM_FEATURE_THUMB2);
         set_feature(env, ARM_FEATURE_VFP);
         set_feature(env, ARM_FEATURE_VFP3);
+        set_feature(env, ARM_FEATURE_VFP4);
         set_feature(env, ARM_FEATURE_VFP_FP16);
         set_feature(env, ARM_FEATURE_NEON);
         set_feature(env, ARM_FEATURE_THUMB2EE);
@@ -3084,6 +3085,19 @@ uint32_t HELPER(rsqrte_u32)(uint32_t a, CPUState *env)
     return 0x80000000 | ((float64_val(f64) >> 21) & 0x7fffffff);
 }
 
+/* VFPv4 fused multiply-accumulate */
+float32 VFP_HELPER(muladd, s)(float32 a, float32 b, float32 c, void *fpstp)
+{
+    float_status *fpst = fpstp;
+    return float32_muladd(a, b, c, 0, fpst);
+}
+
+float64 VFP_HELPER(muladd, d)(float64 a, float64 b, float64 c, void *fpstp)
+{
+    float_status *fpst = fpstp;
+    return float64_muladd(a, b, c, 0, fpst);
+}
+
 void HELPER(set_teecr)(CPUState *env, uint32_t val)
 {
     val &= 1;
diff --git a/target-arm/helper.h b/target-arm/helper.h
index 3ad1cb0..16dd5fc 100644
--- a/target-arm/helper.h
+++ b/target-arm/helper.h
@@ -132,6 +132,9 @@ DEF_HELPER_2(vfp_fcvt_f32_to_f16, i32, f32, env)
 DEF_HELPER_2(neon_fcvt_f16_to_f32, f32, i32, env)
 DEF_HELPER_2(neon_fcvt_f32_to_f16, i32, f32, env)
 
+DEF_HELPER_4(vfp_muladdd, f64, f64, f64, f64, ptr)
+DEF_HELPER_4(vfp_muladds, f32, f32, f32, f32, ptr)
+
 DEF_HELPER_3(recps_f32, f32, f32, f32, env)
 DEF_HELPER_3(rsqrts_f32, f32, f32, f32, env)
 DEF_HELPER_2(recpe_f32, f32, f32, env)
diff --git a/target-arm/translate.c b/target-arm/translate.c
index 812a9e7..0f35b60 100644
--- a/target-arm/translate.c
+++ b/target-arm/translate.c
@@ -3141,6 +3141,57 @@ static int disas_vfp_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 case 8: /* div: fn / fm */
                     gen_vfp_div(dp);
                     break;
+                case 10: /* VFNMA : fd = muladd(-fd,  fn, fm) */
+                case 11: /* VFNMS : fd = muladd(-fd, -fn, fm) */
+                case 12: /* VFMA  : fd = muladd( fd,  fn, fm) */
+                case 13: /* VFMS  : fd = muladd( fd, -fn, fm) */
+                    /* These are fused multiply-add, and must be done as one
+                     * floating point operation with no rounding between the
+                     * multiplication and addition steps.
+                     * NB that doing the negations here as separate steps is
+                     * correct : an input NaN should come out with its sign bit
+                     * flipped if it is a negated-input.
+                     */
+                    if (!arm_feature(env, ARM_FEATURE_VFP4)) {
+                        return 1;
+                    }
+                    if (dp) {
+                        TCGv_ptr fpst;
+                        TCGv_i64 frd;
+                        if (op & 1) {
+                            /* VFNMS, VFMS */
+                            gen_helper_vfp_negd(cpu_F0d, cpu_F0d);
+                        }
+                        frd = tcg_temp_new_i64();
+                        tcg_gen_ld_f64(frd, cpu_env, vfp_reg_offset(dp, rd));
+                        if (op & 2) {
+                            /* VFNMA, VFNMS */
+                            gen_helper_vfp_negd(frd, frd);
+                        }
+                        fpst = get_fpstatus_ptr(0);
+                        gen_helper_vfp_muladdd(cpu_F0d, cpu_F0d,
+                                               cpu_F1d, frd, fpst);
+                        tcg_temp_free_ptr(fpst);
+                        tcg_temp_free_i64(frd);
+                    } else {
+                        TCGv_ptr fpst;
+                        TCGv_i32 frd;
+                        if (op & 1) {
+                            /* VFNMS, VFMS */
+                            gen_helper_vfp_negs(cpu_F0s, cpu_F0s);
+                        }
+                        frd = tcg_temp_new_i32();
+                        tcg_gen_ld_f32(frd, cpu_env, vfp_reg_offset(dp, rd));
+                        if (op & 2) {
+                            gen_helper_vfp_negs(frd, frd);
+                        }
+                        fpst = get_fpstatus_ptr(0);
+                        gen_helper_vfp_muladds(cpu_F0s, cpu_F0s,
+                                               cpu_F1s, frd, fpst);
+                        tcg_temp_free_ptr(fpst);
+                        tcg_temp_free_i32(frd);
+                    }
+                    break;
                 case 14: /* fconst */
                     if (!arm_feature(env, ARM_FEATURE_VFP3))
                       return 1;
@@ -4417,6 +4468,7 @@ static void gen_neon_narrow_op(int op, int u, int size, TCGv dest, TCGv_i64 src)
 #define NEON_3R_VPMIN 21
 #define NEON_3R_VQDMULH_VQRDMULH 22
 #define NEON_3R_VPADD 23
+#define NEON_3R_VFM 25 /* VFMA, VFMS : float fused multiply-add */
 #define NEON_3R_FLOAT_ARITH 26 /* float VADD, VSUB, VPADD, VABD */
 #define NEON_3R_FLOAT_MULTIPLY 27 /* float VMLA, VMLS, VMUL */
 #define NEON_3R_FLOAT_CMP 28 /* float VCEQ, VCGE, VCGT */
@@ -4449,6 +4501,7 @@ static const uint8_t neon_3r_sizes[] = {
     [NEON_3R_VPMIN] = 0x7,
     [NEON_3R_VQDMULH_VQRDMULH] = 0x6,
     [NEON_3R_VPADD] = 0x7,
+    [NEON_3R_VFM] = 0x5, /* size bit 1 encodes op */
     [NEON_3R_FLOAT_ARITH] = 0x5, /* size bit 1 encodes op */
     [NEON_3R_FLOAT_MULTIPLY] = 0x5, /* size bit 1 encodes op */
     [NEON_3R_FLOAT_CMP] = 0x5, /* size bit 1 encodes op */
@@ -4726,6 +4779,11 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
                 return 1;
             }
             break;
+        case NEON_3R_VFM:
+            if (!arm_feature(env, ARM_FEATURE_VFP4) || u) {
+                return 1;
+            }
+            break;
         default:
             break;
         }
@@ -5006,6 +5064,20 @@ static int disas_neon_data_insn(CPUState * env, DisasContext *s, uint32_t insn)
             else
                 gen_helper_rsqrts_f32(tmp, tmp, tmp2, cpu_env);
             break;
+        case NEON_3R_VFM:
+        {
+            /* VFMA, VFMS: fused multiply-add */
+            TCGv_ptr fpstatus = get_fpstatus_ptr(1);
+            TCGv_i32 tmp3 = neon_load_reg(rd, pass);
+            if (size) {
+                /* VFMS */
+                gen_helper_vfp_negs(tmp, tmp);
+            }
+            gen_helper_vfp_muladds(tmp, tmp, tmp2, tmp3, fpstatus);
+            tcg_temp_free_i32(tmp3);
+            tcg_temp_free_ptr(fpstatus);
+            break;
+        }
         default:
             abort();
         }