target/sparc: Implement FMAf extension

Rearrange PDIST so that do_dddd is general purpose and may
be re-used for FMADDd etc.  Add pickNaN and pickNaNMulAdd.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

diff --git a/fpu/softfloat-specialize.c.inc b/fpu/softfloat-specialize.c.inc
index f5730145323e6649db5aebe2c36203e5f7706a55..8f3b97d9bf7b1d1d5fb290eda2d8abf4737d8bd3 100644 (file)
--- a/fpu/softfloat-specialize.c.inc
+++ b/fpu/softfloat-specialize.c.inc
@@ -447,6 +447,17 @@ static int pickNaN(FloatClass a_cls, FloatClass b_cls,
     } else {
         return 1;
     }
+#elif defined(TARGET_SPARC)
+    /* Prefer SNaN over QNaN, order B then A. */
+    if (is_snan(b_cls)) {
+        return 1;
+    } else if (is_snan(a_cls)) {
+        return 0;
+    } else if (is_qnan(b_cls)) {
+        return 1;
+    } else {
+        return 0;
+    }
 #elif defined(TARGET_XTENSA)
     /*
      * Xtensa has two NaN propagation modes.
@@ -624,6 +635,26 @@ static int pickNaNMulAdd(FloatClass a_cls, FloatClass b_cls, FloatClass c_cls,
         float_raise(float_flag_invalid | float_flag_invalid_imz, status);
     }
     return 3; /* default NaN */
+#elif defined(TARGET_SPARC)
+    /* For (inf,0,nan) return c. */
+    if (infzero) {
+        float_raise(float_flag_invalid | float_flag_invalid_imz, status);
+        return 2;
+    }
+    /* Prefer SNaN over QNaN, order C, B, A. */
+    if (is_snan(c_cls)) {
+        return 2;
+    } else if (is_snan(b_cls)) {
+        return 1;
+    } else if (is_snan(a_cls)) {
+        return 0;
+    } else if (is_qnan(c_cls)) {
+        return 2;
+    } else if (is_qnan(b_cls)) {
+        return 1;
+    } else {
+        return 0;
+    }
 #elif defined(TARGET_XTENSA)
     /*
      * For Xtensa, the (inf,zero,nan) case sets InvalidOp and returns

diff --git a/linux-user/elfload.c b/linux-user/elfload.c
index c1e1511ff205d6a6a6b8fb866fec10a46eee470e..6a1457346a290be49e7bd0df4cc2647b8ae5908a 100644 (file)
--- a/linux-user/elfload.c
+++ b/linux-user/elfload.c
@@ -1003,6 +1003,7 @@ static uint32_t get_elf_hwcap(void)
     r |= features & CPU_FEATURE_FSMULD ? HWCAP_SPARC_FSMULD : 0;
     r |= features & CPU_FEATURE_VIS1 ? HWCAP_SPARC_VIS : 0;
     r |= features & CPU_FEATURE_VIS2 ? HWCAP_SPARC_VIS2 : 0;
+    r |= features & CPU_FEATURE_FMAF ? HWCAP_SPARC_FMAF : 0;
 #endif
 
     return r;

diff --git a/target/sparc/cpu-feature.h.inc b/target/sparc/cpu-feature.h.inc
index d800f18c4ed129d54477b9351fc3a8784a5d3860..a30b9255b26cb30c995a9084deb91334277f6390 100644 (file)
--- a/target/sparc/cpu-feature.h.inc
+++ b/target/sparc/cpu-feature.h.inc
@@ -12,3 +12,4 @@ FEATURE(ASR17)
 FEATURE(CACHE_CTRL)
 FEATURE(POWERDOWN)
 FEATURE(CASA)
+FEATURE(FMAF)

diff --git a/target/sparc/cpu.c b/target/sparc/cpu.c
index 5be1592e66a7ae64e669c7bca0efd94b972b5862..ed9238a69df9241079579209dc03cac574f3502b 100644 (file)
--- a/target/sparc/cpu.c
+++ b/target/sparc/cpu.c
@@ -549,6 +549,7 @@ static const char * const feature_name[] = {
     [CPU_FEATURE_BIT_HYPV] = "hypv",
     [CPU_FEATURE_BIT_VIS1] = "vis1",
     [CPU_FEATURE_BIT_VIS2] = "vis2",
+    [CPU_FEATURE_BIT_FMAF] = "fmaf",
 #else
     [CPU_FEATURE_BIT_MUL] = "mul",
     [CPU_FEATURE_BIT_DIV] = "div",
@@ -877,6 +878,8 @@ static Property sparc_cpu_properties[] = {
                     CPU_FEATURE_BIT_VIS1, false),
     DEFINE_PROP_BIT("vis2",     SPARCCPU, env.def.features,
                     CPU_FEATURE_BIT_VIS2, false),
+    DEFINE_PROP_BIT("fmaf",     SPARCCPU, env.def.features,
+                    CPU_FEATURE_BIT_FMAF, false),
 #else
     DEFINE_PROP_BIT("mul",      SPARCCPU, env.def.features,
                     CPU_FEATURE_BIT_MUL, false),

diff --git a/target/sparc/fop_helper.c b/target/sparc/fop_helper.c
index 1205a599ef435881a86b3862c5623d93f4272aac..1de44d79c11e9e895e21c6947764c4a640c186a1 100644 (file)
--- a/target/sparc/fop_helper.c
+++ b/target/sparc/fop_helper.c
@@ -343,6 +343,22 @@ Int128 helper_fsqrtq(CPUSPARCState *env, Int128 src)
     return f128_ret(ret);
 }
 
+float32 helper_fmadds(CPUSPARCState *env, float32 s1,
+                      float32 s2, float32 s3, uint32_t op)
+{
+    float32 ret = float32_muladd(s1, s2, s3, op, &env->fp_status);
+    check_ieee_exceptions(env, GETPC());
+    return ret;
+}
+
+float64 helper_fmaddd(CPUSPARCState *env, float64 s1,
+                      float64 s2, float64 s3, uint32_t op)
+{
+    float64 ret = float64_muladd(s1, s2, s3, op, &env->fp_status);
+    check_ieee_exceptions(env, GETPC());
+    return ret;
+}
+
 static uint32_t finish_fcmp(CPUSPARCState *env, FloatRelation r, uintptr_t ra)
 {
     check_ieee_exceptions(env, ra);

diff --git a/target/sparc/helper.h b/target/sparc/helper.h
index 97fbf6f66c00f7ea33773486f07510e96a449319..f4d3311ac471b8aeaa3f77721a2aae0db5894ffe 100644 (file)
--- a/target/sparc/helper.h
+++ b/target/sparc/helper.h
@@ -56,6 +56,7 @@ DEF_HELPER_FLAGS_3(faddd, TCG_CALL_NO_WG, f64, env, f64, f64)
 DEF_HELPER_FLAGS_3(fsubd, TCG_CALL_NO_WG, f64, env, f64, f64)
 DEF_HELPER_FLAGS_3(fmuld, TCG_CALL_NO_WG, f64, env, f64, f64)
 DEF_HELPER_FLAGS_3(fdivd, TCG_CALL_NO_WG, f64, env, f64, f64)
+DEF_HELPER_FLAGS_5(fmaddd, TCG_CALL_NO_WG, f64, env, f64, f64, f64, i32)
 
 DEF_HELPER_FLAGS_3(faddq, TCG_CALL_NO_WG, i128, env, i128, i128)
 DEF_HELPER_FLAGS_3(fsubq, TCG_CALL_NO_WG, i128, env, i128, i128)
@@ -66,6 +67,7 @@ DEF_HELPER_FLAGS_3(fadds, TCG_CALL_NO_WG, f32, env, f32, f32)
 DEF_HELPER_FLAGS_3(fsubs, TCG_CALL_NO_WG, f32, env, f32, f32)
 DEF_HELPER_FLAGS_3(fmuls, TCG_CALL_NO_WG, f32, env, f32, f32)
 DEF_HELPER_FLAGS_3(fdivs, TCG_CALL_NO_WG, f32, env, f32, f32)
+DEF_HELPER_FLAGS_5(fmadds, TCG_CALL_NO_WG, f32, env, f32, f32, f32, i32)
 
 DEF_HELPER_FLAGS_3(fsmuld, TCG_CALL_NO_WG, f64, env, f32, f32)
 DEF_HELPER_FLAGS_3(fdmulq, TCG_CALL_NO_WG, i128, env, f64, f64)

diff --git a/target/sparc/insns.decode b/target/sparc/insns.decode
index 02fa505b495164ac488c347d6804762ff52289b0..056fba98f93712acfdd5bdee579f6f37471d6d66 100644 (file)
--- a/target/sparc/insns.decode
+++ b/target/sparc/insns.decode
@@ -29,6 +29,7 @@ CALL    01 i:s30
 %dfp_rd     25:5 !function=extract_dfpreg
 %dfp_rs1    14:5 !function=extract_dfpreg
 %dfp_rs2     0:5 !function=extract_dfpreg
+%dfp_rs3     9:5 !function=extract_dfpreg
 
 %qfp_rd     25:5 !function=extract_qfpreg
 %qfp_rs1    14:5 !function=extract_qfpreg
@@ -80,6 +81,11 @@ CALL    01 i:s30
 @q_d2       .. ..... ...... ..... . ........ .....         \
             &r_r rd=%qfp_rd rs=%dfp_rs2
 
+&r_r_r_r    rd rs1 rs2 rs3
+@r_r_r_r    .. rd:5  ...... rs1:5 rs3:5 .... rs2:5         &r_r_r_r
+@d_d_d_d    .. ..... ...... ..... ..... .... .....         \
+            &r_r_r_r rd=%dfp_rd rs1=%dfp_rs1 rs2=%dfp_rs2 rs3=%dfp_rs3
+
 {
   [
     STBAR           10 00000 101000 01111 0 0000000000000
@@ -394,7 +400,8 @@ FCMPEq      10 000 cc:2 110101 .....  0 0101 0111 .....    \
     FPACK32     10 ..... 110110 ..... 0 0011 1010 .....    @d_d_d
     FPACK16     10 ..... 110110 00000 0 0011 1011 .....    @r_d2
     FPACKFIX    10 ..... 110110 00000 0 0011 1101 .....    @r_d2
-    PDIST       10 ..... 110110 ..... 0 0011 1110 .....    @d_d_d
+    PDIST       10 ..... 110110 ..... 0 0011 1110 .....    \
+                &r_r_r_r rd=%dfp_rd rs1=%dfp_rd rs2=%dfp_rs1 rs3=%dfp_rs2
 
     FALIGNDATAg 10 ..... 110110 ..... 0 0100 1000 .....    @d_d_d
     FPMERGE     10 ..... 110110 ..... 0 0100 1011 .....    @d_r_r
@@ -448,7 +455,19 @@ FCMPEq      10 000 cc:2 110101 .....  0 0101 0111 .....    \
   NCP           10 ----- 110110 ----- --------- -----      # v8 CPop1
 }
 
-NCP         10 ----- 110111 ----- --------- -----          # v8 CPop2
+{
+  [
+    FMADDs      10 ..... 110111 ..... ..... 0001 .....     @r_r_r_r
+    FMADDd      10 ..... 110111 ..... ..... 0010 .....     @d_d_d_d
+    FMSUBs      10 ..... 110111 ..... ..... 0101 .....     @r_r_r_r
+    FMSUBd      10 ..... 110111 ..... ..... 0110 .....     @d_d_d_d
+    FNMSUBs     10 ..... 110111 ..... ..... 1001 .....     @r_r_r_r
+    FNMSUBd     10 ..... 110111 ..... ..... 1010 .....     @d_d_d_d
+    FNMADDs     10 ..... 110111 ..... ..... 1101 .....     @r_r_r_r
+    FNMADDd     10 ..... 110111 ..... ..... 1110 .....     @d_d_d_d
+  ]
+  NCP           10 ----- 110111 ----- --------- -----      # v8 CPop2
+}
 
 ##
 ## Major Opcode 11 -- load and store instructions

diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index 7c290293ea0a9a0182ab4f0fff15f0b4bfa51e16..5efd09f4f4cdbd6053e581bdead06c7b0a4878da 100644 (file)
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -28,6 +28,7 @@
 #include "exec/helper-gen.h"
 #include "exec/translator.h"
 #include "exec/log.h"
+#include "fpu/softfloat.h"
 #include "asi.h"
 
 #define HELPER_H "helper.h"
@@ -1142,6 +1143,52 @@ static void gen_op_fabsq(TCGv_i128 dst, TCGv_i128 src)
     tcg_gen_concat_i64_i128(dst, l, h);
 }
 
+static void gen_op_fmadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3)
+{
+    gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(0));
+}
+
+static void gen_op_fmaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3)
+{
+    gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(0));
+}
+
+static void gen_op_fmsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3)
+{
+    int op = float_muladd_negate_c;
+    gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op));
+}
+
+static void gen_op_fmsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3)
+{
+    int op = float_muladd_negate_c;
+    gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op));
+}
+
+static void gen_op_fnmsubs(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3)
+{
+    int op = float_muladd_negate_c | float_muladd_negate_result;
+    gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op));
+}
+
+static void gen_op_fnmsubd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3)
+{
+    int op = float_muladd_negate_c | float_muladd_negate_result;
+    gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op));
+}
+
+static void gen_op_fnmadds(TCGv_i32 d, TCGv_i32 s1, TCGv_i32 s2, TCGv_i32 s3)
+{
+    int op = float_muladd_negate_result;
+    gen_helper_fmadds(d, tcg_env, s1, s2, s3, tcg_constant_i32(op));
+}
+
+static void gen_op_fnmaddd(TCGv_i64 d, TCGv_i64 s1, TCGv_i64 s2, TCGv_i64 s3)
+{
+    int op = float_muladd_negate_result;
+    gen_helper_fmaddd(d, tcg_env, s1, s2, s3, tcg_constant_i32(op));
+}
+
 static void gen_op_fpexception_im(DisasContext *dc, int ftt)
 {
     /*
@@ -2136,6 +2183,7 @@ static int extract_qfpreg(DisasContext *dc, int x)
 # define avail_MUL(C)     true
 # define avail_POWERDOWN(C) false
 # define avail_64(C)      true
+# define avail_FMAF(C)    ((C)->def->features & CPU_FEATURE_FMAF)
 # define avail_GL(C)      ((C)->def->features & CPU_FEATURE_GL)
 # define avail_HYPV(C)    ((C)->def->features & CPU_FEATURE_HYPV)
 # define avail_VIS1(C)    ((C)->def->features & CPU_FEATURE_VIS1)
@@ -2148,6 +2196,7 @@ static int extract_qfpreg(DisasContext *dc, int x)
 # define avail_MUL(C)     ((C)->def->features & CPU_FEATURE_MUL)
 # define avail_POWERDOWN(C) ((C)->def->features & CPU_FEATURE_POWERDOWN)
 # define avail_64(C)      false
+# define avail_FMAF(C)    false
 # define avail_GL(C)      false
 # define avail_HYPV(C)    false
 # define avail_VIS1(C)    false
@@ -4785,25 +4834,52 @@ static bool trans_FsMULd(DisasContext *dc, arg_r_r_r *a)
     return advance_pc(dc);
 }
 
-static bool do_dddd(DisasContext *dc, arg_r_r_r *a,
+static bool do_ffff(DisasContext *dc, arg_r_r_r_r *a,
+                    void (*func)(TCGv_i32, TCGv_i32, TCGv_i32, TCGv_i32))
+{
+    TCGv_i32 dst, src1, src2, src3;
+
+    if (gen_trap_ifnofpu(dc)) {
+        return true;
+    }
+
+    src1 = gen_load_fpr_F(dc, a->rs1);
+    src2 = gen_load_fpr_F(dc, a->rs2);
+    src3 = gen_load_fpr_F(dc, a->rs3);
+    dst = tcg_temp_new_i32();
+    func(dst, src1, src2, src3);
+    gen_store_fpr_F(dc, a->rd, dst);
+    return advance_pc(dc);
+}
+
+TRANS(FMADDs, FMAF, do_ffff, a, gen_op_fmadds)
+TRANS(FMSUBs, FMAF, do_ffff, a, gen_op_fmsubs)
+TRANS(FNMSUBs, FMAF, do_ffff, a, gen_op_fnmsubs)
+TRANS(FNMADDs, FMAF, do_ffff, a, gen_op_fnmadds)
+
+static bool do_dddd(DisasContext *dc, arg_r_r_r_r *a,
                     void (*func)(TCGv_i64, TCGv_i64, TCGv_i64, TCGv_i64))
 {
-    TCGv_i64 dst, src0, src1, src2;
+    TCGv_i64 dst, src1, src2, src3;
 
     if (gen_trap_ifnofpu(dc)) {
         return true;
     }
 
     dst  = tcg_temp_new_i64();
-    src0 = gen_load_fpr_D(dc, a->rd);
     src1 = gen_load_fpr_D(dc, a->rs1);
     src2 = gen_load_fpr_D(dc, a->rs2);
-    func(dst, src0, src1, src2);
+    src3 = gen_load_fpr_D(dc, a->rs3);
+    func(dst, src1, src2, src3);
     gen_store_fpr_D(dc, a->rd, dst);
     return advance_pc(dc);
 }
 
 TRANS(PDIST, VIS1, do_dddd, a, gen_helper_pdist)
+TRANS(FMADDd, FMAF, do_dddd, a, gen_op_fmaddd)
+TRANS(FMSUBd, FMAF, do_dddd, a, gen_op_fmsubd)
+TRANS(FNMSUBd, FMAF, do_dddd, a, gen_op_fnmsubd)
+TRANS(FNMADDd, FMAF, do_dddd, a, gen_op_fnmaddd)
 
 static bool do_env_qqq(DisasContext *dc, arg_r_r_r *a,
                        void (*func)(TCGv_i128, TCGv_env, TCGv_i128, TCGv_i128))