[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PATCH 12/41] target/sparc: Implement FMAf extension
From: |
Richard Henderson |
Subject: |
[PATCH 12/41] target/sparc: Implement FMAf extension |
Date: |
Fri, 1 Mar 2024 19:15:32 -1000 |
Rearrange PDIST so that do_dddd is general purpose and may
be re-used for FMADDd etc.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
target/sparc/helper.h | 2 +
linux-user/elfload.c | 1 +
target/sparc/cpu.c | 3 ++
target/sparc/fop_helper.c | 16 +++++++
target/sparc/translate.c | 84 ++++++++++++++++++++++++++++++++--
target/sparc/cpu-feature.h.inc | 1 +
target/sparc/insns.decode | 23 +++++++++-
7 files changed, 124 insertions(+), 6 deletions(-)
diff --git a/target/sparc/helper.h b/target/sparc/helper.h
index f0576fb748..63ae398841 100644
--- a/target/sparc/helper.h
+++ b/target/sparc/helper.h
@@ -53,6 +53,7 @@ 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)
@@ -63,6 +64,7 @@ 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/linux-user/elfload.c b/linux-user/elfload.c
index 6041270f1c..5ebf2bf789 100644
--- a/linux-user/elfload.c
+++ b/linux-user/elfload.c
@@ -997,6 +997,7 @@ 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.c b/target/sparc/cpu.c
index 313ebc4c11..491e627899 100644
--- a/target/sparc/cpu.c
+++ b/target/sparc/cpu.c
@@ -551,6 +551,7 @@ 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",
@@ -873,6 +874,8 @@ 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 1205a599ef..1de44d79c1 100644
--- 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/translate.c b/target/sparc/translate.c
index ee3da73551..1178fca9e3 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -29,6 +29,7 @@
#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"
@@ -1151,6 +1152,52 @@ 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)
{
/*
@@ -2093,6 +2140,7 @@ 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)
@@ -2105,6 +2153,7 @@ 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
@@ -4762,25 +4811,52 @@ 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))
diff --git a/target/sparc/cpu-feature.h.inc b/target/sparc/cpu-feature.h.inc
index d800f18c4e..a30b9255b2 100644
--- 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/insns.decode b/target/sparc/insns.decode
index 2c23868fc3..6d5fa26e90 100644
--- 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 ..... @d_d2
FPACKFIX 10 ..... 110110 00000 0 0011 1101 ..... @d_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
--
2.34.1
- [PATCH 03/41] target/sparc: Fix FMUL8x16, (continued)
- [PATCH 03/41] target/sparc: Fix FMUL8x16, Richard Henderson, 2024/03/02
- [PATCH 04/41] target/sparc: Fix FMUL8x16A{U,L}, Richard Henderson, 2024/03/02
- [PATCH 01/41] linux-user/sparc: Add more hwcap bits for sparc64, Richard Henderson, 2024/03/02
- [PATCH 05/41] target/sparc: Fix FMULD8*X16, Richard Henderson, 2024/03/02
- [PATCH 02/41] target/sparc: Fix FEXPAND, Richard Henderson, 2024/03/02
- [PATCH 06/41] target/sparc: Fix FPMERGE, Richard Henderson, 2024/03/02
- [PATCH 07/41] target/sparc: Split out do_ms16b, Richard Henderson, 2024/03/02
- [PATCH 09/41] target/sparc: Remove gen_dest_fpr_D, Richard Henderson, 2024/03/02
- [PATCH 08/41] target/sparc: Perform DFPREG/QFPREG in decodetree, Richard Henderson, 2024/03/02
- [PATCH 10/41] target/sparc: Remove cpu_fpr[], Richard Henderson, 2024/03/02
- [PATCH 12/41] target/sparc: Implement FMAf extension,
Richard Henderson <=
- [PATCH 13/41] target/sparc: Add feature bits for VIS 3, Richard Henderson, 2024/03/02
- [PATCH 14/41] target/sparc: Implement ADDXC, ADDXCcc, Richard Henderson, 2024/03/02
- [PATCH 15/41] target/sparc: Implement CMASK instructions, Richard Henderson, 2024/03/02
- [PATCH 16/41] target/sparc: Implement FCHKSM16, Richard Henderson, 2024/03/02
- [PATCH 17/41] target/sparc: Implement FHADD, FHSUB, FNHADD, FNADD, Richard Henderson, 2024/03/02
- [PATCH 11/41] target/sparc: Use gvec for VIS1 parallel add/sub, Richard Henderson, 2024/03/02
- [PATCH 18/41] target/sparc: Implement FNMUL, Richard Henderson, 2024/03/02
- [PATCH 19/41] target/sparc: Implement FLCMP, Richard Henderson, 2024/03/02
- [PATCH 20/41] target/sparc: Implement FMEAN16, Richard Henderson, 2024/03/02
- [PATCH 22/41] target/sparc: Implement FPADDS, FPSUBS, Richard Henderson, 2024/03/02