[PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps

[PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps

[Richard Henderson]

Hi Carl,

While digging through the manual to figure out if we were really
doing the right thing raising the fp exception, I found the fpu
exception state machine.  I'm not sure it's worth emulating the
fp_exception_pending state, but it's certainly easy enough to
emulate the fp_executing/fp_exception states.

In addition, this simplifies the implementation of STDFQ,
restricts FQ to sparc32 system mode, and handles migration.

Would you please double-check this against your Solaris images?


r~


Carl Hauser (2):
  target/sparc: Add FQ and FSR.QNE
  target/sparc: Populate sparc32 FQ when raising fp exception

Richard Henderson (4):
  target/sparc: Restrict STQF to sparcv9
  target/sparc: Add FSR_QNE to tb_flags
  target/sparc: Implement STDFQ
  target/sparc: Add gen_trap_if_nofpu_fpexception

 target/sparc/cpu.h          |  30 ++++++++-
 target/sparc/fop_helper.c   |   4 ++
 target/sparc/int32_helper.c |  32 ++++-----
 target/sparc/machine.c      |  25 +++++++
 target/sparc/translate.c    | 126 ++++++++++++++++++++++++++----------
 target/sparc/insns.decode   |   4 +-
 6 files changed, 169 insertions(+), 52 deletions(-)

-- 
2.43.0

[PATCH v3 1/6] target/sparc: Restrict STQF to sparcv9

[Richard Henderson]

Prior to sparcv9, the same encoding was STDFQ.

Cc: qemu-stable@nongnu.org
Fixes: 06c060d9e5b ("target/sparc: Move simple fp load/store to decodetree")
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/sparc/translate.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index 113639083b..c803e8d1ba 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -4521,7 +4521,7 @@ static bool do_st_fpr(DisasContext *dc, arg_r_r_ri_asi *a, MemOp sz)
 
 TRANS(STF, ALL, do_st_fpr, a, MO_32)
 TRANS(STDF, ALL, do_st_fpr, a, MO_64)
-TRANS(STQF, ALL, do_st_fpr, a, MO_128)
+TRANS(STQF, 64, do_st_fpr, a, MO_128)
 
 TRANS(STFA, 64, do_st_fpr, a, MO_32)
 TRANS(STDFA, 64, do_st_fpr, a, MO_64)
-- 
2.43.0

Re: [PATCH v3 1/6] target/sparc: Restrict STQF to sparcv9

[Philippe Mathieu-Daudé]

On 16/8/24 09:23, Richard Henderson wrote:
> Prior to sparcv9, the same encoding was STDFQ.
> 
> Cc: qemu-stable@nongnu.org
> Fixes: 06c060d9e5b ("target/sparc: Move simple fp load/store to decodetree")
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>   target/sparc/translate.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>

[PATCH v3 2/6] target/sparc: Add FQ and FSR.QNE

[Richard Henderson]

From: Carl Hauser <chauser@pullman.com>

Add support for, and migrate, a single-entry fp
instruction queue for sparc32.

Signed-off-by: Carl Hauser <chauser@pullman.com>
[rth: Split from a larger patch;
      adjust representation with union;
      add migration state]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/sparc/cpu.h        | 22 ++++++++++++++++++++++
 target/sparc/fop_helper.c |  4 ++++
 target/sparc/machine.c    | 25 +++++++++++++++++++++++++
 3 files changed, 51 insertions(+)

diff --git a/target/sparc/cpu.h b/target/sparc/cpu.h
index dfd9512a21..9f2bc44722 100644
--- a/target/sparc/cpu.h
+++ b/target/sparc/cpu.h
@@ -184,6 +184,8 @@ enum {
 #define FSR_FTT_SEQ_ERROR (4ULL << 14)
 #define FSR_FTT_INVAL_FPR (6ULL << 14)
 
+#define FSR_QNE    (1ULL << 13)
+
 #define FSR_FCC0_SHIFT    10
 #define FSR_FCC1_SHIFT    32
 #define FSR_FCC2_SHIFT    34
@@ -438,6 +440,26 @@ struct CPUArchState {
     uint32_t fsr_cexc_ftt;           /* cexc, ftt */
     uint32_t fcc[TARGET_FCCREGS];    /* fcc* */
 
+#if !defined(TARGET_SPARC64) && !defined(CONFIG_USER_ONLY)
+    /*
+     * Single-element FPU fault queue, with address and insn,
+     * packaged into the double-word with which it is stored.
+     */
+    uint32_t fsr_qne;                /* qne */
+    union {
+        uint64_t d;
+        struct {
+#if HOST_BIG_ENDIAN
+            uint32_t addr;
+            uint32_t insn;
+#else
+            uint32_t insn;
+            uint32_t addr;
+#endif
+        } s;
+    } fq;
+#endif
+
     CPU_DoubleU fpr[TARGET_DPREGS];  /* floating point registers */
     uint32_t cwp;      /* index of current register window (extracted
                           from PSR) */
diff --git a/target/sparc/fop_helper.c b/target/sparc/fop_helper.c
index 0b30665b51..b6692382b3 100644
--- a/target/sparc/fop_helper.c
+++ b/target/sparc/fop_helper.c
@@ -545,6 +545,8 @@ target_ulong cpu_get_fsr(CPUSPARCState *env)
     fsr |= (uint64_t)env->fcc[1] << FSR_FCC1_SHIFT;
     fsr |= (uint64_t)env->fcc[2] << FSR_FCC2_SHIFT;
     fsr |= (uint64_t)env->fcc[3] << FSR_FCC3_SHIFT;
+#elif !defined(CONFIG_USER_ONLY)
+    fsr |= env->fsr_qne;
 #endif
 
     /* VER is kept completely separate until re-assembly. */
@@ -591,6 +593,8 @@ void cpu_put_fsr(CPUSPARCState *env, target_ulong fsr)
     env->fcc[1] = extract64(fsr, FSR_FCC1_SHIFT, 2);
     env->fcc[2] = extract64(fsr, FSR_FCC2_SHIFT, 2);
     env->fcc[3] = extract64(fsr, FSR_FCC3_SHIFT, 2);
+#elif !defined(CONFIG_USER_ONLY)
+    env->fsr_qne = fsr & FSR_QNE;
 #endif
 
     set_fsr_nonsplit(env, fsr);
diff --git a/target/sparc/machine.c b/target/sparc/machine.c
index 48e0cf22f3..222e5709c5 100644
--- a/target/sparc/machine.c
+++ b/target/sparc/machine.c
@@ -143,6 +143,24 @@ static const VMStateInfo vmstate_xcc = {
     .get = get_xcc,
     .put = put_xcc,
 };
+#else
+static bool fq_needed(void *opaque)
+{
+    SPARCCPU *cpu = opaque;
+    return cpu->env.fsr_qne;
+}
+
+static const VMStateDescription vmstate_fq = {
+    .name = "cpu/fq",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .needed = fq_needed,
+    .fields = (const VMStateField[]) {
+        VMSTATE_UINT32(env.fq.s.addr, SPARCCPU),
+        VMSTATE_UINT32(env.fq.s.insn, SPARCCPU),
+        VMSTATE_END_OF_LIST()
+    },
+};
 #endif
 
 static int cpu_pre_save(void *opaque)
@@ -265,4 +283,11 @@ const VMStateDescription vmstate_sparc_cpu = {
 #endif
         VMSTATE_END_OF_LIST()
     },
+#ifndef TARGET_SPARC64
+    .subsections = (const VMStateDescription * const []) {
+        &vmstate_fq,
+        NULL
+    },
+#endif
+
 };
-- 
2.43.0

[PATCH v3 3/6] target/sparc: Populate sparc32 FQ when raising fp exception

[Richard Henderson]

From: Carl Hauser <chauser@pullman.com>

Implement a single instruction floating point queue,
populated while delivering an fp exception.

Signed-off-by: Carl Hauser <chauser@pullman.com>
[rth: Split from a larger patch]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/sparc/int32_helper.c | 32 +++++++++++++++++---------------
 1 file changed, 17 insertions(+), 15 deletions(-)

diff --git a/target/sparc/int32_helper.c b/target/sparc/int32_helper.c
index 6b7d65b031..fb6f3799c8 100644
--- a/target/sparc/int32_helper.c
+++ b/target/sparc/int32_helper.c
@@ -21,10 +21,10 @@
 #include "qemu/main-loop.h"
 #include "cpu.h"
 #include "trace.h"
+#include "exec/cpu_ldst.h"
 #include "exec/log.h"
 #include "sysemu/runstate.h"
 
-
 static const char * const excp_names[0x80] = {
     [TT_TFAULT] = "Instruction Access Fault",
     [TT_ILL_INSN] = "Illegal Instruction",
@@ -116,22 +116,9 @@ void sparc_cpu_do_interrupt(CPUState *cs)
 
         qemu_log("%6d: %s (v=%02x)\n", count, name, intno);
         log_cpu_state(cs, 0);
-#if 0
-        {
-            int i;
-            uint8_t *ptr;
-
-            qemu_log("       code=");
-            ptr = (uint8_t *)env->pc;
-            for (i = 0; i < 16; i++) {
-                qemu_log(" %02x", ldub(ptr + i));
-            }
-            qemu_log("\n");
-        }
-#endif
         count++;
     }
-#if !defined(CONFIG_USER_ONLY)
+#ifndef CONFIG_USER_ONLY
     if (env->psret == 0) {
         if (cs->exception_index == 0x80 &&
             env->def.features & CPU_FEATURE_TA0_SHUTDOWN) {
@@ -143,6 +130,21 @@ void sparc_cpu_do_interrupt(CPUState *cs)
         }
         return;
     }
+    if (intno == TT_FP_EXCP) {
+        env->fsr_qne = FSR_QNE;
+        env->fq.s.addr = env->pc;
+        env->fq.s.insn = cpu_ldl_code(env, env->pc);
+        /*
+         * The sparc32 fpu has three states related to exception handling.
+         * The FPop that signals an exception transitions from fp_execute
+         * to fp_exception_pending.  A subsequent FPop transitions from
+         * fp_exception_pending to fp_exception, which forces the trap.
+         * We do not model the fp_exception_pending state, but we do need
+         * to advance pc/npc to mimic the delayed trap delivery.
+         */
+        env->pc = env->npc;
+        env->npc = env->npc + 4;
+    }
 #endif
     env->psret = 0;
     cwp = cpu_cwp_dec(env, env->cwp - 1);
-- 
2.43.0

[PATCH v3 4/6] target/sparc: Add FSR_QNE to tb_flags

[Richard Henderson]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/sparc/cpu.h       |  8 +++++++-
 target/sparc/translate.c | 10 +++++++---
 2 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/target/sparc/cpu.h b/target/sparc/cpu.h
index 9f2bc44722..f517e5a383 100644
--- a/target/sparc/cpu.h
+++ b/target/sparc/cpu.h
@@ -744,6 +744,7 @@ trap_state* cpu_tsptr(CPUSPARCState* env);
 #define TB_FLAG_AM_ENABLED   (1 << 5)
 #define TB_FLAG_SUPER        (1 << 6)
 #define TB_FLAG_HYPER        (1 << 7)
+#define TB_FLAG_FSR_QNE      (1 << 8)
 #define TB_FLAG_ASI_SHIFT    24
 
 static inline void cpu_get_tb_cpu_state(CPUSPARCState *env, vaddr *pc,
@@ -775,7 +776,12 @@ static inline void cpu_get_tb_cpu_state(CPUSPARCState *env, vaddr *pc,
     if (env->psref) {
         flags |= TB_FLAG_FPU_ENABLED;
     }
-#endif
+#ifndef CONFIG_USER_ONLY
+    if (env->fsr_qne) {
+        flags |= TB_FLAG_FSR_QNE;
+    }
+#endif /* !CONFIG_USER_ONLY */
+#endif /* TARGET_SPARC64 */
     *pflags = flags;
 }
 
diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index c803e8d1ba..eb0158a11d 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -185,6 +185,8 @@ typedef struct DisasContext {
     bool supervisor;
 #ifdef TARGET_SPARC64
     bool hypervisor;
+#else
+    bool fsr_qne;
 #endif
 #endif
 
@@ -5596,13 +5598,15 @@ static void sparc_tr_init_disas_context(DisasContextBase *dcbase, CPUState *cs)
     dc->address_mask_32bit = tb_am_enabled(dc->base.tb->flags);
 #ifndef CONFIG_USER_ONLY
     dc->supervisor = (dc->base.tb->flags & TB_FLAG_SUPER) != 0;
+# ifdef TARGET_SPARC64
+    dc->hypervisor = (dc->base.tb->flags & TB_FLAG_HYPER) != 0;
+# else
+    dc->fsr_qne = (dc->base.tb->flags & TB_FLAG_FSR_QNE) != 0;
+# endif
 #endif
 #ifdef TARGET_SPARC64
     dc->fprs_dirty = 0;
     dc->asi = (dc->base.tb->flags >> TB_FLAG_ASI_SHIFT) & 0xff;
-#ifndef CONFIG_USER_ONLY
-    dc->hypervisor = (dc->base.tb->flags & TB_FLAG_HYPER) != 0;
-#endif
 #endif
     /*
      * if we reach a page boundary, we stop generation so that the
-- 
2.43.0

[PATCH v3 5/6] target/sparc: Implement STDFQ

[Richard Henderson]

Invalid encoding of addr should raise TT_ILL_INSN, so
check before supervisor, which might raise TT_PRIV_INSN.
Clear QNE after execution.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/sparc/translate.c  | 24 +++++++++++++++++++++++-
 target/sparc/insns.decode |  4 ++--
 2 files changed, 25 insertions(+), 3 deletions(-)

diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index eb0158a11d..6d4c0e79c9 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -4531,17 +4531,39 @@ TRANS(STQFA, 64, do_st_fpr, a, MO_128)
 
 static bool trans_STDFQ(DisasContext *dc, arg_STDFQ *a)
 {
+    TCGv addr = gen_ldst_addr(dc, a->rs1, a->imm, a->rs2_or_imm);
+
     if (!avail_32(dc)) {
         return false;
     }
+    if (addr == NULL) {
+        return false;
+    }
     if (!supervisor(dc)) {
         return raise_priv(dc);
     }
+#if !defined(TARGET_SPARC64) && !defined(CONFIG_USER_ONLY)
     if (gen_trap_ifnofpu(dc)) {
         return true;
     }
-    gen_op_fpexception_im(dc, FSR_FTT_SEQ_ERROR);
+    if (dc->fsr_qne) {
+        TCGv_i64 fq = tcg_temp_new_i64();
+
+        /* Store the single element from the queue. */
+        tcg_gen_ld_i64(fq, tcg_env, offsetof(CPUSPARCState, fq.d));
+        tcg_gen_qemu_st_i64(fq, addr, dc->mem_idx, MO_TEUQ | MO_ALIGN_4);
+
+        /* Mark the queue empty, transitioning to fp_execute state. */
+        tcg_gen_st_i32(tcg_constant_i32(0), tcg_env,
+                       offsetof(CPUSPARCState, fsr_qne));
+        dc->fsr_qne = 0;
+    } else {
+        gen_op_fpexception_im(dc, FSR_FTT_SEQ_ERROR);
+    }
     return true;
+#else
+    qemu_build_not_reached();
+#endif
 }
 
 static bool trans_LDFSR(DisasContext *dc, arg_r_r_ri *a)
diff --git a/target/sparc/insns.decode b/target/sparc/insns.decode
index fbcb4f7aef..923f348580 100644
--- a/target/sparc/insns.decode
+++ b/target/sparc/insns.decode
@@ -644,8 +644,8 @@ STF         11 ..... 100100 ..... . .............          @r_r_ri_na
 STFSR       11 00000 100101 ..... . .............          @n_r_ri
 STXFSR      11 00001 100101 ..... . .............          @n_r_ri
 {
-  STQF      11 ..... 100110 ..... . .............          @q_r_ri_na
-  STDFQ     11 ----- 100110 ----- - -------------
+  STQF      11 ..... 100110 ..... . .............          @q_r_ri_na # v9
+  STDFQ     11 ..... 100110 ..... . .............          @r_r_ri    # v7,v8
 }
 STDF        11 ..... 100111 ..... . .............          @d_r_ri_na
 
-- 
2.43.0

[PATCH v3 6/6] target/sparc: Add gen_trap_if_nofpu_fpexception

[Richard Henderson]

Model fp_exception state, in which only fp stores are allowed
until such time as the FQ has been flushed.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/sparc/translate.c | 90 +++++++++++++++++++++++++++-------------
 1 file changed, 61 insertions(+), 29 deletions(-)

diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index 6d4c0e79c9..46eb27c497 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -1465,15 +1465,48 @@ static void gen_op_fpexception_im(DisasContext *dc, int ftt)
     gen_exception(dc, TT_FP_EXCP);
 }
 
-static int gen_trap_ifnofpu(DisasContext *dc)
+static bool gen_trap_ifnofpu(DisasContext *dc)
 {
 #if !defined(CONFIG_USER_ONLY)
     if (!dc->fpu_enabled) {
         gen_exception(dc, TT_NFPU_INSN);
-        return 1;
+        return true;
     }
 #endif
-    return 0;
+    return false;
+}
+
+static bool gen_trap_iffpexception(DisasContext *dc)
+{
+#if !defined(TARGET_SPARC64) && !defined(CONFIG_USER_ONLY)
+    /*
+     * There are 3 states for the sparc32 fpu:
+     * Normally the fpu is in fp_execute, and all insns are allowed.
+     * When an exception is signaled, it moves to fp_exception_pending state.
+     * Upon seeing the next FPop, the fpu moves to fp_exception state,
+     * populates the FQ, and generates an fp_exception trap.
+     * The fpu remains in fp_exception state until FQ becomes empty
+     * after execution of a STDFQ instruction.  While the fpu is in
+     * fp_exception state, and FPop, fp load or fp branch insn will
+     * return to fp_exception_pending state, set FSR.FTT to sequence_error,
+     * and the insn will not be entered into the FQ.
+     *
+     * In QEMU, we do not model the fp_exception_pending state and
+     * instead populate FQ and raise the exception immediately.
+     * But we can still honor fp_exception state by noticing when
+     * the FQ is not empty.
+     */
+    if (dc->fsr_qne) {
+        gen_op_fpexception_im(dc, FSR_FTT_SEQ_ERROR);
+        return true;
+    }
+#endif
+    return false;
+}
+
+static bool gen_trap_if_nofpu_fpexception(DisasContext *dc)
+{
+    return gen_trap_ifnofpu(dc) || gen_trap_iffpexception(dc);
 }
 
 /* asi moves */
@@ -2643,7 +2676,7 @@ static bool do_fbpfcc(DisasContext *dc, arg_bcc *a)
 {
     DisasCompare cmp;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     gen_fcompare(&cmp, a->cc, a->cond);
@@ -4482,7 +4515,7 @@ static bool do_ld_fpr(DisasContext *dc, arg_r_r_ri_asi *a, MemOp sz)
     if (addr == NULL) {
         return false;
     }
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (sz == MO_128 && gen_trap_float128(dc)) {
@@ -4510,6 +4543,7 @@ static bool do_st_fpr(DisasContext *dc, arg_r_r_ri_asi *a, MemOp sz)
     if (addr == NULL) {
         return false;
     }
+    /* Store insns are ok in fp_exception_pending state. */
     if (gen_trap_ifnofpu(dc)) {
         return true;
     }
@@ -4574,7 +4608,7 @@ static bool trans_LDFSR(DisasContext *dc, arg_r_r_ri *a)
     if (addr == NULL) {
         return false;
     }
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4598,7 +4632,7 @@ static bool do_ldxfsr(DisasContext *dc, arg_r_r_ri *a, bool entire)
     if (addr == NULL) {
         return false;
     }
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4635,6 +4669,7 @@ static bool do_stfsr(DisasContext *dc, arg_r_r_ri *a, MemOp mop)
     if (addr == NULL) {
         return false;
     }
+    /* Store insns are ok in fp_exception_pending state. */
     if (gen_trap_ifnofpu(dc)) {
         return true;
     }
@@ -4677,7 +4712,7 @@ static bool do_ff(DisasContext *dc, arg_r_r *a,
 {
     TCGv_i32 tmp;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4718,7 +4753,7 @@ static bool do_env_ff(DisasContext *dc, arg_r_r *a,
 {
     TCGv_i32 tmp;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4738,7 +4773,7 @@ static bool do_env_fd(DisasContext *dc, arg_r_r *a,
     TCGv_i32 dst;
     TCGv_i64 src;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4758,7 +4793,7 @@ static bool do_dd(DisasContext *dc, arg_r_r *a,
 {
     TCGv_i64 dst, src;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4780,7 +4815,7 @@ static bool do_env_dd(DisasContext *dc, arg_r_r *a,
 {
     TCGv_i64 dst, src;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4820,7 +4855,7 @@ static bool do_env_df(DisasContext *dc, arg_r_r *a,
     TCGv_i64 dst;
     TCGv_i32 src;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -4863,7 +4898,7 @@ static bool do_env_qq(DisasContext *dc, arg_r_r *a,
 {
     TCGv_i128 t;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (gen_trap_float128(dc)) {
@@ -4884,7 +4919,7 @@ static bool do_env_fq(DisasContext *dc, arg_r_r *a,
     TCGv_i128 src;
     TCGv_i32 dst;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (gen_trap_float128(dc)) {
@@ -4907,7 +4942,7 @@ static bool do_env_dq(DisasContext *dc, arg_r_r *a,
     TCGv_i128 src;
     TCGv_i64 dst;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (gen_trap_float128(dc)) {
@@ -4930,7 +4965,7 @@ static bool do_env_qf(DisasContext *dc, arg_r_r *a,
     TCGv_i32 src;
     TCGv_i128 dst;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (gen_trap_float128(dc)) {
@@ -4953,10 +4988,7 @@ static bool do_env_qd(DisasContext *dc, arg_r_r *a,
     TCGv_i64 src;
     TCGv_i128 dst;
 
-    if (gen_trap_ifnofpu(dc)) {
-        return true;
-    }
-    if (gen_trap_float128(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -5013,7 +5045,7 @@ static bool do_env_fff(DisasContext *dc, arg_r_r_r *a,
 {
     TCGv_i32 src1, src2;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -5222,7 +5254,7 @@ static bool do_env_ddd(DisasContext *dc, arg_r_r_r *a,
 {
     TCGv_i64 dst, src1, src2;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -5246,7 +5278,7 @@ static bool trans_FsMULd(DisasContext *dc, arg_r_r_r *a)
     TCGv_i64 dst;
     TCGv_i32 src1, src2;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (!(dc->def->features & CPU_FEATURE_FSMULD)) {
@@ -5355,7 +5387,7 @@ static bool do_env_qqq(DisasContext *dc, arg_r_r_r *a,
 {
     TCGv_i128 src1, src2;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (gen_trap_float128(dc)) {
@@ -5379,7 +5411,7 @@ static bool trans_FdMULq(DisasContext *dc, arg_r_r_r *a)
     TCGv_i64 src1, src2;
     TCGv_i128 dst;
 
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (gen_trap_float128(dc)) {
@@ -5469,7 +5501,7 @@ static bool do_fcmps(DisasContext *dc, arg_FCMPs *a, bool e)
     if (avail_32(dc) && a->cc != 0) {
         return false;
     }
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -5493,7 +5525,7 @@ static bool do_fcmpd(DisasContext *dc, arg_FCMPd *a, bool e)
     if (avail_32(dc) && a->cc != 0) {
         return false;
     }
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
 
@@ -5517,7 +5549,7 @@ static bool do_fcmpq(DisasContext *dc, arg_FCMPq *a, bool e)
     if (avail_32(dc) && a->cc != 0) {
         return false;
     }
-    if (gen_trap_ifnofpu(dc)) {
+    if (gen_trap_if_nofpu_fpexception(dc)) {
         return true;
     }
     if (gen_trap_float128(dc)) {
-- 
2.43.0

Re: [PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps

[Carl Hauser]

[-- Attachment #1: Type: text/html, Size: 4169 bytes --]

Re: [PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps

[Carl Hauser]

[-- Attachment #1: Type: text/html, Size: 4711 bytes --]

Re: [PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps

[Richard Henderson]

On 8/17/24 07:46, Carl Hauser wrote:
> OK, I think the problem is the handling of dc->fsr_qne in trans_STDFQ, lines 4583 and 4593 
> -- the code is evaluating dc->fsr_qne at translation time and not at runtime.

That's what patch 4 does, ensure that the runtime value is available at translation time.


r~

Re: [PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps

[Richard Henderson]

On 8/17/24 08:58, Carl Hauser wrote:
> Yes, but ...
> 
> isn't the state of dc->fsr_qne at translation time irrelevant?

No, because patch 4 made it part of the hashed TB state.
It's checked and verified, generating a new TB if state does not match.

> And changing it at 
> translation time (line 4593) is dangerous (because it pertains to runtime, not translation 
> time); i.e. why is 0 stored at both translation time (4593) and at runtime (4591)?

That keeps the translation time state in sync with the runtime state until the end of the TB.


r~

Re: [PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps

[Richard Henderson]

On 8/17/24 09:48, Carl Hauser wrote:
> netbsd panics in the kernel trap handler; unfortunately it does not include the fsr in the 
> panic message, but I expect it will be the same as on Solaris.

Ok, I have reproduced this with netbsd 9.3.
I'll have a look.


r~

Re: [PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps -- CORRECTION

[Richard Henderson]

On 8/18/24 10:03, Carl Hauser wrote:
> I changed translate.c:4597 from return true; to return advance_pc(dc); and it worked.

Whoops, yes indeed.  Thanks for the catch.


r~

Re: [PATCH for-9.2 v3 0/6] target/sparc: emulate floating point queue when raising fp traps -- CORRECTION

[Richard Henderson]

On 8/21/24 02:59, Carl Hauser wrote:
> Do you want me to submit a patch set fixing this or will you?

I will.


r~