b/sysdeps/x86_64/fpu/multiarch/svml_d_cos8_core_avx512.S
@@ -278,7 +278,13 @@ WRAPPER_IMPL_AVX512 _ZGVdN4v_cos
vmovups __dC7(%rax), %zmm9
/* Get absolute argument value: X' = |X'| */
- vandpd __dAbsMask(%rax), %zmm6, %zmm1
+/* Encoded vandpd __dAbsMask(%rax), %zmm6, %zmm1 (with __dAbsMask = 0). */
+ .byte 0x62
+ .byte 0xf1
+ .byte 0xcd
+ .byte 0x48
+ .byte 0x54
+ .byte 0x08
/* Y = X'*InvPi + RS : right shifter add */
vfmadd213pd %zmm4, %zmm3, %zmm6
@@ -324,7 +330,14 @@ WRAPPER_IMPL_AVX512 _ZGVdN4v_cos
RECONSTRUCTION:
Final sign setting: Res = Poly^SignRes
*/
- vxorpd %zmm13, %zmm12, %zmm1
+/* Encoded vxorpd %zmm13, %zmm12, %zmm1. */
+ .byte 0x62
+ .byte 0xd1
+ .byte 0x9d
+ .byte 0x48
+ .byte 0x57
+ .byte 0xcd
+
kmovw %k0, %ecx
testl %ecx, %ecx
jne .LBL_2_3
b/sysdeps/x86_64/fpu/multiarch/svml_s_cosf16_core_avx512.S
@@ -287,7 +287,14 @@ WRAPPER_IMPL_AVX512 _ZGVdN8v_cosf
vaddps __sHalfPI(%rax), %zmm0, %zmm2
/* Check for large and special arguments */
- vandps __sAbsMask(%rax), %zmm0, %zmm1
+/* Encoded vandps __sAbsMask(%rax), %zmm0, %zmm1 (__sAbsMask = 0x1000). */
+ .byte 0x62
+ .byte 0xf1
+ .byte 0x7c
+ .byte 0x48
+ .byte 0x54
+ .byte 0x48
+ .byte 0x40
/*
1) Range reduction to [-Pi/2; +Pi/2] interval
@@ -320,7 +327,13 @@ WRAPPER_IMPL_AVX512 _ZGVdN8v_cosf
a) Set shifted destination sign using XOR operation:
R = XOR( R, S );
*/
- vxorps %zmm8, %zmm7, %zmm11
+/* Encoded vxorps %zmm8, %zmm7, %zmm11. */
+ .byte 0x62
+ .byte 0x51
+ .byte 0x44
+ .byte 0x48
+ .byte 0x57
+ .byte 0xd8
/*
b) Calculate polynomial: