Skip to contentSkip to conversation

target/riscv: Enhanced ISA extension checks

PATCH Applied Tsukasa OI 4 participants 27 messages May 13, 2022 – May 24, 2022 RISC-V

lore
Share & tools
lore

Tsukasa OI

author
#0 lore raw

cover target/riscv: Enhanced ISA extension checks

Hello,

This is another patchset for RISC-V ISA extension / feature handling.

Aside from coding style fix / refactoring patch (PATCH 1 and 5), this patchset contains two changes:

1. "G" extension handling

1.A. "G" extension expansion (PATCH 3)

On ISA version 20190608 or later, "G" expands to "IMAFD_Zicsr_Zifencei", not just "IMAFD" (this is because "Zicsr" and "Zifencei" extensions are splitted from "I"). Because both "Zicsr" and "Zifencei" are enabled by default, it should be safe to change "G" extension expansion.

1.B. Disable "G" by default (PATCH 2)

This seems quite odd but I have a good reason. While "G" is enabled by default, all "I", "M", "A", "F", "D", "Zicsr" and "Zifencei" are also enabled by default, making default "G" expansion useless.

There's more. If we want to change detailed configuration of a RV32/RV64 CPU and want to disable some, for example, "F" and "D", we must also disable "G". This is obviously bloat.

This doesn't work:

    -cpu rv64,f=off,d=off

This does work (but bloat):

    -cpu rv64,g=off,f=off,d=off

Disabling "G" suppresses such problem and mostly harmless, too.

2. Floating point arithmetic consistency (PATCH 4)

With -cpu option, we can configure details of RISC-V CPU emulated by QEMU. However, we can set some invalid combinations that would make FP arithmetic effectively unusable (and invalid on RISC-V ISA specification).

-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D

Reproducing this kind of problem requires manually disabling certain extensions (because all "Zicsr", "F" and "D" are enabled by default). So, I consider that just making an error message (when unsatisfied) should be enough, not implying related extensions like "zk*" properties.

Note that this checking only works on any, rv32 and rv64. On other CPUs (for example, sifive-u54), it sets nonzero misa' value on corresponding set_misa' call (c.f. in rv64_sifive_u_cpu_init in target/riscv/cpu.c) and consistency checks are skipped (because nonzero `misa' value is set prior to RISC-V CPU realization process).

I think we generally use generic "rv32" or "rv64" on heavy customizing so I don't think this is not a big problem. Still, we could fix this later (e.g. by setting properties on CPU init function or by checking some consistency problems even if previously-set `misa' is nonzero).

Tsukasa OI (5):
target/riscv: Fix "G" extension expansion typing target/riscv: Disable "G" by default target/riscv: Change "G" expansion target/riscv: FP extension requirements target/riscv: Move/refactor ISA extension checks

 target/riscv/cpu.c | 62 +++++++++++++++++++++++++++++++++-------------
 1 file changed, 45 insertions(+), 17 deletions(-)

base-commit: 178bacb66d98d9ee7a702b9f2a4dfcd88b72a9ab

Signature · 1 line
2.34.1
author
#1 lore raw

PATCH 1/5 target/riscv: Fix "G" extension expansion typing

Because ext_? members are in bool type, operator &&' should be used instead of &'.

---
 target/riscv/cpu.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
target/riscv/cpu.c +2 -2
index ccacdee215..00bf26ec8b 100644
@@ -596,8 +596,8 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             return;
         }
 
-        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i & cpu->cfg.ext_m &
-                                cpu->cfg.ext_a & cpu->cfg.ext_f &
+        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
+                                cpu->cfg.ext_a && cpu->cfg.ext_f &&
                                 cpu->cfg.ext_d)) {
             warn_report("Setting G will also set IMAFD");
             cpu->cfg.ext_i = true;
Signature · 1 line
2.34.1
Reviewed-by on v1 1/5
#2 lore raw

Re: [PATCH 1/5] target/riscv: Fix "G" extension expansion typing

Quoted 26 lines · Tsukasa OI

[E-MAIL EXTERNO] Não clique em links ou abra anexos, a menos que você possa confirmar o remetente e saber que o conteúdo é seguro. Em caso de e-mail suspeito entre imediatamente em contato com o DTI.

Because ext_? members are in bool type, operator &&' should be used instead of &'.

---
  target/riscv/cpu.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
target/riscv/cpu.c +2 -3
index ccacdee215..00bf26ec8b 100644
@@ -596,8 +596,8 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
              return;
          }
 
-        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i & cpu->cfg.ext_m &
-                                cpu->cfg.ext_a & cpu->cfg.ext_f &
+        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
+                                cpu->cfg.ext_a && cpu->cfg.ext_f &&
                                  cpu->cfg.ext_d)) {
              warn_report("Setting G will also set IMAFD");
              cpu->cfg.ext_i = true;
--
Quoted 3 lines

2.34.1

--
Víctor Cora Colombo
Instituto de Pesquisas ELDORADO
Aviso Legal - Disclaimer <https://www.eldorado.org.br/disclaimer.html>

author
#3 lore raw

PATCH 2/5 target/riscv: Disable "G" by default

In reply to Tsukasa OI

Because "G" virtual extension expands to "IMAFD", we cannot separately disable extensions like "F" or "D" without disabling "G". Because all "IMAFD" are enabled by default, it's harmless to disable "G" by default.

---
 target/riscv/cpu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/cpu.c +1 -1
index 00bf26ec8b..3ea68d5cd7 100644
@@ -812,7 +812,7 @@ static Property riscv_cpu_properties[] = {
     /* Defaults for standard extensions */
     DEFINE_PROP_BOOL("i", RISCVCPU, cfg.ext_i, true),
     DEFINE_PROP_BOOL("e", RISCVCPU, cfg.ext_e, false),
-    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, true),
+    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, false),
     DEFINE_PROP_BOOL("m", RISCVCPU, cfg.ext_m, true),
     DEFINE_PROP_BOOL("a", RISCVCPU, cfg.ext_a, true),
     DEFINE_PROP_BOOL("f", RISCVCPU, cfg.ext_f, true),
Signature · 1 line
2.34.1
Reviewed-by on v1 2/5
#4 lore raw

Re: [PATCH 2/5] target/riscv: Disable "G" by default

Quoted 6 lines · Tsukasa OI

Because "G" virtual extension expands to "IMAFD", we cannot separately disable extensions like "F" or "D" without disabling "G". Because all "IMAFD" are enabled by default, it's harmless to disable "G" by default.

Alistair

Quoted 20 lines
---
 target/riscv/cpu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/cpu.c +1 -2
index 00bf26ec8b..3ea68d5cd7 100644
@@ -812,7 +812,7 @@ static Property riscv_cpu_properties[] = {
     /* Defaults for standard extensions */
     DEFINE_PROP_BOOL("i", RISCVCPU, cfg.ext_i, true),
     DEFINE_PROP_BOOL("e", RISCVCPU, cfg.ext_e, false),
-    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, true),
+    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, false),
     DEFINE_PROP_BOOL("m", RISCVCPU, cfg.ext_m, true),
     DEFINE_PROP_BOOL("a", RISCVCPU, cfg.ext_a, true),
     DEFINE_PROP_BOOL("f", RISCVCPU, cfg.ext_f, true),
--

2.34.1

author
#5 lore raw

PATCH 3/5 target/riscv: Change "G" expansion

In reply to Tsukasa OI

On ISA version 20190608 or later, "G" expands to "IMAFD_Zicsr_Zifencei". Both "Zicsr" and "Zifencei" are enabled by default and "G" is supposed to be (virtually) enabled as well, it should be safe to change its expansion.

---
 target/riscv/cpu.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
target/riscv/cpu.c +5 -2
index 3ea68d5cd7..0854ca9103 100644
@@ -598,13 +598,16 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
 
         if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
                                 cpu->cfg.ext_a && cpu->cfg.ext_f &&
-                                cpu->cfg.ext_d)) {
-            warn_report("Setting G will also set IMAFD");
+                                cpu->cfg.ext_d &&
+                                cpu->cfg.ext_icsr && cpu->cfg.ext_ifencei)) {
+            warn_report("Setting G will also set IMAFD_Zicsr_Zifencei");
             cpu->cfg.ext_i = true;
             cpu->cfg.ext_m = true;
             cpu->cfg.ext_a = true;
             cpu->cfg.ext_f = true;
             cpu->cfg.ext_d = true;
+            cpu->cfg.ext_icsr = true;
+            cpu->cfg.ext_ifencei = true;
         }
 
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
Signature · 1 line
2.34.1
author
#6 lore raw

PATCH 4/5 target/riscv: FP extension requirements

In reply to Tsukasa OI

QEMU allowed inconsistent configurations that made floating point arithmetic effectively unusable.

This commit adds certain checks for consistent FP arithmetic:

-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D

Because F/D/Zicsr are enabled by default (and an error will not occur unless we manually disable one or more of prerequisites), this commit just enforces the user to give consistent combinations.

---
 target/riscv/cpu.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
target/riscv/cpu.c +25 -0
index 0854ca9103..5371b0fd17 100644
@@ -610,6 +610,31 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             cpu->cfg.ext_ifencei = true;
         }
 
+        if (cpu->cfg.ext_f && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "F extension requires Zicsr");
+            return;
+        }
+
+        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "Zfinx extension requires Zicsr");
+            return;
+        }
+
+        if ((cpu->cfg.ext_zfh || cpu->cfg.ext_zfhmin) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zfh/Zfhmin extensions require F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_d && !cpu->cfg.ext_f) {
+            error_setg(errp, "D extension requires F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_v && !cpu->cfg.ext_d) {
+            error_setg(errp, "V extension requires D extension");
+            return;
+        }
+
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
             cpu->cfg.ext_zhinxmin) {
             cpu->cfg.ext_zfinx = true;
Signature · 1 line
2.34.1
author
#7 lore raw

PATCH 5/5 target/riscv: Move/refactor ISA extension checks

In reply to Tsukasa OI

We should separate "check" and "configure" steps as possible. This commit separates both steps except vector/Zfinx-related checks.

---
 target/riscv/cpu.c | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)
target/riscv/cpu.c +12 -12
index 5371b0fd17..f654a6727f 100644
@@ -635,11 +635,23 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             return;
         }
 
+        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zve32f/Zve64f extensions require F extension");
+            return;
+        }
+
+        /* Set the ISA extensions, checks should have happened above */
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
             cpu->cfg.ext_zhinxmin) {
             cpu->cfg.ext_zfinx = true;
         }
 
+        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_f) {
+            error_setg(errp,
+                    "Zfinx cannot be supported together with F extension");
+            return;
+        }
+
         if (cpu->cfg.ext_zk) {
             cpu->cfg.ext_zkn = true;
             cpu->cfg.ext_zkr = true;
@@ -663,7 +675,6 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             cpu->cfg.ext_zksh = true;
         }
 
-        /* Set the ISA extensions, checks should have happened above */
         if (cpu->cfg.ext_i) {
             ext |= RVI;
         }
@@ -734,20 +745,9 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             }
             set_vext_version(env, vext_version);
         }
-        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
-            error_setg(errp, "Zve32f/Zve64f extension depends upon RVF.");
-            return;
-        }
         if (cpu->cfg.ext_j) {
             ext |= RVJ;
         }
-        if (cpu->cfg.ext_zfinx && ((ext & (RVF | RVD)) || cpu->cfg.ext_zfh ||
-                                   cpu->cfg.ext_zfhmin)) {
-            error_setg(errp,
-                    "'Zfinx' cannot be supported together with 'F', 'D', 'Zfh',"
-                    " 'Zfhmin'");
-            return;
-        }
 
         set_misa(env, env->misa_mxl, ext);
     }
Signature · 1 line
2.34.1
author
#8 lore raw

v2 cover target/riscv: Enhanced ISA extension checks

In reply to Tsukasa OI

c.f.
<https://lists.gnu.org/archive/html/qemu-riscv/2022-05/msg00221.html>

I was obviously drunk when finalizing PATCH v1.

[BUGS in PATCH v1 (fixed in v2)]
PATCH 1: My English was (or, "is"?) broken
(commit subject and message is rewritten)
PATCH 4: Zfinx requirement test were in the wrong place PATCH 5: Zfinx/F exclusivity test was completely wrong

         I meant Zfinx&&F but when finalizing my patchset, I somehow
         changed this place to Zfinx&&!F.

Tsukasa OI (5):
target/riscv: Fix coding style on "G" expansion target/riscv: Disable "G" by default target/riscv: Change "G" expansion target/riscv: FP extension requirements target/riscv: Move/refactor ISA extension checks

 target/riscv/cpu.c | 63 +++++++++++++++++++++++++++++++++-------------
 1 file changed, 46 insertions(+), 17 deletions(-)

base-commit: 178bacb66d98d9ee7a702b9f2a4dfcd88b72a9ab

Signature · 1 line
2.34.1
author
#9 lore raw

PATCH v2 1/5 target/riscv: Fix coding style on "G" expansion

Because ext_? members are boolean variables, operator &&' should be used instead of &'.

---
 target/riscv/cpu.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
target/riscv/cpu.c +2 -2
index ccacdee215..00bf26ec8b 100644
@@ -596,8 +596,8 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             return;
         }
 
-        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i & cpu->cfg.ext_m &
-                                cpu->cfg.ext_a & cpu->cfg.ext_f &
+        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
+                                cpu->cfg.ext_a && cpu->cfg.ext_f &&
                                 cpu->cfg.ext_d)) {
             warn_report("Setting G will also set IMAFD");
             cpu->cfg.ext_i = true;
Signature · 1 line
2.34.1
Reviewed-by on v2 1/5
#10 lore raw

Re: [PATCH v2 1/5] target/riscv: Fix coding style on "G" expansion

Quoted 26 lines · Tsukasa OI

[E-MAIL EXTERNO] Não clique em links ou abra anexos, a menos que você possa confirmar o remetente e saber que o conteúdo é seguro. Em caso de e-mail suspeito entre imediatamente em contato com o DTI.

Because ext_? members are boolean variables, operator &&' should be used instead of &'.

---
  target/riscv/cpu.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)
target/riscv/cpu.c +2 -3
index ccacdee215..00bf26ec8b 100644
@@ -596,8 +596,8 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
              return;
          }
 
-        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i & cpu->cfg.ext_m &
-                                cpu->cfg.ext_a & cpu->cfg.ext_f &
+        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
+                                cpu->cfg.ext_a && cpu->cfg.ext_f &&
                                  cpu->cfg.ext_d)) {
              warn_report("Setting G will also set IMAFD");
              cpu->cfg.ext_i = true;
--
Quoted 3 lines

2.34.1

Sorry, looks like I mistakenly reviewed v1 earlier

Signature · 3 lines
Víctor Cora Colombo
Instituto de Pesquisas ELDORADO
Aviso Legal - Disclaimer <https://www.eldorado.org.br/disclaimer.html>
Reviewed-by on v2 1/5
#11 lore raw

Re: [PATCH v2 1/5] target/riscv: Fix coding style on "G" expansion

In reply to Tsukasa OI
Quoted 5 lines · Tsukasa OI

Because ext_? members are boolean variables, operator &&' should be used instead of &'.

Alistair

Quoted 22 lines
---
 target/riscv/cpu.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
target/riscv/cpu.c +2 -3
index ccacdee215..00bf26ec8b 100644
@@ -596,8 +596,8 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             return;
         }
 
-        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i & cpu->cfg.ext_m &
-                                cpu->cfg.ext_a & cpu->cfg.ext_f &
+        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
+                                cpu->cfg.ext_a && cpu->cfg.ext_f &&
                                 cpu->cfg.ext_d)) {
             warn_report("Setting G will also set IMAFD");
             cpu->cfg.ext_i = true;
--

2.34.1

author
#12 lore raw

PATCH v2 2/5 target/riscv: Disable "G" by default

In reply to Tsukasa OI

Because "G" virtual extension expands to "IMAFD", we cannot separately disable extensions like "F" or "D" without disabling "G". Because all "IMAFD" are enabled by default, it's harmless to disable "G" by default.

---
 target/riscv/cpu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/cpu.c +1 -1
index 00bf26ec8b..3ea68d5cd7 100644
@@ -812,7 +812,7 @@ static Property riscv_cpu_properties[] = {
     /* Defaults for standard extensions */
     DEFINE_PROP_BOOL("i", RISCVCPU, cfg.ext_i, true),
     DEFINE_PROP_BOOL("e", RISCVCPU, cfg.ext_e, false),
-    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, true),
+    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, false),
     DEFINE_PROP_BOOL("m", RISCVCPU, cfg.ext_m, true),
     DEFINE_PROP_BOOL("a", RISCVCPU, cfg.ext_a, true),
     DEFINE_PROP_BOOL("f", RISCVCPU, cfg.ext_f, true),
Signature · 1 line
2.34.1
#13 lore raw

Re: [PATCH v2 2/5] target/riscv: Disable "G" by default

Quoted 23 lines · Tsukasa OI

Because "G" virtual extension expands to "IMAFD", we cannot separately disable extensions like "F" or "D" without disabling "G". Because all "IMAFD" are enabled by default, it's harmless to disable "G" by default.

---
  target/riscv/cpu.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/cpu.c +1 -2
index 00bf26ec8b..3ea68d5cd7 100644
@@ -812,7 +812,7 @@ static Property riscv_cpu_properties[] = {
      /* Defaults for standard extensions */
      DEFINE_PROP_BOOL("i", RISCVCPU, cfg.ext_i, true),
      DEFINE_PROP_BOOL("e", RISCVCPU, cfg.ext_e, false),
-    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, true),
+    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, false),
      DEFINE_PROP_BOOL("m", RISCVCPU, cfg.ext_m, true),
      DEFINE_PROP_BOOL("a", RISCVCPU, cfg.ext_a, true),
      DEFINE_PROP_BOOL("f", RISCVCPU, cfg.ext_f, true),
--
Quoted 3 lines

2.34.1

I think the logic looks ok, and (with my limited understanding of the code) I agree on the reasoning for the changes in patches 2 and 3. Just some clarification needed: where is the value of 'g' checked? can the behavior change in this patch cause a situation where IMAFD_Zicsr_Zifencei is set but 'g' is not, or does patch 3 guarantee that in this case 'g' will be set?

Thanks!

Signature · 3 lines
Víctor Cora Colombo
Instituto de Pesquisas ELDORADO
Aviso Legal - Disclaimer <https://www.eldorado.org.br/disclaimer.html>
author
#14 lore raw

Re: [PATCH v2 2/5] target/riscv: Disable "G" by default

Quoted 36 lines · Víctor Colombo
Quoted 26 lines · Tsukasa OI

Because "G" virtual extension expands to "IMAFD", we cannot separately disable extensions like "F" or "D" without disabling "G".  Because all "IMAFD" are enabled by default, it's harmless to disable "G" by default.

---
  target/riscv/cpu.c | 2 +-

  1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/cpu.c +1 -1
index 00bf26ec8b..3ea68d5cd7 100644
@@ -812,7 +812,7 @@ static Property riscv_cpu_properties[] = {
       /* Defaults for standard extensions */
       DEFINE_PROP_BOOL("i", RISCVCPU, cfg.ext_i, true),
       DEFINE_PROP_BOOL("e", RISCVCPU, cfg.ext_e, false),
-    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, true),
+    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, false),
       DEFINE_PROP_BOOL("m", RISCVCPU, cfg.ext_m, true),
       DEFINE_PROP_BOOL("a", RISCVCPU, cfg.ext_a, true),
       DEFINE_PROP_BOOL("f", RISCVCPU, cfg.ext_f, true),
Signature · 1 line
2.34.1

I think the logic looks ok, and (with my limited understanding of the code) I agree on the reasoning for the changes in patches 2 and 3. Just some clarification needed: where is the value of 'g' checked? can the behavior change in this patch cause a situation where IMAFD_Zicsr_Zifencei is set but 'g' is not, or does patch 3 guarantee that in this case 'g' will be set?

Quoted 2 lines

Thanks!

Probably too late to answer but on Alistair's riscv-to-apply.next tree...

target/riscv/cpu.c (19f13a9) line 599-611:
if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&

                        cpu->cfg.ext_a && cpu->cfg.ext_f &&
                        cpu->cfg.ext_d &&
                        cpu->cfg.ext_icsr && cpu->cfg.ext_ifencei)) {
    warn_report("Setting G will also set IMAFD_Zicsr_Zifencei");
    cpu->cfg.ext_i = true;
    cpu->cfg.ext_m = true;
    cpu->cfg.ext_a = true;
    cpu->cfg.ext_f = true;
    cpu->cfg.ext_d = true;
    cpu->cfg.ext_icsr = true;
    cpu->cfg.ext_ifencei = true;
}

This is the only place where "G" (ext_g) is read. Here, if G is enabled and not all IMAFD_Zicsr_Zifencei are enabled, it enables them with a warning message.

So, even if "G" is disabled alone, it's harmless. Note that IMAFD_Zicsr_Zifencei are enabled by default.

Thanks,
Tsukasa

#15 lore raw

Re: [PATCH v2 2/5] target/riscv: Disable "G" by default

Quoted 61 lines · Tsukasa OI
Quoted 36 lines · Víctor Colombo
Quoted 23 lines · Tsukasa OI

Because "G" virtual extension expands to "IMAFD", we cannot separately disable extensions like "F" or "D" without disabling "G". Because all "IMAFD" are enabled by default, it's harmless to disable "G" by default.

---
   target/riscv/cpu.c | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
target/riscv/cpu.c +1 -2
index 00bf26ec8b..3ea68d5cd7 100644
@@ -812,7 +812,7 @@ static Property riscv_cpu_properties[] = {
       /* Defaults for standard extensions */
       DEFINE_PROP_BOOL("i", RISCVCPU, cfg.ext_i, true),
       DEFINE_PROP_BOOL("e", RISCVCPU, cfg.ext_e, false),
-    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, true),
+    DEFINE_PROP_BOOL("g", RISCVCPU, cfg.ext_g, false),
       DEFINE_PROP_BOOL("m", RISCVCPU, cfg.ext_m, true),
       DEFINE_PROP_BOOL("a", RISCVCPU, cfg.ext_a, true),
       DEFINE_PROP_BOOL("f", RISCVCPU, cfg.ext_f, true),
--
Quoted 3 lines

2.34.1

I think the logic looks ok, and (with my limited understanding of the code) I agree on the reasoning for the changes in patches 2 and 3. Just some clarification needed: where is the value of 'g' checked? can the behavior change in this patch cause a situation where IMAFD_Zicsr_Zifencei is set but 'g' is not, or does patch 3 guarantee that in this case 'g' will be set?

Quoted 2 lines

Thanks!

Probably too late to answer but on Alistair's riscv-to-apply.next tree...

target/riscv/cpu.c (19f13a9) line 599-611:
if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&

                         cpu->cfg.ext_a && cpu->cfg.ext_f &&
                         cpu->cfg.ext_d &&
                         cpu->cfg.ext_icsr && cpu->cfg.ext_ifencei)) {
     warn_report("Setting G will also set IMAFD_Zicsr_Zifencei");
     cpu->cfg.ext_i = true;
     cpu->cfg.ext_m = true;
     cpu->cfg.ext_a = true;
     cpu->cfg.ext_f = true;
     cpu->cfg.ext_d = true;
     cpu->cfg.ext_icsr = true;
     cpu->cfg.ext_ifencei = true;
}

This is the only place where "G" (ext_g) is read. Here, if G is enabled and not all IMAFD_Zicsr_Zifencei are enabled, it enables them with a warning message.

Quoted 5 lines

So, even if "G" is disabled alone, it's harmless. Note that IMAFD_Zicsr_Zifencei are enabled by default.

Thanks,
Tsukasa

Hello!
Thank you very much for the clarification, it was helpful. Still getting used to the riscv code base in QEMU

Best regards,

Signature · 3 lines
Víctor Cora Colombo
Instituto de Pesquisas ELDORADO
Aviso Legal - Disclaimer <https://www.eldorado.org.br/disclaimer.html>
author
#16 lore raw

PATCH v2 3/5 target/riscv: Change "G" expansion

In reply to Tsukasa OI

On ISA version 20190608 or later, "G" expands to "IMAFD_Zicsr_Zifencei". Both "Zicsr" and "Zifencei" are enabled by default and "G" is supposed to be (virtually) enabled as well, it should be safe to change its expansion.

---
 target/riscv/cpu.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
target/riscv/cpu.c +5 -2
index 3ea68d5cd7..0854ca9103 100644
@@ -598,13 +598,16 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
 
         if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
                                 cpu->cfg.ext_a && cpu->cfg.ext_f &&
-                                cpu->cfg.ext_d)) {
-            warn_report("Setting G will also set IMAFD");
+                                cpu->cfg.ext_d &&
+                                cpu->cfg.ext_icsr && cpu->cfg.ext_ifencei)) {
+            warn_report("Setting G will also set IMAFD_Zicsr_Zifencei");
             cpu->cfg.ext_i = true;
             cpu->cfg.ext_m = true;
             cpu->cfg.ext_a = true;
             cpu->cfg.ext_f = true;
             cpu->cfg.ext_d = true;
+            cpu->cfg.ext_icsr = true;
+            cpu->cfg.ext_ifencei = true;
         }
 
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
Signature · 1 line
2.34.1
Reviewed-by on v2 3/5
#17 lore raw

Re: [PATCH v2 3/5] target/riscv: Change "G" expansion

Quoted 6 lines · Tsukasa OI

On ISA version 20190608 or later, "G" expands to "IMAFD_Zicsr_Zifencei". Both "Zicsr" and "Zifencei" are enabled by default and "G" is supposed to be (virtually) enabled as well, it should be safe to change its expansion.

Alistair

Quoted 30 lines
---
 target/riscv/cpu.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
target/riscv/cpu.c +5 -3
index 3ea68d5cd7..0854ca9103 100644
@@ -598,13 +598,16 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
 
         if (cpu->cfg.ext_g && !(cpu->cfg.ext_i && cpu->cfg.ext_m &&
                                 cpu->cfg.ext_a && cpu->cfg.ext_f &&
-                                cpu->cfg.ext_d)) {
-            warn_report("Setting G will also set IMAFD");
+                                cpu->cfg.ext_d &&
+                                cpu->cfg.ext_icsr && cpu->cfg.ext_ifencei)) {
+            warn_report("Setting G will also set IMAFD_Zicsr_Zifencei");
             cpu->cfg.ext_i = true;
             cpu->cfg.ext_m = true;
             cpu->cfg.ext_a = true;
             cpu->cfg.ext_f = true;
             cpu->cfg.ext_d = true;
+            cpu->cfg.ext_icsr = true;
+            cpu->cfg.ext_ifencei = true;
         }
 
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
--

2.34.1

author
#18 lore raw

PATCH v2 4/5 target/riscv: FP extension requirements

In reply to Tsukasa OI

QEMU allowed inconsistent configurations that made floating point arithmetic effectively unusable.

This commit adds certain checks for consistent FP arithmetic:

-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D

Because F/D/Zicsr are enabled by default (and an error will not occur unless we manually disable one or more of prerequisites), this commit just enforces the user to give consistent combinations.

---
 target/riscv/cpu.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
target/riscv/cpu.c +25 -0
index 0854ca9103..f910a41407 100644
@@ -610,11 +610,36 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             cpu->cfg.ext_ifencei = true;
         }
 
+        if (cpu->cfg.ext_f && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "F extension requires Zicsr");
+            return;
+        }
+
+        if ((cpu->cfg.ext_zfh || cpu->cfg.ext_zfhmin) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zfh/Zfhmin extensions require F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_d && !cpu->cfg.ext_f) {
+            error_setg(errp, "D extension requires F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_v && !cpu->cfg.ext_d) {
+            error_setg(errp, "V extension requires D extension");
+            return;
+        }
+
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
             cpu->cfg.ext_zhinxmin) {
             cpu->cfg.ext_zfinx = true;
         }
 
+        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "Zfinx extension requires Zicsr");
+            return;
+        }
+
         if (cpu->cfg.ext_zk) {
             cpu->cfg.ext_zkn = true;
             cpu->cfg.ext_zkr = true;
Signature · 1 line
2.34.1
#19 lore raw

Re: [PATCH v2 4/5] target/riscv: FP extension requirements

Review

Quoted 10 lines · Tsukasa OI

QEMU allowed inconsistent configurations that made floating point arithmetic effectively unusable.

This commit adds certain checks for consistent FP arithmetic:

(quoted diff) +0 -5
-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D

Why 'V requires D'? I know partial vector instructions require D, However,  I think they  just like c.fsd

instruction requires D, not 'C requires D' or 'D requires C'. Is there any rvv spec change I don't know?

Regards.

Weiwei Li

Quoted 51 lines

Because F/D/Zicsr are enabled by default (and an error will not occur unless we manually disable one or more of prerequisites), this commit just enforces the user to give consistent combinations.

---
  target/riscv/cpu.c | 25 +++++++++++++++++++++++++
  1 file changed, 25 insertions(+)
target/riscv/cpu.c +25 -0
index 0854ca9103..f910a41407 100644
@@ -610,11 +610,36 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
              cpu->cfg.ext_ifencei = true;
          }
 
+        if (cpu->cfg.ext_f && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "F extension requires Zicsr");
+            return;
+        }
+
+        if ((cpu->cfg.ext_zfh || cpu->cfg.ext_zfhmin) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zfh/Zfhmin extensions require F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_d && !cpu->cfg.ext_f) {
+            error_setg(errp, "D extension requires F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_v && !cpu->cfg.ext_d) {
+            error_setg(errp, "V extension requires D extension");
+            return;
+        }
+
          if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
              cpu->cfg.ext_zhinxmin) {
              cpu->cfg.ext_zfinx = true;
          }
 
+        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "Zfinx extension requires Zicsr");
+            return;
+        }
+
          if (cpu->cfg.ext_zk) {
              cpu->cfg.ext_zkn = true;
              cpu->cfg.ext_zkr = true;
author
#20 lore raw

Re: [PATCH v2 4/5] target/riscv: FP extension requirements

Quoted 13 lines · Weiwei Li
Quoted 10 lines · Tsukasa OI

QEMU allowed inconsistent configurations that made floating point arithmetic effectively unusable.

This commit adds certain checks for consistent FP arithmetic:

-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D

Quoted 3 lines

Why 'V requires D'? I know partial vector instructions require D, However,  I think they  just like c.fsd

instruction requires D, not 'C requires D' or 'D requires C'. Is there any rvv spec change I don't know?

I thought it was crystal-clear.

RISC-V "V" Vector Extension Version 1.0 (riscv-v-spec-1.0.pdf) 18.3. V: Vector Extension for Application Processors Page 94:

Quoted 5 lines

The V extension requires the scalar processor implements the F and D extensions, and implements all vector floating-point instructions (Section Vector Floating-Point Instructions) for floating-point operands with EEW=32 or EEW=64 (including widening instructions and conversions between FP32 and FP64).

c.f. <https://github.com/riscv/riscv-v-spec/releases/tag/v1.0>

Thanks,

Tsukasa

Quoted 55 lines

Regards.

Weiwei Li

Quoted 49 lines

Because F/D/Zicsr are enabled by default (and an error will not occur unless we manually disable one or more of prerequisites), this commit just enforces the user to give consistent combinations.

---
  target/riscv/cpu.c | 25 +++++++++++++++++++++++++

  1 file changed, 25 insertions(+)
target/riscv/cpu.c +19 -0
index 0854ca9103..f910a41407 100644
@@ -610,11 +610,36 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
               cpu->cfg.ext_ifencei = true;
           }
   +        if (cpu->cfg.ext_f && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "F extension requires Zicsr");
+            return;
+        }
+
+        if ((cpu->cfg.ext_zfh || cpu->cfg.ext_zfhmin) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zfh/Zfhmin extensions require F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_d && !cpu->cfg.ext_f) {
+            error_setg(errp, "D extension requires F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_v && !cpu->cfg.ext_d) {
+            error_setg(errp, "V extension requires D extension");
+            return;
+        }
+

          if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx || cpu->cfg.ext_zhinxmin) { cpu->cfg.ext_zfinx = true;
          }
  +        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) { +            error_setg(errp, "Zfinx extension requires Zicsr"); +            return;
+        }
+
          if (cpu->cfg.ext_zk) {
              cpu->cfg.ext_zkn = true; cpu->cfg.ext_zkr = true;

#21 lore raw

Re: [PATCH v2 4/5] target/riscv: FP extension requirements

Quoted 26 lines · Tsukasa OI
Quoted 6 lines · Weiwei Li
Quoted 5 lines · Tsukasa OI

QEMU allowed inconsistent configurations that made floating point arithmetic effectively unusable.

This commit adds certain checks for consistent FP arithmetic:

Quoted 8 lines
Quoted 5 lines

-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D

Why 'V requires D'? I know partial vector instructions require D, However,  I think they  just like c.fsd

instruction requires D, not 'C requires D' or 'D requires C'. Is there any rvv spec change I don't know?

I thought it was crystal-clear.

RISC-V "V" Vector Extension Version 1.0 (riscv-v-spec-1.0.pdf) 18.3. V: Vector Extension for Application Processors Page 94:

Quoted 5 lines

The V extension requires the scalar processor implements the F and D extensions, and implements all vector floating-point instructions (Section Vector Floating-Point Instructions) for floating-point operands with EEW=32 or EEW=64 (including widening instructions and conversions between FP32 and FP64).

Quoted 5 lines

c.f. <https://github.com/riscv/riscv-v-spec/releases/tag/v1.0>

Thanks,

Tsukasa

OK. Thanks.

Regards,

Weiwei Li

Quoted 53 lines
Quoted 53 lines

Regards.

Weiwei Li

Quoted 48 lines

Because F/D/Zicsr are enabled by default (and an error will not occur unless we manually disable one or more of prerequisites), this commit just enforces the user to give consistent combinations.

---
   target/riscv/cpu.c | 25 +++++++++++++++++++++++++
   1 file changed, 25 insertions(+)
target/riscv/cpu.c +23 -0
index 0854ca9103..f910a41407 100644
@@ -610,11 +610,36 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
               cpu->cfg.ext_ifencei = true;
           }
   +        if (cpu->cfg.ext_f && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "F extension requires Zicsr");
+            return;
+        }
+
+        if ((cpu->cfg.ext_zfh || cpu->cfg.ext_zfhmin) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zfh/Zfhmin extensions require F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_d && !cpu->cfg.ext_f) {
+            error_setg(errp, "D extension requires F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_v && !cpu->cfg.ext_d) {
+            error_setg(errp, "V extension requires D extension");
+            return;
+        }
+
           if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
               cpu->cfg.ext_zhinxmin) {
               cpu->cfg.ext_zfinx = true;
           }
   +        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "Zfinx extension requires Zicsr");
+            return;
+        }
+
           if (cpu->cfg.ext_zk) {
               cpu->cfg.ext_zkn = true;
               cpu->cfg.ext_zkr = true;
Reviewed-by on v2 4/5
#22 lore raw

Re: [PATCH v2 4/5] target/riscv: FP extension requirements

In reply to Tsukasa OI
Quoted 12 lines · Tsukasa OI

QEMU allowed inconsistent configurations that made floating point arithmetic effectively unusable.

This commit adds certain checks for consistent FP arithmetic:

(quoted diff) +0 -5
-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D
Quoted 5 lines

Because F/D/Zicsr are enabled by default (and an error will not occur unless we manually disable one or more of prerequisites), this commit just enforces the user to give consistent combinations.

Alistair

Quoted 48 lines
---
 target/riscv/cpu.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
target/riscv/cpu.c +25 -1
index 0854ca9103..f910a41407 100644
@@ -610,11 +610,36 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             cpu->cfg.ext_ifencei = true;
         }
 
+        if (cpu->cfg.ext_f && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "F extension requires Zicsr");
+            return;
+        }
+
+        if ((cpu->cfg.ext_zfh || cpu->cfg.ext_zfhmin) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zfh/Zfhmin extensions require F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_d && !cpu->cfg.ext_f) {
+            error_setg(errp, "D extension requires F extension");
+            return;
+        }
+
+        if (cpu->cfg.ext_v && !cpu->cfg.ext_d) {
+            error_setg(errp, "V extension requires D extension");
+            return;
+        }
+
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
             cpu->cfg.ext_zhinxmin) {
             cpu->cfg.ext_zfinx = true;
         }
 
+        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
+            error_setg(errp, "Zfinx extension requires Zicsr");
+            return;
+        }
+
         if (cpu->cfg.ext_zk) {
             cpu->cfg.ext_zkn = true;
             cpu->cfg.ext_zkr = true;
--

2.34.1

author
#23 lore raw

PATCH v2 5/5 target/riscv: Move/refactor ISA extension checks

In reply to Tsukasa OI

We should separate "check" and "configure" steps as possible. This commit separates both steps except vector/Zfinx-related checks.

---
 target/riscv/cpu.c | 31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)
target/riscv/cpu.c +16 -15
index f910a41407..5ab246bf63 100644
@@ -630,14 +630,27 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             return;
         }
 
+        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zve32f/Zve64f extensions require F extension");
+            return;
+        }
+
+        /* Set the ISA extensions, checks should have happened above */
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
             cpu->cfg.ext_zhinxmin) {
             cpu->cfg.ext_zfinx = true;
         }
 
-        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
-            error_setg(errp, "Zfinx extension requires Zicsr");
-            return;
+        if (cpu->cfg.ext_zfinx) {
+            if (!cpu->cfg.ext_icsr) {
+                error_setg(errp, "Zfinx extension requires Zicsr");
+                return;
+            }
+            if (cpu->cfg.ext_f) {
+                error_setg(errp,
+                    "Zfinx cannot be supported together with F extension");
+                return;
+            }
         }
 
         if (cpu->cfg.ext_zk) {
@@ -663,7 +676,6 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             cpu->cfg.ext_zksh = true;
         }
 
-        /* Set the ISA extensions, checks should have happened above */
         if (cpu->cfg.ext_i) {
             ext |= RVI;
         }
@@ -734,20 +746,9 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             }
             set_vext_version(env, vext_version);
         }
-        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
-            error_setg(errp, "Zve32f/Zve64f extension depends upon RVF.");
-            return;
-        }
         if (cpu->cfg.ext_j) {
             ext |= RVJ;
         }
-        if (cpu->cfg.ext_zfinx && ((ext & (RVF | RVD)) || cpu->cfg.ext_zfh ||
-                                   cpu->cfg.ext_zfhmin)) {
-            error_setg(errp,
-                    "'Zfinx' cannot be supported together with 'F', 'D', 'Zfh',"
-                    " 'Zfhmin'");
-            return;
-        }
 
         set_misa(env, env->misa_mxl, ext);
     }
Signature · 1 line
2.34.1
#24 lore raw

Re: [PATCH v2 5/5] target/riscv: Move/refactor ISA extension checks

Quoted 31 lines · Tsukasa OI

We should separate "check" and "configure" steps as possible. This commit separates both steps except vector/Zfinx-related checks.

---
  target/riscv/cpu.c | 31 ++++++++++++++++---------------
  1 file changed, 16 insertions(+), 15 deletions(-)
target/riscv/cpu.c +7 -3
index f910a41407..5ab246bf63 100644
@@ -630,14 +630,27 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
              return;
          }
 
+        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zve32f/Zve64f extensions require F extension");
+            return;
+        }
+
+        /* Set the ISA extensions, checks should have happened above */
          if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
              cpu->cfg.ext_zhinxmin) {
              cpu->cfg.ext_zfinx = true;
          }
 
-        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
-            error_setg(errp, "Zfinx extension requires Zicsr");
-            return;
+        if (cpu->cfg.ext_zfinx) {

target/riscv/cpu.c static void riscv_cpu_realize(DeviceState *dev, Error **errp) · line 654

Quoted 10 lines
target/riscv/cpu.c +9 -0
+            if (!cpu->cfg.ext_icsr) {
+                error_setg(errp, "Zfinx extension requires Zicsr");
+                return;
+            }
+            if (cpu->cfg.ext_f) {
+                error_setg(errp,
+                    "Zfinx cannot be supported together with F extension");
+                return;
+            }
          }

I think these checks for non-single-letter extensions are  better to move  out of the 'if (env->misa_ext == 0)) { ...}', since they are enabled

directly by cfg property, such as we can set cpu option to sifive-u34 with zfinx=true. This may not be a proper way to set cpu option,

However it's truly a legal command option, but  configure an illegal supported ISA which enable both f and zfinx.

Regards,

Weiwei Li

Quoted 31 lines
target/riscv/cpu.c +0 -12
          if (cpu->cfg.ext_zk) {
@@ -663,7 +676,6 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
              cpu->cfg.ext_zksh = true;
          }
 
-        /* Set the ISA extensions, checks should have happened above */
          if (cpu->cfg.ext_i) {
              ext |= RVI;
          }
@@ -734,20 +746,9 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
              }
              set_vext_version(env, vext_version);
          }
-        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
-            error_setg(errp, "Zve32f/Zve64f extension depends upon RVF.");
-            return;
-        }
          if (cpu->cfg.ext_j) {
              ext |= RVJ;
          }
-        if (cpu->cfg.ext_zfinx && ((ext & (RVF | RVD)) || cpu->cfg.ext_zfh ||
-                                   cpu->cfg.ext_zfhmin)) {
-            error_setg(errp,
-                    "'Zfinx' cannot be supported together with 'F', 'D', 'Zfh',"
-                    " 'Zfhmin'");
-            return;
-        }
 
          set_misa(env, env->misa_mxl, ext);
      }
Reviewed-by on v2 5/5
#25 lore raw

Re: [PATCH v2 5/5] target/riscv: Move/refactor ISA extension checks

In reply to Tsukasa OI
Quoted 5 lines · Tsukasa OI

We should separate "check" and "configure" steps as possible. This commit separates both steps except vector/Zfinx-related checks.

Alistair

Quoted 71 lines
---
 target/riscv/cpu.c | 31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)
target/riscv/cpu.c +16 -16
index f910a41407..5ab246bf63 100644
@@ -630,14 +630,27 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             return;
         }
 
+        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
+            error_setg(errp, "Zve32f/Zve64f extensions require F extension");
+            return;
+        }
+
+        /* Set the ISA extensions, checks should have happened above */
         if (cpu->cfg.ext_zdinx || cpu->cfg.ext_zhinx ||
             cpu->cfg.ext_zhinxmin) {
             cpu->cfg.ext_zfinx = true;
         }
 
-        if (cpu->cfg.ext_zfinx && !cpu->cfg.ext_icsr) {
-            error_setg(errp, "Zfinx extension requires Zicsr");
-            return;
+        if (cpu->cfg.ext_zfinx) {
+            if (!cpu->cfg.ext_icsr) {
+                error_setg(errp, "Zfinx extension requires Zicsr");
+                return;
+            }
+            if (cpu->cfg.ext_f) {
+                error_setg(errp,
+                    "Zfinx cannot be supported together with F extension");
+                return;
+            }
         }
 
         if (cpu->cfg.ext_zk) {
@@ -663,7 +676,6 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             cpu->cfg.ext_zksh = true;
         }
 
-        /* Set the ISA extensions, checks should have happened above */
         if (cpu->cfg.ext_i) {
             ext |= RVI;
         }
@@ -734,20 +746,9 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
             }
             set_vext_version(env, vext_version);
         }
-        if ((cpu->cfg.ext_zve32f || cpu->cfg.ext_zve64f) && !cpu->cfg.ext_f) {
-            error_setg(errp, "Zve32f/Zve64f extension depends upon RVF.");
-            return;
-        }
         if (cpu->cfg.ext_j) {
             ext |= RVJ;
         }
-        if (cpu->cfg.ext_zfinx && ((ext & (RVF | RVD)) || cpu->cfg.ext_zfh ||
-                                   cpu->cfg.ext_zfhmin)) {
-            error_setg(errp,
-                    "'Zfinx' cannot be supported together with 'F', 'D', 'Zfh',"
-                    " 'Zfhmin'");
-            return;
-        }
 
         set_misa(env, env->misa_mxl, ext);
     }
--

2.34.1

#26 lore raw
In reply to Tsukasa OI
Quoted 71 lines · Tsukasa OI

Hello,

This is another patchset for RISC-V ISA extension / feature handling.

Aside from coding style fix / refactoring patch (PATCH 1 and 5), this patchset contains two changes:

1. "G" extension handling

1.A. "G" extension expansion (PATCH 3)

On ISA version 20190608 or later, "G" expands to "IMAFD_Zicsr_Zifencei", not just "IMAFD" (this is because "Zicsr" and "Zifencei" extensions are splitted from "I"). Because both "Zicsr" and "Zifencei" are enabled by default, it should be safe to change "G" extension expansion.

1.B. Disable "G" by default (PATCH 2)

This seems quite odd but I have a good reason. While "G" is enabled by default, all "I", "M", "A", "F", "D", "Zicsr" and "Zifencei" are also enabled by default, making default "G" expansion useless.

There's more. If we want to change detailed configuration of a RV32/RV64 CPU and want to disable some, for example, "F" and "D", we must also disable "G". This is obviously bloat.

This doesn't work:

    -cpu rv64,f=off,d=off

This does work (but bloat):

    -cpu rv64,g=off,f=off,d=off

Disabling "G" suppresses such problem and mostly harmless, too.

2. Floating point arithmetic consistency (PATCH 4)

With -cpu option, we can configure details of RISC-V CPU emulated by QEMU. However, we can set some invalid combinations that would make FP arithmetic effectively unusable (and invalid on RISC-V ISA specification).

(quoted diff) +0 -5
-   F requires Zicsr
-   Zfinx requires Zicsr
-   Zfh/Zfhmin require F
-   D requires F
-   V requires D

Reproducing this kind of problem requires manually disabling certain extensions (because all "Zicsr", "F" and "D" are enabled by default). So, I consider that just making an error message (when unsatisfied) should be enough, not implying related extensions like "zk*" properties.

Note that this checking only works on any, rv32 and rv64. On other CPUs (for example, sifive-u54), it sets nonzero misa' value on corresponding set_misa' call (c.f. in rv64_sifive_u_cpu_init in target/riscv/cpu.c) and consistency checks are skipped (because nonzero `misa' value is set prior to RISC-V CPU realization process).

I think we generally use generic "rv32" or "rv64" on heavy customizing so I don't think this is not a big problem. Still, we could fix this later (e.g. by setting properties on CPU init function or by checking some consistency problems even if previously-set `misa' is nonzero).

Quoted 6 lines

Tsukasa OI (5):
target/riscv: Fix "G" extension expansion typing target/riscv: Disable "G" by default target/riscv: Change "G" expansion target/riscv: FP extension requirements target/riscv: Move/refactor ISA extension checks

Thanks!

Applied to riscv-to-apply.next

Alistair

Quoted 9 lines
 target/riscv/cpu.c | 62 +++++++++++++++++++++++++++++++++-------------
 1 file changed, 45 insertions(+), 17 deletions(-)

base-commit: 178bacb66d98d9ee7a702b9f2a4dfcd88b72a9ab --
2.34.1