* [RFC PATCH v3 1/6] riscv/cmpxchg: Deduplicate cmpxchg() asm functions
2023-04-04 16:37 [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Leonardo Bras
@ 2023-04-04 16:37 ` Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 2/6] riscv/cmpxchg: Deduplicate cmpxchg() macros Leonardo Bras
` (5 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Leonardo Bras @ 2023-04-04 16:37 UTC (permalink / raw)
To: Paul Walmsley, Palmer Dabbelt, Albert Ou, Leonardo Bras, Guo Ren,
Andrea Parri, Conor Dooley
Cc: linux-riscv, linux-kernel
In this header every cmpxchg define (_relaxed, _acquire, _release,
vanilla) contain it's own asm file, both for 4-byte variables an 8-byte
variables, on a total of 8 versions of mostly the same asm.
This is usually bad, as it means any change may be done in up to 8
different places.
Unify those versions by creating a new define with enough parameters to
generate any version of the previous 8.
(This did not cause any change in generated asm)
Signed-off-by: Leonardo Bras <leobras@redhat.com>
---
arch/riscv/include/asm/cmpxchg.h | 110 +++++++++----------------------
1 file changed, 32 insertions(+), 78 deletions(-)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index 12debce235e52..22f2ad13cad71 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -163,6 +163,22 @@
* store NEW in MEM. Return the initial value in MEM. Success is
* indicated by comparing RETURN with OLD.
*/
+
+#define ___cmpxchg(lr_sfx, sc_sfx, prepend, append, r, rc, p, co, o, n) \
+{ \
+ __asm__ __volatile__ ( \
+ prepend \
+ "0: lr" lr_sfx " %0, %2\n" \
+ " bne %0, %z3, 1f\n" \
+ " sc" sc_sfx " %1, %z4, %2\n" \
+ " bnez %1, 0b\n" \
+ append \
+ "1:\n" \
+ : "=&r" (r), "=&r" (rc), "+A" (*(p)) \
+ : "rJ" (co o), "rJ" (n) \
+ : "memory"); \
+}
+
#define __cmpxchg_relaxed(ptr, old, new, size) \
({ \
__typeof__(ptr) __ptr = (ptr); \
@@ -172,26 +188,12 @@
register unsigned int __rc; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- "0: lr.w %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.w %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" ((long)__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".w", ".w", "", "", \
+ __ret, __rc, __ptr, (long), __old, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- "0: lr.d %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.d %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" (__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".d", ".d", "", "", \
+ __ret, __rc, __ptr, /**/, __old, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -216,28 +218,12 @@
register unsigned int __rc; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- "0: lr.w %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.w %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- RISCV_ACQUIRE_BARRIER \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" ((long)__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".w", ".w", "", RISCV_ACQUIRE_BARRIER, \
+ __ret, __rc, __ptr, (long), __old, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- "0: lr.d %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.d %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- RISCV_ACQUIRE_BARRIER \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" (__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".d", ".d", "", RISCV_ACQUIRE_BARRIER, \
+ __ret, __rc, __ptr, /**/, __old, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -262,28 +248,12 @@
register unsigned int __rc; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- RISCV_RELEASE_BARRIER \
- "0: lr.w %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.w %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" ((long)__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".w", ".w", RISCV_RELEASE_BARRIER, "", \
+ __ret, __rc, __ptr, (long), __old, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- RISCV_RELEASE_BARRIER \
- "0: lr.d %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.d %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" (__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".d", ".d", RISCV_RELEASE_BARRIER, "", \
+ __ret, __rc, __ptr, /**/, __old, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -308,28 +278,12 @@
register unsigned int __rc; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- "0: lr.w %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.w.rl %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- " fence rw, rw\n" \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" ((long)__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".w", ".w.rl", "", " fence rw, rw\n", \
+ __ret, __rc, __ptr, (long), __old, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- "0: lr.d %0, %2\n" \
- " bne %0, %z3, 1f\n" \
- " sc.d.rl %1, %z4, %2\n" \
- " bnez %1, 0b\n" \
- " fence rw, rw\n" \
- "1:\n" \
- : "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr) \
- : "rJ" (__old), "rJ" (__new) \
- : "memory"); \
+ ___cmpxchg(".d", ".d.rl", "", " fence rw, rw\n", \
+ __ret, __rc, __ptr, /**/, __old, __new); \
break; \
default: \
BUILD_BUG(); \
--
2.40.0
^ permalink raw reply [flat|nested] 9+ messages in thread* [RFC PATCH v3 2/6] riscv/cmpxchg: Deduplicate cmpxchg() macros
2023-04-04 16:37 [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 1/6] riscv/cmpxchg: Deduplicate cmpxchg() asm functions Leonardo Bras
@ 2023-04-04 16:37 ` Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 3/6] riscv/cmpxchg: Deduplicate arch_cmpxchg() macros Leonardo Bras
` (4 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Leonardo Bras @ 2023-04-04 16:37 UTC (permalink / raw)
To: Paul Walmsley, Palmer Dabbelt, Albert Ou, Leonardo Bras, Guo Ren,
Andrea Parri, Conor Dooley
Cc: linux-riscv, linux-kernel
Every cmpxchg define (_relaxed, _acquire, _release, vanilla) contain it's
own define for creating tmp variables and selecting the correct asm code
for give variable size.
All those defines are mostly the same code (other than specific barriers),
so there is no need to keep the 4 copies.
Unify those under a more general define, that can reproduce the previous 4
versions.
(This did not cause any change in generated asm)
Signed-off-by: Leonardo Bras <leobras@redhat.com>
---
arch/riscv/include/asm/cmpxchg.h | 82 ++++++--------------------------
1 file changed, 14 insertions(+), 68 deletions(-)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index 22f2ad13cad71..f76a82274a4fd 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -164,8 +164,8 @@
* indicated by comparing RETURN with OLD.
*/
-#define ___cmpxchg(lr_sfx, sc_sfx, prepend, append, r, rc, p, co, o, n) \
-{ \
+#define ____cmpxchg(lr_sfx, sc_sfx, prepend, append, r, rc, p, co, o, n)\
+({ \
__asm__ __volatile__ ( \
prepend \
"0: lr" lr_sfx " %0, %2\n" \
@@ -177,9 +177,9 @@
: "=&r" (r), "=&r" (rc), "+A" (*(p)) \
: "rJ" (co o), "rJ" (n) \
: "memory"); \
-}
+})
-#define __cmpxchg_relaxed(ptr, old, new, size) \
+#define ___cmpxchg(ptr, old, new, size, sc_sfx, prepend, append) \
({ \
__typeof__(ptr) __ptr = (ptr); \
__typeof__(*(ptr)) __old = (old); \
@@ -188,12 +188,12 @@
register unsigned int __rc; \
switch (size) { \
case 4: \
- ___cmpxchg(".w", ".w", "", "", \
- __ret, __rc, __ptr, (long), __old, __new); \
+ ____cmpxchg(".w", ".w" sc_sfx, prepend, append, \
+ __ret, __rc, __ptr, (long), __old, __new); \
break; \
case 8: \
- ___cmpxchg(".d", ".d", "", "", \
- __ret, __rc, __ptr, /**/, __old, __new); \
+ ____cmpxchg(".d", ".d" sc_sfx, prepend, append, \
+ __ret, __rc, __ptr, /**/, __old, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -201,6 +201,9 @@
__ret; \
})
+#define __cmpxchg_relaxed(ptr, old, new, size) \
+ ___cmpxchg(ptr, old, new, size, "", "", "")
+
#define arch_cmpxchg_relaxed(ptr, o, n) \
({ \
__typeof__(*(ptr)) _o_ = (o); \
@@ -210,26 +213,7 @@
})
#define __cmpxchg_acquire(ptr, old, new, size) \
-({ \
- __typeof__(ptr) __ptr = (ptr); \
- __typeof__(*(ptr)) __old = (old); \
- __typeof__(*(ptr)) __new = (new); \
- __typeof__(*(ptr)) __ret; \
- register unsigned int __rc; \
- switch (size) { \
- case 4: \
- ___cmpxchg(".w", ".w", "", RISCV_ACQUIRE_BARRIER, \
- __ret, __rc, __ptr, (long), __old, __new); \
- break; \
- case 8: \
- ___cmpxchg(".d", ".d", "", RISCV_ACQUIRE_BARRIER, \
- __ret, __rc, __ptr, /**/, __old, __new); \
- break; \
- default: \
- BUILD_BUG(); \
- } \
- __ret; \
-})
+ ___cmpxchg(ptr, old, new, size, "", "", RISCV_ACQUIRE_BARRIER)
#define arch_cmpxchg_acquire(ptr, o, n) \
({ \
@@ -240,26 +224,7 @@
})
#define __cmpxchg_release(ptr, old, new, size) \
-({ \
- __typeof__(ptr) __ptr = (ptr); \
- __typeof__(*(ptr)) __old = (old); \
- __typeof__(*(ptr)) __new = (new); \
- __typeof__(*(ptr)) __ret; \
- register unsigned int __rc; \
- switch (size) { \
- case 4: \
- ___cmpxchg(".w", ".w", RISCV_RELEASE_BARRIER, "", \
- __ret, __rc, __ptr, (long), __old, __new); \
- break; \
- case 8: \
- ___cmpxchg(".d", ".d", RISCV_RELEASE_BARRIER, "", \
- __ret, __rc, __ptr, /**/, __old, __new); \
- break; \
- default: \
- BUILD_BUG(); \
- } \
- __ret; \
-})
+ ___cmpxchg(ptr, old, new, size, "", RISCV_RELEASE_BARRIER, "")
#define arch_cmpxchg_release(ptr, o, n) \
({ \
@@ -270,26 +235,7 @@
})
#define __cmpxchg(ptr, old, new, size) \
-({ \
- __typeof__(ptr) __ptr = (ptr); \
- __typeof__(*(ptr)) __old = (old); \
- __typeof__(*(ptr)) __new = (new); \
- __typeof__(*(ptr)) __ret; \
- register unsigned int __rc; \
- switch (size) { \
- case 4: \
- ___cmpxchg(".w", ".w.rl", "", " fence rw, rw\n", \
- __ret, __rc, __ptr, (long), __old, __new); \
- break; \
- case 8: \
- ___cmpxchg(".d", ".d.rl", "", " fence rw, rw\n", \
- __ret, __rc, __ptr, /**/, __old, __new); \
- break; \
- default: \
- BUILD_BUG(); \
- } \
- __ret; \
-})
+ ___cmpxchg(ptr, old, new, size, ".rl", "", " fence rw, rw\n")
#define arch_cmpxchg(ptr, o, n) \
({ \
--
2.40.0
^ permalink raw reply [flat|nested] 9+ messages in thread* [RFC PATCH v3 3/6] riscv/cmpxchg: Deduplicate arch_cmpxchg() macros
2023-04-04 16:37 [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 1/6] riscv/cmpxchg: Deduplicate cmpxchg() asm functions Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 2/6] riscv/cmpxchg: Deduplicate cmpxchg() macros Leonardo Bras
@ 2023-04-04 16:37 ` Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 4/6] riscv/cmpxchg: Deduplicate xchg() asm functions Leonardo Bras
` (3 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Leonardo Bras @ 2023-04-04 16:37 UTC (permalink / raw)
To: Paul Walmsley, Palmer Dabbelt, Albert Ou, Leonardo Bras, Guo Ren,
Andrea Parri, Conor Dooley
Cc: linux-riscv, linux-kernel
Every arch_cmpxchg define (_relaxed, _acquire, _release, vanilla) contain
it's own define for creating tmp variables and calling the correct internal
macro for the desired version.
Those defines are mostly the same code, so there is no need to keep the 4
copies.
Create a helper define to avoid code duplication.
(This did not cause any change in generated asm)
Signed-off-by: Leonardo Bras <leobras@redhat.com>
---
arch/riscv/include/asm/cmpxchg.h | 30 +++++++++---------------------
1 file changed, 9 insertions(+), 21 deletions(-)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index f76a82274a4fd..f88fae357071c 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -204,46 +204,34 @@
#define __cmpxchg_relaxed(ptr, old, new, size) \
___cmpxchg(ptr, old, new, size, "", "", "")
-#define arch_cmpxchg_relaxed(ptr, o, n) \
+#define _arch_cmpxchg(order, ptr, o, n) \
({ \
__typeof__(*(ptr)) _o_ = (o); \
__typeof__(*(ptr)) _n_ = (n); \
- (__typeof__(*(ptr))) __cmpxchg_relaxed((ptr), \
- _o_, _n_, sizeof(*(ptr))); \
+ (__typeof__(*(ptr))) __cmpxchg ## order((ptr), _o_, _n_, \
+ sizeof(*(ptr))); \
})
+#define arch_cmpxchg_relaxed(ptr, o, n) \
+ _arch_cmpxchg(_relaxed, ptr, o, n)
+
#define __cmpxchg_acquire(ptr, old, new, size) \
___cmpxchg(ptr, old, new, size, "", "", RISCV_ACQUIRE_BARRIER)
#define arch_cmpxchg_acquire(ptr, o, n) \
-({ \
- __typeof__(*(ptr)) _o_ = (o); \
- __typeof__(*(ptr)) _n_ = (n); \
- (__typeof__(*(ptr))) __cmpxchg_acquire((ptr), \
- _o_, _n_, sizeof(*(ptr))); \
-})
+ _arch_cmpxchg(_acquire, ptr, o, n)
#define __cmpxchg_release(ptr, old, new, size) \
___cmpxchg(ptr, old, new, size, "", RISCV_RELEASE_BARRIER, "")
#define arch_cmpxchg_release(ptr, o, n) \
-({ \
- __typeof__(*(ptr)) _o_ = (o); \
- __typeof__(*(ptr)) _n_ = (n); \
- (__typeof__(*(ptr))) __cmpxchg_release((ptr), \
- _o_, _n_, sizeof(*(ptr))); \
-})
+ _arch_cmpxchg(_release, ptr, o, n)
#define __cmpxchg(ptr, old, new, size) \
___cmpxchg(ptr, old, new, size, ".rl", "", " fence rw, rw\n")
#define arch_cmpxchg(ptr, o, n) \
-({ \
- __typeof__(*(ptr)) _o_ = (o); \
- __typeof__(*(ptr)) _n_ = (n); \
- (__typeof__(*(ptr))) __cmpxchg((ptr), \
- _o_, _n_, sizeof(*(ptr))); \
-})
+ _arch_cmpxchg(, ptr, o, n)
#define arch_cmpxchg_local(ptr, o, n) \
(__cmpxchg_relaxed((ptr), (o), (n), sizeof(*(ptr))))
--
2.40.0
^ permalink raw reply [flat|nested] 9+ messages in thread* [RFC PATCH v3 4/6] riscv/cmpxchg: Deduplicate xchg() asm functions
2023-04-04 16:37 [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Leonardo Bras
` (2 preceding siblings ...)
2023-04-04 16:37 ` [RFC PATCH v3 3/6] riscv/cmpxchg: Deduplicate arch_cmpxchg() macros Leonardo Bras
@ 2023-04-04 16:37 ` Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 5/6] riscv/cmpxchg: Deduplicate xchg() macros Leonardo Bras
` (2 subsequent siblings)
6 siblings, 0 replies; 9+ messages in thread
From: Leonardo Bras @ 2023-04-04 16:37 UTC (permalink / raw)
To: Paul Walmsley, Palmer Dabbelt, Albert Ou, Leonardo Bras, Guo Ren,
Andrea Parri, Conor Dooley
Cc: linux-riscv, linux-kernel
In this header every xchg define (_relaxed, _acquire, _release, vanilla)
contain it's own asm file, both for 4-byte variables an 8-byte variables,
on a total of 8 versions of mostly the same asm.
This is usually bad, as it means any change may be done in up to 8
different places.
Unify those versions by creating a new define with enough parameters to
generate any version of the previous 8.
(This did not cause any change in generated asm)
Signed-off-by: Leonardo Bras <leobras@redhat.com>
---
arch/riscv/include/asm/cmpxchg.h | 71 ++++++++++++--------------------
1 file changed, 27 insertions(+), 44 deletions(-)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index f88fae357071c..a524484946815 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -11,6 +11,17 @@
#include <asm/barrier.h>
#include <asm/fence.h>
+#define ___xchg(sfx, prepend, append, r, p, n) \
+({ \
+ __asm__ __volatile__ ( \
+ prepend \
+ " amoswap" sfx " %0, %2, %1\n" \
+ append \
+ : "=r" (r), "+A" (*(p)) \
+ : "r" (n) \
+ : "memory"); \
+})
+
#define __xchg_relaxed(ptr, new, size) \
({ \
__typeof__(ptr) __ptr = (ptr); \
@@ -18,18 +29,12 @@
__typeof__(*(ptr)) __ret; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- " amoswap.w %0, %2, %1\n" \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".w", "", "", \
+ __ret, __ptr, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- " amoswap.d %0, %2, %1\n" \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".d", "", "", \
+ __ret, __ptr, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -51,20 +56,12 @@
__typeof__(*(ptr)) __ret; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- " amoswap.w %0, %2, %1\n" \
- RISCV_ACQUIRE_BARRIER \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".w", "", RISCV_ACQUIRE_BARRIER, \
+ __ret, __ptr, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- " amoswap.d %0, %2, %1\n" \
- RISCV_ACQUIRE_BARRIER \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".d", "", RISCV_ACQUIRE_BARRIER, \
+ __ret, __ptr, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -86,20 +83,12 @@
__typeof__(*(ptr)) __ret; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- RISCV_RELEASE_BARRIER \
- " amoswap.w %0, %2, %1\n" \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".w", RISCV_RELEASE_BARRIER, "", \
+ __ret, __ptr, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- RISCV_RELEASE_BARRIER \
- " amoswap.d %0, %2, %1\n" \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".d", RISCV_RELEASE_BARRIER, "", \
+ __ret, __ptr, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -121,18 +110,12 @@
__typeof__(*(ptr)) __ret; \
switch (size) { \
case 4: \
- __asm__ __volatile__ ( \
- " amoswap.w.aqrl %0, %2, %1\n" \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".w.aqrl", "", "", \
+ __ret, __ptr, __new); \
break; \
case 8: \
- __asm__ __volatile__ ( \
- " amoswap.d.aqrl %0, %2, %1\n" \
- : "=r" (__ret), "+A" (*__ptr) \
- : "r" (__new) \
- : "memory"); \
+ ___xchg(".d.aqrl", "", "", \
+ __ret, __ptr, __new); \
break; \
default: \
BUILD_BUG(); \
--
2.40.0
^ permalink raw reply [flat|nested] 9+ messages in thread* [RFC PATCH v3 5/6] riscv/cmpxchg: Deduplicate xchg() macros
2023-04-04 16:37 [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Leonardo Bras
` (3 preceding siblings ...)
2023-04-04 16:37 ` [RFC PATCH v3 4/6] riscv/cmpxchg: Deduplicate xchg() asm functions Leonardo Bras
@ 2023-04-04 16:37 ` Leonardo Bras
2023-04-04 16:37 ` [RFC PATCH v3 6/6] riscv/cmpxchg: Deduplicate arch_xchg() macros Leonardo Bras
2023-04-06 2:39 ` [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Guo Ren
6 siblings, 0 replies; 9+ messages in thread
From: Leonardo Bras @ 2023-04-04 16:37 UTC (permalink / raw)
To: Paul Walmsley, Palmer Dabbelt, Albert Ou, Leonardo Bras, Guo Ren,
Andrea Parri, Conor Dooley
Cc: linux-riscv, linux-kernel
Every xchg define (_relaxed, _acquire, _release, vanilla) contain it's own
define for creating tmp variables and selecting the correct asm code for
given variable size.
All those defines are mostly the same code (other than specific barriers),
so there is no need to keep the 4 copies.
Unify those under a more general define, that can reproduce the previous 4
versions.
(This did not cause any change in generated asm)
Signed-off-by: Leonardo Bras <leobras@redhat.com>
---
arch/riscv/include/asm/cmpxchg.h | 72 ++++++--------------------------
1 file changed, 12 insertions(+), 60 deletions(-)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index a524484946815..480234719b78c 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -11,7 +11,7 @@
#include <asm/barrier.h>
#include <asm/fence.h>
-#define ___xchg(sfx, prepend, append, r, p, n) \
+#define ____xchg(sfx, prepend, append, r, p, n) \
({ \
__asm__ __volatile__ ( \
prepend \
@@ -22,19 +22,19 @@
: "memory"); \
})
-#define __xchg_relaxed(ptr, new, size) \
+#define ___xchg(ptr, new, size, sfx, prepend, append) \
({ \
__typeof__(ptr) __ptr = (ptr); \
__typeof__(new) __new = (new); \
__typeof__(*(ptr)) __ret; \
switch (size) { \
case 4: \
- ___xchg(".w", "", "", \
- __ret, __ptr, __new); \
+ ____xchg(".w" sfx, prepend, append, \
+ __ret, __ptr, __new); \
break; \
case 8: \
- ___xchg(".d", "", "", \
- __ret, __ptr, __new); \
+ ____xchg(".d" sfx, prepend, append, \
+ __ret, __ptr, __new); \
break; \
default: \
BUILD_BUG(); \
@@ -42,6 +42,9 @@
__ret; \
})
+#define __xchg_relaxed(ptr, new, size) \
+ ___xchg(ptr, new, size, "", "", "")
+
#define arch_xchg_relaxed(ptr, x) \
({ \
__typeof__(*(ptr)) _x_ = (x); \
@@ -50,24 +53,7 @@
})
#define __xchg_acquire(ptr, new, size) \
-({ \
- __typeof__(ptr) __ptr = (ptr); \
- __typeof__(new) __new = (new); \
- __typeof__(*(ptr)) __ret; \
- switch (size) { \
- case 4: \
- ___xchg(".w", "", RISCV_ACQUIRE_BARRIER, \
- __ret, __ptr, __new); \
- break; \
- case 8: \
- ___xchg(".d", "", RISCV_ACQUIRE_BARRIER, \
- __ret, __ptr, __new); \
- break; \
- default: \
- BUILD_BUG(); \
- } \
- __ret; \
-})
+ ___xchg(ptr, new, size, "", "", RISCV_ACQUIRE_BARRIER)
#define arch_xchg_acquire(ptr, x) \
({ \
@@ -77,24 +63,7 @@
})
#define __xchg_release(ptr, new, size) \
-({ \
- __typeof__(ptr) __ptr = (ptr); \
- __typeof__(new) __new = (new); \
- __typeof__(*(ptr)) __ret; \
- switch (size) { \
- case 4: \
- ___xchg(".w", RISCV_RELEASE_BARRIER, "", \
- __ret, __ptr, __new); \
- break; \
- case 8: \
- ___xchg(".d", RISCV_RELEASE_BARRIER, "", \
- __ret, __ptr, __new); \
- break; \
- default: \
- BUILD_BUG(); \
- } \
- __ret; \
-})
+ ___xchg(ptr, new, size, "", RISCV_RELEASE_BARRIER, "")
#define arch_xchg_release(ptr, x) \
({ \
@@ -104,24 +73,7 @@
})
#define __xchg(ptr, new, size) \
-({ \
- __typeof__(ptr) __ptr = (ptr); \
- __typeof__(new) __new = (new); \
- __typeof__(*(ptr)) __ret; \
- switch (size) { \
- case 4: \
- ___xchg(".w.aqrl", "", "", \
- __ret, __ptr, __new); \
- break; \
- case 8: \
- ___xchg(".d.aqrl", "", "", \
- __ret, __ptr, __new); \
- break; \
- default: \
- BUILD_BUG(); \
- } \
- __ret; \
-})
+ ___xchg(ptr, new, size, ".aqrl", "", "")
#define arch_xchg(ptr, x) \
({ \
--
2.40.0
^ permalink raw reply [flat|nested] 9+ messages in thread* [RFC PATCH v3 6/6] riscv/cmpxchg: Deduplicate arch_xchg() macros
2023-04-04 16:37 [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Leonardo Bras
` (4 preceding siblings ...)
2023-04-04 16:37 ` [RFC PATCH v3 5/6] riscv/cmpxchg: Deduplicate xchg() macros Leonardo Bras
@ 2023-04-04 16:37 ` Leonardo Bras
2023-04-06 2:39 ` [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Guo Ren
6 siblings, 0 replies; 9+ messages in thread
From: Leonardo Bras @ 2023-04-04 16:37 UTC (permalink / raw)
To: Paul Walmsley, Palmer Dabbelt, Albert Ou, Leonardo Bras, Guo Ren,
Andrea Parri, Conor Dooley
Cc: linux-riscv, linux-kernel
Every arch_xchg define (_relaxed, _acquire, _release, vanilla) contain it's
own define for creating tmp variables and calling the correct internal
macro for the desired version.
Those defines are mostly the same code, so there is no need to keep the 4
copies.
Create a helper define to avoid code duplication.
(This did not cause any change in generated asm)
Signed-off-by: Leonardo Bras <leobras@redhat.com>
---
arch/riscv/include/asm/cmpxchg.h | 26 +++++++++-----------------
1 file changed, 9 insertions(+), 17 deletions(-)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index 480234719b78c..905a888d8b04d 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -45,41 +45,33 @@
#define __xchg_relaxed(ptr, new, size) \
___xchg(ptr, new, size, "", "", "")
-#define arch_xchg_relaxed(ptr, x) \
+#define _arch_xchg(order, ptr, x) \
({ \
__typeof__(*(ptr)) _x_ = (x); \
- (__typeof__(*(ptr))) __xchg_relaxed((ptr), \
- _x_, sizeof(*(ptr))); \
+ (__typeof__(*(ptr))) __xchg ## order((ptr), \
+ _x_, sizeof(*(ptr))); \
})
+#define arch_xchg_relaxed(ptr, x) \
+ _arch_xchg(_relaxed, ptr, x)
+
#define __xchg_acquire(ptr, new, size) \
___xchg(ptr, new, size, "", "", RISCV_ACQUIRE_BARRIER)
#define arch_xchg_acquire(ptr, x) \
-({ \
- __typeof__(*(ptr)) _x_ = (x); \
- (__typeof__(*(ptr))) __xchg_acquire((ptr), \
- _x_, sizeof(*(ptr))); \
-})
+ _arch_xchg(_acquire, ptr, x)
#define __xchg_release(ptr, new, size) \
___xchg(ptr, new, size, "", RISCV_RELEASE_BARRIER, "")
#define arch_xchg_release(ptr, x) \
-({ \
- __typeof__(*(ptr)) _x_ = (x); \
- (__typeof__(*(ptr))) __xchg_release((ptr), \
- _x_, sizeof(*(ptr))); \
-})
+ _arch_xchg(_release, ptr, x)
#define __xchg(ptr, new, size) \
___xchg(ptr, new, size, ".aqrl", "", "")
#define arch_xchg(ptr, x) \
-({ \
- __typeof__(*(ptr)) _x_ = (x); \
- (__typeof__(*(ptr))) __xchg((ptr), _x_, sizeof(*(ptr))); \
-})
+ _arch_xchg(, ptr, x)
#define xchg32(ptr, x) \
({ \
--
2.40.0
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros
2023-04-04 16:37 [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Leonardo Bras
` (5 preceding siblings ...)
2023-04-04 16:37 ` [RFC PATCH v3 6/6] riscv/cmpxchg: Deduplicate arch_xchg() macros Leonardo Bras
@ 2023-04-06 2:39 ` Guo Ren
2023-04-06 8:04 ` Leonardo Bras Soares Passos
6 siblings, 1 reply; 9+ messages in thread
From: Guo Ren @ 2023-04-06 2:39 UTC (permalink / raw)
To: Leonardo Bras
Cc: Paul Walmsley, Palmer Dabbelt, Albert Ou, Andrea Parri,
Conor Dooley, linux-riscv, linux-kernel
Thx Leonardo,
It's a good point for cleaning up.
On Wed, Apr 5, 2023 at 12:37 AM Leonardo Bras <leobras@redhat.com> wrote:
>
> While studying riscv's cmpxchg.h file, I got really interested in
> understanding how RISCV asm implemented the different versions of
> {cmp,}xchg.
>
> When I understood the pattern, it made sense for me to remove the
> duplications and create macros to make it easier to understand what exactly
> changes between the versions: Instruction sufixes & barriers.
>
> I split those changes in 3 levels for each cmpxchg and xchg, resulting a
But how about merging 3 levels into one? (I mean one for cmpxchg and
one for xchg, resulting in 2 patches.)
> total of 6 patches. I did this so it becomes easier to review and remove
> the last levels if desired, but I have no issue squashing them if it's
> better.
>
> Please provide comments.
>
> Thanks!
> Leo
>
> Changes since v2:
> - Fixed macros that depend on having a local variable with a magic name
> - Previous cast to (long) is now only applied on 4-bytes cmpxchg
>
> Changes since v1:
> - Fixed patch 4/6 suffix from 'w.aqrl' to '.w.aqrl', to avoid build error
>
>
> Leonardo Bras (6):
> riscv/cmpxchg: Deduplicate cmpxchg() asm functions
> riscv/cmpxchg: Deduplicate cmpxchg() macros
> riscv/cmpxchg: Deduplicate arch_cmpxchg() macros
> riscv/cmpxchg: Deduplicate xchg() asm functions
> riscv/cmpxchg: Deduplicate xchg() macros
> riscv/cmpxchg: Deduplicate arch_xchg() macros
>
> arch/riscv/include/asm/cmpxchg.h | 319 +++++++------------------------
> 1 file changed, 67 insertions(+), 252 deletions(-)
>
> --
> 2.40.0
>
--
Best Regards
Guo Ren
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros
2023-04-06 2:39 ` [RFC PATCH v3 0/6] Deduplicating RISCV cmpxchg.h macros Guo Ren
@ 2023-04-06 8:04 ` Leonardo Bras Soares Passos
0 siblings, 0 replies; 9+ messages in thread
From: Leonardo Bras Soares Passos @ 2023-04-06 8:04 UTC (permalink / raw)
To: Guo Ren
Cc: Paul Walmsley, Palmer Dabbelt, Albert Ou, Andrea Parri,
Conor Dooley, linux-riscv, linux-kernel
On Wed, Apr 5, 2023 at 11:40 PM Guo Ren <guoren@kernel.org> wrote:
>
> Thx Leonardo,
>
> It's a good point for cleaning up.
Hello Guo,
Thanks for the feedback!
>
> On Wed, Apr 5, 2023 at 12:37 AM Leonardo Bras <leobras@redhat.com> wrote:
> >
> > While studying riscv's cmpxchg.h file, I got really interested in
> > understanding how RISCV asm implemented the different versions of
> > {cmp,}xchg.
> >
> > When I understood the pattern, it made sense for me to remove the
> > duplications and create macros to make it easier to understand what exactly
> > changes between the versions: Instruction sufixes & barriers.
> >
> > I split those changes in 3 levels for each cmpxchg and xchg, resulting a
> But how about merging 3 levels into one? (I mean one for cmpxchg and
> one for xchg, resulting in 2 patches.)
Sure, as mentioned below, I have no issue squashing the patches, so 2
should work fine.
I was previously worried about the changes, so I split them in 3 so
it's easier to understand my thought process.
I will resend it in a couple minutes.
>
> > total of 6 patches. I did this so it becomes easier to review and remove
> > the last levels if desired, but I have no issue squashing them if it's
> > better.
> >
> > Please provide comments.
> >
> > Thanks!
> > Leo
> >
> > Changes since v2:
> > - Fixed macros that depend on having a local variable with a magic name
> > - Previous cast to (long) is now only applied on 4-bytes cmpxchg
> >
> > Changes since v1:
> > - Fixed patch 4/6 suffix from 'w.aqrl' to '.w.aqrl', to avoid build error
> >
> >
> > Leonardo Bras (6):
> > riscv/cmpxchg: Deduplicate cmpxchg() asm functions
> > riscv/cmpxchg: Deduplicate cmpxchg() macros
> > riscv/cmpxchg: Deduplicate arch_cmpxchg() macros
> > riscv/cmpxchg: Deduplicate xchg() asm functions
> > riscv/cmpxchg: Deduplicate xchg() macros
> > riscv/cmpxchg: Deduplicate arch_xchg() macros
> >
> > arch/riscv/include/asm/cmpxchg.h | 319 +++++++------------------------
> > 1 file changed, 67 insertions(+), 252 deletions(-)
> >
> > --
> > 2.40.0
> >
>
>
> --
> Best Regards
> Guo Ren
>
Best regards,
Leo
^ permalink raw reply [flat|nested] 9+ messages in thread