diff mbox series

[v4,44/47] target/ppc: Implement xvcvbf16spn and xvcvspbf16 instructions

Message ID 20220222143646.1268606-45-matheus.ferst@eldorado.org.br (mailing list archive)
State New, archived
Headers show
Series target/ppc: PowerISA Vector/VSX instruction batch | expand

Commit Message

Matheus K. Ferst Feb. 22, 2022, 2:36 p.m. UTC 
From: Víctor Colombo <victor.colombo@eldorado.org.br>

Signed-off-by: Víctor Colombo <victor.colombo@eldorado.org.br>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
---
 target/ppc/fpu_helper.c             | 21 +++++++++++++++++++
 target/ppc/helper.h                 |  1 +
 target/ppc/insn32.decode            | 11 +++++++---
 target/ppc/translate/vsx-impl.c.inc | 31 ++++++++++++++++++++++++++++-
 4 files changed, 60 insertions(+), 4 deletions(-)

Comments

Richard Henderson Feb. 23, 2022, 3:08 a.m. UTC | #1 
On 2/22/22 04:36, matheus.ferst@eldorado.org.br wrote:
> From: Víctor Colombo <victor.colombo@eldorado.org.br>
> 
> Signed-off-by: Víctor Colombo <victor.colombo@eldorado.org.br>
> Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
> ---
>   target/ppc/fpu_helper.c             | 21 +++++++++++++++++++
>   target/ppc/helper.h                 |  1 +
>   target/ppc/insn32.decode            | 11 +++++++---
>   target/ppc/translate/vsx-impl.c.inc | 31 ++++++++++++++++++++++++++++-
>   4 files changed, 60 insertions(+), 4 deletions(-)
> 
> diff --git a/target/ppc/fpu_helper.c b/target/ppc/fpu_helper.c
> index 7773333bd7..d77900fff1 100644
> --- a/target/ppc/fpu_helper.c
> +++ b/target/ppc/fpu_helper.c
> @@ -2790,6 +2790,27 @@ VSX_CVT_FP_TO_FP_HP(xscvhpdp, 1, float16, float64, VsrH(3), VsrD(0), 1)
>   VSX_CVT_FP_TO_FP_HP(xvcvsphp, 4, float32, float16, VsrW(i), VsrH(2 * i  + 1), 0)
>   VSX_CVT_FP_TO_FP_HP(xvcvhpsp, 4, float16, float32, VsrH(2 * i + 1), VsrW(i), 0)
>   
> +void helper_XVCVSPBF16(CPUPPCState *env, ppc_vsr_t *xt, ppc_vsr_t *xb)
> +{
> +    ppc_vsr_t t = { };
> +    int i;
> +
> +    helper_reset_fpstatus(env);
> +    for (i = 0; i < 4; i++) {
> +        if (unlikely(float32_is_signaling_nan(xb->VsrW(i), &env->fp_status))) {
> +            float_invalid_op_vxsnan(env, GETPC());
> +            t.VsrH(2 * i + 1) = float32_to_bfloat16(
> +                float32_snan_to_qnan(xb->VsrW(i)), &env->fp_status);
> +        } else {
> +            t.VsrH(2 * i + 1) =
> +                float32_to_bfloat16(xb->VsrW(i), &env->fp_status);
> +        }
> +    }

Do not check for snan first; use float_flag_invalid_snan.
And you can move that check outside the loop, before the
writeback of t to *xt.


r~
diff mbox series

Patch

diff --git a/target/ppc/fpu_helper.c b/target/ppc/fpu_helper.c
index 7773333bd7..d77900fff1 100644
--- a/target/ppc/fpu_helper.c
+++ b/target/ppc/fpu_helper.c
@@ -2790,6 +2790,27 @@  VSX_CVT_FP_TO_FP_HP(xscvhpdp, 1, float16, float64, VsrH(3), VsrD(0), 1)
 VSX_CVT_FP_TO_FP_HP(xvcvsphp, 4, float32, float16, VsrW(i), VsrH(2 * i  + 1), 0)
 VSX_CVT_FP_TO_FP_HP(xvcvhpsp, 4, float16, float32, VsrH(2 * i + 1), VsrW(i), 0)
 
+void helper_XVCVSPBF16(CPUPPCState *env, ppc_vsr_t *xt, ppc_vsr_t *xb)
+{
+    ppc_vsr_t t = { };
+    int i;
+
+    helper_reset_fpstatus(env);
+    for (i = 0; i < 4; i++) {
+        if (unlikely(float32_is_signaling_nan(xb->VsrW(i), &env->fp_status))) {
+            float_invalid_op_vxsnan(env, GETPC());
+            t.VsrH(2 * i + 1) = float32_to_bfloat16(
+                float32_snan_to_qnan(xb->VsrW(i)), &env->fp_status);
+        } else {
+            t.VsrH(2 * i + 1) =
+                float32_to_bfloat16(xb->VsrW(i), &env->fp_status);
+        }
+    }
+
+    *xt = t;
+    do_float_check_status(env, GETPC());
+}
+
 void helper_XSCVQPDP(CPUPPCState *env, uint32_t ro, ppc_vsr_t *xt,
                      ppc_vsr_t *xb)
 {
diff --git a/target/ppc/helper.h b/target/ppc/helper.h
index d3af130dc2..805a5046d8 100644
--- a/target/ppc/helper.h
+++ b/target/ppc/helper.h
@@ -494,6 +494,7 @@  DEF_HELPER_FLAGS_4(xvcmpnesp, TCG_CALL_NO_RWG, i32, env, vsr, vsr, vsr)
 DEF_HELPER_3(xvcvspdp, void, env, vsr, vsr)
 DEF_HELPER_3(xvcvsphp, void, env, vsr, vsr)
 DEF_HELPER_3(xvcvhpsp, void, env, vsr, vsr)
+DEF_HELPER_3(XVCVSPBF16, void, env, vsr, vsr)
 DEF_HELPER_3(xvcvspsxds, void, env, vsr, vsr)
 DEF_HELPER_3(xvcvspsxws, void, env, vsr, vsr)
 DEF_HELPER_3(xvcvspuxds, void, env, vsr, vsr)
diff --git a/target/ppc/insn32.decode b/target/ppc/insn32.decode
index 892d4bfd84..8964898f20 100644
--- a/target/ppc/insn32.decode
+++ b/target/ppc/insn32.decode
@@ -152,8 +152,11 @@ 
 %xx_xb          1:1 11:5
 %xx_xa          2:1 16:5
 %xx_xc          3:1 6:5
-&XX2            xt xb uim:uint8_t
-@XX2            ...... ..... ... uim:2 ..... ......... ..       &XX2 xt=%xx_xt xb=%xx_xb
+&XX2            xt xb
+@XX2            ...... ..... ..... ..... ......... ..           &XX2 xt=%xx_xt xb=%xx_xb
+
+&XX2_uim2       xt xb uim:uint8_t
+@XX2_uim2       ...... ..... ... uim:2 ..... ......... ..       &XX2_uim2 xt=%xx_xt xb=%xx_xb
 
 &XX2_bf_xb      bf xb
 @XX2_bf_xb      ...... bf:3 .. ..... ..... ......... . .        &XX2_bf_xb xb=%xx_xb
@@ -635,7 +638,7 @@  XSNMSUBQP       111111 ..... ..... ..... 0111100100 .   @X_rc
 ## VSX splat instruction
 
 XXSPLTIB        111100 ..... 00 ........ 0101101000 .   @X_imm8
-XXSPLTW         111100 ..... ---.. ..... 010100100 . .  @XX2
+XXSPLTW         111100 ..... ---.. ..... 010100100 . .  @XX2_uim2
 
 ## VSX Permute Instructions
 
@@ -675,6 +678,8 @@  XSCMPGTQP       111111 ..... ..... ..... 0011100100 -   @X
 ## VSX Binary Floating-Point Convert Instructions
 
 XSCVQPDP        111111 ..... 10100 ..... 1101000100 .   @X_tb_rc
+XVCVBF16SPN     111100 ..... 10000 ..... 111011011 ..   @XX2
+XVCVSPBF16      111100 ..... 10001 ..... 111011011 ..   @XX2
 
 ## VSX Vector Test Least-Significant Bit by Byte Instruction
 
diff --git a/target/ppc/translate/vsx-impl.c.inc b/target/ppc/translate/vsx-impl.c.inc
index 0546dc736e..2930537b8e 100644
--- a/target/ppc/translate/vsx-impl.c.inc
+++ b/target/ppc/translate/vsx-impl.c.inc
@@ -1576,7 +1576,7 @@  static bool trans_XXSEL(DisasContext *ctx, arg_XX4 *a)
     return true;
 }
 
-static bool trans_XXSPLTW(DisasContext *ctx, arg_XX2 *a)
+static bool trans_XXSPLTW(DisasContext *ctx, arg_XX2_uim2 *a)
 {
     int tofs, bofs;
 
@@ -2532,6 +2532,35 @@  TRANS(XSCMPGTQP, do_xscmpqp, gen_helper_XSCMPGTQP)
 TRANS(XSMAXCQP, do_xscmpqp, gen_helper_XSMAXCQP)
 TRANS(XSMINCQP, do_xscmpqp, gen_helper_XSMINCQP)
 
+static bool trans_XVCVSPBF16(DisasContext *ctx, arg_XX2 *a)
+{
+    TCGv_ptr xt, xb;
+
+    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
+    REQUIRE_VSX(ctx);
+
+    xt = gen_vsr_ptr(a->xt);
+    xb = gen_vsr_ptr(a->xb);
+
+    gen_helper_XVCVSPBF16(cpu_env, xt, xb);
+
+    tcg_temp_free_ptr(xt);
+    tcg_temp_free_ptr(xb);
+
+    return true;
+}
+
+static bool trans_XVCVBF16SPN(DisasContext *ctx, arg_XX2 *a)
+{
+    REQUIRE_INSNS_FLAGS2(ctx, ISA310);
+    REQUIRE_VSX(ctx);
+
+    tcg_gen_gvec_shli(MO_32, vsr_full_offset(a->xt), vsr_full_offset(a->xb),
+                      16, 16, 16);
+
+    return true;
+}
+
 #undef GEN_XX2FORM
 #undef GEN_XX3FORM
 #undef GEN_XX2IFORM