Test COMPARE HALFWORD RELATIVE LONG instructions.
Test that the bytes following the second operand do not affect the
instruction.
Test the sign extension performed on the second operand.

Signed-off-by: Nina Schoetterl-Glausch <nsg@linux.ibm.com>
---


I don't know what the coding style is for inline asm.
checkpatch.sh complains about the tabs inside the asm, which I find a
bit surprising given they're inside a string.
IMO emitting tabs makes sense in order to be consistent with gcc output.
I left the tabs in for now, but can remove them.


 tests/tcg/s390x/chrl.c          | 76 +++++++++++++++++++++++++++++++++
 tests/tcg/s390x/Makefile.target |  1 +
 2 files changed, 77 insertions(+)
 create mode 100644 tests/tcg/s390x/chrl.c

diff --git a/tests/tcg/s390x/chrl.c b/tests/tcg/s390x/chrl.c
new file mode 100644
index 0000000000..fa33d43c59
--- /dev/null
+++ b/tests/tcg/s390x/chrl.c
@@ -0,0 +1,76 @@
+#include <stdlib.h>
+#include <assert.h>
+#include <stdint.h>
+
+static void test_chrl(void)
+{
+    uint32_t program_mask, cc;
+
+    asm volatile (
+               ".pushsection .rodata\n"
+        "0:	.short	1,0x8000\n"
+        "	.popsection\n"
+
+        "	chrl	%[r],0b\n"
+        "	ipm	%[program_mask]\n"
+        : [program_mask] "=r" (program_mask)
+        : [r] "r" (1)
+    );
+
+    cc = program_mask >> 28;
+    assert(!cc);
+
+    asm volatile (
+               ".pushsection .rodata\n"
+        "0:	.short	-1,0x8000\n"
+        "	.popsection\n"
+
+        "	chrl	%[r],0b\n"
+        "	ipm	%[program_mask]\n"
+        : [program_mask] "=r" (program_mask)
+        : [r] "r" (-1)
+    );
+
+    cc = program_mask >> 28;
+    assert(!cc);
+}
+
+static void test_cghrl(void)
+{
+    uint32_t program_mask, cc;
+
+    asm volatile (
+               ".pushsection .rodata\n"
+        "0:	.short	1,0x8000,0,0\n"
+        "	.popsection\n"
+
+        "	cghrl	%[r],0b\n"
+        "	ipm	%[program_mask]\n"
+        : [program_mask] "=r" (program_mask)
+        : [r] "r" (1L)
+    );
+
+    cc = program_mask >> 28;
+    assert(!cc);
+
+    asm volatile (
+               ".pushsection .rodata\n"
+        "0:	.short	-1,0x8000,0,0\n"
+        "	.popsection\n"
+
+        "	cghrl	%[r],0b\n"
+        "	ipm	%[program_mask]\n"
+        : [program_mask] "=r" (program_mask)
+        : [r] "r" (-1L)
+    );
+
+    cc = program_mask >> 28;
+    assert(!cc);
+}
+
+int main(void)
+{
+    test_chrl();
+    test_cghrl();
+    return EXIT_SUCCESS;
+}
diff --git a/tests/tcg/s390x/Makefile.target b/tests/tcg/s390x/Makefile.target
index 72ad309b27..a3d3beeffe 100644
--- a/tests/tcg/s390x/Makefile.target
+++ b/tests/tcg/s390x/Makefile.target
@@ -28,6 +28,7 @@ TESTS+=div
 TESTS+=clst
 TESTS+=long-double
 TESTS+=cdsg
+TESTS+=chrl
 
 cdsg: CFLAGS+=-pthread
 cdsg: LDFLAGS+=-pthread
-- 
2.37.2

On 08/03/2023 22.02, Nina Schoetterl-Glausch wrote:
> Test COMPARE HALFWORD RELATIVE LONG instructions.
> Test that the bytes following the second operand do not affect the
> instruction.
> Test the sign extension performed on the second operand.
> 
> Signed-off-by: Nina Schoetterl-Glausch <nsg@linux.ibm.com>
> ---
> 
> 
> I don't know what the coding style is for inline asm.
> checkpatch.sh complains about the tabs inside the asm, which I find a
> bit surprising given they're inside a string.
> IMO emitting tabs makes sense in order to be consistent with gcc output.
> I left the tabs in for now, but can remove them.

I don't mind too much, but all the other files use spaces, not tabs, so I 
think it's maybe best to also use spaces here for consistency?

  Thomas

On Thu, 2023-03-09 at 14:47 +0100, Thomas Huth wrote:
> On 08/03/2023 22.02, Nina Schoetterl-Glausch wrote:
> > Test COMPARE HALFWORD RELATIVE LONG instructions.
> > Test that the bytes following the second operand do not affect the
> > instruction.
> > Test the sign extension performed on the second operand.
> > 
> > Signed-off-by: Nina Schoetterl-Glausch <nsg@linux.ibm.com>
> > ---
> > 
> > 
> > I don't know what the coding style is for inline asm.
> > checkpatch.sh complains about the tabs inside the asm, which I find a
> > bit surprising given they're inside a string.
> > IMO emitting tabs makes sense in order to be consistent with gcc output.
> > I left the tabs in for now, but can remove them.
> 
> I don't mind too much, but all the other files use spaces, not tabs, so I 
> think it's maybe best to also use spaces here for consistency?

Seems pretty inconsistent. Sometimes \t is used everywhere, sometimes some number of spaces.
I'll put \n\t at the end of lines, put the labels on a separate line and use a single space
between the mnemonic and the operands.
> 
>   Thomas
> 

On 3/8/23 13:02, Nina Schoetterl-Glausch wrote:
> Test COMPARE HALFWORD RELATIVE LONG instructions.
> Test that the bytes following the second operand do not affect the
> instruction.
> Test the sign extension performed on the second operand.
> 
> Signed-off-by: Nina Schoetterl-Glausch<nsg@linux.ibm.com>
> ---

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~