[compiler-rt] dd369c7 - [compiler-rt] Fix tests of __aeabi_(idivmod|uidivmod|uldivmod) to support big endian (#126277)

[via llvm-commits]

Author: Victor Campos
Date: 2025-02-11T09:49:56Z
New Revision: dd369c771e215ebf7bf2d09e874643788e49d334

URL: https://github.com/llvm/llvm-project/commit/dd369c771e215ebf7bf2d09e874643788e49d334
DIFF: https://github.com/llvm/llvm-project/commit/dd369c771e215ebf7bf2d09e874643788e49d334.diff

LOG: [compiler-rt] Fix tests of __aeabi_(idivmod|uidivmod|uldivmod) to support big endian (#126277)

This patch makes these functions' tests work in big endian mode:
 - `__aeabi_idivmod`.
 - `__aeabi_uidivmod`.
 -  `__aeabi_uldivmod`.

The three functions return a struct containing two fields, quotient and
remainder, via *value in regs* calling convention. They differ in the
integer type of each field.

In the tests of the first two, a 64-bit integer is used as the return
type of the call. And as consequence of the ABI rules for structs
(Composite Types), the quotient resides in `r0` and the remainder in
`r1` regardless of endianness. So, in order to access each component
from the 64-bit integer in the caller code, care must be taken to access
the correct bits as they do depend on endianness in this case.

In the test of the third one, the caller code has inline assembly to
access the components. This assembly code assumed little endian, so it
had to be made flexible for big endian as well.

`_YUGA_BIG_ENDIAN` is defined in `int_endianness.h`. It's a macro
internal to compiler-rt that's in theory compatible with more toolchains
than gcc and clang.

Added: 
    

Modified: 
    compiler-rt/test/builtins/Unit/arm/aeabi_idivmod_test.c
    compiler-rt/test/builtins/Unit/arm/aeabi_uidivmod_test.c
    compiler-rt/test/builtins/Unit/arm/aeabi_uldivmod_test.c

Removed: 
    


################################################################################
diff  --git a/compiler-rt/test/builtins/Unit/arm/aeabi_idivmod_test.c b/compiler-rt/test/builtins/Unit/arm/aeabi_idivmod_test.c
index ff767d754eb974c..e4953bf51b037d6 100644
--- a/compiler-rt/test/builtins/Unit/arm/aeabi_idivmod_test.c
+++ b/compiler-rt/test/builtins/Unit/arm/aeabi_idivmod_test.c
@@ -14,8 +14,19 @@ int test__aeabi_idivmod(si_int a, si_int b,
 {
 	  si_int rem;
     du_int ret = __aeabi_idivmod(a, b);
+    // __aeabi_idivmod actually returns a struct { quotient; remainder; } using
+    // value_in_regs calling convention. Due to the ABI rules, struct fields
+    // come in the same order regardless of endianness. However since the
+    // result is received here as a 64-bit integer, in which endianness does
+    // matter, the position of each component (quotient and remainder) varies
+    // depending on endianness.
+#  if _YUGA_BIG_ENDIAN
+    rem = ret & 0xFFFFFFFF;
+    si_int result = ret >> 32;
+#  else
     rem = ret >> 32;
     si_int result = ret & 0xFFFFFFFF;
+#  endif
     if (result != expected_result) {
         printf("error in __aeabi_idivmod: %d / %d = %d, expected %d\n",
                a, b, result, expected_result);

diff  --git a/compiler-rt/test/builtins/Unit/arm/aeabi_uidivmod_test.c b/compiler-rt/test/builtins/Unit/arm/aeabi_uidivmod_test.c
index de5a43d5b8143cd..a507267c446023b 100644
--- a/compiler-rt/test/builtins/Unit/arm/aeabi_uidivmod_test.c
+++ b/compiler-rt/test/builtins/Unit/arm/aeabi_uidivmod_test.c
@@ -13,8 +13,19 @@ int test__aeabi_uidivmod(su_int a, su_int b,
 						su_int expected_result, su_int expected_rem)
 {
     du_int ret = __aeabi_uidivmod(a, b);
+    // __aeabi_uidivmod actually returns a struct { quotient; remainder; }
+    // using value_in_regs calling convention. Due to the ABI rules, struct
+    // fields come in the same order regardless of endianness. However since
+    // the result is received here as a 64-bit integer, in which endianness
+    // does matter, the position of each component (quotient and remainder)
+    // varies depending on endianness.
+#  if _YUGA_BIG_ENDIAN
+    su_int rem = ret & 0xFFFFFFFF;
+    si_int result = ret >> 32;
+#  else
     su_int rem = ret >> 32;
     si_int result = ret & 0xFFFFFFFF;
+#  endif
 
     if (result != expected_result) {
         printf("error in __aeabi_uidivmod: %u / %u = %u, expected %u\n",

diff  --git a/compiler-rt/test/builtins/Unit/arm/aeabi_uldivmod_test.c b/compiler-rt/test/builtins/Unit/arm/aeabi_uldivmod_test.c
index 2ff65a8b9ec3fa8..02cfa2f38713b90 100644
--- a/compiler-rt/test/builtins/Unit/arm/aeabi_uldivmod_test.c
+++ b/compiler-rt/test/builtins/Unit/arm/aeabi_uldivmod_test.c
@@ -12,20 +12,37 @@ COMPILER_RT_ABI void /* __value_in_regs */ __aeabi_uldivmod(du_int a, du_int b);
 int test_aeabi_uldivmod(du_int a, du_int b, du_int expected_q, du_int expected_r)
 {
     du_int q, r;
+    // __aeabi_uldivmod returns a struct { quotient; remainder; } using
+    // value_in_regs calling convention. Each field is a 64-bit integer, so the
+    // quotient resides in r0 and r1, while the remainder in r2 and r3. The
+    // byte order however depends on the endianness.
     __asm__(
+#  if _YUGA_BIG_ENDIAN
+        "movs r1, %Q[a] \n"
+        "movs r0, %R[a] \n"
+        "movs r3, %Q[b] \n"
+        "movs r2, %R[b] \n"
+#  else
         "movs r0, %Q[a] \n"
         "movs r1, %R[a] \n"
         "movs r2, %Q[b] \n"
         "movs r3, %R[b] \n"
+#  endif
         "bl __aeabi_uldivmod \n"
+#  if _YUGA_BIG_ENDIAN
+        "movs %Q[q], r1\n"
+        "movs %R[q], r0\n"
+        "movs %Q[r], r3\n"
+        "movs %R[r], r2\n"
+#  else
         "movs %Q[q], r0\n"
         "movs %R[q], r1\n"
         "movs %Q[r], r2\n"
         "movs %R[r], r3\n"
-        : [q] "=r" (q), [r] "=r"(r)
+#  endif
+        : [q] "=r"(q), [r] "=r"(r)
         : [a] "r"(a), [b] "r"(b)
-        : "lr", "r0", "r1", "r2", "r3"
-        );
+        : "lr", "r0", "r1", "r2", "r3");
     if (q != expected_q || r != expected_r)
         printf("error in aeabi_uldivmod: %llX / %llX = %llX, R = %llX, expected %llX, %llX\n",
                a, b, q, r, expected_q, expected_r);