x86 assembly: addpd

userland/arch/aarch64/simd.S

@@ -35,7 +35,7 @@ ENTRY
     u64_1:          .quad 0x1555555515555555, 0x1666666616666666
     u64_sum_expect: .quad 0x0666666706666666, 0x0888888908888888
 .bss
-    u64_sum: .skip 16
+    u64_sum:        .skip 16
 .text
     adr x0, u64_0
     ld1 {v0.2d}, [x0]
@@ -56,7 +56,7 @@ ENTRY
     f32_1:          .float 5.5, 6.5,  7.5,  8.5
     f32_sum_expect: .float 7.0, 9.0, 11.0, 13.0
 .bss
-    f32_sum: .skip 16
+    f32_sum:        .skip 16
 .text
     adr x0, f32_0
     ld1 {v0.4s}, [x0]

userland/arch/x86_64/addpd.S

@@ -0,0 +1,32 @@
+/* https://github.com/cirosantilli/linux-kernel-module-cheat#x86-simd
+ *
+ * Add a bunch of floating point numbers in one go.
+ */
+
+#include "common.h"
+
+ENTRY
+.bss
+    output: .skip 16
+
+#define t
+
+.data
+    addps_input0: .float 1.5, 2.5,  3.5,  4.5
+    addps_input1: .float 5.5, 6.5,  7.5,  8.5
+    addps_expect: .float 7.0, 9.0, 11.0, 13.0
+    addpd_input0: .double 1.5, 2.5
+    addpd_input1: .double 5.5, 6.5
+    addpd_expect: .double 7.0, 9.0
+.text
+#define TEST(size) \
+    movups addp ## size ## _input0, %xmm0; \
+    movups addp ## size ## _input1, %xmm1; \
+    addp ## size %xmm1, %xmm0; \
+    movups %xmm0, output; \
+    ASSERT_MEMCMP(output, addp ## size ## _expect, $0x10)
+
+    TEST(s)
+    TEST(d)
+#undef TEST
+EXIT

userland/arch/x86_64/paddq.S

@@ -1,6 +1,8 @@
 /* https://github.com/cirosantilli/linux-kernel-module-cheat#x86-simd
  *
  * Add a bunch of integers in one go.
+ *
+ * The different variants basically determine if carries get forwarded or not.
  */
 
 #include "common.h"
@@ -14,14 +16,14 @@ ENTRY
     paddd_expect: .long 0x04040403, 0x06060605, 0x08080807, 0x0A0A0A09
     paddq_expect: .long 0x04040403, 0x06060606, 0x08080807, 0x0A0A0A0A
 .bss
-    result: .skip 16
+    output:       .skip 16
 .text
     movups input1, %xmm1
 #define TEST(size) \
     movups input0, %xmm0; \
     padd ## size %xmm1, %xmm0; \
-    movups %xmm0, result; \
-    ASSERT_MEMCMP(result, padd ## size ## _expect, $0x10)
+    movups %xmm0, output; \
+    ASSERT_MEMCMP(output, padd ## size ## _expect, $0x10)
 
     TEST(b)
     TEST(w)