@@ -49,14 +49,6 @@ typedef double __m512d __attribute__ ((__vector_size__ (64), __may_alias__));
typedef unsigned char __mmask8;
typedef unsigned short __mmask16;
-/* Rounding mode macros. */
-#define _MM_FROUND_TO_NEAREST_INT 0x00
-#define _MM_FROUND_TO_NEG_INF 0x01
-#define _MM_FROUND_TO_POS_INF 0x02
-#define _MM_FROUND_TO_ZERO 0x03
-#define _MM_FROUND_CUR_DIRECTION 0x04
-#define _MM_FROUND_NO_EXC 0x05
-
extern __inline __m512i
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_set_epi64 (long long __A, long long __B, long long __C,
@@ -14706,7 +14706,8 @@ get_some_local_dynamic_name (void)
F,f -- likewise, but for floating-point.
O -- if HAVE_AS_IX86_CMOV_SUN_SYNTAX, expand to "w.", "l." or "q.",
otherwise nothing
- R -- print the prefix for register names.
+ R -- print embeded rounding and sae.
+ r -- print only sae.
z -- print the opcode suffix for the size of the current operand.
Z -- likewise, with special suffixes for x87 instructions.
* -- print a star (in certain assembler syntax)
@@ -15156,6 +15157,20 @@ ix86_print_operand (FILE *file, rtx x, int code)
fputs ("{z}", file);
return;
+ case 'r':
+ gcc_assert (CONST_INT_P (x));
+ gcc_assert (INTVAL (x) == ROUND_SAE);
+
+ if (ASSEMBLER_DIALECT == ASM_INTEL)
+ fputs (", ", file);
+
+ fputs ("{sae}", file);
+
+ if (ASSEMBLER_DIALECT == ASM_ATT)
+ fputs (", ", file);
+
+ return;
+
case 'R':
gcc_assert (CONST_INT_P (x));
@@ -15164,21 +15179,18 @@ ix86_print_operand (FILE *file, rtx x, int code)
switch (INTVAL (x))
{
- case ROUND_NEAREST_INT:
+ case ROUND_NEAREST_INT | ROUND_SAE:
fputs ("{rn-sae}", file);
break;
- case ROUND_NEG_INF:
+ case ROUND_NEG_INF | ROUND_SAE:
fputs ("{rd-sae}", file);
break;
- case ROUND_POS_INF:
+ case ROUND_POS_INF | ROUND_SAE:
fputs ("{ru-sae}", file);
break;
- case ROUND_ZERO:
+ case ROUND_ZERO | ROUND_SAE:
fputs ("{rz-sae}", file);
break;
- case ROUND_SAE:
- fputs ("{sae}", file);
- break;
default:
gcc_unreachable ();
}
@@ -251,7 +251,7 @@
(ROUND_POS_INF 2)
(ROUND_ZERO 3)
(NO_ROUND 4)
- (ROUND_SAE 5)
+ (ROUND_SAE 8)
])
;; Constants to represent pcomtrue/pcomfalse variants
@@ -676,6 +676,22 @@
return i == 3 || i == 5 || i == 9;
})
+;; Match 4 or 8 to 11. Used for embeded rounding.
+(define_predicate "const_4_or_8_to_11_operand"
+ (match_code "const_int")
+{
+ HOST_WIDE_INT i = INTVAL (op);
+ return i == 4 || (i >= 8 && i <= 11);
+})
+
+;; Match 4 or 8. Used for SAE.
+(define_predicate "const48_operand"
+ (match_code "const_int")
+{
+ HOST_WIDE_INT i = INTVAL (op);
+ return i == 4 || i == 8;
+})
+
;; Match 0 or 1.
(define_predicate "const_0_to_1_operand"
(and (match_code "const_int")
@@ -128,18 +128,18 @@
[(parallel[
(set (match_dup 0)
(match_dup 1))
- (unspec [(match_operand:SI 2 "const_0_to_4_operand")] UNSPEC_EMBEDDED_ROUNDING)])])
+ (unspec [(match_operand:SI 2 "const_4_or_8_to_11_operand")] UNSPEC_EMBEDDED_ROUNDING)])])
(define_subst_attr "round_saeonly_name" "round_saeonly" "" "_round")
-(define_subst_attr "round_saeonly_mask_operand2" "mask" "%R2" "%R4")
-(define_subst_attr "round_saeonly_mask_operand3" "mask" "%R3" "%R5")
-(define_subst_attr "round_saeonly_mask_scalar_merge_operand4" "mask_scalar_merge" "%R4" "%R5")
-(define_subst_attr "round_saeonly_sd_mask_operand5" "sd" "%R5" "%R7")
-(define_subst_attr "round_saeonly_op2" "round_saeonly" "" "%R2")
-(define_subst_attr "round_saeonly_op3" "round_saeonly" "" "%R3")
-(define_subst_attr "round_saeonly_op4" "round_saeonly" "" "%R4")
-(define_subst_attr "round_saeonly_op5" "round_saeonly" "" "%R5")
-(define_subst_attr "round_saeonly_op6" "round_saeonly" "" "%R6")
+(define_subst_attr "round_saeonly_mask_operand2" "mask" "%r2" "%r4")
+(define_subst_attr "round_saeonly_mask_operand3" "mask" "%r3" "%r5")
+(define_subst_attr "round_saeonly_mask_scalar_merge_operand4" "mask_scalar_merge" "%r4" "%r5")
+(define_subst_attr "round_saeonly_sd_mask_operand5" "sd" "%r5" "%r7")
+(define_subst_attr "round_saeonly_op2" "round_saeonly" "" "%r2")
+(define_subst_attr "round_saeonly_op3" "round_saeonly" "" "%r3")
+(define_subst_attr "round_saeonly_op4" "round_saeonly" "" "%r4")
+(define_subst_attr "round_saeonly_op5" "round_saeonly" "" "%r5")
+(define_subst_attr "round_saeonly_op6" "round_saeonly" "" "%r6")
(define_subst_attr "round_saeonly_prefix" "round_saeonly" "vex" "evex")
(define_subst_attr "round_saeonly_mask_op2" "round_saeonly" "" "<round_saeonly_mask_operand2>")
(define_subst_attr "round_saeonly_mask_op3" "round_saeonly" "" "<round_saeonly_mask_operand3>")
@@ -157,7 +157,7 @@
[(parallel[
(set (match_dup 0)
(match_dup 1))
- (unspec [(match_operand:SI 2 "const_4_to_5_operand")] UNSPEC_EMBEDDED_ROUNDING)])])
+ (unspec [(match_operand:SI 2 "const48_operand")] UNSPEC_EMBEDDED_ROUNDING)])])
(define_subst_attr "round_expand_name" "round_expand" "" "_round")
(define_subst_attr "round_expand_nimm_predicate" "round_expand" "nonimmediate_operand" "register_operand")
@@ -175,7 +175,7 @@
(match_dup 2)
(match_dup 3)
(match_dup 4)
- (unspec [(match_operand:SI 5 "const_0_to_4_operand")] UNSPEC_EMBEDDED_ROUNDING)])
+ (unspec [(match_operand:SI 5 "const_4_or_8_to_11_operand")] UNSPEC_EMBEDDED_ROUNDING)])
(define_subst_attr "round_saeonly_expand_name" "round_saeonly_expand" "" "_round")
(define_subst_attr "round_saeonly_expand_nimm_predicate" "round_saeonly_expand" "nonimmediate_operand" "register_operand")
@@ -195,4 +195,4 @@
(match_dup 3)
(match_dup 4)
(match_dup 5)
- (unspec [(match_operand:SI 6 "const_4_to_5_operand")] UNSPEC_EMBEDDED_ROUNDING)])
+ (unspec [(match_operand:SI 6 "const48_operand")] UNSPEC_EMBEDDED_ROUNDING)])
@@ -167,86 +167,86 @@
#define __builtin_ia32_xabort(I) __builtin_ia32_xabort(0)
/* avx512fintrin.h */
-#define __builtin_ia32_addpd512_mask(A, B, C, D, E) __builtin_ia32_addpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_addps512_mask(A, B, C, D, E) __builtin_ia32_addps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_addsd_round(A, B, C) __builtin_ia32_addsd_round(A, B, 1)
-#define __builtin_ia32_addss_round(A, B, C) __builtin_ia32_addss_round(A, B, 1)
+#define __builtin_ia32_addpd512_mask(A, B, C, D, E) __builtin_ia32_addpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_addps512_mask(A, B, C, D, E) __builtin_ia32_addps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_addsd_round(A, B, C) __builtin_ia32_addsd_round(A, B, 8)
+#define __builtin_ia32_addss_round(A, B, C) __builtin_ia32_addss_round(A, B, 8)
#define __builtin_ia32_alignd512_mask(A, B, F, D, E) __builtin_ia32_alignd512_mask(A, B, 1, D, E)
#define __builtin_ia32_alignq512_mask(A, B, F, D, E) __builtin_ia32_alignq512_mask(A, B, 1, D, E)
#define __builtin_ia32_cmpd512_mask(A, B, E, D) __builtin_ia32_cmpd512_mask(A, B, 1, D)
-#define __builtin_ia32_cmppd512_mask(A, B, F, D, E) __builtin_ia32_cmppd512_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cmpps512_mask(A, B, F, D, E) __builtin_ia32_cmpps512_mask(A, B, 1, D, 5)
+#define __builtin_ia32_cmppd512_mask(A, B, F, D, E) __builtin_ia32_cmppd512_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cmpps512_mask(A, B, F, D, E) __builtin_ia32_cmpps512_mask(A, B, 1, D, 8)
#define __builtin_ia32_cmpq512_mask(A, B, E, D) __builtin_ia32_cmpq512_mask(A, B, 1, D)
-#define __builtin_ia32_cmpsd_mask(A, B, F, D, E) __builtin_ia32_cmpsd_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cmpss_mask(A, B, F, D, E) __builtin_ia32_cmpss_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cvtdq2ps512_mask(A, B, C, D) __builtin_ia32_cvtdq2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2dq512_mask(A, B, C, D) __builtin_ia32_cvtpd2dq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2ps512_mask(A, B, C, D) __builtin_ia32_cvtpd2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2udq512_mask(A, B, C, D) __builtin_ia32_cvtpd2udq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtps2dq512_mask(A, B, C, D) __builtin_ia32_cvtps2dq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtps2pd512_mask(A, B, C, D) __builtin_ia32_cvtps2pd512_mask(A, B, C, 5)
-#define __builtin_ia32_cvtps2udq512_mask(A, B, C, D) __builtin_ia32_cvtps2udq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtsd2ss_round(A, B, C) __builtin_ia32_cvtsd2ss_round(A, B, 1)
+#define __builtin_ia32_cmpsd_mask(A, B, F, D, E) __builtin_ia32_cmpsd_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cmpss_mask(A, B, F, D, E) __builtin_ia32_cmpss_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cvtdq2ps512_mask(A, B, C, D) __builtin_ia32_cvtdq2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2dq512_mask(A, B, C, D) __builtin_ia32_cvtpd2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2ps512_mask(A, B, C, D) __builtin_ia32_cvtpd2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2udq512_mask(A, B, C, D) __builtin_ia32_cvtpd2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2dq512_mask(A, B, C, D) __builtin_ia32_cvtps2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2pd512_mask(A, B, C, D) __builtin_ia32_cvtps2pd512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2udq512_mask(A, B, C, D) __builtin_ia32_cvtps2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtsd2ss_round(A, B, C) __builtin_ia32_cvtsd2ss_round(A, B, 8)
#define __builtin_ia32_cvtss2sd_round(A, B, C) __builtin_ia32_cvtss2sd_round(A, B, 4)
-#define __builtin_ia32_cvtsi2sd64(A, B, C) __builtin_ia32_cvtsi2sd64(A, B, 1)
-#define __builtin_ia32_cvtsi2ss32(A, B, C) __builtin_ia32_cvtsi2ss32(A, B, 1)
-#define __builtin_ia32_cvtsi2ss64(A, B, C) __builtin_ia32_cvtsi2ss64(A, B, 1)
-#define __builtin_ia32_cvttpd2dq512_mask(A, B, C, D) __builtin_ia32_cvttpd2dq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttpd2udq512_mask(A, B, C, D) __builtin_ia32_cvttpd2udq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttps2dq512_mask(A, B, C, D) __builtin_ia32_cvttps2dq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttps2udq512_mask(A, B, C, D) __builtin_ia32_cvttps2udq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvtudq2ps512_mask(A, B, C, D) __builtin_ia32_cvtudq2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtusi2sd64(A, B, C) __builtin_ia32_cvtusi2sd64(A, B, 1)
-#define __builtin_ia32_cvtusi2ss32(A, B, C) __builtin_ia32_cvtusi2ss32(A, B, 1)
-#define __builtin_ia32_cvtusi2ss64(A, B, C) __builtin_ia32_cvtusi2ss64(A, B, 1)
-#define __builtin_ia32_divpd512_mask(A, B, C, D, E) __builtin_ia32_divpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_divps512_mask(A, B, C, D, E) __builtin_ia32_divps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_divsd_round(A, B, C) __builtin_ia32_divsd_round(A, B, 1)
-#define __builtin_ia32_divss_round(A, B, C) __builtin_ia32_divss_round(A, B, 1)
+#define __builtin_ia32_cvtsi2sd64(A, B, C) __builtin_ia32_cvtsi2sd64(A, B, 8)
+#define __builtin_ia32_cvtsi2ss32(A, B, C) __builtin_ia32_cvtsi2ss32(A, B, 8)
+#define __builtin_ia32_cvtsi2ss64(A, B, C) __builtin_ia32_cvtsi2ss64(A, B, 8)
+#define __builtin_ia32_cvttpd2dq512_mask(A, B, C, D) __builtin_ia32_cvttpd2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttpd2udq512_mask(A, B, C, D) __builtin_ia32_cvttpd2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttps2dq512_mask(A, B, C, D) __builtin_ia32_cvttps2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttps2udq512_mask(A, B, C, D) __builtin_ia32_cvttps2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtudq2ps512_mask(A, B, C, D) __builtin_ia32_cvtudq2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtusi2sd64(A, B, C) __builtin_ia32_cvtusi2sd64(A, B, 8)
+#define __builtin_ia32_cvtusi2ss32(A, B, C) __builtin_ia32_cvtusi2ss32(A, B, 8)
+#define __builtin_ia32_cvtusi2ss64(A, B, C) __builtin_ia32_cvtusi2ss64(A, B, 8)
+#define __builtin_ia32_divpd512_mask(A, B, C, D, E) __builtin_ia32_divpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_divps512_mask(A, B, C, D, E) __builtin_ia32_divps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_divsd_round(A, B, C) __builtin_ia32_divsd_round(A, B, 8)
+#define __builtin_ia32_divss_round(A, B, C) __builtin_ia32_divss_round(A, B, 8)
#define __builtin_ia32_extractf32x4_mask(A, E, C, D) __builtin_ia32_extractf32x4_mask(A, 1, C, D)
#define __builtin_ia32_extractf64x4_mask(A, E, C, D) __builtin_ia32_extractf64x4_mask(A, 1, C, D)
#define __builtin_ia32_extracti32x4_mask(A, E, C, D) __builtin_ia32_extracti32x4_mask(A, 1, C, D)
#define __builtin_ia32_extracti64x4_mask(A, E, C, D) __builtin_ia32_extracti64x4_mask(A, 1, C, D)
-#define __builtin_ia32_fixupimmpd512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmpd512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmps512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmps512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmsd_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmsd_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmss_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmss_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmss_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmss_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_gatherdiv8df(A, B, C, D, F) __builtin_ia32_gatherdiv8df(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv8di(A, B, C, D, F) __builtin_ia32_gatherdiv8di(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv16sf(A, B, C, D, F) __builtin_ia32_gatherdiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv16si(A, B, C, D, F) __builtin_ia32_gatherdiv16si(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv16sf(A, B, C, D, F) __builtin_ia32_gathersiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv16si(A, B, C, D, F) __builtin_ia32_gathersiv16si(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv8df(A, B, C, D, F) __builtin_ia32_gathersiv8df(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv8di(A, B, C, D, F) __builtin_ia32_gathersiv8di(A, B, C, D, 1)
-#define __builtin_ia32_getexppd512_mask(A, B, C, D) __builtin_ia32_getexppd512_mask(A, B, C, 5)
-#define __builtin_ia32_getexpps512_mask(A, B, C, D) __builtin_ia32_getexpps512_mask(A, B, C, 5)
+#define __builtin_ia32_fixupimmpd512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmpd512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmps512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmps512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmsd_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmsd_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmss_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmss_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmss_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmss_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_gatherdiv8df(A, B, C, D, F) __builtin_ia32_gatherdiv8df(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv8di(A, B, C, D, F) __builtin_ia32_gatherdiv8di(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv16sf(A, B, C, D, F) __builtin_ia32_gatherdiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv16si(A, B, C, D, F) __builtin_ia32_gatherdiv16si(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv16sf(A, B, C, D, F) __builtin_ia32_gathersiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv16si(A, B, C, D, F) __builtin_ia32_gathersiv16si(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv8df(A, B, C, D, F) __builtin_ia32_gathersiv8df(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv8di(A, B, C, D, F) __builtin_ia32_gathersiv8di(A, B, C, D, 8)
+#define __builtin_ia32_getexppd512_mask(A, B, C, D) __builtin_ia32_getexppd512_mask(A, B, C, 8)
+#define __builtin_ia32_getexpps512_mask(A, B, C, D) __builtin_ia32_getexpps512_mask(A, B, C, 8)
#define __builtin_ia32_getexpsd128_round(A, B, C) __builtin_ia32_getexpsd128_round(A, B, 4)
#define __builtin_ia32_getexpss128_round(A, B, C) __builtin_ia32_getexpss128_round(A, B, 4)
-#define __builtin_ia32_getmantpd512_mask(A, F, C, D, E) __builtin_ia32_getmantpd512_mask(A, 1, C, D, 5)
-#define __builtin_ia32_getmantps512_mask(A, F, C, D, E) __builtin_ia32_getmantps512_mask(A, 1, C, D, 5)
+#define __builtin_ia32_getmantpd512_mask(A, F, C, D, E) __builtin_ia32_getmantpd512_mask(A, 1, C, D, 8)
+#define __builtin_ia32_getmantps512_mask(A, F, C, D, E) __builtin_ia32_getmantps512_mask(A, 1, C, D, 8)
#define __builtin_ia32_getmantsd_round(A, B, C, D) __builtin_ia32_getmantsd_round(A, B, 1, 4)
#define __builtin_ia32_getmantss_round(A, B, C, D) __builtin_ia32_getmantss_round(A, B, 1, 4)
#define __builtin_ia32_insertf32x4_mask(A, B, F, D, E) __builtin_ia32_insertf32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_insertf64x4_mask(A, B, F, D, E) __builtin_ia32_insertf64x4_mask(A, B, 1, D, E)
#define __builtin_ia32_inserti32x4_mask(A, B, F, D, E) __builtin_ia32_inserti32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_inserti64x4_mask(A, B, F, D, E) __builtin_ia32_inserti64x4_mask(A, B, 1, D, E)
-#define __builtin_ia32_maxpd512_mask(A, B, C, D, E) __builtin_ia32_maxpd512_mask(A, B, C, D, 5)
-#define __builtin_ia32_maxps512_mask(A, B, C, D, E) __builtin_ia32_maxps512_mask(A, B, C, D, 5)
+#define __builtin_ia32_maxpd512_mask(A, B, C, D, E) __builtin_ia32_maxpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_maxps512_mask(A, B, C, D, E) __builtin_ia32_maxps512_mask(A, B, C, D, 8)
#define __builtin_ia32_maxsd_round(A, B, C) __builtin_ia32_maxsd_round(A, B, 4)
#define __builtin_ia32_maxss_round(A, B, C) __builtin_ia32_maxss_round(A, B, 4)
-#define __builtin_ia32_minpd512_mask(A, B, C, D, E) __builtin_ia32_minpd512_mask(A, B, C, D, 5)
-#define __builtin_ia32_minps512_mask(A, B, C, D, E) __builtin_ia32_minps512_mask(A, B, C, D, 5)
+#define __builtin_ia32_minpd512_mask(A, B, C, D, E) __builtin_ia32_minpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_minps512_mask(A, B, C, D, E) __builtin_ia32_minps512_mask(A, B, C, D, 8)
#define __builtin_ia32_minsd_round(A, B, C) __builtin_ia32_minsd_round(A, B, 4)
#define __builtin_ia32_minss_round(A, B, C) __builtin_ia32_minss_round(A, B, 4)
-#define __builtin_ia32_mulpd512_mask(A, B, C, D, E) __builtin_ia32_mulpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_mulps512_mask(A, B, C, D, E) __builtin_ia32_mulps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_mulsd_round(A, B, C) __builtin_ia32_mulsd_round(A, B, 1)
-#define __builtin_ia32_mulss_round(A, B, C) __builtin_ia32_mulss_round(A, B, 1)
+#define __builtin_ia32_mulpd512_mask(A, B, C, D, E) __builtin_ia32_mulpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulps512_mask(A, B, C, D, E) __builtin_ia32_mulps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulsd_round(A, B, C) __builtin_ia32_mulsd_round(A, B, 8)
+#define __builtin_ia32_mulss_round(A, B, C) __builtin_ia32_mulss_round(A, B, 8)
#define __builtin_ia32_permdf512_mask(A, E, C, D) __builtin_ia32_permdf512_mask(A, 1, C, D)
#define __builtin_ia32_permdi512_mask(A, E, C, D) __builtin_ia32_permdi512_mask(A, 1, C, D)
#define __builtin_ia32_prold512_mask(A, E, C, D) __builtin_ia32_prold512_mask(A, 1, C, D)
@@ -264,98 +264,98 @@
#define __builtin_ia32_pternlogd512_maskz(A, B, C, F, E) __builtin_ia32_pternlogd512_maskz(A, B, C, 1, E)
#define __builtin_ia32_pternlogq512_mask(A, B, C, F, E) __builtin_ia32_pternlogq512_mask(A, B, C, 1, E)
#define __builtin_ia32_pternlogq512_maskz(A, B, C, F, E) __builtin_ia32_pternlogq512_maskz(A, B, C, 1, E)
-#define __builtin_ia32_rndscalepd_mask(A, F, C, D, E) __builtin_ia32_rndscalepd_mask(A, 1, C, D, 5)
-#define __builtin_ia32_rndscaleps_mask(A, F, C, D, E) __builtin_ia32_rndscaleps_mask(A, 1, C, D, 5)
+#define __builtin_ia32_rndscalepd_mask(A, F, C, D, E) __builtin_ia32_rndscalepd_mask(A, 1, C, D, 8)
+#define __builtin_ia32_rndscaleps_mask(A, F, C, D, E) __builtin_ia32_rndscaleps_mask(A, 1, C, D, 8)
#define __builtin_ia32_rndscalesd_round(A, B, C, D) __builtin_ia32_rndscalesd_round(A, B, 1, 4)
#define __builtin_ia32_rndscaless_round(A, B, C, D) __builtin_ia32_rndscaless_round(A, B, 1, 4)
-#define __builtin_ia32_scalefpd512_mask(A, B, C, D, E) __builtin_ia32_scalefpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_scalefps512_mask(A, B, C, D, E) __builtin_ia32_scalefps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_scalefsd_round(A, B, C) __builtin_ia32_scalefsd_round(A, B, 1)
-#define __builtin_ia32_scalefss_round(A, B, C) __builtin_ia32_scalefss_round(A, B, 1)
-#define __builtin_ia32_scatterdiv8df(A, B, C, D, F) __builtin_ia32_scatterdiv8df(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv8di(A, B, C, D, F) __builtin_ia32_scatterdiv8di(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv16sf(A, B, C, D, F) __builtin_ia32_scatterdiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv16si(A, B, C, D, F) __builtin_ia32_scatterdiv16si(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv16sf(A, B, C, D, F) __builtin_ia32_scattersiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv16si(A, B, C, D, F) __builtin_ia32_scattersiv16si(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv8df(A, B, C, D, F) __builtin_ia32_scattersiv8df(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv8di(A, B, C, D, F) __builtin_ia32_scattersiv8di(A, B, C, D, 1)
+#define __builtin_ia32_scalefpd512_mask(A, B, C, D, E) __builtin_ia32_scalefpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_scalefps512_mask(A, B, C, D, E) __builtin_ia32_scalefps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_scalefsd_round(A, B, C) __builtin_ia32_scalefsd_round(A, B, 8)
+#define __builtin_ia32_scalefss_round(A, B, C) __builtin_ia32_scalefss_round(A, B, 8)
+#define __builtin_ia32_scatterdiv8df(A, B, C, D, F) __builtin_ia32_scatterdiv8df(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv8di(A, B, C, D, F) __builtin_ia32_scatterdiv8di(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv16sf(A, B, C, D, F) __builtin_ia32_scatterdiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv16si(A, B, C, D, F) __builtin_ia32_scatterdiv16si(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv16sf(A, B, C, D, F) __builtin_ia32_scattersiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv16si(A, B, C, D, F) __builtin_ia32_scattersiv16si(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv8df(A, B, C, D, F) __builtin_ia32_scattersiv8df(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv8di(A, B, C, D, F) __builtin_ia32_scattersiv8di(A, B, C, D, 8)
#define __builtin_ia32_shuf_f32x4_mask(A, B, F, D, E) __builtin_ia32_shuf_f32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_f64x2_mask(A, B, F, D, E) __builtin_ia32_shuf_f64x2_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_i32x4_mask(A, B, F, D, E) __builtin_ia32_shuf_i32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_i64x2_mask(A, B, F, D, E) __builtin_ia32_shuf_i64x2_mask(A, B, 1, D, E)
#define __builtin_ia32_shufpd512_mask(A, B, F, D, E) __builtin_ia32_shufpd512_mask(A, B, 1, D, E)
#define __builtin_ia32_shufps512_mask(A, B, F, D, E) __builtin_ia32_shufps512_mask(A, B, 1, D, E)
-#define __builtin_ia32_sqrtpd512_mask(A, B, C, D) __builtin_ia32_sqrtpd512_mask(A, B, C, 1)
-#define __builtin_ia32_sqrtps512_mask(A, B, C, D) __builtin_ia32_sqrtps512_mask(A, B, C, 1)
-#define __builtin_ia32_sqrtss_round(A, B, C) __builtin_ia32_sqrtss_round(A, B, 1)
-#define __builtin_ia32_sqrtsd_round(A, B, C) __builtin_ia32_sqrtsd_round(A, B, 1)
-#define __builtin_ia32_subpd512_mask(A, B, C, D, E) __builtin_ia32_subpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_subps512_mask(A, B, C, D, E) __builtin_ia32_subps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_subsd_round(A, B, C) __builtin_ia32_subsd_round(A, B, 1)
-#define __builtin_ia32_subss_round(A, B, C) __builtin_ia32_subss_round(A, B, 1)
+#define __builtin_ia32_sqrtpd512_mask(A, B, C, D) __builtin_ia32_sqrtpd512_mask(A, B, C, 8)
+#define __builtin_ia32_sqrtps512_mask(A, B, C, D) __builtin_ia32_sqrtps512_mask(A, B, C, 8)
+#define __builtin_ia32_sqrtss_round(A, B, C) __builtin_ia32_sqrtss_round(A, B, 8)
+#define __builtin_ia32_sqrtsd_round(A, B, C) __builtin_ia32_sqrtsd_round(A, B, 8)
+#define __builtin_ia32_subpd512_mask(A, B, C, D, E) __builtin_ia32_subpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_subps512_mask(A, B, C, D, E) __builtin_ia32_subps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_subsd_round(A, B, C) __builtin_ia32_subsd_round(A, B, 8)
+#define __builtin_ia32_subss_round(A, B, C) __builtin_ia32_subss_round(A, B, 8)
#define __builtin_ia32_ucmpd512_mask(A, B, E, D) __builtin_ia32_ucmpd512_mask(A, B, 1, D)
#define __builtin_ia32_ucmpq512_mask(A, B, E, D) __builtin_ia32_ucmpq512_mask(A, B, 1, D)
-#define __builtin_ia32_vcomisd(A, B, C, D) __builtin_ia32_vcomisd(A, B, 1, 5)
-#define __builtin_ia32_vcomiss(A, B, C, D) __builtin_ia32_vcomiss(A, B, 1, 5)
-#define __builtin_ia32_vcvtph2ps512_mask(A, B, C, D) __builtin_ia32_vcvtph2ps512_mask(A, B, C, 5)
+#define __builtin_ia32_vcomisd(A, B, C, D) __builtin_ia32_vcomisd(A, B, 1, 8)
+#define __builtin_ia32_vcomiss(A, B, C, D) __builtin_ia32_vcomiss(A, B, 1, 8)
+#define __builtin_ia32_vcvtph2ps512_mask(A, B, C, D) __builtin_ia32_vcvtph2ps512_mask(A, B, C, 8)
#define __builtin_ia32_vcvtps2ph512_mask(A, E, C, D) __builtin_ia32_vcvtps2ph512_mask(A, 1, C, D)
-#define __builtin_ia32_vcvtsd2si32(A, B) __builtin_ia32_vcvtsd2si32(A, 1)
-#define __builtin_ia32_vcvtsd2si64(A, B) __builtin_ia32_vcvtsd2si64(A, 1)
-#define __builtin_ia32_vcvtsd2usi32(A, B) __builtin_ia32_vcvtsd2usi32(A, 1)
-#define __builtin_ia32_vcvtsd2usi64(A, B) __builtin_ia32_vcvtsd2usi64(A, 1)
-#define __builtin_ia32_vcvtss2si32(A, B) __builtin_ia32_vcvtss2si32(A, 1)
-#define __builtin_ia32_vcvtss2si64(A, B) __builtin_ia32_vcvtss2si64(A, 1)
-#define __builtin_ia32_vcvtss2usi32(A, B) __builtin_ia32_vcvtss2usi32(A, 1)
-#define __builtin_ia32_vcvtss2usi64(A, B) __builtin_ia32_vcvtss2usi64(A, 1)
-#define __builtin_ia32_vcvttsd2si32(A, B) __builtin_ia32_vcvttsd2si32(A, 5)
-#define __builtin_ia32_vcvttsd2si64(A, B) __builtin_ia32_vcvttsd2si64(A, 5)
-#define __builtin_ia32_vcvttsd2usi32(A, B) __builtin_ia32_vcvttsd2usi32(A, 5)
-#define __builtin_ia32_vcvttsd2usi64(A, B) __builtin_ia32_vcvttsd2usi64(A, 5)
-#define __builtin_ia32_vcvttss2si32(A, B) __builtin_ia32_vcvttss2si32(A, 5)
-#define __builtin_ia32_vcvttss2si64(A, B) __builtin_ia32_vcvttss2si64(A, 5)
-#define __builtin_ia32_vcvttss2usi32(A, B) __builtin_ia32_vcvttss2usi32(A, 5)
-#define __builtin_ia32_vcvttss2usi64(A, B) __builtin_ia32_vcvttss2usi64(A, 5)
-#define __builtin_ia32_vfmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddps512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsd3_round(A, B, C, D) __builtin_ia32_vfmaddsd3_round(A, B, C, 1)
-#define __builtin_ia32_vfmaddss3_round(A, B, C, D) __builtin_ia32_vfmaddss3_round(A, B, C, 1)
-#define __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubss3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubss3_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubps512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, 1)
+#define __builtin_ia32_vcvtsd2si32(A, B) __builtin_ia32_vcvtsd2si32(A, 8)
+#define __builtin_ia32_vcvtsd2si64(A, B) __builtin_ia32_vcvtsd2si64(A, 8)
+#define __builtin_ia32_vcvtsd2usi32(A, B) __builtin_ia32_vcvtsd2usi32(A, 8)
+#define __builtin_ia32_vcvtsd2usi64(A, B) __builtin_ia32_vcvtsd2usi64(A, 8)
+#define __builtin_ia32_vcvtss2si32(A, B) __builtin_ia32_vcvtss2si32(A, 8)
+#define __builtin_ia32_vcvtss2si64(A, B) __builtin_ia32_vcvtss2si64(A, 8)
+#define __builtin_ia32_vcvtss2usi32(A, B) __builtin_ia32_vcvtss2usi32(A, 8)
+#define __builtin_ia32_vcvtss2usi64(A, B) __builtin_ia32_vcvtss2usi64(A, 8)
+#define __builtin_ia32_vcvttsd2si32(A, B) __builtin_ia32_vcvttsd2si32(A, 8)
+#define __builtin_ia32_vcvttsd2si64(A, B) __builtin_ia32_vcvttsd2si64(A, 8)
+#define __builtin_ia32_vcvttsd2usi32(A, B) __builtin_ia32_vcvttsd2usi32(A, 8)
+#define __builtin_ia32_vcvttsd2usi64(A, B) __builtin_ia32_vcvttsd2usi64(A, 8)
+#define __builtin_ia32_vcvttss2si32(A, B) __builtin_ia32_vcvttss2si32(A, 8)
+#define __builtin_ia32_vcvttss2si64(A, B) __builtin_ia32_vcvttss2si64(A, 8)
+#define __builtin_ia32_vcvttss2usi32(A, B) __builtin_ia32_vcvttss2usi32(A, 8)
+#define __builtin_ia32_vcvttss2usi64(A, B) __builtin_ia32_vcvttss2usi64(A, 8)
+#define __builtin_ia32_vfmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddps512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsd3_round(A, B, C, D) __builtin_ia32_vfmaddsd3_round(A, B, C, 8)
+#define __builtin_ia32_vfmaddss3_round(A, B, C, D) __builtin_ia32_vfmaddss3_round(A, B, C, 8)
+#define __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubss3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubss3_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubps512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, 8)
#define __builtin_ia32_vpermilpd512_mask(A, E, C, D) __builtin_ia32_vpermilpd512_mask(A, 1, C, D)
#define __builtin_ia32_vpermilps512_mask(A, E, C, D) __builtin_ia32_vpermilps512_mask(A, 1, C, D)
/* avx512erintrin.h */
-#define __builtin_ia32_exp2ps_mask(A, B, C, D) __builtin_ia32_exp2ps_mask(A, B, C, 5)
-#define __builtin_ia32_exp2pd_mask(A, B, C, D) __builtin_ia32_exp2pd_mask(A, B, C, 5)
-#define __builtin_ia32_rcp28ps_mask(A, B, C, D) __builtin_ia32_rcp28ps_mask(A, B, C, 5)
-#define __builtin_ia32_rcp28pd_mask(A, B, C, D) __builtin_ia32_rcp28pd_mask(A, B, C, 5)
-#define __builtin_ia32_rsqrt28ps_mask(A, B, C, D) __builtin_ia32_rsqrt28ps_mask(A, B, C, 5)
-#define __builtin_ia32_rsqrt28pd_mask(A, B, C, D) __builtin_ia32_rsqrt28pd_mask(A, B, C, 5)
-#define __builtin_ia32_rcp28ss_round(A, B, C) __builtin_ia32_rcp28ss_round(A, B, 5)
-#define __builtin_ia32_rcp28sd_round(A, B, C) __builtin_ia32_rcp28sd_round(A, B, 5)
-#define __builtin_ia32_rsqrt28ss_round(A, B, C) __builtin_ia32_rsqrt28ss_round(A, B, 5)
-#define __builtin_ia32_rsqrt28sd_round(A, B, C) __builtin_ia32_rsqrt28sd_round(A, B, 5)
+#define __builtin_ia32_exp2ps_mask(A, B, C, D) __builtin_ia32_exp2ps_mask(A, B, C, 8)
+#define __builtin_ia32_exp2pd_mask(A, B, C, D) __builtin_ia32_exp2pd_mask(A, B, C, 8)
+#define __builtin_ia32_rcp28ps_mask(A, B, C, D) __builtin_ia32_rcp28ps_mask(A, B, C, 8)
+#define __builtin_ia32_rcp28pd_mask(A, B, C, D) __builtin_ia32_rcp28pd_mask(A, B, C, 8)
+#define __builtin_ia32_rsqrt28ps_mask(A, B, C, D) __builtin_ia32_rsqrt28ps_mask(A, B, C, 8)
+#define __builtin_ia32_rsqrt28pd_mask(A, B, C, D) __builtin_ia32_rsqrt28pd_mask(A, B, C, 8)
+#define __builtin_ia32_rcp28ss_round(A, B, C) __builtin_ia32_rcp28ss_round(A, B, 8)
+#define __builtin_ia32_rcp28sd_round(A, B, C) __builtin_ia32_rcp28sd_round(A, B, 8)
+#define __builtin_ia32_rsqrt28ss_round(A, B, C) __builtin_ia32_rsqrt28ss_round(A, B, 8)
+#define __builtin_ia32_rsqrt28sd_round(A, B, C) __builtin_ia32_rsqrt28sd_round(A, B, 8)
/* avx512pfintrin.h */
#define __builtin_ia32_gatherpfdps(A, B, C, D, E) __builtin_ia32_gatherpfdps(A, B, C, 1, 1)
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_add_pd (x, x);
x = _mm512_mask_add_pd (x, m, x, x);
x = _mm512_maskz_add_pd (m, x, x);
- x = _mm512_add_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_add_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF);
- x = _mm512_maskz_add_round_pd (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_add_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_add_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_add_round_pd (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_add_ps (x, x);
x = _mm512_mask_add_ps (x, m, x, x);
x = _mm512_maskz_add_ps (m, x, x);
- x = _mm512_add_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_add_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF);
- x = _mm512_maskz_add_round_ps (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_add_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_add_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_add_round_ps (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_add_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_add_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_add_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_add_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -19,7 +19,7 @@ avx512f_test (void)
res = _mm512_cvtepi32_ps (s);
res = _mm512_mask_cvtepi32_ps (res, m, s);
res = _mm512_maskz_cvtepi32_ps (m, s);
- res = _mm512_cvt_roundepi32_ps (s, _MM_FROUND_TO_NEAREST_INT);
- res = _mm512_mask_cvt_roundepi32_ps (res, m, s, _MM_FROUND_TO_POS_INF);
- res = _mm512_maskz_cvt_roundepi32_ps (m, s, _MM_FROUND_TO_ZERO);
+ res = _mm512_cvt_roundepi32_ps (s, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ res = _mm512_mask_cvt_roundepi32_ps (res, m, s, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ res = _mm512_maskz_cvt_roundepi32_ps (m, s, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -19,7 +19,7 @@ avx512f_test (void)
res = _mm512_cvtpd_epi32 (s);
res = _mm512_mask_cvtpd_epi32 (res, m, s);
res = _mm512_maskz_cvtpd_epi32 (m, s);
- res = _mm512_cvt_roundpd_epi32 (s, _MM_FROUND_TO_NEAREST_INT);
- res = _mm512_mask_cvt_roundpd_epi32 (res, m, s, _MM_FROUND_TO_POS_INF);
- res = _mm512_maskz_cvt_roundpd_epi32 (m, s, _MM_FROUND_TO_ZERO);
+ res = _mm512_cvt_roundpd_epi32 (s, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ res = _mm512_mask_cvt_roundpd_epi32 (res, m, s, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ res = _mm512_maskz_cvt_roundpd_epi32 (m, s, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
y = _mm512_cvtpd_ps (x);
y = _mm512_mask_cvtpd_ps (y, 4, x);
y = _mm512_maskz_cvtpd_ps (6, x);
- y = _mm512_cvt_roundpd_ps (x, _MM_FROUND_TO_NEAREST_INT);
- y = _mm512_mask_cvt_roundpd_ps (y, 4, x, _MM_FROUND_TO_NEG_INF);
- y = _mm512_maskz_cvt_roundpd_ps (6, x, _MM_FROUND_TO_ZERO);
+ y = _mm512_cvt_roundpd_ps (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ y = _mm512_mask_cvt_roundpd_ps (y, 4, x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ y = _mm512_maskz_cvt_roundpd_ps (6, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -19,7 +19,7 @@ avx512f_test (void)
res = _mm512_cvtpd_epu32 (s);
res = _mm512_mask_cvtpd_epu32 (res, m, s);
res = _mm512_maskz_cvtpd_epu32 (m, s);
- res = _mm512_cvt_roundpd_epu32 (s, _MM_FROUND_TO_NEAREST_INT);
- res = _mm512_mask_cvt_roundpd_epu32 (res, m, s, _MM_FROUND_TO_POS_INF);
- res = _mm512_maskz_cvt_roundpd_epu32 (m, s, _MM_FROUND_TO_ZERO);
+ res = _mm512_cvt_roundpd_epu32 (s, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ res = _mm512_mask_cvt_roundpd_epu32 (res, m, s, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ res = _mm512_maskz_cvt_roundpd_epu32 (m, s, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -19,7 +19,7 @@ avx512f_test (void)
res = _mm512_cvtps_epi32 (s);
res = _mm512_mask_cvtps_epi32 (res, m, s);
res = _mm512_maskz_cvtps_epi32 (m, s);
- res = _mm512_cvt_roundps_epi32 (s, _MM_FROUND_TO_NEAREST_INT);
- res = _mm512_mask_cvt_roundps_epi32 (res, m, s, _MM_FROUND_TO_POS_INF);
- res = _mm512_maskz_cvt_roundps_epi32 (m, s, _MM_FROUND_TO_ZERO);
+ res = _mm512_cvt_roundps_epi32 (s, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ res = _mm512_mask_cvt_roundps_epi32 (res, m, s, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ res = _mm512_maskz_cvt_roundps_epi32 (m, s, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -19,7 +19,7 @@ avx512f_test (void)
res = _mm512_cvtps_epu32 (s);
res = _mm512_mask_cvtps_epu32 (res, m, s);
res = _mm512_maskz_cvtps_epu32 (m, s);
- res = _mm512_cvt_roundps_epu32 (s, _MM_FROUND_TO_NEAREST_INT);
- res = _mm512_mask_cvt_roundps_epu32 (res, m, s, _MM_FROUND_TO_NEG_INF);
- res = _mm512_maskz_cvt_roundps_epu32 (m, s, _MM_FROUND_TO_ZERO);
+ res = _mm512_cvt_roundps_epu32 (s, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ res = _mm512_mask_cvt_roundps_epu32 (res, m, s, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ res = _mm512_maskz_cvt_roundps_epu32 (m, s, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile unsigned y;
void extern
avx512f_test (void)
{
- y = _mm_cvt_roundsd_i32 (x, _MM_FROUND_TO_NEAREST_INT);
+ y = _mm_cvt_roundsd_i32 (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -10,5 +10,5 @@ volatile unsigned long long y;
void extern
avx512f_test (void)
{
- y = _mm_cvt_roundsd_i64 (x, _MM_FROUND_TO_ZERO);
+ y = _mm_cvt_roundsd_i64 (x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -10,5 +10,5 @@ volatile __m128d s2;
void extern
avx512f_test (void)
{
- r = _mm_cvt_roundsd_ss (s1, s2, _MM_FROUND_TO_NEAREST_INT);
+ r = _mm_cvt_roundsd_ss (s1, s2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -11,5 +11,5 @@ void extern
avx512f_test (void)
{
y = _mm_cvtsd_u32 (x);
- y = _mm_cvt_roundsd_u32 (x, _MM_FROUND_TO_NEG_INF);
+ y = _mm_cvt_roundsd_u32 (x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
}
@@ -12,5 +12,5 @@ void extern
avx512f_test (void)
{
y = _mm_cvtsd_u64 (x);
- y = _mm_cvt_roundsd_u64 (x, _MM_FROUND_TO_POS_INF);
+ y = _mm_cvt_roundsd_u64 (x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
}
@@ -10,5 +10,5 @@ volatile long long n;
void extern
avx512f_test (void)
{
- x = _mm_cvt_roundi64_sd (x, n, _MM_FROUND_TO_POS_INF);
+ x = _mm_cvt_roundi64_sd (x, n, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
}
@@ -10,5 +10,5 @@ volatile int n;
void extern
avx512f_test (void)
{
- x = _mm_cvt_roundi32_ss (x, n, _MM_FROUND_TO_NEAREST_INT);
+ x = _mm_cvt_roundi32_ss (x, n, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -10,5 +10,5 @@ volatile long long n;
void extern
avx512f_test (void)
{
- x = _mm_cvt_roundi64_ss (x, n, _MM_FROUND_TO_ZERO);
+ x = _mm_cvt_roundi64_ss (x, n, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile unsigned y;
void extern
avx512f_test (void)
{
- y = _mm_cvt_roundss_i32 (x, _MM_FROUND_TO_NEAREST_INT);
+ y = _mm_cvt_roundss_i32 (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -10,5 +10,5 @@ volatile unsigned long long y;
void extern
avx512f_test (void)
{
- y = _mm_cvt_roundss_i64 (x, _MM_FROUND_TO_ZERO);
+ y = _mm_cvt_roundss_i64 (x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -11,5 +11,5 @@ void extern
avx512f_test (void)
{
y = _mm_cvtss_u32 (x);
- y = _mm_cvt_roundss_u32 (x, _MM_FROUND_TO_NEG_INF);
+ y = _mm_cvt_roundss_u32 (x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
}
@@ -12,5 +12,5 @@ void extern
avx512f_test (void)
{
y = _mm_cvtss_u64 (x);
- y = _mm_cvt_roundss_u64 (x, _MM_FROUND_TO_POS_INF);
+ y = _mm_cvt_roundss_u64 (x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
}
@@ -19,7 +19,7 @@ avx512f_test (void)
res = _mm512_cvtepu32_ps (s);
res = _mm512_mask_cvtepu32_ps (res, m, s);
res = _mm512_maskz_cvtepu32_ps (m, s);
- res = _mm512_cvt_roundepu32_ps (s, _MM_FROUND_TO_NEAREST_INT);
- res = _mm512_mask_cvt_roundepu32_ps (res, m, s, _MM_FROUND_TO_NEG_INF);
- res = _mm512_maskz_cvt_roundepu32_ps (m, s, _MM_FROUND_TO_ZERO);
+ res = _mm512_cvt_roundepu32_ps (s, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ res = _mm512_mask_cvt_roundepu32_ps (res, m, s, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ res = _mm512_maskz_cvt_roundepu32_ps (m, s, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -12,5 +12,5 @@ void extern
avx512f_test (void)
{
x = _mm_cvtu64_sd (x, n);
- x = _mm_cvt_roundu64_sd (x, n, _MM_FROUND_TO_POS_INF);
+ x = _mm_cvt_roundu64_sd (x, n, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
}
@@ -12,5 +12,5 @@ void extern
avx512f_test (void)
{
x = _mm_cvtu32_ss (x, n);
- x = _mm_cvt_roundu32_ss (x, n, _MM_FROUND_TO_NEAREST_INT);
+ x = _mm_cvt_roundu32_ss (x, n, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -12,5 +12,5 @@ void extern
avx512f_test (void)
{
x = _mm_cvtu64_ss (x, n);
- x = _mm_cvt_roundu64_ss (x, n, _MM_FROUND_TO_ZERO);
+ x = _mm_cvt_roundu64_ss (x, n, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_div_pd (x, x);
x = _mm512_mask_div_pd (x, m, x, x);
x = _mm512_maskz_div_pd (m, x, x);
- x = _mm512_div_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_div_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF);
- x = _mm512_maskz_div_round_pd (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_div_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_div_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_div_round_pd (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_div_ps (x, x);
x = _mm512_mask_div_ps (x, m, x, x);
x = _mm512_maskz_div_ps (m, x, x);
- x = _mm512_div_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_div_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF);
- x = _mm512_maskz_div_round_ps (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_div_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_div_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_div_round_ps (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -10,5 +10,5 @@ volatile __mmask8 m;
void extern
avx512f_test (void)
{
- x1 = _mm_div_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_div_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_div_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_div_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmadd_pd (x1, m, x2, x3);
x3 = _mm512_mask3_fmadd_pd (x1, x2, x3, m);
x1 = _mm512_maskz_fmadd_pd (m, x1, x2, x3);
- x1 = _mm512_fmadd_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmadd_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmadd_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmadd_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmadd_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmadd_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmadd_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmadd_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmadd_ps (x1, m, x2, x3);
x3 = _mm512_mask3_fmadd_ps (x1, x2, x3, m);
x1 = _mm512_maskz_fmadd_ps (m, x1, x2, x3);
- x1 = _mm512_fmadd_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmadd_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmadd_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmadd_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmadd_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmadd_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmadd_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmadd_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fmadd_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fmadd_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fmadd_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fmadd_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmaddsub_pd (x1, m, x2, x3);
x3 = _mm512_mask3_fmaddsub_pd (x1, x2, x3, m);
x1 = _mm512_maskz_fmaddsub_pd (m, x1, x2, x3);
- x1 = _mm512_fmaddsub_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmaddsub_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmaddsub_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmaddsub_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmaddsub_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmaddsub_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmaddsub_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmaddsub_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmaddsub_ps (x1, m, x2, x3);
x3 = _mm512_mask3_fmaddsub_ps (x1, x2, x3, m);
x1 = _mm512_maskz_fmaddsub_ps (m, x1, x2, x3);
- x1 = _mm512_fmaddsub_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmaddsub_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmaddsub_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmaddsub_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmaddsub_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmaddsub_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmaddsub_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmaddsub_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmsub_pd (x1, m, x2, x3);
x3 = _mm512_mask3_fmsub_pd (x1, x2, x3, m);
x1 = _mm512_maskz_fmsub_pd (m, x1, x2, x3);
- x1 = _mm512_fmsub_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmsub_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmsub_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmsub_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmsub_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmsub_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmsub_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmsub_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmsub_ps (x1, m, x2, x3);
x3 = _mm512_mask3_fmsub_ps (x1, x2, x3, m);
x1 = _mm512_maskz_fmsub_ps (m, x1, x2, x3);
- x1 = _mm512_fmsub_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmsub_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmsub_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmsub_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmsub_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmsub_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmsub_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmsub_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fmsub_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fmsub_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fmsub_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fmsub_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmsubadd_pd (x1, m, x2, x3);
x3 = _mm512_mask3_fmsubadd_pd (x1, x2, x3, m);
x1 = _mm512_maskz_fmsubadd_pd (m, x1, x2, x3);
- x1 = _mm512_fmsubadd_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmsubadd_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmsubadd_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmsubadd_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmsubadd_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmsubadd_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmsubadd_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmsubadd_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fmsubadd_ps (x1, m, x2, x3);
x3 = _mm512_mask3_fmsubadd_ps (x1, x2, x3, m);
x1 = _mm512_maskz_fmsubadd_ps (m, x1, x2, x3);
- x1 = _mm512_fmsubadd_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fmsubadd_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fmsubadd_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fmsubadd_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fmsubadd_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fmsubadd_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fmsubadd_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fmsubadd_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fnmadd_pd (x1, m, x2, x3);
x3 = _mm512_mask3_fnmadd_pd (x1, x2, x3, m);
x1 = _mm512_maskz_fnmadd_pd (m, x1, x2, x3);
- x1 = _mm512_fnmadd_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fnmadd_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fnmadd_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fnmadd_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fnmadd_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fnmadd_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fnmadd_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fnmadd_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fnmadd_ps (x1, m, x2, x3);
x3 = _mm512_mask3_fnmadd_ps (x1, x2, x3, m);
x1 = _mm512_maskz_fnmadd_ps (m, x1, x2, x3);
- x1 = _mm512_fnmadd_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fnmadd_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fnmadd_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fnmadd_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fnmadd_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fnmadd_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fnmadd_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fnmadd_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fnmadd_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fnmadd_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fnmadd_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fnmadd_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fnmsub_pd (x1, m, x2, x3);
x3 = _mm512_mask3_fnmsub_pd (x1, x2, x3, m);
x1 = _mm512_maskz_fnmsub_pd (m, x1, x2, x3);
- x1 = _mm512_fnmsub_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fnmsub_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fnmsub_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fnmsub_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fnmsub_round_pd (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fnmsub_round_pd (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fnmsub_round_pd (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fnmsub_round_pd (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -21,8 +21,8 @@ avx512f_test (void)
x1 = _mm512_mask_fnmsub_ps (x1, m, x2, x3);
x3 = _mm512_mask3_fnmsub_ps (x1, x2, x3, m);
x1 = _mm512_maskz_fnmsub_ps (m, x1, x2, x3);
- x1 = _mm512_fnmsub_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT);
- x1 = _mm512_mask_fnmsub_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF);
- x3 = _mm512_mask3_fnmsub_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF);
- x1 = _mm512_maskz_fnmsub_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO);
+ x1 = _mm512_fnmsub_round_ps (x1, x2, x3, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x1 = _mm512_mask_fnmsub_round_ps (x1, m, x2, x3, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x3 = _mm512_mask3_fnmsub_round_ps (x1, x2, x3, m, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x1 = _mm512_maskz_fnmsub_round_ps (m, x1, x2, x3, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fnmsub_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fnmsub_round_sd (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 a, b, c;
void extern
avx512f_test (void)
{
- a = _mm_fnmsub_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT);
+ a = _mm_fnmsub_round_ss (a, b, c, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_mul_pd (x, x);
x = _mm512_mask_mul_pd (x, m, x, x);
x = _mm512_maskz_mul_pd (m, x, x);
- x = _mm512_mul_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_mul_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF);
- x = _mm512_maskz_mul_round_pd (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_mul_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_mul_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_mul_round_pd (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_mul_ps (x, x);
x = _mm512_mask_mul_ps (x, m, x, x);
x = _mm512_maskz_mul_ps (m, x, x);
- x = _mm512_mul_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_mul_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF);
- x = _mm512_maskz_mul_round_ps (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_mul_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_mul_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_mul_round_ps (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_mul_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_mul_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_mul_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_mul_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_scalef_pd (x, x);
x = _mm512_mask_scalef_pd (x, m, x, x);
x = _mm512_maskz_scalef_pd (m, x, x);
- x = _mm512_scalef_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_scalef_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF);
- x = _mm512_maskz_scalef_round_pd (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_scalef_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_scalef_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_scalef_round_pd (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_scalef_ps (x, x);
x = _mm512_mask_scalef_ps (x, m, x, x);
x = _mm512_maskz_scalef_ps (m, x, x);
- x = _mm512_scalef_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_scalef_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF);
- x = _mm512_maskz_scalef_round_ps (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_scalef_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_scalef_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_scalef_round_ps (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -11,5 +11,5 @@ void extern
avx512f_test (void)
{
x = _mm_scalef_sd (x, x);
- x = _mm_scalef_round_sd (x, x, _MM_FROUND_TO_NEAREST_INT);
+ x = _mm_scalef_round_sd (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -11,5 +11,5 @@ void extern
avx512f_test (void)
{
x = _mm_scalef_ss (x, x);
- x = _mm_scalef_round_ss (x, x, _MM_FROUND_TO_NEAREST_INT);
+ x = _mm_scalef_round_ss (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_sqrt_pd (x);
x = _mm512_mask_sqrt_pd (x, m, x);
x = _mm512_maskz_sqrt_pd (m, x);
- x = _mm512_sqrt_round_pd (x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_sqrt_round_pd (x, m, x, _MM_FROUND_TO_NEG_INF);
- x = _mm512_maskz_sqrt_round_pd (m, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_sqrt_round_pd (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_sqrt_round_pd (x, m, x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_sqrt_round_pd (m, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_sqrt_ps (x);
x = _mm512_mask_sqrt_ps (x, m, x);
x = _mm512_maskz_sqrt_ps (m, x);
- x = _mm512_sqrt_round_ps (x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_sqrt_round_ps (x, m, x, _MM_FROUND_TO_POS_INF);
- x = _mm512_maskz_sqrt_round_ps (m, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_sqrt_round_ps (x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_sqrt_round_ps (x, m, x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_sqrt_round_ps (m, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_sqrt_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_sqrt_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_sqrt_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_sqrt_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_sub_pd (x, x);
x = _mm512_mask_sub_pd (x, m, x, x);
x = _mm512_maskz_sub_pd (m, x, x);
- x = _mm512_sub_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_sub_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF);
- x = _mm512_maskz_sub_round_pd (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_sub_round_pd (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_sub_round_pd (x, m, x, x, _MM_FROUND_TO_NEG_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_sub_round_pd (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -18,7 +18,7 @@ avx512f_test (void)
x = _mm512_sub_ps (x, x);
x = _mm512_mask_sub_ps (x, m, x, x);
x = _mm512_maskz_sub_ps (m, x, x);
- x = _mm512_sub_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT);
- x = _mm512_mask_sub_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF);
- x = _mm512_maskz_sub_round_ps (m, x, x, _MM_FROUND_TO_ZERO);
+ x = _mm512_sub_round_ps (x, x, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
+ x = _mm512_mask_sub_round_ps (x, m, x, x, _MM_FROUND_TO_POS_INF | _MM_FROUND_NO_EXC);
+ x = _mm512_maskz_sub_round_ps (m, x, x, _MM_FROUND_TO_ZERO | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128d x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_sub_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_sub_round_sd (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -9,5 +9,5 @@ volatile __m128 x1, x2;
void extern
avx512f_test (void)
{
- x1 = _mm_sub_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT);
+ x1 = _mm_sub_round_ss (x1, x2, _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC);
}
@@ -198,86 +198,86 @@
#define __builtin_ia32_xabort (N) __builtin_ia32_xabort (1)
/* avx512fintrin.h */
-#define __builtin_ia32_addpd512_mask(A, B, C, D, E) __builtin_ia32_addpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_addps512_mask(A, B, C, D, E) __builtin_ia32_addps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_addsd_mask(A, B, C, D, E) __builtin_ia32_addsd_mask(A, B, C, D, 1)
-#define __builtin_ia32_addss_mask(A, B, C, D, E) __builtin_ia32_addss_mask(A, B, C, D, 1)
+#define __builtin_ia32_addpd512_mask(A, B, C, D, E) __builtin_ia32_addpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_addps512_mask(A, B, C, D, E) __builtin_ia32_addps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_addsd_mask(A, B, C, D, E) __builtin_ia32_addsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_addss_mask(A, B, C, D, E) __builtin_ia32_addss_mask(A, B, C, D, 8)
#define __builtin_ia32_alignd512_mask(A, B, F, D, E) __builtin_ia32_alignd512_mask(A, B, 1, D, E)
#define __builtin_ia32_alignq512_mask(A, B, F, D, E) __builtin_ia32_alignq512_mask(A, B, 1, D, E)
#define __builtin_ia32_cmpd512_mask(A, B, E, D) __builtin_ia32_cmpd512_mask(A, B, 1, D)
-#define __builtin_ia32_cmppd512_mask(A, B, F, D, E) __builtin_ia32_cmppd512_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cmpps512_mask(A, B, F, D, E) __builtin_ia32_cmpps512_mask(A, B, 1, D, 5)
+#define __builtin_ia32_cmppd512_mask(A, B, F, D, E) __builtin_ia32_cmppd512_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cmpps512_mask(A, B, F, D, E) __builtin_ia32_cmpps512_mask(A, B, 1, D, 8)
#define __builtin_ia32_cmpq512_mask(A, B, E, D) __builtin_ia32_cmpq512_mask(A, B, 1, D)
-#define __builtin_ia32_cmpsd_mask(A, B, F, D, E) __builtin_ia32_cmpsd_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cmpss_mask(A, B, F, D, E) __builtin_ia32_cmpss_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cvtdq2ps512_mask(A, B, C, D) __builtin_ia32_cvtdq2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2dq512_mask(A, B, C, D) __builtin_ia32_cvtpd2dq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2ps512_mask(A, B, C, D) __builtin_ia32_cvtpd2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2udq512_mask(A, B, C, D) __builtin_ia32_cvtpd2udq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtps2dq512_mask(A, B, C, D) __builtin_ia32_cvtps2dq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtps2pd512_mask(A, B, C, D) __builtin_ia32_cvtps2pd512_mask(A, B, C, 5)
-#define __builtin_ia32_cvtps2udq512_mask(A, B, C, D) __builtin_ia32_cvtps2udq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtsd2ss_mask(A, B, C, D, E) __builtin_ia32_cvtsd2ss_mask(A, B, C, D, 1)
-#define __builtin_ia32_cvtsi2sd64(A, B, C) __builtin_ia32_cvtsi2sd64(A, B, 1)
-#define __builtin_ia32_cvtsi2ss32(A, B, C) __builtin_ia32_cvtsi2ss32(A, B, 1)
-#define __builtin_ia32_cvtsi2ss64(A, B, C) __builtin_ia32_cvtsi2ss64(A, B, 1)
-#define __builtin_ia32_cvtss2sd_mask(A, B, C, D, E) __builtin_ia32_cvtss2sd_mask(A, B, C, D, 5)
-#define __builtin_ia32_cvttpd2dq512_mask(A, B, C, D) __builtin_ia32_cvttpd2dq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttpd2udq512_mask(A, B, C, D) __builtin_ia32_cvttpd2udq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttps2dq512_mask(A, B, C, D) __builtin_ia32_cvttps2dq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttps2udq512_mask(A, B, C, D) __builtin_ia32_cvttps2udq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvtudq2ps512_mask(A, B, C, D) __builtin_ia32_cvtudq2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtusi2sd64(A, B, C) __builtin_ia32_cvtusi2sd64(A, B, 1)
-#define __builtin_ia32_cvtusi2ss32(A, B, C) __builtin_ia32_cvtusi2ss32(A, B, 1)
-#define __builtin_ia32_cvtusi2ss64(A, B, C) __builtin_ia32_cvtusi2ss64(A, B, 1)
-#define __builtin_ia32_divpd512_mask(A, B, C, D, E) __builtin_ia32_divpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_divps512_mask(A, B, C, D, E) __builtin_ia32_divps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_divsd_mask(A, B, C, D, E) __builtin_ia32_divsd_mask(A, B, C, D, 1)
-#define __builtin_ia32_divss_mask(A, B, C, D, E) __builtin_ia32_divss_mask(A, B, C, D, 1)
+#define __builtin_ia32_cmpsd_mask(A, B, F, D, E) __builtin_ia32_cmpsd_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cmpss_mask(A, B, F, D, E) __builtin_ia32_cmpss_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cvtdq2ps512_mask(A, B, C, D) __builtin_ia32_cvtdq2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2dq512_mask(A, B, C, D) __builtin_ia32_cvtpd2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2ps512_mask(A, B, C, D) __builtin_ia32_cvtpd2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2udq512_mask(A, B, C, D) __builtin_ia32_cvtpd2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2dq512_mask(A, B, C, D) __builtin_ia32_cvtps2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2pd512_mask(A, B, C, D) __builtin_ia32_cvtps2pd512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2udq512_mask(A, B, C, D) __builtin_ia32_cvtps2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtsd2ss_mask(A, B, C, D, E) __builtin_ia32_cvtsd2ss_mask(A, B, C, D, 8)
+#define __builtin_ia32_cvtsi2sd64(A, B, C) __builtin_ia32_cvtsi2sd64(A, B, 8)
+#define __builtin_ia32_cvtsi2ss32(A, B, C) __builtin_ia32_cvtsi2ss32(A, B, 8)
+#define __builtin_ia32_cvtsi2ss64(A, B, C) __builtin_ia32_cvtsi2ss64(A, B, 8)
+#define __builtin_ia32_cvtss2sd_mask(A, B, C, D, E) __builtin_ia32_cvtss2sd_mask(A, B, C, D, 8)
+#define __builtin_ia32_cvttpd2dq512_mask(A, B, C, D) __builtin_ia32_cvttpd2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttpd2udq512_mask(A, B, C, D) __builtin_ia32_cvttpd2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttps2dq512_mask(A, B, C, D) __builtin_ia32_cvttps2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttps2udq512_mask(A, B, C, D) __builtin_ia32_cvttps2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtudq2ps512_mask(A, B, C, D) __builtin_ia32_cvtudq2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtusi2sd64(A, B, C) __builtin_ia32_cvtusi2sd64(A, B, 8)
+#define __builtin_ia32_cvtusi2ss32(A, B, C) __builtin_ia32_cvtusi2ss32(A, B, 8)
+#define __builtin_ia32_cvtusi2ss64(A, B, C) __builtin_ia32_cvtusi2ss64(A, B, 8)
+#define __builtin_ia32_divpd512_mask(A, B, C, D, E) __builtin_ia32_divpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_divps512_mask(A, B, C, D, E) __builtin_ia32_divps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_divsd_mask(A, B, C, D, E) __builtin_ia32_divsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_divss_mask(A, B, C, D, E) __builtin_ia32_divss_mask(A, B, C, D, 8)
#define __builtin_ia32_extractf32x4_mask(A, E, C, D) __builtin_ia32_extractf32x4_mask(A, 1, C, D)
#define __builtin_ia32_extractf64x4_mask(A, E, C, D) __builtin_ia32_extractf64x4_mask(A, 1, C, D)
#define __builtin_ia32_extracti32x4_mask(A, E, C, D) __builtin_ia32_extracti32x4_mask(A, 1, C, D)
#define __builtin_ia32_extracti64x4_mask(A, E, C, D) __builtin_ia32_extracti64x4_mask(A, 1, C, D)
-#define __builtin_ia32_fixupimmpd512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmpd512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmps512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmps512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmsd_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmsd_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmss_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmss_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmss_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmss_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_gatherdiv8df(A, B, C, D, F) __builtin_ia32_gatherdiv8df(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv8di(A, B, C, D, F) __builtin_ia32_gatherdiv8di(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv16sf(A, B, C, D, F) __builtin_ia32_gatherdiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv16si(A, B, C, D, F) __builtin_ia32_gatherdiv16si(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv16sf(A, B, C, D, F) __builtin_ia32_gathersiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv16si(A, B, C, D, F) __builtin_ia32_gathersiv16si(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv8df(A, B, C, D, F) __builtin_ia32_gathersiv8df(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv8di(A, B, C, D, F) __builtin_ia32_gathersiv8di(A, B, C, D, 1)
-#define __builtin_ia32_getexppd512_mask(A, B, C, D) __builtin_ia32_getexppd512_mask(A, B, C, 5)
-#define __builtin_ia32_getexpps512_mask(A, B, C, D) __builtin_ia32_getexpps512_mask(A, B, C, 5)
-#define __builtin_ia32_getexpsd128_mask(A, B, C, D, E) __builtin_ia32_getexpsd128_mask(A, B, C, D, 5)
-#define __builtin_ia32_getexpss128_mask(A, B, C, D, E) __builtin_ia32_getexpss128_mask(A, B, C, D, 5)
-#define __builtin_ia32_getmantpd512_mask(A, F, C, D, E) __builtin_ia32_getmantpd512_mask(A, 1, C, D, 5)
-#define __builtin_ia32_getmantps512_mask(A, F, C, D, E) __builtin_ia32_getmantps512_mask(A, 1, C, D, 5)
-#define __builtin_ia32_getmantsd_mask(A, B, I, D, E, F) __builtin_ia32_getmantsd_mask(A, B, 1, D, E, 5)
-#define __builtin_ia32_getmantss_mask(A, B, I, D, E, F) __builtin_ia32_getmantss_mask(A, B, 1, D, E, 5)
+#define __builtin_ia32_fixupimmpd512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmpd512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmps512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmps512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmsd_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmsd_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmss_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmss_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmss_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmss_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_gatherdiv8df(A, B, C, D, F) __builtin_ia32_gatherdiv8df(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv8di(A, B, C, D, F) __builtin_ia32_gatherdiv8di(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv16sf(A, B, C, D, F) __builtin_ia32_gatherdiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv16si(A, B, C, D, F) __builtin_ia32_gatherdiv16si(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv16sf(A, B, C, D, F) __builtin_ia32_gathersiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv16si(A, B, C, D, F) __builtin_ia32_gathersiv16si(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv8df(A, B, C, D, F) __builtin_ia32_gathersiv8df(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv8di(A, B, C, D, F) __builtin_ia32_gathersiv8di(A, B, C, D, 8)
+#define __builtin_ia32_getexppd512_mask(A, B, C, D) __builtin_ia32_getexppd512_mask(A, B, C, 8)
+#define __builtin_ia32_getexpps512_mask(A, B, C, D) __builtin_ia32_getexpps512_mask(A, B, C, 8)
+#define __builtin_ia32_getexpsd128_mask(A, B, C, D, E) __builtin_ia32_getexpsd128_mask(A, B, C, D, 8)
+#define __builtin_ia32_getexpss128_mask(A, B, C, D, E) __builtin_ia32_getexpss128_mask(A, B, C, D, 8)
+#define __builtin_ia32_getmantpd512_mask(A, F, C, D, E) __builtin_ia32_getmantpd512_mask(A, 1, C, D, 8)
+#define __builtin_ia32_getmantps512_mask(A, F, C, D, E) __builtin_ia32_getmantps512_mask(A, 1, C, D, 8)
+#define __builtin_ia32_getmantsd_mask(A, B, I, D, E, F) __builtin_ia32_getmantsd_mask(A, B, 1, D, E, 8)
+#define __builtin_ia32_getmantss_mask(A, B, I, D, E, F) __builtin_ia32_getmantss_mask(A, B, 1, D, E, 8)
#define __builtin_ia32_insertf32x4_mask(A, B, F, D, E) __builtin_ia32_insertf32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_insertf64x4_mask(A, B, F, D, E) __builtin_ia32_insertf64x4_mask(A, B, 1, D, E)
#define __builtin_ia32_inserti32x4_mask(A, B, F, D, E) __builtin_ia32_inserti32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_inserti64x4_mask(A, B, F, D, E) __builtin_ia32_inserti64x4_mask(A, B, 1, D, E)
-#define __builtin_ia32_maxpd512_mask(A, B, C, D, E) __builtin_ia32_maxpd512_mask(A, B, C, D, 5)
-#define __builtin_ia32_maxps512_mask(A, B, C, D, E) __builtin_ia32_maxps512_mask(A, B, C, D, 5)
-#define __builtin_ia32_maxsd_mask(A, B, C, D, E) __builtin_ia32_maxsd_mask(A, B, C, D, 5)
-#define __builtin_ia32_maxss_mask(A, B, C, D, E) __builtin_ia32_maxss_mask(A, B, C, D, 5)
-#define __builtin_ia32_minpd512_mask(A, B, C, D, E) __builtin_ia32_minpd512_mask(A, B, C, D, 5)
-#define __builtin_ia32_minps512_mask(A, B, C, D, E) __builtin_ia32_minps512_mask(A, B, C, D, 5)
-#define __builtin_ia32_minsd_mask(A, B, C, D, E) __builtin_ia32_minsd_mask(A, B, C, D, 5)
-#define __builtin_ia32_minss_mask(A, B, C, D, E) __builtin_ia32_minss_mask(A, B, C, D, 5)
-#define __builtin_ia32_mulpd512_mask(A, B, C, D, E) __builtin_ia32_mulpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_mulps512_mask(A, B, C, D, E) __builtin_ia32_mulps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_mulsd_mask(A, B, C, D, E) __builtin_ia32_mulsd_mask(A, B, C, D, 1)
-#define __builtin_ia32_mulss_mask(A, B, C, D, E) __builtin_ia32_mulss_mask(A, B, C, D, 1)
+#define __builtin_ia32_maxpd512_mask(A, B, C, D, E) __builtin_ia32_maxpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_maxps512_mask(A, B, C, D, E) __builtin_ia32_maxps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_maxsd_mask(A, B, C, D, E) __builtin_ia32_maxsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_maxss_mask(A, B, C, D, E) __builtin_ia32_maxss_mask(A, B, C, D, 8)
+#define __builtin_ia32_minpd512_mask(A, B, C, D, E) __builtin_ia32_minpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_minps512_mask(A, B, C, D, E) __builtin_ia32_minps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_minsd_mask(A, B, C, D, E) __builtin_ia32_minsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_minss_mask(A, B, C, D, E) __builtin_ia32_minss_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulpd512_mask(A, B, C, D, E) __builtin_ia32_mulpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulps512_mask(A, B, C, D, E) __builtin_ia32_mulps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulsd_mask(A, B, C, D, E) __builtin_ia32_mulsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulss_mask(A, B, C, D, E) __builtin_ia32_mulss_mask(A, B, C, D, 8)
#define __builtin_ia32_permdf512_mask(A, E, C, D) __builtin_ia32_permdf512_mask(A, 1, C, D)
#define __builtin_ia32_permdi512_mask(A, E, C, D) __builtin_ia32_permdi512_mask(A, 1, C, D)
#define __builtin_ia32_prold512_mask(A, E, C, D) __builtin_ia32_prold512_mask(A, 1, C, D)
@@ -295,88 +295,88 @@
#define __builtin_ia32_pternlogd512_maskz(A, B, C, F, E) __builtin_ia32_pternlogd512_maskz(A, B, C, 1, E)
#define __builtin_ia32_pternlogq512_mask(A, B, C, F, E) __builtin_ia32_pternlogq512_mask(A, B, C, 1, E)
#define __builtin_ia32_pternlogq512_maskz(A, B, C, F, E) __builtin_ia32_pternlogq512_maskz(A, B, C, 1, E)
-#define __builtin_ia32_rndscalepd_mask(A, F, C, D, E) __builtin_ia32_rndscalepd_mask(A, 1, C, D, 5)
-#define __builtin_ia32_rndscaleps_mask(A, F, C, D, E) __builtin_ia32_rndscaleps_mask(A, 1, C, D, 5)
-#define __builtin_ia32_rndscalesd_mask(A, B, I, D, E, F) __builtin_ia32_rndscalesd_mask(A, B, 1, D, E, 5)
-#define __builtin_ia32_rndscaless_mask(A, B, I, D, E, F) __builtin_ia32_rndscaless_mask(A, B, 1, D, E, 5)
-#define __builtin_ia32_scalefpd512_mask(A, B, C, D, E) __builtin_ia32_scalefpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_scalefps512_mask(A, B, C, D, E) __builtin_ia32_scalefps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_scalefsd_mask(A, B, C, D, E) __builtin_ia32_scalefsd_mask(A, B, C, D, 1)
-#define __builtin_ia32_scalefss_mask(A, B, C, D, E) __builtin_ia32_scalefss_mask(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv8df(A, B, C, D, F) __builtin_ia32_scatterdiv8df(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv8di(A, B, C, D, F) __builtin_ia32_scatterdiv8di(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv16sf(A, B, C, D, F) __builtin_ia32_scatterdiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv16si(A, B, C, D, F) __builtin_ia32_scatterdiv16si(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv16sf(A, B, C, D, F) __builtin_ia32_scattersiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv16si(A, B, C, D, F) __builtin_ia32_scattersiv16si(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv8df(A, B, C, D, F) __builtin_ia32_scattersiv8df(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv8di(A, B, C, D, F) __builtin_ia32_scattersiv8di(A, B, C, D, 1)
+#define __builtin_ia32_rndscalepd_mask(A, F, C, D, E) __builtin_ia32_rndscalepd_mask(A, 1, C, D, 8)
+#define __builtin_ia32_rndscaleps_mask(A, F, C, D, E) __builtin_ia32_rndscaleps_mask(A, 1, C, D, 8)
+#define __builtin_ia32_rndscalesd_mask(A, B, I, D, E, F) __builtin_ia32_rndscalesd_mask(A, B, 1, D, E, 8)
+#define __builtin_ia32_rndscaless_mask(A, B, I, D, E, F) __builtin_ia32_rndscaless_mask(A, B, 1, D, E, 8)
+#define __builtin_ia32_scalefpd512_mask(A, B, C, D, E) __builtin_ia32_scalefpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_scalefps512_mask(A, B, C, D, E) __builtin_ia32_scalefps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_scalefsd_mask(A, B, C, D, E) __builtin_ia32_scalefsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_scalefss_mask(A, B, C, D, E) __builtin_ia32_scalefss_mask(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv8df(A, B, C, D, F) __builtin_ia32_scatterdiv8df(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv8di(A, B, C, D, F) __builtin_ia32_scatterdiv8di(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv16sf(A, B, C, D, F) __builtin_ia32_scatterdiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv16si(A, B, C, D, F) __builtin_ia32_scatterdiv16si(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv16sf(A, B, C, D, F) __builtin_ia32_scattersiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv16si(A, B, C, D, F) __builtin_ia32_scattersiv16si(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv8df(A, B, C, D, F) __builtin_ia32_scattersiv8df(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv8di(A, B, C, D, F) __builtin_ia32_scattersiv8di(A, B, C, D, 8)
#define __builtin_ia32_shuf_f32x4_mask(A, B, F, D, E) __builtin_ia32_shuf_f32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_f64x2_mask(A, B, F, D, E) __builtin_ia32_shuf_f64x2_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_i32x4_mask(A, B, F, D, E) __builtin_ia32_shuf_i32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_i64x2_mask(A, B, F, D, E) __builtin_ia32_shuf_i64x2_mask(A, B, 1, D, E)
#define __builtin_ia32_shufpd512_mask(A, B, F, D, E) __builtin_ia32_shufpd512_mask(A, B, 1, D, E)
#define __builtin_ia32_shufps512_mask(A, B, F, D, E) __builtin_ia32_shufps512_mask(A, B, 1, D, E)
-#define __builtin_ia32_sqrtpd512_mask(A, B, C, D) __builtin_ia32_sqrtpd512_mask(A, B, C, 1)
-#define __builtin_ia32_sqrtps512_mask(A, B, C, D) __builtin_ia32_sqrtps512_mask(A, B, C, 1)
-#define __builtin_ia32_sqrtsd_mask(A, B, C, D, E) __builtin_ia32_sqrtsd_mask(A, B, C, D, 1)
-#define __builtin_ia32_sqrtss_mask(A, B, C, D, E) __builtin_ia32_sqrtss_mask(A, B, C, D, 1)
-#define __builtin_ia32_subpd512_mask(A, B, C, D, E) __builtin_ia32_subpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_subps512_mask(A, B, C, D, E) __builtin_ia32_subps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_subsd_mask(A, B, C, D, E) __builtin_ia32_subsd_mask(A, B, C, D, 1)
-#define __builtin_ia32_subss_mask(A, B, C, D, E) __builtin_ia32_subss_mask(A, B, C, D, 1)
+#define __builtin_ia32_sqrtpd512_mask(A, B, C, D) __builtin_ia32_sqrtpd512_mask(A, B, C, 8)
+#define __builtin_ia32_sqrtps512_mask(A, B, C, D) __builtin_ia32_sqrtps512_mask(A, B, C, 8)
+#define __builtin_ia32_sqrtsd_mask(A, B, C, D, E) __builtin_ia32_sqrtsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_sqrtss_mask(A, B, C, D, E) __builtin_ia32_sqrtss_mask(A, B, C, D, 8)
+#define __builtin_ia32_subpd512_mask(A, B, C, D, E) __builtin_ia32_subpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_subps512_mask(A, B, C, D, E) __builtin_ia32_subps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_subsd_mask(A, B, C, D, E) __builtin_ia32_subsd_mask(A, B, C, D, 8)
+#define __builtin_ia32_subss_mask(A, B, C, D, E) __builtin_ia32_subss_mask(A, B, C, D, 8)
#define __builtin_ia32_ucmpd512_mask(A, B, E, D) __builtin_ia32_ucmpd512_mask(A, B, 1, D)
#define __builtin_ia32_ucmpq512_mask(A, B, E, D) __builtin_ia32_ucmpq512_mask(A, B, 1, D)
-#define __builtin_ia32_vcomisd(A, B, C, D) __builtin_ia32_vcomisd(A, B, 1, 5)
-#define __builtin_ia32_vcomiss(A, B, C, D) __builtin_ia32_vcomiss(A, B, 1, 5)
-#define __builtin_ia32_vcvtph2ps512_mask(A, B, C, D) __builtin_ia32_vcvtph2ps512_mask(A, B, C, 5)
+#define __builtin_ia32_vcomisd(A, B, C, D) __builtin_ia32_vcomisd(A, B, 1, 8)
+#define __builtin_ia32_vcomiss(A, B, C, D) __builtin_ia32_vcomiss(A, B, 1, 8)
+#define __builtin_ia32_vcvtph2ps512_mask(A, B, C, D) __builtin_ia32_vcvtph2ps512_mask(A, B, C, 8)
#define __builtin_ia32_vcvtps2ph512_mask(A, E, C, D) __builtin_ia32_vcvtps2ph512_mask(A, 1, C, D)
-#define __builtin_ia32_vcvtsd2si32(A, B) __builtin_ia32_vcvtsd2si32(A, 1)
-#define __builtin_ia32_vcvtsd2si64(A, B) __builtin_ia32_vcvtsd2si64(A, 1)
-#define __builtin_ia32_vcvtsd2usi32(A, B) __builtin_ia32_vcvtsd2usi32(A, 1)
-#define __builtin_ia32_vcvtsd2usi64(A, B) __builtin_ia32_vcvtsd2usi64(A, 1)
-#define __builtin_ia32_vcvtss2si32(A, B) __builtin_ia32_vcvtss2si32(A, 1)
-#define __builtin_ia32_vcvtss2si64(A, B) __builtin_ia32_vcvtss2si64(A, 1)
-#define __builtin_ia32_vcvtss2usi32(A, B) __builtin_ia32_vcvtss2usi32(A, 1)
-#define __builtin_ia32_vcvtss2usi64(A, B) __builtin_ia32_vcvtss2usi64(A, 1)
-#define __builtin_ia32_vcvttsd2si32(A, B) __builtin_ia32_vcvttsd2si32(A, 5)
-#define __builtin_ia32_vcvttsd2si64(A, B) __builtin_ia32_vcvttsd2si64(A, 5)
-#define __builtin_ia32_vcvttsd2usi32(A, B) __builtin_ia32_vcvttsd2usi32(A, 5)
-#define __builtin_ia32_vcvttsd2usi64(A, B) __builtin_ia32_vcvttsd2usi64(A, 5)
-#define __builtin_ia32_vcvttss2si32(A, B) __builtin_ia32_vcvttss2si32(A, 5)
-#define __builtin_ia32_vcvttss2si64(A, B) __builtin_ia32_vcvttss2si64(A, 5)
-#define __builtin_ia32_vcvttss2usi32(A, B) __builtin_ia32_vcvttss2usi32(A, 5)
-#define __builtin_ia32_vcvttss2usi64(A, B) __builtin_ia32_vcvttss2usi64(A, 5)
-#define __builtin_ia32_vfmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddps512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsd3_mask(A, B, C, D, E) __builtin_ia32_vfmaddsd3_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsd3_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsd3_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsd3_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsd3_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddss3_mask(A, B, C, D, E) __builtin_ia32_vfmaddss3_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddss3_mask3(A, B, C, D, E) __builtin_ia32_vfmaddss3_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddss3_maskz(A, B, C, D, E) __builtin_ia32_vfmaddss3_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubss3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubss3_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubps512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, 1)
+#define __builtin_ia32_vcvtsd2si32(A, B) __builtin_ia32_vcvtsd2si32(A, 8)
+#define __builtin_ia32_vcvtsd2si64(A, B) __builtin_ia32_vcvtsd2si64(A, 8)
+#define __builtin_ia32_vcvtsd2usi32(A, B) __builtin_ia32_vcvtsd2usi32(A, 8)
+#define __builtin_ia32_vcvtsd2usi64(A, B) __builtin_ia32_vcvtsd2usi64(A, 8)
+#define __builtin_ia32_vcvtss2si32(A, B) __builtin_ia32_vcvtss2si32(A, 8)
+#define __builtin_ia32_vcvtss2si64(A, B) __builtin_ia32_vcvtss2si64(A, 8)
+#define __builtin_ia32_vcvtss2usi32(A, B) __builtin_ia32_vcvtss2usi32(A, 8)
+#define __builtin_ia32_vcvtss2usi64(A, B) __builtin_ia32_vcvtss2usi64(A, 8)
+#define __builtin_ia32_vcvttsd2si32(A, B) __builtin_ia32_vcvttsd2si32(A, 8)
+#define __builtin_ia32_vcvttsd2si64(A, B) __builtin_ia32_vcvttsd2si64(A, 8)
+#define __builtin_ia32_vcvttsd2usi32(A, B) __builtin_ia32_vcvttsd2usi32(A, 8)
+#define __builtin_ia32_vcvttsd2usi64(A, B) __builtin_ia32_vcvttsd2usi64(A, 8)
+#define __builtin_ia32_vcvttss2si32(A, B) __builtin_ia32_vcvttss2si32(A, 8)
+#define __builtin_ia32_vcvttss2si64(A, B) __builtin_ia32_vcvttss2si64(A, 8)
+#define __builtin_ia32_vcvttss2usi32(A, B) __builtin_ia32_vcvttss2usi32(A, 8)
+#define __builtin_ia32_vcvttss2usi64(A, B) __builtin_ia32_vcvttss2usi64(A, 8)
+#define __builtin_ia32_vfmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddps512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsd3_mask(A, B, C, D, E) __builtin_ia32_vfmaddsd3_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsd3_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsd3_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsd3_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsd3_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddss3_mask(A, B, C, D, E) __builtin_ia32_vfmaddss3_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddss3_mask3(A, B, C, D, E) __builtin_ia32_vfmaddss3_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddss3_maskz(A, B, C, D, E) __builtin_ia32_vfmaddss3_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubsd3_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubss3_mask3(A, B, C, D, E) __builtin_ia32_vfmsubss3_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubps512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, 8)
#define __builtin_ia32_vpermilpd512_mask(A, E, C, D) __builtin_ia32_vpermilpd512_mask(A, 1, C, D)
#define __builtin_ia32_vpermilps512_mask(A, E, C, D) __builtin_ia32_vpermilps512_mask(A, 1, C, D)
@@ -142,34 +142,34 @@ test_2 (_mm256_insert_epi32, __m256i, __m256i, int, 3)
#ifdef __x86_64__
test_2 (_mm256_insert_epi64, __m256i, __m256i, long long, 1)
#endif
-test_1 (_mm256_round_pd, __m256d, __m256d, 1)
-test_1 (_mm256_round_ps, __m256, __m256, 1)
+test_1 (_mm256_round_pd, __m256d, __m256d, 9)
+test_1 (_mm256_round_ps, __m256, __m256, 9)
test_1 (_cvtss_sh, unsigned short, float, 1)
test_1 (_mm_cvtps_ph, __m128i, __m128, 1)
test_1 (_mm256_cvtps_ph, __m128i, __m256, 1)
test_0 (_xabort, void, 1)
-test_1 (_mm512_cvt_roundepi32_ps, __m512, __m512i, 1)
-test_1 (_mm512_cvt_roundepu32_ps, __m512, __m512i, 1)
-test_1 (_mm512_cvt_roundpd_epi32, __m256i, __m512d, 1)
-test_1 (_mm512_cvt_roundpd_epu32, __m256i, __m512d, 1)
-test_1 (_mm512_cvt_roundpd_ps, __m256, __m512d, 1)
-test_1 (_mm512_cvt_roundph_ps, __m512, __m256i, 5)
-test_1 (_mm512_cvt_roundps_epi32, __m512i, __m512, 1)
-test_1 (_mm512_cvt_roundps_epu32, __m512i, __m512, 1)
-test_1 (_mm512_cvt_roundps_pd, __m512d, __m256, 5)
+test_1 (_mm512_cvt_roundepi32_ps, __m512, __m512i, 9)
+test_1 (_mm512_cvt_roundepu32_ps, __m512, __m512i, 9)
+test_1 (_mm512_cvt_roundpd_epi32, __m256i, __m512d, 9)
+test_1 (_mm512_cvt_roundpd_epu32, __m256i, __m512d, 9)
+test_1 (_mm512_cvt_roundpd_ps, __m256, __m512d, 9)
+test_1 (_mm512_cvt_roundph_ps, __m512, __m256i, 8)
+test_1 (_mm512_cvt_roundps_epi32, __m512i, __m512, 9)
+test_1 (_mm512_cvt_roundps_epu32, __m512i, __m512, 9)
+test_1 (_mm512_cvt_roundps_pd, __m512d, __m256, 8)
test_1 (_mm512_cvtps_ph, __m256i, __m512, 1)
-test_1 (_mm512_cvtt_roundpd_epi32, __m256i, __m512d, 5)
-test_1 (_mm512_cvtt_roundpd_epu32, __m256i, __m512d, 5)
-test_1 (_mm512_cvtt_roundps_epi32, __m512i, __m512, 5)
-test_1 (_mm512_cvtt_roundps_epu32, __m512i, __m512, 5)
+test_1 (_mm512_cvtt_roundpd_epi32, __m256i, __m512d, 8)
+test_1 (_mm512_cvtt_roundpd_epu32, __m256i, __m512d, 8)
+test_1 (_mm512_cvtt_roundps_epi32, __m512i, __m512, 8)
+test_1 (_mm512_cvtt_roundps_epu32, __m512i, __m512, 8)
test_1 (_mm512_extractf32x4_ps, __m128, __m512, 1)
test_1 (_mm512_extractf64x4_pd, __m256d, __m512d, 1)
test_1 (_mm512_extracti32x4_epi32, __m128i, __m512i, 1)
test_1 (_mm512_extracti64x4_epi64, __m256i, __m512i, 1)
-test_1 (_mm512_getexp_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_getexp_round_ps, __m512, __m512, 5)
-test_1y (_mm512_getmant_round_pd, __m512d, __m512d, 1, 1, 5)
-test_1y (_mm512_getmant_round_ps, __m512, __m512, 1, 1, 5)
+test_1 (_mm512_getexp_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_getexp_round_ps, __m512, __m512, 8)
+test_1y (_mm512_getmant_round_pd, __m512d, __m512d, 1, 1, 8)
+test_1y (_mm512_getmant_round_ps, __m512, __m512, 1, 1, 8)
test_1 (_mm512_permute_pd, __m512d, __m512d, 1)
test_1 (_mm512_permute_ps, __m512, __m512, 1)
test_1 (_mm512_permutex_epi64, __m512i, __m512i, 1)
@@ -181,27 +181,27 @@ test_1 (_mm512_ror_epi64, __m512i, __m512i, 1)
test_1 (_mm512_shuffle_epi32, __m512i, __m512i, 1)
test_1 (_mm512_slli_epi32, __m512i, __m512i, 1)
test_1 (_mm512_slli_epi64, __m512i, __m512i, 1)
-test_1 (_mm512_sqrt_round_pd, __m512d, __m512d, 1)
-test_1 (_mm512_sqrt_round_ps, __m512, __m512, 1)
+test_1 (_mm512_sqrt_round_pd, __m512d, __m512d, 9)
+test_1 (_mm512_sqrt_round_ps, __m512, __m512, 9)
test_1 (_mm512_srai_epi32, __m512i, __m512i, 1)
test_1 (_mm512_srai_epi64, __m512i, __m512i, 1)
test_1 (_mm512_srli_epi32, __m512i, __m512i, 1)
test_1 (_mm512_srli_epi64, __m512i, __m512i, 1)
-test_1 (_mm_cvt_roundsd_i32, int, __m128d, 1)
-test_1 (_mm_cvt_roundsd_u32, unsigned, __m128d, 1)
-test_1 (_mm_cvt_roundss_i32, int, __m128, 1)
-test_1 (_mm_cvt_roundss_u32, unsigned, __m128, 1)
-test_1 (_mm_cvtt_roundsd_i32, int, __m128d, 5)
-test_1 (_mm_cvtt_roundsd_u32, unsigned, __m128d, 5)
-test_1 (_mm_cvtt_roundss_i32, int, __m128, 5)
-test_1 (_mm_cvtt_roundss_u32, unsigned, __m128, 5)
+test_1 (_mm_cvt_roundsd_i32, int, __m128d, 9)
+test_1 (_mm_cvt_roundsd_u32, unsigned, __m128d, 9)
+test_1 (_mm_cvt_roundss_i32, int, __m128, 9)
+test_1 (_mm_cvt_roundss_u32, unsigned, __m128, 9)
+test_1 (_mm_cvtt_roundsd_i32, int, __m128d, 8)
+test_1 (_mm_cvtt_roundsd_u32, unsigned, __m128d, 8)
+test_1 (_mm_cvtt_roundss_i32, int, __m128, 8)
+test_1 (_mm_cvtt_roundss_u32, unsigned, __m128, 8)
test_1x (_mm512_getmant_pd, __m512d, __m512d, 1, 1)
test_1x (_mm512_getmant_ps, __m512, __m512, 1, 1)
-test_1x (_mm512_roundscale_round_pd, __m512d, __m512d, 1, 5)
-test_1x (_mm512_roundscale_round_ps, __m512, __m512, 1, 5)
-test_1x (_mm_cvt_roundi32_ss, __m128, __m128, 1, 1)
-test_2 (_mm512_add_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_add_round_ps, __m512, __m512, __m512, 1)
+test_1x (_mm512_roundscale_round_pd, __m512d, __m512d, 1, 8)
+test_1x (_mm512_roundscale_round_ps, __m512, __m512, 1, 8)
+test_1x (_mm_cvt_roundi32_ss, __m128, __m128, 1, 9)
+test_2 (_mm512_add_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_add_round_ps, __m512, __m512, __m512, 9)
test_2 (_mm512_alignr_epi32, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_alignr_epi64, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_cmp_epi32_mask, __mmask16, __m512i, __m512i, 1)
@@ -210,8 +210,8 @@ test_2 (_mm512_cmp_epu32_mask, __mmask16, __m512i, __m512i, 1)
test_2 (_mm512_cmp_epu64_mask, __mmask8, __m512i, __m512i, 1)
test_2 (_mm512_cmp_pd_mask, __mmask8, __m512d, __m512d, 1)
test_2 (_mm512_cmp_ps_mask, __mmask16, __m512, __m512, 1)
-test_2 (_mm512_div_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_div_round_ps, __m512, __m512, __m512, 1)
+test_2 (_mm512_div_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_div_round_ps, __m512, __m512, __m512, 9)
test_2 (_mm512_i32gather_epi32, __m512i, __m512i, void const *, 1)
test_2 (_mm512_i32gather_epi64, __m512i, __m256i, void const *, 1)
test_2 (_mm512_i32gather_pd, __m512d, __m256i, void const *, 1)
@@ -224,28 +224,28 @@ test_2 (_mm512_insertf32x4, __m512, __m512, __m128, 1)
test_2 (_mm512_insertf64x4, __m512d, __m512d, __m256d, 1)
test_2 (_mm512_inserti32x4, __m512i, __m512i, __m128i, 1)
test_2 (_mm512_inserti64x4, __m512i, __m512i, __m256i, 1)
-test_2 (_mm512_maskz_cvt_roundepi32_ps, __m512, __mmask16, __m512i, 1)
-test_2 (_mm512_maskz_cvt_roundepu32_ps, __m512, __mmask16, __m512i, 1)
-test_2 (_mm512_maskz_cvt_roundpd_epi32, __m256i, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_cvt_roundpd_epu32, __m256i, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_cvt_roundpd_ps, __m256, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_cvt_roundph_ps, __m512, __mmask16, __m256i, 5)
-test_2 (_mm512_maskz_cvt_roundps_epi32, __m512i, __mmask16, __m512, 1)
-test_2 (_mm512_maskz_cvt_roundps_epu32, __m512i, __mmask16, __m512, 1)
-test_2 (_mm512_maskz_cvt_roundps_pd, __m512d, __mmask8, __m256, 5)
+test_2 (_mm512_maskz_cvt_roundepi32_ps, __m512, __mmask16, __m512i, 9)
+test_2 (_mm512_maskz_cvt_roundepu32_ps, __m512, __mmask16, __m512i, 9)
+test_2 (_mm512_maskz_cvt_roundpd_epi32, __m256i, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_cvt_roundpd_epu32, __m256i, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_cvt_roundpd_ps, __m256, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_cvt_roundph_ps, __m512, __mmask16, __m256i, 8)
+test_2 (_mm512_maskz_cvt_roundps_epi32, __m512i, __mmask16, __m512, 9)
+test_2 (_mm512_maskz_cvt_roundps_epu32, __m512i, __mmask16, __m512, 9)
+test_2 (_mm512_maskz_cvt_roundps_pd, __m512d, __mmask8, __m256, 8)
test_2 (_mm512_maskz_cvtps_ph, __m256i, __mmask16, __m512, 1)
-test_2 (_mm512_maskz_cvtt_roundpd_epi32, __m256i, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_cvtt_roundpd_epu32, __m256i, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_cvtt_roundps_epi32, __m512i, __mmask16, __m512, 5)
-test_2 (_mm512_maskz_cvtt_roundps_epu32, __m512i, __mmask16, __m512, 5)
+test_2 (_mm512_maskz_cvtt_roundpd_epi32, __m256i, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_cvtt_roundpd_epu32, __m256i, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_cvtt_roundps_epi32, __m512i, __mmask16, __m512, 8)
+test_2 (_mm512_maskz_cvtt_roundps_epu32, __m512i, __mmask16, __m512, 8)
test_2 (_mm512_maskz_extractf32x4_ps, __m128, __mmask8, __m512, 1)
test_2 (_mm512_maskz_extractf64x4_pd, __m256d, __mmask8, __m512d, 1)
test_2 (_mm512_maskz_extracti32x4_epi32, __m128i, __mmask8, __m512i, 1)
test_2 (_mm512_maskz_extracti64x4_epi64, __m256i, __mmask8, __m512i, 1)
-test_2 (_mm512_maskz_getexp_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_getexp_round_ps, __m512, __mmask16, __m512, 5)
-test_2y (_mm512_maskz_getmant_round_pd, __m512d, __mmask8, __m512d, 1, 1, 5)
-test_2y (_mm512_maskz_getmant_round_ps, __m512, __mmask16, __m512, 1, 1, 5)
+test_2 (_mm512_maskz_getexp_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_getexp_round_ps, __m512, __mmask16, __m512, 8)
+test_2y (_mm512_maskz_getmant_round_pd, __m512d, __mmask8, __m512d, 1, 1, 8)
+test_2y (_mm512_maskz_getmant_round_ps, __m512, __mmask16, __m512, 1, 1, 8)
test_2 (_mm512_maskz_permute_pd, __m512d, __mmask8, __m512d, 1)
test_2 (_mm512_maskz_permute_ps, __m512, __mmask16, __m512, 1)
test_2 (_mm512_maskz_permutex_epi64, __m512i, __mmask8, __m512i, 1)
@@ -257,107 +257,107 @@ test_2 (_mm512_maskz_ror_epi64, __m512i, __mmask8, __m512i, 1)
test_2 (_mm512_maskz_shuffle_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_slli_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_slli_epi64, __m512i, __mmask8, __m512i, 1)
-test_2 (_mm512_maskz_sqrt_round_pd, __m512d, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_sqrt_round_ps, __m512, __mmask16, __m512, 1)
+test_2 (_mm512_maskz_sqrt_round_pd, __m512d, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_sqrt_round_ps, __m512, __mmask16, __m512, 9)
test_2 (_mm512_maskz_srai_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_srai_epi64, __m512i, __mmask8, __m512i, 1)
test_2 (_mm512_maskz_srli_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_srli_epi64, __m512i, __mmask8, __m512i, 1)
-test_2 (_mm512_max_round_pd, __m512d, __m512d, __m512d, 5)
-test_2 (_mm512_max_round_ps, __m512, __m512, __m512, 5)
-test_2 (_mm512_min_round_pd, __m512d, __m512d, __m512d, 5)
-test_2 (_mm512_min_round_ps, __m512, __m512, __m512, 5)
-test_2 (_mm512_mul_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_mul_round_ps, __m512, __m512, __m512, 1)
-test_2 (_mm512_scalef_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_scalef_round_ps, __m512, __m512, __m512, 1)
+test_2 (_mm512_max_round_pd, __m512d, __m512d, __m512d, 8)
+test_2 (_mm512_max_round_ps, __m512, __m512, __m512, 8)
+test_2 (_mm512_min_round_pd, __m512d, __m512d, __m512d, 8)
+test_2 (_mm512_min_round_ps, __m512, __m512, __m512, 8)
+test_2 (_mm512_mul_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_mul_round_ps, __m512, __m512, __m512, 9)
+test_2 (_mm512_scalef_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_scalef_round_ps, __m512, __m512, __m512, 9)
test_2 (_mm512_shuffle_f32x4, __m512, __m512, __m512, 1)
test_2 (_mm512_shuffle_f64x2, __m512d, __m512d, __m512d, 1)
test_2 (_mm512_shuffle_i32x4, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_shuffle_i64x2, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_shuffle_pd, __m512d, __m512d, __m512d, 1)
test_2 (_mm512_shuffle_ps, __m512, __m512, __m512, 1)
-test_2 (_mm512_sub_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_sub_round_ps, __m512, __m512, __m512, 1)
-test_2 (_mm_add_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_add_round_ss, __m128, __m128, __m128, 1)
+test_2 (_mm512_sub_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_sub_round_ps, __m512, __m512, __m512, 9)
+test_2 (_mm_add_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_add_round_ss, __m128, __m128, __m128, 9)
test_2 (_mm_cmp_sd_mask, __mmask8, __m128d, __m128d, 1)
test_2 (_mm_cmp_ss_mask, __mmask8, __m128, __m128, 1)
#ifdef __x86_64__
-test_2 (_mm_cvt_roundi64_sd, __m128d, __m128d, long long, 1)
-test_2 (_mm_cvt_roundi64_ss, __m128, __m128, long long, 1)
+test_2 (_mm_cvt_roundi64_sd, __m128d, __m128d, long long, 9)
+test_2 (_mm_cvt_roundi64_ss, __m128, __m128, long long, 9)
#endif
-test_2 (_mm_cvt_roundsd_ss, __m128, __m128, __m128d, 1)
-test_2 (_mm_cvt_roundss_sd, __m128d, __m128d, __m128, 5)
-test_2 (_mm_cvt_roundu32_ss, __m128, __m128, unsigned, 1)
+test_2 (_mm_cvt_roundsd_ss, __m128, __m128, __m128d, 9)
+test_2 (_mm_cvt_roundss_sd, __m128d, __m128d, __m128, 8)
+test_2 (_mm_cvt_roundu32_ss, __m128, __m128, unsigned, 9)
#ifdef __x86_64__
-test_2 (_mm_cvt_roundu64_sd, __m128d, __m128d, unsigned long long, 1)
-test_2 (_mm_cvt_roundu64_ss, __m128, __m128, unsigned long long, 1)
+test_2 (_mm_cvt_roundu64_sd, __m128d, __m128d, unsigned long long, 9)
+test_2 (_mm_cvt_roundu64_ss, __m128, __m128, unsigned long long, 9)
#endif
-test_2 (_mm_div_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_div_round_ss, __m128, __m128, __m128, 1)
-test_2 (_mm_getexp_round_sd, __m128d, __m128d, __m128d, 5)
-test_2 (_mm_getexp_round_ss, __m128, __m128, __m128, 5)
-test_2y (_mm_getmant_round_sd, __m128d, __m128d, __m128d, 1, 1, 5)
-test_2y (_mm_getmant_round_ss, __m128, __m128, __m128, 1, 1, 5)
-test_2 (_mm_mul_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_mul_round_ss, __m128, __m128, __m128, 1)
-test_2 (_mm_scalef_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_scalef_round_ss, __m128, __m128, __m128, 1)
-test_2 (_mm_sqrt_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_sqrt_round_ss, __m128, __m128, __m128, 1)
-test_2 (_mm_sub_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_sub_round_ss, __m128, __m128, __m128, 1)
-test_2x (_mm512_cmp_round_pd_mask, __mmask8, __m512d, __m512d, 1, 5)
-test_2x (_mm512_cmp_round_ps_mask, __mmask16, __m512, __m512, 1, 5)
-test_2x (_mm512_maskz_roundscale_round_pd, __m512d, __mmask8, __m512d, 1, 5)
-test_2x (_mm512_maskz_roundscale_round_ps, __m512, __mmask16, __m512, 1, 5)
-test_2x (_mm_cmp_round_sd_mask, __mmask8, __m128d, __m128d, 1, 5)
-test_2x (_mm_cmp_round_ss_mask, __mmask8, __m128, __m128, 1, 5)
-test_2x (_mm_comi_round_sd, int, __m128d, __m128d, 1, 5)
-test_2x (_mm_comi_round_ss, int, __m128, __m128, 1, 5)
-test_2x (_mm_roundscale_round_sd, __m128d, __m128d, __m128d, 1, 5)
-test_2x (_mm_roundscale_round_ss, __m128, __m128, __m128, 1, 5)
-test_3 (_mm512_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmadd_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmaddsub_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmsub_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmsubadd_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fnmadd_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fnmsub_round_ps, __m512, __m512, __m512, __m512, 1)
+test_2 (_mm_div_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_div_round_ss, __m128, __m128, __m128, 9)
+test_2 (_mm_getexp_round_sd, __m128d, __m128d, __m128d, 8)
+test_2 (_mm_getexp_round_ss, __m128, __m128, __m128, 8)
+test_2y (_mm_getmant_round_sd, __m128d, __m128d, __m128d, 1, 1, 8)
+test_2y (_mm_getmant_round_ss, __m128, __m128, __m128, 1, 1, 8)
+test_2 (_mm_mul_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_mul_round_ss, __m128, __m128, __m128, 9)
+test_2 (_mm_scalef_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_scalef_round_ss, __m128, __m128, __m128, 9)
+test_2 (_mm_sqrt_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_sqrt_round_ss, __m128, __m128, __m128, 9)
+test_2 (_mm_sub_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_sub_round_ss, __m128, __m128, __m128, 9)
+test_2x (_mm512_cmp_round_pd_mask, __mmask8, __m512d, __m512d, 1, 8)
+test_2x (_mm512_cmp_round_ps_mask, __mmask16, __m512, __m512, 1, 8)
+test_2x (_mm512_maskz_roundscale_round_pd, __m512d, __mmask8, __m512d, 1, 8)
+test_2x (_mm512_maskz_roundscale_round_ps, __m512, __mmask16, __m512, 1, 8)
+test_2x (_mm_cmp_round_sd_mask, __mmask8, __m128d, __m128d, 1, 8)
+test_2x (_mm_cmp_round_ss_mask, __mmask8, __m128, __m128, 1, 8)
+test_2x (_mm_comi_round_sd, int, __m128d, __m128d, 1, 8)
+test_2x (_mm_comi_round_ss, int, __m128, __m128, 1, 8)
+test_2x (_mm_roundscale_round_sd, __m128d, __m128d, __m128d, 1, 8)
+test_2x (_mm_roundscale_round_ss, __m128, __m128, __m128, 1, 8)
+test_3 (_mm512_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmadd_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmaddsub_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmsub_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmsubadd_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fnmadd_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fnmsub_round_ps, __m512, __m512, __m512, __m512, 9)
test_3 (_mm512_mask_cmp_epi32_mask, __mmask16, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_epi64_mask, __mmask8, __mmask8, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_epu32_mask, __mmask16, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_epu64_mask, __mmask8, __mmask8, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_pd_mask, __mmask8, __mmask8, __m512d, __m512d, 1)
test_3 (_mm512_mask_cmp_ps_mask, __mmask16, __mmask16, __m512, __m512, 1)
-test_3 (_mm512_mask_cvt_roundepi32_ps, __m512, __m512, __mmask16, __m512i, 1)
-test_3 (_mm512_mask_cvt_roundepu32_ps, __m512, __m512, __mmask16, __m512i, 1)
-test_3 (_mm512_mask_cvt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_cvt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_cvt_roundpd_ps, __m256, __m256, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_cvt_roundph_ps, __m512, __m512, __mmask16, __m256i, 5)
-test_3 (_mm512_mask_cvt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 1)
-test_3 (_mm512_mask_cvt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 1)
-test_3 (_mm512_mask_cvt_roundps_pd, __m512d, __m512d, __mmask8, __m256, 5)
+test_3 (_mm512_mask_cvt_roundepi32_ps, __m512, __m512, __mmask16, __m512i, 9)
+test_3 (_mm512_mask_cvt_roundepu32_ps, __m512, __m512, __mmask16, __m512i, 9)
+test_3 (_mm512_mask_cvt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_cvt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_cvt_roundpd_ps, __m256, __m256, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_cvt_roundph_ps, __m512, __m512, __mmask16, __m256i, 8)
+test_3 (_mm512_mask_cvt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 9)
+test_3 (_mm512_mask_cvt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 9)
+test_3 (_mm512_mask_cvt_roundps_pd, __m512d, __m512d, __mmask8, __m256, 8)
test_3 (_mm512_mask_cvtps_ph, __m256i, __m256i, __mmask16, __m512, 1)
-test_3 (_mm512_mask_cvtt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_cvtt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_cvtt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 5)
-test_3 (_mm512_mask_cvtt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 5)
+test_3 (_mm512_mask_cvtt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_cvtt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_cvtt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 8)
+test_3 (_mm512_mask_cvtt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 8)
test_3 (_mm512_mask_extractf32x4_ps, __m128, __m128, __mmask8, __m512, 1)
test_3 (_mm512_mask_extractf64x4_pd, __m256d, __m256d, __mmask8, __m512d, 1)
test_3 (_mm512_mask_extracti32x4_epi32, __m128i, __m128i, __mmask8, __m512i, 1)
test_3 (_mm512_mask_extracti64x4_epi64, __m256i, __m256i, __mmask8, __m512i, 1)
-test_3 (_mm512_mask_getexp_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_getexp_round_ps, __m512, __m512, __mmask16, __m512, 5)
-test_3y (_mm512_mask_getmant_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 1, 5)
-test_3y (_mm512_mask_getmant_round_ps, __m512, __m512, __mmask16, __m512, 1, 1, 5)
+test_3 (_mm512_mask_getexp_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_getexp_round_ps, __m512, __m512, __mmask16, __m512, 8)
+test_3y (_mm512_mask_getmant_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 1, 8)
+test_3y (_mm512_mask_getmant_round_ps, __m512, __m512, __mmask16, __m512, 1, 1, 8)
test_3 (_mm512_mask_permute_pd, __m512d, __m512d, __mmask8, __m512d, 1)
test_3 (_mm512_mask_permute_ps, __m512, __m512, __mmask16, __m512, 1)
test_3 (_mm512_mask_permutex_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
@@ -369,48 +369,48 @@ test_3 (_mm512_mask_ror_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
test_3 (_mm512_mask_shuffle_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_slli_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_slli_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
-test_3 (_mm512_mask_sqrt_round_pd, __m512d, __m512d, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_sqrt_round_ps, __m512, __m512, __mmask16, __m512, 1)
+test_3 (_mm512_mask_sqrt_round_pd, __m512d, __m512d, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_sqrt_round_ps, __m512, __m512, __mmask16, __m512, 9)
test_3 (_mm512_mask_srai_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_srai_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
test_3 (_mm512_mask_srli_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_srli_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
-test_3 (_mm512_maskz_add_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_add_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_add_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_add_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_maskz_alignr_epi32, __m512i, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_maskz_alignr_epi64, __m512i, __mmask8, __m512i, __m512i, 1)
-test_3 (_mm512_maskz_div_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_div_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_div_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_div_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_maskz_insertf32x4, __m512, __mmask16, __m512, __m128, 1)
test_3 (_mm512_maskz_insertf64x4, __m512d, __mmask8, __m512d, __m256d, 1)
test_3 (_mm512_maskz_inserti32x4, __m512i, __mmask16, __m512i, __m128i, 1)
test_3 (_mm512_maskz_inserti64x4, __m512i, __mmask8, __m512i, __m256i, 1)
-test_3 (_mm512_maskz_max_round_pd, __m512d, __mmask8, __m512d, __m512d, 5)
-test_3 (_mm512_maskz_max_round_ps, __m512, __mmask16, __m512, __m512, 5)
-test_3 (_mm512_maskz_min_round_pd, __m512d, __mmask8, __m512d, __m512d, 5)
-test_3 (_mm512_maskz_min_round_ps, __m512, __mmask16, __m512, __m512, 5)
-test_3 (_mm512_maskz_mul_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_mul_round_ps, __m512, __mmask16, __m512, __m512, 1)
-test_3 (_mm512_maskz_scalef_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_scalef_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_max_round_pd, __m512d, __mmask8, __m512d, __m512d, 8)
+test_3 (_mm512_maskz_max_round_ps, __m512, __mmask16, __m512, __m512, 8)
+test_3 (_mm512_maskz_min_round_pd, __m512d, __mmask8, __m512d, __m512d, 8)
+test_3 (_mm512_maskz_min_round_ps, __m512, __mmask16, __m512, __m512, 8)
+test_3 (_mm512_maskz_mul_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_mul_round_ps, __m512, __mmask16, __m512, __m512, 9)
+test_3 (_mm512_maskz_scalef_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_scalef_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_maskz_shuffle_f32x4, __m512, __mmask16, __m512, __m512, 1)
test_3 (_mm512_maskz_shuffle_f64x2, __m512d, __mmask8, __m512d, __m512d, 1)
test_3 (_mm512_maskz_shuffle_i32x4, __m512i, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_maskz_shuffle_i64x2, __m512i, __mmask8, __m512i, __m512i, 1)
test_3 (_mm512_maskz_shuffle_pd, __m512d, __mmask8, __m512d, __m512d, 1)
test_3 (_mm512_maskz_shuffle_ps, __m512, __mmask16, __m512, __m512, 1)
-test_3 (_mm512_maskz_sub_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_sub_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_sub_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_sub_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_ternarylogic_epi32, __m512i, __m512i, __m512i, __m512i, 1)
test_3 (_mm512_ternarylogic_epi64, __m512i, __m512i, __m512i, __m512i, 1)
-test_3 (_mm_fmadd_round_sd, __m128d, __m128d, __m128d, __m128d, 1)
-test_3 (_mm_fmadd_round_ss, __m128, __m128, __m128, __m128, 1)
-test_3 (_mm_fmsub_round_sd, __m128d, __m128d, __m128d, __m128d, 1)
-test_3 (_mm_fmsub_round_ss, __m128, __m128, __m128, __m128, 1)
-test_3 (_mm_fnmadd_round_sd, __m128d, __m128d, __m128d, __m128d, 1)
-test_3 (_mm_fnmadd_round_ss, __m128, __m128, __m128, __m128, 1)
-test_3 (_mm_fnmsub_round_sd, __m128d, __m128d, __m128d, __m128d, 1)
-test_3 (_mm_fnmsub_round_ss, __m128, __m128, __m128, __m128, 1)
+test_3 (_mm_fmadd_round_sd, __m128d, __m128d, __m128d, __m128d, 9)
+test_3 (_mm_fmadd_round_ss, __m128, __m128, __m128, __m128, 9)
+test_3 (_mm_fmsub_round_sd, __m128d, __m128d, __m128d, __m128d, 9)
+test_3 (_mm_fmsub_round_ss, __m128, __m128, __m128, __m128, 9)
+test_3 (_mm_fnmadd_round_sd, __m128d, __m128d, __m128d, __m128d, 9)
+test_3 (_mm_fnmadd_round_ss, __m128, __m128, __m128, __m128, 9)
+test_3 (_mm_fnmsub_round_sd, __m128d, __m128d, __m128d, __m128d, 9)
+test_3 (_mm_fnmsub_round_ss, __m128, __m128, __m128, __m128, 9)
test_3 (_mm_mask_cmp_sd_mask, __mmask8, __mmask8, __m128d, __m128d, 1)
test_3 (_mm_mask_cmp_ss_mask, __mmask8, __mmask8, __m128, __m128, 1)
test_3v (_mm512_i32scatter_epi32, void *, __m512i, __m512i, 1)
@@ -421,42 +421,42 @@ test_3v (_mm512_i64scatter_epi32, void *, __m512i, __m256i, 1)
test_3v (_mm512_i64scatter_epi64, void *, __m512i, __m512i, 1)
test_3v (_mm512_i64scatter_pd, void *, __m512i, __m512d, 1)
test_3v (_mm512_i64scatter_ps, void *, __m512i, __m256, 1)
-test_3x (_mm512_mask_roundscale_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 5)
-test_3x (_mm512_mask_roundscale_round_ps, __m512, __m512, __mmask16, __m512, 1, 5)
-test_3x (_mm_fixupimm_round_sd, __m128d, __m128d, __m128d, __m128i, 1, 5)
-test_3x (_mm_fixupimm_round_ss, __m128, __m128, __m128, __m128i, 1, 5)
-test_3x (_mm_mask_cmp_round_sd_mask, __mmask8, __mmask8, __m128d, __m128d, 1, 5)
-test_3x (_mm_mask_cmp_round_ss_mask, __mmask8, __mmask8, __m128, __m128, 1, 5)
-test_4 (_mm512_mask3_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmaddsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmsubadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fnmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fnmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask_add_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_add_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_3x (_mm512_mask_roundscale_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 8)
+test_3x (_mm512_mask_roundscale_round_ps, __m512, __m512, __mmask16, __m512, 1, 8)
+test_3x (_mm_fixupimm_round_sd, __m128d, __m128d, __m128d, __m128i, 1, 8)
+test_3x (_mm_fixupimm_round_ss, __m128, __m128, __m128, __m128i, 1, 8)
+test_3x (_mm_mask_cmp_round_sd_mask, __mmask8, __mmask8, __m128d, __m128d, 1, 8)
+test_3x (_mm_mask_cmp_round_ss_mask, __mmask8, __mmask8, __m128, __m128, 1, 8)
+test_4 (_mm512_mask3_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmaddsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmsubadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fnmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fnmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask_add_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_add_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_alignr_epi32, __m512i, __m512i, __mmask16, __m512i, __m512i, 1)
test_4 (_mm512_mask_alignr_epi64, __m512i, __m512i, __mmask8, __m512i, __m512i, 1)
-test_4 (_mm512_mask_div_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_div_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmaddsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmaddsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmsubadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmsubadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fnmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fnmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fnmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fnmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_4 (_mm512_mask_div_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_div_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmaddsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmaddsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmsubadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmsubadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fnmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fnmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fnmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fnmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_i32gather_epi32, __m512i, __m512i, __mmask16, __m512i, void const *, 1)
test_4 (_mm512_mask_i32gather_epi64, __m512i, __m512i, __mmask8, __m256i, void const *, 1)
test_4 (_mm512_mask_i32gather_pd, __m512d, __m512d, __mmask8, __m256i, void const *, 1)
@@ -469,36 +469,36 @@ test_4 (_mm512_mask_insertf32x4, __m512, __m512, __mmask16, __m512, __m128, 1)
test_4 (_mm512_mask_insertf64x4, __m512d, __m512d, __mmask8, __m512d, __m256d, 1)
test_4 (_mm512_mask_inserti32x4, __m512i, __m512i, __mmask16, __m512i, __m128i, 1)
test_4 (_mm512_mask_inserti64x4, __m512i, __m512i, __mmask8, __m512i, __m256i, 1)
-test_4 (_mm512_mask_max_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 5)
-test_4 (_mm512_mask_max_round_ps, __m512, __m512, __mmask16, __m512, __m512, 5)
-test_4 (_mm512_mask_min_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 5)
-test_4 (_mm512_mask_min_round_ps, __m512, __m512, __mmask16, __m512, __m512, 5)
-test_4 (_mm512_mask_mul_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_mul_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_scalef_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_scalef_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_4 (_mm512_mask_max_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 8)
+test_4 (_mm512_mask_max_round_ps, __m512, __m512, __mmask16, __m512, __m512, 8)
+test_4 (_mm512_mask_min_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 8)
+test_4 (_mm512_mask_min_round_ps, __m512, __m512, __mmask16, __m512, __m512, 8)
+test_4 (_mm512_mask_mul_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_mul_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_scalef_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_scalef_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_shuffle_f32x4, __m512, __m512, __mmask16, __m512, __m512, 1)
test_4 (_mm512_mask_shuffle_f64x2, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
test_4 (_mm512_mask_shuffle_i32x4, __m512i, __m512i, __mmask16, __m512i, __m512i, 1)
test_4 (_mm512_mask_shuffle_i64x2, __m512i, __m512i, __mmask8, __m512i, __m512i, 1)
test_4 (_mm512_mask_shuffle_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
test_4 (_mm512_mask_shuffle_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_sub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_sub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_4 (_mm512_mask_sub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_sub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_ternarylogic_epi32, __m512i, __m512i, __mmask16, __m512i, __m512i, 1)
test_4 (_mm512_mask_ternarylogic_epi64, __m512i, __m512i, __mmask8, __m512i, __m512i, 1)
-test_4 (_mm512_maskz_fmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fmaddsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmaddsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fmsubadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmsubadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fnmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fnmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fnmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fnmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
+test_4 (_mm512_maskz_fmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fmaddsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmaddsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fmsubadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmsubadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fnmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fnmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fnmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fnmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
test_4 (_mm512_maskz_ternarylogic_epi32, __m512i, __mmask16, __m512i, __m512i, __m512i, 1)
test_4 (_mm512_maskz_ternarylogic_epi64, __m512i, __mmask8, __m512i, __m512i, __m512i, 1)
test_4v (_mm512_mask_i32scatter_epi32, void *, __mmask16, __m512i, __m512i, 1)
@@ -509,14 +509,14 @@ test_4v (_mm512_mask_i64scatter_epi32, void *, __mmask8, __m512i, __m256i, 1)
test_4v (_mm512_mask_i64scatter_epi64, void *, __mmask8, __m512i, __m512i, 1)
test_4v (_mm512_mask_i64scatter_pd, void *, __mmask8, __m512i, __m512d, 1)
test_4v (_mm512_mask_i64scatter_ps, void *, __mmask8, __m512i, __m256, 1)
-test_4x (_mm512_mask_fixupimm_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512i, 1, 5)
-test_4x (_mm512_mask_fixupimm_round_ps, __m512, __m512, __mmask16, __m512, __m512i, 1, 5)
-test_4x (_mm512_maskz_fixupimm_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512i, 1, 5)
-test_4x (_mm512_maskz_fixupimm_round_ps, __m512, __mmask16, __m512, __m512, __m512i, 1, 5)
-test_4x (_mm_mask_fixupimm_round_sd, __m128d, __m128d, __mmask8, __m128d, __m128i, 1, 5)
-test_4x (_mm_mask_fixupimm_round_ss, __m128, __m128, __mmask8, __m128, __m128i, 1, 5)
-test_4x (_mm_maskz_fixupimm_round_sd, __m128d, __mmask8, __m128d, __m128d, __m128i, 1, 5)
-test_4x (_mm_maskz_fixupimm_round_ss, __m128, __mmask8, __m128, __m128, __m128i, 1, 5)
+test_4x (_mm512_mask_fixupimm_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512i, 1, 8)
+test_4x (_mm512_mask_fixupimm_round_ps, __m512, __m512, __mmask16, __m512, __m512i, 1, 8)
+test_4x (_mm512_maskz_fixupimm_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512i, 1, 8)
+test_4x (_mm512_maskz_fixupimm_round_ps, __m512, __mmask16, __m512, __m512, __m512i, 1, 8)
+test_4x (_mm_mask_fixupimm_round_sd, __m128d, __m128d, __mmask8, __m128d, __m128i, 1, 8)
+test_4x (_mm_mask_fixupimm_round_ss, __m128, __m128, __mmask8, __m128, __m128i, 1, 8)
+test_4x (_mm_maskz_fixupimm_round_sd, __m128d, __mmask8, __m128d, __m128d, __m128i, 1, 8)
+test_4x (_mm_maskz_fixupimm_round_ss, __m128, __mmask8, __m128, __m128, __m128i, 1, 8)
/* avx512pfintrin.h */
test_3vx (_mm512_mask_prefetch_i32gather_ps, __m512i, __mmask16, void const *, 1, 1)
@@ -529,24 +529,24 @@ test_3vx (_mm512_mask_prefetch_i64gather_pd, __m512i, __mmask8, void const *, 1,
test_3vx (_mm512_mask_prefetch_i64scatter_pd, void const *, __mmask8, __m512i, 1, 1)
/* avx512erintrin.h */
-test_1 (_mm512_exp2a23_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_exp2a23_round_ps, __m512, __m512, 5)
-test_1 (_mm512_rcp28_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_rcp28_round_ps, __m512, __m512, 5)
-test_1 (_mm512_rsqrt28_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_rsqrt28_round_ps, __m512, __m512, 5)
-test_2 (_mm512_maskz_exp2a23_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_exp2a23_round_ps, __m512, __mmask16, __m512, 5)
-test_2 (_mm512_maskz_rcp28_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_rcp28_round_ps, __m512, __mmask16, __m512, 5)
-test_2 (_mm512_maskz_rsqrt28_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_rsqrt28_round_ps, __m512, __mmask16, __m512, 5)
-test_3 (_mm512_mask_exp2a23_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_exp2a23_round_ps, __m512, __m512, __mmask16, __m512, 5)
-test_3 (_mm512_mask_rcp28_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_rcp28_round_ps, __m512, __m512, __mmask16, __m512, 5)
-test_3 (_mm512_mask_rsqrt28_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_rsqrt28_round_ps, __m512, __m512, __mmask16, __m512, 5)
+test_1 (_mm512_exp2a23_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_exp2a23_round_ps, __m512, __m512, 8)
+test_1 (_mm512_rcp28_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_rcp28_round_ps, __m512, __m512, 8)
+test_1 (_mm512_rsqrt28_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_rsqrt28_round_ps, __m512, __m512, 8)
+test_2 (_mm512_maskz_exp2a23_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_exp2a23_round_ps, __m512, __mmask16, __m512, 8)
+test_2 (_mm512_maskz_rcp28_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_rcp28_round_ps, __m512, __mmask16, __m512, 8)
+test_2 (_mm512_maskz_rsqrt28_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_rsqrt28_round_ps, __m512, __mmask16, __m512, 8)
+test_3 (_mm512_mask_exp2a23_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_exp2a23_round_ps, __m512, __m512, __mmask16, __m512, 8)
+test_3 (_mm512_mask_rcp28_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_rcp28_round_ps, __m512, __m512, __mmask16, __m512, 8)
+test_3 (_mm512_mask_rsqrt28_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_rsqrt28_round_ps, __m512, __m512, __mmask16, __m512, 8)
/* shaintrin.h */
test_2 (_mm_sha1rnds4_epu32, __m128i, __m128i, __m128i, 1)
@@ -556,10 +556,10 @@ test_1 (_mm_aeskeygenassist_si128, __m128i, __m128i, 1)
test_2 (_mm_clmulepi64_si128, __m128i, __m128i, __m128i, 1)
/* smmintrin.h */
-test_1 (_mm_round_pd, __m128d, __m128d, 1)
-test_1 (_mm_round_ps, __m128, __m128, 1)
-test_2 (_mm_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_round_ss, __m128, __m128, __m128, 1)
+test_1 (_mm_round_pd, __m128d, __m128d, 9)
+test_1 (_mm_round_ps, __m128, __m128, 9)
+test_2 (_mm_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_round_ss, __m128, __m128, __m128, 9)
test_2 (_mm_blend_epi16, __m128i, __m128i, __m128i, 1)
test_2 (_mm_blend_ps, __m128, __m128, __m128, 1)
@@ -174,10 +174,10 @@ test_2x (_mm_inserti_si64, __m128i, __m128i, __m128i, 1, 1)
#endif
#include <nmmintrin.h>
/* smmintrin.h (SSE4.2). */
-test_1 (_mm_round_pd, __m128d, __m128d, 1)
-test_1 (_mm_round_ps, __m128, __m128, 1)
-test_2 (_mm_round_sd, __m128d, __m128d, __m128d, 1)
-test_2 (_mm_round_ss, __m128, __m128, __m128, 1)
+test_1 (_mm_round_pd, __m128d, __m128d, 9)
+test_1 (_mm_round_ps, __m128, __m128, 9)
+test_2 (_mm_round_sd, __m128d, __m128d, __m128d, 9)
+test_2 (_mm_round_ss, __m128, __m128, __m128, 9)
test_2 (_mm_blend_epi16, __m128i, __m128i, __m128i, 1)
test_2 (_mm_blend_ps, __m128, __m128, __m128, 1)
@@ -258,8 +258,8 @@ test_2 (_mm256_insert_epi32, __m256i, __m256i, int, 3)
#ifdef __x86_64__
test_2 (_mm256_insert_epi64, __m256i, __m256i, long long, 1)
#endif
-test_1 (_mm256_round_pd, __m256d, __m256d, 1)
-test_1 (_mm256_round_ps, __m256, __m256, 1)
+test_1 (_mm256_round_pd, __m256d, __m256d, 9)
+test_1 (_mm256_round_ps, __m256, __m256, 9)
/* avx2intrin.h */
test_2 ( _mm256_mpsadbw_epu8, __m256i, __m256i, __m256i, 1)
@@ -298,28 +298,28 @@ test_2 ( _mm256_i64gather_epi32, __m128i, int const *, __m256i, 1)
test_0 ( _xabort, void, 1)
/* avx512fintrin.h */
-test_1 (_mm512_cvt_roundepi32_ps, __m512, __m512i, 1)
-test_1 (_mm512_cvt_roundepu32_ps, __m512, __m512i, 1)
-test_1 (_mm512_cvt_roundpd_epi32, __m256i, __m512d, 1)
-test_1 (_mm512_cvt_roundpd_epu32, __m256i, __m512d, 1)
-test_1 (_mm512_cvt_roundpd_ps, __m256, __m512d, 1)
-test_1 (_mm512_cvt_roundph_ps, __m512, __m256i, 5)
-test_1 (_mm512_cvt_roundps_epi32, __m512i, __m512, 1)
-test_1 (_mm512_cvt_roundps_epu32, __m512i, __m512, 1)
-test_1 (_mm512_cvt_roundps_pd, __m512d, __m256, 5)
+test_1 (_mm512_cvt_roundepi32_ps, __m512, __m512i, 9)
+test_1 (_mm512_cvt_roundepu32_ps, __m512, __m512i, 9)
+test_1 (_mm512_cvt_roundpd_epi32, __m256i, __m512d, 9)
+test_1 (_mm512_cvt_roundpd_epu32, __m256i, __m512d, 9)
+test_1 (_mm512_cvt_roundpd_ps, __m256, __m512d, 9)
+test_1 (_mm512_cvt_roundph_ps, __m512, __m256i, 8)
+test_1 (_mm512_cvt_roundps_epi32, __m512i, __m512, 9)
+test_1 (_mm512_cvt_roundps_epu32, __m512i, __m512, 9)
+test_1 (_mm512_cvt_roundps_pd, __m512d, __m256, 8)
test_1 (_mm512_cvtps_ph, __m256i, __m512, 1)
-test_1 (_mm512_cvtt_roundpd_epi32, __m256i, __m512d, 5)
-test_1 (_mm512_cvtt_roundpd_epu32, __m256i, __m512d, 5)
-test_1 (_mm512_cvtt_roundps_epi32, __m512i, __m512, 5)
-test_1 (_mm512_cvtt_roundps_epu32, __m512i, __m512, 5)
+test_1 (_mm512_cvtt_roundpd_epi32, __m256i, __m512d, 8)
+test_1 (_mm512_cvtt_roundpd_epu32, __m256i, __m512d, 8)
+test_1 (_mm512_cvtt_roundps_epi32, __m512i, __m512, 8)
+test_1 (_mm512_cvtt_roundps_epu32, __m512i, __m512, 8)
test_1 (_mm512_extractf32x4_ps, __m128, __m512, 1)
test_1 (_mm512_extractf64x4_pd, __m256d, __m512d, 1)
test_1 (_mm512_extracti32x4_epi32, __m128i, __m512i, 1)
test_1 (_mm512_extracti64x4_epi64, __m256i, __m512i, 1)
-test_1 (_mm512_getexp_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_getexp_round_ps, __m512, __m512, 5)
-test_1y (_mm512_getmant_round_pd, __m512d, __m512d, 1, 1, 5)
-test_1y (_mm512_getmant_round_ps, __m512, __m512, 1, 1, 5)
+test_1 (_mm512_getexp_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_getexp_round_ps, __m512, __m512, 8)
+test_1y (_mm512_getmant_round_pd, __m512d, __m512d, 1, 1, 8)
+test_1y (_mm512_getmant_round_ps, __m512, __m512, 1, 1, 8)
test_1 (_mm512_permute_pd, __m512d, __m512d, 1)
test_1 (_mm512_permute_ps, __m512, __m512, 1)
test_1 (_mm512_permutex_epi64, __m512i, __m512i, 1)
@@ -331,25 +331,25 @@ test_1 (_mm512_ror_epi64, __m512i, __m512i, 1)
test_1 (_mm512_shuffle_epi32, __m512i, __m512i, 1)
test_1 (_mm512_slli_epi32, __m512i, __m512i, 1)
test_1 (_mm512_slli_epi64, __m512i, __m512i, 1)
-test_1 (_mm512_sqrt_round_pd, __m512d, __m512d, 1)
-test_1 (_mm512_sqrt_round_ps, __m512, __m512, 1)
+test_1 (_mm512_sqrt_round_pd, __m512d, __m512d, 9)
+test_1 (_mm512_sqrt_round_ps, __m512, __m512, 9)
test_1 (_mm512_srai_epi32, __m512i, __m512i, 1)
test_1 (_mm512_srai_epi64, __m512i, __m512i, 1)
test_1 (_mm512_srli_epi32, __m512i, __m512i, 1)
test_1 (_mm512_srli_epi64, __m512i, __m512i, 1)
-test_1 (_mm_cvt_roundsd_i32, int, __m128d, 1)
-test_1 (_mm_cvt_roundsd_u32, unsigned, __m128d, 1)
-test_1 (_mm_cvt_roundss_i32, int, __m128, 1)
-test_1 (_mm_cvt_roundss_u32, unsigned, __m128, 1)
-test_1 (_mm_cvtt_roundsd_i32, int, __m128d, 5)
-test_1 (_mm_cvtt_roundsd_u32, unsigned, __m128d, 5)
-test_1 (_mm_cvtt_roundss_i32, int, __m128, 5)
-test_1 (_mm_cvtt_roundss_u32, unsigned, __m128, 5)
+test_1 (_mm_cvt_roundsd_i32, int, __m128d, 9)
+test_1 (_mm_cvt_roundsd_u32, unsigned, __m128d, 9)
+test_1 (_mm_cvt_roundss_i32, int, __m128, 9)
+test_1 (_mm_cvt_roundss_u32, unsigned, __m128, 9)
+test_1 (_mm_cvtt_roundsd_i32, int, __m128d, 8)
+test_1 (_mm_cvtt_roundsd_u32, unsigned, __m128d, 8)
+test_1 (_mm_cvtt_roundss_i32, int, __m128, 8)
+test_1 (_mm_cvtt_roundss_u32, unsigned, __m128, 8)
test_1x (_mm512_getmant_pd, __m512d, __m512d, 1, 1)
test_1x (_mm512_getmant_ps, __m512, __m512, 1, 1)
-test_1x (_mm_cvt_roundi32_ss, __m128, __m128, 1, 1)
-test_2 (_mm512_add_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_add_round_ps, __m512, __m512, __m512, 1)
+test_1x (_mm_cvt_roundi32_ss, __m128, __m128, 1, 9)
+test_2 (_mm512_add_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_add_round_ps, __m512, __m512, __m512, 9)
test_2 (_mm512_alignr_epi32, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_alignr_epi64, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_cmp_epi32_mask, __mmask16, __m512i, __m512i, 1)
@@ -358,8 +358,8 @@ test_2 (_mm512_cmp_epu32_mask, __mmask16, __m512i, __m512i, 1)
test_2 (_mm512_cmp_epu64_mask, __mmask8, __m512i, __m512i, 1)
test_2 (_mm512_cmp_pd_mask, __mmask8, __m512d, __m512d, 1)
test_2 (_mm512_cmp_ps_mask, __mmask16, __m512, __m512, 1)
-test_2 (_mm512_div_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_div_round_ps, __m512, __m512, __m512, 1)
+test_2 (_mm512_div_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_div_round_ps, __m512, __m512, __m512, 9)
test_2 (_mm512_i32gather_epi32, __m512i, __m512i, void const *, 1)
test_2 (_mm512_i32gather_epi64, __m512i, __m256i, void const *, 1)
test_2 (_mm512_i32gather_pd, __m512d, __m256i, void const *, 1)
@@ -372,28 +372,28 @@ test_2 (_mm512_insertf32x4, __m512, __m512, __m128, 1)
test_2 (_mm512_insertf64x4, __m512d, __m512d, __m256d, 1)
test_2 (_mm512_inserti32x4, __m512i, __m512i, __m128i, 1)
test_2 (_mm512_inserti64x4, __m512i, __m512i, __m256i, 1)
-test_2 (_mm512_maskz_cvt_roundepi32_ps, __m512, __mmask16, __m512i, 1)
-test_2 (_mm512_maskz_cvt_roundepu32_ps, __m512, __mmask16, __m512i, 1)
-test_2 (_mm512_maskz_cvt_roundpd_epi32, __m256i, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_cvt_roundpd_epu32, __m256i, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_cvt_roundpd_ps, __m256, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_cvt_roundph_ps, __m512, __mmask16, __m256i, 5)
-test_2 (_mm512_maskz_cvt_roundps_epi32, __m512i, __mmask16, __m512, 1)
-test_2 (_mm512_maskz_cvt_roundps_epu32, __m512i, __mmask16, __m512, 1)
-test_2 (_mm512_maskz_cvt_roundps_pd, __m512d, __mmask8, __m256, 5)
+test_2 (_mm512_maskz_cvt_roundepi32_ps, __m512, __mmask16, __m512i, 9)
+test_2 (_mm512_maskz_cvt_roundepu32_ps, __m512, __mmask16, __m512i, 9)
+test_2 (_mm512_maskz_cvt_roundpd_epi32, __m256i, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_cvt_roundpd_epu32, __m256i, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_cvt_roundpd_ps, __m256, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_cvt_roundph_ps, __m512, __mmask16, __m256i, 8)
+test_2 (_mm512_maskz_cvt_roundps_epi32, __m512i, __mmask16, __m512, 9)
+test_2 (_mm512_maskz_cvt_roundps_epu32, __m512i, __mmask16, __m512, 9)
+test_2 (_mm512_maskz_cvt_roundps_pd, __m512d, __mmask8, __m256, 8)
test_2 (_mm512_maskz_cvtps_ph, __m256i, __mmask16, __m512, 1)
-test_2 (_mm512_maskz_cvtt_roundpd_epi32, __m256i, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_cvtt_roundpd_epu32, __m256i, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_cvtt_roundps_epi32, __m512i, __mmask16, __m512, 5)
-test_2 (_mm512_maskz_cvtt_roundps_epu32, __m512i, __mmask16, __m512, 5)
+test_2 (_mm512_maskz_cvtt_roundpd_epi32, __m256i, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_cvtt_roundpd_epu32, __m256i, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_cvtt_roundps_epi32, __m512i, __mmask16, __m512, 8)
+test_2 (_mm512_maskz_cvtt_roundps_epu32, __m512i, __mmask16, __m512, 8)
test_2 (_mm512_maskz_extractf32x4_ps, __m128, __mmask8, __m512, 1)
test_2 (_mm512_maskz_extractf64x4_pd, __m256d, __mmask8, __m512d, 1)
test_2 (_mm512_maskz_extracti32x4_epi32, __m128i, __mmask8, __m512i, 1)
test_2 (_mm512_maskz_extracti64x4_epi64, __m256i, __mmask8, __m512i, 1)
-test_2 (_mm512_maskz_getexp_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_getexp_round_ps, __m512, __mmask16, __m512, 5)
-test_2y (_mm512_maskz_getmant_round_pd, __m512d, __mmask8, __m512d, 1, 1, 5)
-test_2y (_mm512_maskz_getmant_round_ps, __m512, __mmask16, __m512, 1, 1, 5)
+test_2 (_mm512_maskz_getexp_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_getexp_round_ps, __m512, __mmask16, __m512, 8)
+test_2y (_mm512_maskz_getmant_round_pd, __m512d, __mmask8, __m512d, 1, 1, 8)
+test_2y (_mm512_maskz_getmant_round_ps, __m512, __mmask16, __m512, 1, 1, 8)
test_2 (_mm512_maskz_permute_pd, __m512d, __mmask8, __m512d, 1)
test_2 (_mm512_maskz_permute_ps, __m512, __mmask16, __m512, 1)
test_2 (_mm512_maskz_permutex_epi64, __m512i, __mmask8, __m512i, 1)
@@ -405,87 +405,87 @@ test_2 (_mm512_maskz_ror_epi64, __m512i, __mmask8, __m512i, 1)
test_2 (_mm512_maskz_shuffle_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_slli_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_slli_epi64, __m512i, __mmask8, __m512i, 1)
-test_2 (_mm512_maskz_sqrt_round_pd, __m512d, __mmask8, __m512d, 1)
-test_2 (_mm512_maskz_sqrt_round_ps, __m512, __mmask16, __m512, 1)
+test_2 (_mm512_maskz_sqrt_round_pd, __m512d, __mmask8, __m512d, 9)
+test_2 (_mm512_maskz_sqrt_round_ps, __m512, __mmask16, __m512, 9)
test_2 (_mm512_maskz_srai_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_srai_epi64, __m512i, __mmask8, __m512i, 1)
test_2 (_mm512_maskz_srli_epi32, __m512i, __mmask16, __m512i, 1)
test_2 (_mm512_maskz_srli_epi64, __m512i, __mmask8, __m512i, 1)
-test_2 (_mm512_max_round_pd, __m512d, __m512d, __m512d, 5)
-test_2 (_mm512_max_round_ps, __m512, __m512, __m512, 5)
-test_2 (_mm512_min_round_pd, __m512d, __m512d, __m512d, 5)
-test_2 (_mm512_min_round_ps, __m512, __m512, __m512, 5)
-test_2 (_mm512_mul_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_mul_round_ps, __m512, __m512, __m512, 1)
-test_2 (_mm512_scalef_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_scalef_round_ps, __m512, __m512, __m512, 1)
+test_2 (_mm512_max_round_pd, __m512d, __m512d, __m512d, 8)
+test_2 (_mm512_max_round_ps, __m512, __m512, __m512, 8)
+test_2 (_mm512_min_round_pd, __m512d, __m512d, __m512d, 8)
+test_2 (_mm512_min_round_ps, __m512, __m512, __m512, 8)
+test_2 (_mm512_mul_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_mul_round_ps, __m512, __m512, __m512, 9)
+test_2 (_mm512_scalef_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_scalef_round_ps, __m512, __m512, __m512, 9)
test_2 (_mm512_shuffle_f32x4, __m512, __m512, __m512, 1)
test_2 (_mm512_shuffle_f64x2, __m512d, __m512d, __m512d, 1)
test_2 (_mm512_shuffle_i32x4, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_shuffle_i64x2, __m512i, __m512i, __m512i, 1)
test_2 (_mm512_shuffle_pd, __m512d, __m512d, __m512d, 1)
test_2 (_mm512_shuffle_ps, __m512, __m512, __m512, 1)
-test_2 (_mm512_sub_round_pd, __m512d, __m512d, __m512d, 1)
-test_2 (_mm512_sub_round_ps, __m512, __m512, __m512, 1)
+test_2 (_mm512_sub_round_pd, __m512d, __m512d, __m512d, 9)
+test_2 (_mm512_sub_round_ps, __m512, __m512, __m512, 9)
test_2 (_mm_cmp_sd_mask, __mmask8, __m128d, __m128d, 1)
test_2 (_mm_cmp_ss_mask, __mmask8, __m128, __m128, 1)
#ifdef __x86_64__
-test_2 (_mm_cvt_roundi64_sd, __m128d, __m128d, long long, 1)
-test_2 (_mm_cvt_roundi64_ss, __m128, __m128, long long, 1)
+test_2 (_mm_cvt_roundi64_sd, __m128d, __m128d, long long, 9)
+test_2 (_mm_cvt_roundi64_ss, __m128, __m128, long long, 9)
#endif
-test_2 (_mm_cvt_roundu32_ss, __m128, __m128, unsigned, 1)
+test_2 (_mm_cvt_roundu32_ss, __m128, __m128, unsigned, 9)
#ifdef __x86_64__
-test_2 (_mm_cvt_roundu64_sd, __m128d, __m128d, unsigned long long, 1)
-test_2 (_mm_cvt_roundu64_ss, __m128, __m128, unsigned long long, 1)
+test_2 (_mm_cvt_roundu64_sd, __m128d, __m128d, unsigned long long, 9)
+test_2 (_mm_cvt_roundu64_ss, __m128, __m128, unsigned long long, 9)
#endif
-test_2x (_mm512_cmp_round_pd_mask, __mmask8, __m512d, __m512d, 1, 5)
-test_2x (_mm512_cmp_round_ps_mask, __mmask16, __m512, __m512, 1, 5)
-test_2x (_mm512_maskz_roundscale_round_pd, __m512d, __mmask8, __m512d, 1, 5)
-test_2x (_mm512_maskz_roundscale_round_ps, __m512, __mmask16, __m512, 1, 5)
-test_2x (_mm_cmp_round_sd_mask, __mmask8, __m128d, __m128d, 1, 5)
-test_2x (_mm_cmp_round_ss_mask, __mmask8, __m128, __m128, 1, 5)
-test_2x (_mm_comi_round_sd, int, __m128d, __m128d, 1, 5)
-test_2x (_mm_comi_round_ss, int, __m128, __m128, 1, 5)
-test_3 (_mm512_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmadd_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmaddsub_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmsub_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fmsubadd_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fnmadd_round_ps, __m512, __m512, __m512, __m512, 1)
-test_3 (_mm512_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 1)
-test_3 (_mm512_fnmsub_round_ps, __m512, __m512, __m512, __m512, 1)
+test_2x (_mm512_cmp_round_pd_mask, __mmask8, __m512d, __m512d, 1, 8)
+test_2x (_mm512_cmp_round_ps_mask, __mmask16, __m512, __m512, 1, 8)
+test_2x (_mm512_maskz_roundscale_round_pd, __m512d, __mmask8, __m512d, 1, 8)
+test_2x (_mm512_maskz_roundscale_round_ps, __m512, __mmask16, __m512, 1, 8)
+test_2x (_mm_cmp_round_sd_mask, __mmask8, __m128d, __m128d, 1, 8)
+test_2x (_mm_cmp_round_ss_mask, __mmask8, __m128, __m128, 1, 8)
+test_2x (_mm_comi_round_sd, int, __m128d, __m128d, 1, 8)
+test_2x (_mm_comi_round_ss, int, __m128, __m128, 1, 8)
+test_3 (_mm512_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmadd_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmaddsub_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmsub_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fmsubadd_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fnmadd_round_ps, __m512, __m512, __m512, __m512, 9)
+test_3 (_mm512_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, 9)
+test_3 (_mm512_fnmsub_round_ps, __m512, __m512, __m512, __m512, 9)
test_3 (_mm512_mask_cmp_epi32_mask, __mmask16, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_epi64_mask, __mmask8, __mmask8, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_epu32_mask, __mmask16, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_epu64_mask, __mmask8, __mmask8, __m512i, __m512i, 1)
test_3 (_mm512_mask_cmp_pd_mask, __mmask8, __mmask8, __m512d, __m512d, 1)
test_3 (_mm512_mask_cmp_ps_mask, __mmask16, __mmask16, __m512, __m512, 1)
-test_3 (_mm512_mask_cvt_roundepi32_ps, __m512, __m512, __mmask16, __m512i, 1)
-test_3 (_mm512_mask_cvt_roundepu32_ps, __m512, __m512, __mmask16, __m512i, 1)
-test_3 (_mm512_mask_cvt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_cvt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_cvt_roundpd_ps, __m256, __m256, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_cvt_roundph_ps, __m512, __m512, __mmask16, __m256i, 5)
-test_3 (_mm512_mask_cvt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 1)
-test_3 (_mm512_mask_cvt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 1)
-test_3 (_mm512_mask_cvt_roundps_pd, __m512d, __m512d, __mmask8, __m256, 5)
+test_3 (_mm512_mask_cvt_roundepi32_ps, __m512, __m512, __mmask16, __m512i, 9)
+test_3 (_mm512_mask_cvt_roundepu32_ps, __m512, __m512, __mmask16, __m512i, 9)
+test_3 (_mm512_mask_cvt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_cvt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_cvt_roundpd_ps, __m256, __m256, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_cvt_roundph_ps, __m512, __m512, __mmask16, __m256i, 8)
+test_3 (_mm512_mask_cvt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 9)
+test_3 (_mm512_mask_cvt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 9)
+test_3 (_mm512_mask_cvt_roundps_pd, __m512d, __m512d, __mmask8, __m256, 8)
test_3 (_mm512_mask_cvtps_ph, __m256i, __m256i, __mmask16, __m512, 1)
-test_3 (_mm512_mask_cvtt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_cvtt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_cvtt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 5)
-test_3 (_mm512_mask_cvtt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 5)
+test_3 (_mm512_mask_cvtt_roundpd_epi32, __m256i, __m256i, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_cvtt_roundpd_epu32, __m256i, __m256i, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_cvtt_roundps_epi32, __m512i, __m512i, __mmask16, __m512, 8)
+test_3 (_mm512_mask_cvtt_roundps_epu32, __m512i, __m512i, __mmask16, __m512, 8)
test_3 (_mm512_mask_extractf32x4_ps, __m128, __m128, __mmask8, __m512, 1)
test_3 (_mm512_mask_extractf64x4_pd, __m256d, __m256d, __mmask8, __m512d, 1)
test_3 (_mm512_mask_extracti32x4_epi32, __m128i, __m128i, __mmask8, __m512i, 1)
test_3 (_mm512_mask_extracti64x4_epi64, __m256i, __m256i, __mmask8, __m512i, 1)
-test_3 (_mm512_mask_getexp_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_getexp_round_ps, __m512, __m512, __mmask16, __m512, 5)
-test_3y (_mm512_mask_getmant_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 1, 5)
-test_3y (_mm512_mask_getmant_round_ps, __m512, __m512, __mmask16, __m512, 1, 1, 5)
+test_3 (_mm512_mask_getexp_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_getexp_round_ps, __m512, __m512, __mmask16, __m512, 8)
+test_3y (_mm512_mask_getmant_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 1, 8)
+test_3y (_mm512_mask_getmant_round_ps, __m512, __m512, __mmask16, __m512, 1, 1, 8)
test_3 (_mm512_mask_permute_pd, __m512d, __m512d, __mmask8, __m512d, 1)
test_3 (_mm512_mask_permute_ps, __m512, __m512, __mmask16, __m512, 1)
test_3 (_mm512_mask_permutex_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
@@ -497,38 +497,38 @@ test_3 (_mm512_mask_ror_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
test_3 (_mm512_mask_shuffle_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_slli_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_slli_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
-test_3 (_mm512_mask_sqrt_round_pd, __m512d, __m512d, __mmask8, __m512d, 1)
-test_3 (_mm512_mask_sqrt_round_ps, __m512, __m512, __mmask16, __m512, 1)
+test_3 (_mm512_mask_sqrt_round_pd, __m512d, __m512d, __mmask8, __m512d, 9)
+test_3 (_mm512_mask_sqrt_round_ps, __m512, __m512, __mmask16, __m512, 9)
test_3 (_mm512_mask_srai_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_srai_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
test_3 (_mm512_mask_srli_epi32, __m512i, __m512i, __mmask16, __m512i, 1)
test_3 (_mm512_mask_srli_epi64, __m512i, __m512i, __mmask8, __m512i, 1)
-test_3 (_mm512_maskz_add_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_add_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_add_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_add_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_maskz_alignr_epi32, __m512i, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_maskz_alignr_epi64, __m512i, __mmask8, __m512i, __m512i, 1)
-test_3 (_mm512_maskz_div_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_div_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_div_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_div_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_maskz_insertf32x4, __m512, __mmask16, __m512, __m128, 1)
test_3 (_mm512_maskz_insertf64x4, __m512d, __mmask8, __m512d, __m256d, 1)
test_3 (_mm512_maskz_inserti32x4, __m512i, __mmask16, __m512i, __m128i, 1)
test_3 (_mm512_maskz_inserti64x4, __m512i, __mmask8, __m512i, __m256i, 1)
-test_3 (_mm512_maskz_max_round_pd, __m512d, __mmask8, __m512d, __m512d, 5)
-test_3 (_mm512_maskz_max_round_ps, __m512, __mmask16, __m512, __m512, 5)
-test_3 (_mm512_maskz_min_round_pd, __m512d, __mmask8, __m512d, __m512d, 5)
-test_3 (_mm512_maskz_min_round_ps, __m512, __mmask16, __m512, __m512, 5)
-test_3 (_mm512_maskz_mul_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_mul_round_ps, __m512, __mmask16, __m512, __m512, 1)
-test_3 (_mm512_maskz_scalef_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_scalef_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_max_round_pd, __m512d, __mmask8, __m512d, __m512d, 8)
+test_3 (_mm512_maskz_max_round_ps, __m512, __mmask16, __m512, __m512, 8)
+test_3 (_mm512_maskz_min_round_pd, __m512d, __mmask8, __m512d, __m512d, 8)
+test_3 (_mm512_maskz_min_round_ps, __m512, __mmask16, __m512, __m512, 8)
+test_3 (_mm512_maskz_mul_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_mul_round_ps, __m512, __mmask16, __m512, __m512, 9)
+test_3 (_mm512_maskz_scalef_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_scalef_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_maskz_shuffle_f32x4, __m512, __mmask16, __m512, __m512, 1)
test_3 (_mm512_maskz_shuffle_f64x2, __m512d, __mmask8, __m512d, __m512d, 1)
test_3 (_mm512_maskz_shuffle_i32x4, __m512i, __mmask16, __m512i, __m512i, 1)
test_3 (_mm512_maskz_shuffle_i64x2, __m512i, __mmask8, __m512i, __m512i, 1)
test_3 (_mm512_maskz_shuffle_pd, __m512d, __mmask8, __m512d, __m512d, 1)
test_3 (_mm512_maskz_shuffle_ps, __m512, __mmask16, __m512, __m512, 1)
-test_3 (_mm512_maskz_sub_round_pd, __m512d, __mmask8, __m512d, __m512d, 1)
-test_3 (_mm512_maskz_sub_round_ps, __m512, __mmask16, __m512, __m512, 1)
+test_3 (_mm512_maskz_sub_round_pd, __m512d, __mmask8, __m512d, __m512d, 9)
+test_3 (_mm512_maskz_sub_round_ps, __m512, __mmask16, __m512, __m512, 9)
test_3 (_mm512_ternarylogic_epi32, __m512i, __m512i, __m512i, __m512i, 1)
test_3 (_mm512_ternarylogic_epi64, __m512i, __m512i, __m512i, __m512i, 1)
test_3 (_mm_mask_cmp_sd_mask, __mmask8, __mmask8, __m128d, __m128d, 1)
@@ -541,44 +541,44 @@ test_3v (_mm512_i64scatter_epi32, void *, __m512i, __m256i, 1)
test_3v (_mm512_i64scatter_epi64, void *, __m512i, __m512i, 1)
test_3v (_mm512_i64scatter_pd, void *, __m512i, __m512d, 1)
test_3v (_mm512_i64scatter_ps, void *, __m512i, __m256, 1)
-test_3x (_mm512_mask_roundscale_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 5)
-test_3x (_mm512_mask_roundscale_round_ps, __m512, __m512, __mmask16, __m512, 1, 5)
-test_3x (_mm512_mask_cmp_round_pd_mask, __mmask8, __mmask8, __m512d, __m512d, 1, 5)
-test_3x (_mm512_mask_cmp_round_ps_mask, __mmask16, __mmask16, __m512, __m512, 1, 5)
-test_3x (_mm_fixupimm_round_sd, __m128d, __m128d, __m128d, __m128i, 1, 5)
-test_3x (_mm_fixupimm_round_ss, __m128, __m128, __m128, __m128i, 1, 5)
-test_3x (_mm_mask_cmp_round_sd_mask, __mmask8, __mmask8, __m128d, __m128d, 1, 5)
-test_3x (_mm_mask_cmp_round_ss_mask, __mmask8, __mmask8, __m128, __m128, 1, 5)
-test_4 (_mm512_mask3_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmaddsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fmsubadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fnmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask3_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 1)
-test_4 (_mm512_mask3_fnmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 1)
-test_4 (_mm512_mask_add_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_add_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_3x (_mm512_mask_roundscale_round_pd, __m512d, __m512d, __mmask8, __m512d, 1, 8)
+test_3x (_mm512_mask_roundscale_round_ps, __m512, __m512, __mmask16, __m512, 1, 8)
+test_3x (_mm512_mask_cmp_round_pd_mask, __mmask8, __mmask8, __m512d, __m512d, 1, 8)
+test_3x (_mm512_mask_cmp_round_ps_mask, __mmask16, __mmask16, __m512, __m512, 1, 8)
+test_3x (_mm_fixupimm_round_sd, __m128d, __m128d, __m128d, __m128i, 1, 8)
+test_3x (_mm_fixupimm_round_ss, __m128, __m128, __m128, __m128i, 1, 8)
+test_3x (_mm_mask_cmp_round_sd_mask, __mmask8, __mmask8, __m128d, __m128d, 1, 8)
+test_3x (_mm_mask_cmp_round_ss_mask, __mmask8, __mmask8, __m128, __m128, 1, 8)
+test_4 (_mm512_mask3_fmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fmaddsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmaddsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fmsubadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fmsubadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fnmadd_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fnmadd_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask3_fnmsub_round_pd, __m512d, __m512d, __m512d, __m512d, __mmask8, 9)
+test_4 (_mm512_mask3_fnmsub_round_ps, __m512, __m512, __m512, __m512, __mmask16, 9)
+test_4 (_mm512_mask_add_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_add_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_alignr_epi32, __m512i, __m512i, __mmask16, __m512i, __m512i, 1)
test_4 (_mm512_mask_alignr_epi64, __m512i, __m512i, __mmask8, __m512i, __m512i, 1)
-test_4 (_mm512_mask_div_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_div_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmaddsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmaddsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fmsubadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fmsubadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fnmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fnmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_fnmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_fnmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_4 (_mm512_mask_div_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_div_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmaddsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmaddsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fmsubadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fmsubadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fnmadd_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fnmadd_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_fnmsub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_fnmsub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_i32gather_epi32, __m512i, __m512i, __mmask16, __m512i, void const *, 1)
test_4 (_mm512_mask_i32gather_epi64, __m512i, __m512i, __mmask8, __m256i, void const *, 1)
test_4 (_mm512_mask_i32gather_pd, __m512d, __m512d, __mmask8, __m256i, void const *, 1)
@@ -591,36 +591,36 @@ test_4 (_mm512_mask_insertf32x4, __m512, __m512, __mmask16, __m512, __m128, 1)
test_4 (_mm512_mask_insertf64x4, __m512d, __m512d, __mmask8, __m512d, __m256d, 1)
test_4 (_mm512_mask_inserti32x4, __m512i, __m512i, __mmask16, __m512i, __m128i, 1)
test_4 (_mm512_mask_inserti64x4, __m512i, __m512i, __mmask8, __m512i, __m256i, 1)
-test_4 (_mm512_mask_max_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 5)
-test_4 (_mm512_mask_max_round_ps, __m512, __m512, __mmask16, __m512, __m512, 5)
-test_4 (_mm512_mask_min_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 5)
-test_4 (_mm512_mask_min_round_ps, __m512, __m512, __mmask16, __m512, __m512, 5)
-test_4 (_mm512_mask_mul_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_mul_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_scalef_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_scalef_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_4 (_mm512_mask_max_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 8)
+test_4 (_mm512_mask_max_round_ps, __m512, __m512, __mmask16, __m512, __m512, 8)
+test_4 (_mm512_mask_min_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 8)
+test_4 (_mm512_mask_min_round_ps, __m512, __m512, __mmask16, __m512, __m512, 8)
+test_4 (_mm512_mask_mul_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_mul_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
+test_4 (_mm512_mask_scalef_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_scalef_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_shuffle_f32x4, __m512, __m512, __mmask16, __m512, __m512, 1)
test_4 (_mm512_mask_shuffle_f64x2, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
test_4 (_mm512_mask_shuffle_i32x4, __m512i, __m512i, __mmask16, __m512i, __m512i, 1)
test_4 (_mm512_mask_shuffle_i64x2, __m512i, __m512i, __mmask8, __m512i, __m512i, 1)
test_4 (_mm512_mask_shuffle_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
test_4 (_mm512_mask_shuffle_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
-test_4 (_mm512_mask_sub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 1)
-test_4 (_mm512_mask_sub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 1)
+test_4 (_mm512_mask_sub_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512d, 9)
+test_4 (_mm512_mask_sub_round_ps, __m512, __m512, __mmask16, __m512, __m512, 9)
test_4 (_mm512_mask_ternarylogic_epi32, __m512i, __m512i, __mmask16, __m512i, __m512i, 1)
test_4 (_mm512_mask_ternarylogic_epi64, __m512i, __m512i, __mmask8, __m512i, __m512i, 1)
-test_4 (_mm512_maskz_fmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fmaddsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmaddsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fmsubadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fmsubadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fnmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fnmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
-test_4 (_mm512_maskz_fnmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 1)
-test_4 (_mm512_maskz_fnmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 1)
+test_4 (_mm512_maskz_fmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fmaddsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmaddsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fmsubadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fmsubadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fnmadd_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fnmadd_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
+test_4 (_mm512_maskz_fnmsub_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512d, 9)
+test_4 (_mm512_maskz_fnmsub_round_ps, __m512, __mmask16, __m512, __m512, __m512, 9)
test_4 (_mm512_maskz_ternarylogic_epi32, __m512i, __mmask16, __m512i, __m512i, __m512i, 1)
test_4 (_mm512_maskz_ternarylogic_epi64, __m512i, __mmask8, __m512i, __m512i, __m512i, 1)
test_4v (_mm512_mask_i32scatter_epi32, void *, __mmask16, __m512i, __m512i, 1)
@@ -631,14 +631,14 @@ test_4v (_mm512_mask_i64scatter_epi32, void *, __mmask8, __m512i, __m256i, 1)
test_4v (_mm512_mask_i64scatter_epi64, void *, __mmask8, __m512i, __m512i, 1)
test_4v (_mm512_mask_i64scatter_pd, void *, __mmask8, __m512i, __m512d, 1)
test_4v (_mm512_mask_i64scatter_ps, void *, __mmask8, __m512i, __m256, 1)
-test_4x (_mm512_mask_fixupimm_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512i, 1, 5)
-test_4x (_mm512_mask_fixupimm_round_ps, __m512, __m512, __mmask16, __m512, __m512i, 1, 5)
-test_4x (_mm512_maskz_fixupimm_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512i, 1, 5)
-test_4x (_mm512_maskz_fixupimm_round_ps, __m512, __mmask16, __m512, __m512, __m512i, 1, 5)
-test_4x (_mm_mask_fixupimm_round_sd, __m128d, __m128d, __mmask8, __m128d, __m128i, 1, 5)
-test_4x (_mm_mask_fixupimm_round_ss, __m128, __m128, __mmask8, __m128, __m128i, 1, 5)
-test_4x (_mm_maskz_fixupimm_round_sd, __m128d, __mmask8, __m128d, __m128d, __m128i, 1, 5)
-test_4x (_mm_maskz_fixupimm_round_ss, __m128, __mmask8, __m128, __m128, __m128i, 1, 5)
+test_4x (_mm512_mask_fixupimm_round_pd, __m512d, __m512d, __mmask8, __m512d, __m512i, 1, 8)
+test_4x (_mm512_mask_fixupimm_round_ps, __m512, __m512, __mmask16, __m512, __m512i, 1, 8)
+test_4x (_mm512_maskz_fixupimm_round_pd, __m512d, __mmask8, __m512d, __m512d, __m512i, 1, 8)
+test_4x (_mm512_maskz_fixupimm_round_ps, __m512, __mmask16, __m512, __m512, __m512i, 1, 8)
+test_4x (_mm_mask_fixupimm_round_sd, __m128d, __m128d, __mmask8, __m128d, __m128i, 1, 8)
+test_4x (_mm_mask_fixupimm_round_ss, __m128, __m128, __mmask8, __m128, __m128i, 1, 8)
+test_4x (_mm_maskz_fixupimm_round_sd, __m128d, __mmask8, __m128d, __m128d, __m128i, 1, 8)
+test_4x (_mm_maskz_fixupimm_round_ss, __m128, __mmask8, __m128, __m128, __m128i, 1, 8)
/* avx512pfintrin.h */
test_3vx (_mm512_mask_prefetch_i32gather_ps, __m512i, __mmask16, void const *, 1, 1)
@@ -652,28 +652,28 @@ test_3vx (_mm512_mask_prefetch_i64gather_pd, __m512i, __mmask8, long long *, 1,
test_3vx (_mm512_mask_prefetch_i64scatter_pd, void const *, __mmask8, __m512i, 1, 1)
/* avx512erintrin.h */
-test_1 (_mm512_exp2a23_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_exp2a23_round_ps, __m512, __m512, 5)
-test_1 (_mm512_rcp28_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_rcp28_round_ps, __m512, __m512, 5)
-test_1 (_mm512_rsqrt28_round_pd, __m512d, __m512d, 5)
-test_1 (_mm512_rsqrt28_round_ps, __m512, __m512, 5)
-test_2 (_mm512_maskz_exp2a23_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_exp2a23_round_ps, __m512, __mmask16, __m512, 5)
-test_2 (_mm512_maskz_rcp28_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_rcp28_round_ps, __m512, __mmask16, __m512, 5)
-test_2 (_mm512_maskz_rsqrt28_round_pd, __m512d, __mmask8, __m512d, 5)
-test_2 (_mm512_maskz_rsqrt28_round_ps, __m512, __mmask16, __m512, 5)
-test_3 (_mm512_mask_exp2a23_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_exp2a23_round_ps, __m512, __m512, __mmask16, __m512, 5)
-test_3 (_mm512_mask_rcp28_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_rcp28_round_ps, __m512, __m512, __mmask16, __m512, 5)
-test_3 (_mm512_mask_rsqrt28_round_pd, __m512d, __m512d, __mmask8, __m512d, 5)
-test_3 (_mm512_mask_rsqrt28_round_ps, __m512, __m512, __mmask16, __m512, 5)
-test_2 (_mm_rcp28_round_sd, __m128d, __m128d, __m128d, 5)
-test_2 (_mm_rcp28_round_ss, __m128, __m128, __m128, 5)
-test_2 (_mm_rsqrt28_round_sd, __m128d, __m128d, __m128d, 5)
-test_2 (_mm_rsqrt28_round_ss, __m128, __m128, __m128, 5)
+test_1 (_mm512_exp2a23_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_exp2a23_round_ps, __m512, __m512, 8)
+test_1 (_mm512_rcp28_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_rcp28_round_ps, __m512, __m512, 8)
+test_1 (_mm512_rsqrt28_round_pd, __m512d, __m512d, 8)
+test_1 (_mm512_rsqrt28_round_ps, __m512, __m512, 8)
+test_2 (_mm512_maskz_exp2a23_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_exp2a23_round_ps, __m512, __mmask16, __m512, 8)
+test_2 (_mm512_maskz_rcp28_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_rcp28_round_ps, __m512, __mmask16, __m512, 8)
+test_2 (_mm512_maskz_rsqrt28_round_pd, __m512d, __mmask8, __m512d, 8)
+test_2 (_mm512_maskz_rsqrt28_round_ps, __m512, __mmask16, __m512, 8)
+test_3 (_mm512_mask_exp2a23_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_exp2a23_round_ps, __m512, __m512, __mmask16, __m512, 8)
+test_3 (_mm512_mask_rcp28_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_rcp28_round_ps, __m512, __m512, __mmask16, __m512, 8)
+test_3 (_mm512_mask_rsqrt28_round_pd, __m512d, __m512d, __mmask8, __m512d, 8)
+test_3 (_mm512_mask_rsqrt28_round_ps, __m512, __m512, __mmask16, __m512, 8)
+test_2 (_mm_rcp28_round_sd, __m128d, __m128d, __m128d, 8)
+test_2 (_mm_rcp28_round_ss, __m128, __m128, __m128, 8)
+test_2 (_mm_rsqrt28_round_sd, __m128d, __m128d, __m128d, 8)
+test_2 (_mm_rsqrt28_round_ss, __m128, __m128, __m128, 8)
/* shaintrin.h */
test_2 (_mm_sha1rnds4_epu32, __m128i, __m128i, __m128i, 1)
@@ -184,86 +184,86 @@
#define __builtin_ia32_xabort(M) __builtin_ia32_xabort(1)
/* avx512fintrin.h */
-#define __builtin_ia32_addpd512_mask(A, B, C, D, E) __builtin_ia32_addpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_addps512_mask(A, B, C, D, E) __builtin_ia32_addps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_addsd_round(A, B, C) __builtin_ia32_addsd_round(A, B, 1)
-#define __builtin_ia32_addss_round(A, B, C) __builtin_ia32_addss_round(A, B, 1)
+#define __builtin_ia32_addpd512_mask(A, B, C, D, E) __builtin_ia32_addpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_addps512_mask(A, B, C, D, E) __builtin_ia32_addps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_addsd_round(A, B, C) __builtin_ia32_addsd_round(A, B, 8)
+#define __builtin_ia32_addss_round(A, B, C) __builtin_ia32_addss_round(A, B, 8)
#define __builtin_ia32_alignd512_mask(A, B, F, D, E) __builtin_ia32_alignd512_mask(A, B, 1, D, E)
#define __builtin_ia32_alignq512_mask(A, B, F, D, E) __builtin_ia32_alignq512_mask(A, B, 1, D, E)
#define __builtin_ia32_cmpd512_mask(A, B, E, D) __builtin_ia32_cmpd512_mask(A, B, 1, D)
-#define __builtin_ia32_cmppd512_mask(A, B, F, D, E) __builtin_ia32_cmppd512_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cmpps512_mask(A, B, F, D, E) __builtin_ia32_cmpps512_mask(A, B, 1, D, 5)
+#define __builtin_ia32_cmppd512_mask(A, B, F, D, E) __builtin_ia32_cmppd512_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cmpps512_mask(A, B, F, D, E) __builtin_ia32_cmpps512_mask(A, B, 1, D, 8)
#define __builtin_ia32_cmpq512_mask(A, B, E, D) __builtin_ia32_cmpq512_mask(A, B, 1, D)
-#define __builtin_ia32_cmpsd_mask(A, B, F, D, E) __builtin_ia32_cmpsd_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cmpss_mask(A, B, F, D, E) __builtin_ia32_cmpss_mask(A, B, 1, D, 5)
-#define __builtin_ia32_cvtdq2ps512_mask(A, B, C, D) __builtin_ia32_cvtdq2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2dq512_mask(A, B, C, D) __builtin_ia32_cvtpd2dq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2ps512_mask(A, B, C, D) __builtin_ia32_cvtpd2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtpd2udq512_mask(A, B, C, D) __builtin_ia32_cvtpd2udq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtps2dq512_mask(A, B, C, D) __builtin_ia32_cvtps2dq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtps2pd512_mask(A, B, C, D) __builtin_ia32_cvtps2pd512_mask(A, B, C, 5)
-#define __builtin_ia32_cvtps2udq512_mask(A, B, C, D) __builtin_ia32_cvtps2udq512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtsd2ss_round(A, B, C) __builtin_ia32_cvtsd2ss_round(A, B, 1)
+#define __builtin_ia32_cmpsd_mask(A, B, F, D, E) __builtin_ia32_cmpsd_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cmpss_mask(A, B, F, D, E) __builtin_ia32_cmpss_mask(A, B, 1, D, 8)
+#define __builtin_ia32_cvtdq2ps512_mask(A, B, C, D) __builtin_ia32_cvtdq2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2dq512_mask(A, B, C, D) __builtin_ia32_cvtpd2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2ps512_mask(A, B, C, D) __builtin_ia32_cvtpd2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtpd2udq512_mask(A, B, C, D) __builtin_ia32_cvtpd2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2dq512_mask(A, B, C, D) __builtin_ia32_cvtps2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2pd512_mask(A, B, C, D) __builtin_ia32_cvtps2pd512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtps2udq512_mask(A, B, C, D) __builtin_ia32_cvtps2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtsd2ss_round(A, B, C) __builtin_ia32_cvtsd2ss_round(A, B, 8)
#define __builtin_ia32_cvtss2sd_round(A, B, C) __builtin_ia32_cvtss2sd_round(A, B, 4)
-#define __builtin_ia32_cvtsi2sd64(A, B, C) __builtin_ia32_cvtsi2sd64(A, B, 1)
-#define __builtin_ia32_cvtsi2ss32(A, B, C) __builtin_ia32_cvtsi2ss32(A, B, 1)
-#define __builtin_ia32_cvtsi2ss64(A, B, C) __builtin_ia32_cvtsi2ss64(A, B, 1)
-#define __builtin_ia32_cvttpd2dq512_mask(A, B, C, D) __builtin_ia32_cvttpd2dq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttpd2udq512_mask(A, B, C, D) __builtin_ia32_cvttpd2udq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttps2dq512_mask(A, B, C, D) __builtin_ia32_cvttps2dq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvttps2udq512_mask(A, B, C, D) __builtin_ia32_cvttps2udq512_mask(A, B, C, 5)
-#define __builtin_ia32_cvtudq2ps512_mask(A, B, C, D) __builtin_ia32_cvtudq2ps512_mask(A, B, C, 1)
-#define __builtin_ia32_cvtusi2sd64(A, B, C) __builtin_ia32_cvtusi2sd64(A, B, 1)
-#define __builtin_ia32_cvtusi2ss32(A, B, C) __builtin_ia32_cvtusi2ss32(A, B, 1)
-#define __builtin_ia32_cvtusi2ss64(A, B, C) __builtin_ia32_cvtusi2ss64(A, B, 1)
-#define __builtin_ia32_divpd512_mask(A, B, C, D, E) __builtin_ia32_divpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_divps512_mask(A, B, C, D, E) __builtin_ia32_divps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_divsd_round(A, B, C) __builtin_ia32_divsd_round(A, B, 1)
-#define __builtin_ia32_divss_round(A, B, C) __builtin_ia32_divss_round(A, B, 1)
+#define __builtin_ia32_cvtsi2sd64(A, B, C) __builtin_ia32_cvtsi2sd64(A, B, 8)
+#define __builtin_ia32_cvtsi2ss32(A, B, C) __builtin_ia32_cvtsi2ss32(A, B, 8)
+#define __builtin_ia32_cvtsi2ss64(A, B, C) __builtin_ia32_cvtsi2ss64(A, B, 8)
+#define __builtin_ia32_cvttpd2dq512_mask(A, B, C, D) __builtin_ia32_cvttpd2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttpd2udq512_mask(A, B, C, D) __builtin_ia32_cvttpd2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttps2dq512_mask(A, B, C, D) __builtin_ia32_cvttps2dq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvttps2udq512_mask(A, B, C, D) __builtin_ia32_cvttps2udq512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtudq2ps512_mask(A, B, C, D) __builtin_ia32_cvtudq2ps512_mask(A, B, C, 8)
+#define __builtin_ia32_cvtusi2sd64(A, B, C) __builtin_ia32_cvtusi2sd64(A, B, 8)
+#define __builtin_ia32_cvtusi2ss32(A, B, C) __builtin_ia32_cvtusi2ss32(A, B, 8)
+#define __builtin_ia32_cvtusi2ss64(A, B, C) __builtin_ia32_cvtusi2ss64(A, B, 8)
+#define __builtin_ia32_divpd512_mask(A, B, C, D, E) __builtin_ia32_divpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_divps512_mask(A, B, C, D, E) __builtin_ia32_divps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_divsd_round(A, B, C) __builtin_ia32_divsd_round(A, B, 8)
+#define __builtin_ia32_divss_round(A, B, C) __builtin_ia32_divss_round(A, B, 8)
#define __builtin_ia32_extractf32x4_mask(A, E, C, D) __builtin_ia32_extractf32x4_mask(A, 1, C, D)
#define __builtin_ia32_extractf64x4_mask(A, E, C, D) __builtin_ia32_extractf64x4_mask(A, 1, C, D)
#define __builtin_ia32_extracti32x4_mask(A, E, C, D) __builtin_ia32_extracti32x4_mask(A, 1, C, D)
#define __builtin_ia32_extracti64x4_mask(A, E, C, D) __builtin_ia32_extracti64x4_mask(A, 1, C, D)
-#define __builtin_ia32_fixupimmpd512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmpd512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmps512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmps512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmsd_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmsd_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmss_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmss_mask(A, B, C, 1, E, 5)
-#define __builtin_ia32_fixupimmss_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmss_maskz(A, B, C, 1, E, 5)
-#define __builtin_ia32_gatherdiv8df(A, B, C, D, F) __builtin_ia32_gatherdiv8df(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv8di(A, B, C, D, F) __builtin_ia32_gatherdiv8di(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv16sf(A, B, C, D, F) __builtin_ia32_gatherdiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_gatherdiv16si(A, B, C, D, F) __builtin_ia32_gatherdiv16si(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv16sf(A, B, C, D, F) __builtin_ia32_gathersiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv16si(A, B, C, D, F) __builtin_ia32_gathersiv16si(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv8df(A, B, C, D, F) __builtin_ia32_gathersiv8df(A, B, C, D, 1)
-#define __builtin_ia32_gathersiv8di(A, B, C, D, F) __builtin_ia32_gathersiv8di(A, B, C, D, 1)
-#define __builtin_ia32_getexppd512_mask(A, B, C, D) __builtin_ia32_getexppd512_mask(A, B, C, 5)
-#define __builtin_ia32_getexpps512_mask(A, B, C, D) __builtin_ia32_getexpps512_mask(A, B, C, 5)
+#define __builtin_ia32_fixupimmpd512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmpd512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmpd512_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmps512_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmps512_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmps512_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmsd_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmsd_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmsd_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmss_mask(A, B, C, I, E, F) __builtin_ia32_fixupimmss_mask(A, B, C, 1, E, 8)
+#define __builtin_ia32_fixupimmss_maskz(A, B, C, I, E, F) __builtin_ia32_fixupimmss_maskz(A, B, C, 1, E, 8)
+#define __builtin_ia32_gatherdiv8df(A, B, C, D, F) __builtin_ia32_gatherdiv8df(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv8di(A, B, C, D, F) __builtin_ia32_gatherdiv8di(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv16sf(A, B, C, D, F) __builtin_ia32_gatherdiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_gatherdiv16si(A, B, C, D, F) __builtin_ia32_gatherdiv16si(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv16sf(A, B, C, D, F) __builtin_ia32_gathersiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv16si(A, B, C, D, F) __builtin_ia32_gathersiv16si(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv8df(A, B, C, D, F) __builtin_ia32_gathersiv8df(A, B, C, D, 8)
+#define __builtin_ia32_gathersiv8di(A, B, C, D, F) __builtin_ia32_gathersiv8di(A, B, C, D, 8)
+#define __builtin_ia32_getexppd512_mask(A, B, C, D) __builtin_ia32_getexppd512_mask(A, B, C, 8)
+#define __builtin_ia32_getexpps512_mask(A, B, C, D) __builtin_ia32_getexpps512_mask(A, B, C, 8)
#define __builtin_ia32_getexpsd128_round(A, B, C) __builtin_ia32_getexpsd128_round(A, B, 4)
#define __builtin_ia32_getexpss128_round(A, B, C) __builtin_ia32_getexpss128_round(A, B, 4)
-#define __builtin_ia32_getmantpd512_mask(A, F, C, D, E) __builtin_ia32_getmantpd512_mask(A, 1, C, D, 5)
-#define __builtin_ia32_getmantps512_mask(A, F, C, D, E) __builtin_ia32_getmantps512_mask(A, 1, C, D, 5)
+#define __builtin_ia32_getmantpd512_mask(A, F, C, D, E) __builtin_ia32_getmantpd512_mask(A, 1, C, D, 8)
+#define __builtin_ia32_getmantps512_mask(A, F, C, D, E) __builtin_ia32_getmantps512_mask(A, 1, C, D, 8)
#define __builtin_ia32_getmantsd_round(A, B, C, D) __builtin_ia32_getmantsd_round(A, B, 1, 4)
#define __builtin_ia32_getmantss_round(A, B, C, D) __builtin_ia32_getmantss_round(A, B, 1, 4)
#define __builtin_ia32_insertf32x4_mask(A, B, F, D, E) __builtin_ia32_insertf32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_insertf64x4_mask(A, B, F, D, E) __builtin_ia32_insertf64x4_mask(A, B, 1, D, E)
#define __builtin_ia32_inserti32x4_mask(A, B, F, D, E) __builtin_ia32_inserti32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_inserti64x4_mask(A, B, F, D, E) __builtin_ia32_inserti64x4_mask(A, B, 1, D, E)
-#define __builtin_ia32_maxpd512_mask(A, B, C, D, E) __builtin_ia32_maxpd512_mask(A, B, C, D, 5)
-#define __builtin_ia32_maxps512_mask(A, B, C, D, E) __builtin_ia32_maxps512_mask(A, B, C, D, 5)
+#define __builtin_ia32_maxpd512_mask(A, B, C, D, E) __builtin_ia32_maxpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_maxps512_mask(A, B, C, D, E) __builtin_ia32_maxps512_mask(A, B, C, D, 8)
#define __builtin_ia32_maxsd_round(A, B, C) __builtin_ia32_maxsd_round(A, B, 4)
#define __builtin_ia32_maxss_round(A, B, C) __builtin_ia32_maxss_round(A, B, 4)
-#define __builtin_ia32_minpd512_mask(A, B, C, D, E) __builtin_ia32_minpd512_mask(A, B, C, D, 5)
-#define __builtin_ia32_minps512_mask(A, B, C, D, E) __builtin_ia32_minps512_mask(A, B, C, D, 5)
+#define __builtin_ia32_minpd512_mask(A, B, C, D, E) __builtin_ia32_minpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_minps512_mask(A, B, C, D, E) __builtin_ia32_minps512_mask(A, B, C, D, 8)
#define __builtin_ia32_minsd_round(A, B, C) __builtin_ia32_minsd_round(A, B, 4)
#define __builtin_ia32_minss_round(A, B, C) __builtin_ia32_minss_round(A, B, 4)
-#define __builtin_ia32_mulpd512_mask(A, B, C, D, E) __builtin_ia32_mulpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_mulps512_mask(A, B, C, D, E) __builtin_ia32_mulps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_mulsd_round(A, B, C) __builtin_ia32_mulsd_round(A, B, 1)
-#define __builtin_ia32_mulss_round(A, B, C) __builtin_ia32_mulss_round(A, B, 1)
+#define __builtin_ia32_mulpd512_mask(A, B, C, D, E) __builtin_ia32_mulpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulps512_mask(A, B, C, D, E) __builtin_ia32_mulps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_mulsd_round(A, B, C) __builtin_ia32_mulsd_round(A, B, 8)
+#define __builtin_ia32_mulss_round(A, B, C) __builtin_ia32_mulss_round(A, B, 8)
#define __builtin_ia32_permdf512_mask(A, E, C, D) __builtin_ia32_permdf512_mask(A, 1, C, D)
#define __builtin_ia32_permdi512_mask(A, E, C, D) __builtin_ia32_permdi512_mask(A, 1, C, D)
#define __builtin_ia32_prold512_mask(A, E, C, D) __builtin_ia32_prold512_mask(A, 1, C, D)
@@ -281,82 +281,82 @@
#define __builtin_ia32_pternlogd512_maskz(A, B, C, F, E) __builtin_ia32_pternlogd512_maskz(A, B, C, 1, E)
#define __builtin_ia32_pternlogq512_mask(A, B, C, F, E) __builtin_ia32_pternlogq512_mask(A, B, C, 1, E)
#define __builtin_ia32_pternlogq512_maskz(A, B, C, F, E) __builtin_ia32_pternlogq512_maskz(A, B, C, 1, E)
-#define __builtin_ia32_rndscalepd_mask(A, F, C, D, E) __builtin_ia32_rndscalepd_mask(A, 1, C, D, 5)
-#define __builtin_ia32_rndscaleps_mask(A, F, C, D, E) __builtin_ia32_rndscaleps_mask(A, 1, C, D, 5)
+#define __builtin_ia32_rndscalepd_mask(A, F, C, D, E) __builtin_ia32_rndscalepd_mask(A, 1, C, D, 8)
+#define __builtin_ia32_rndscaleps_mask(A, F, C, D, E) __builtin_ia32_rndscaleps_mask(A, 1, C, D, 8)
#define __builtin_ia32_rndscalesd_round(A, B, C, D) __builtin_ia32_rndscalesd_round(A, B, 1, 4)
#define __builtin_ia32_rndscaless_round(A, B, C, D) __builtin_ia32_rndscaless_round(A, B, 1, 4)
-#define __builtin_ia32_scalefpd512_mask(A, B, C, D, E) __builtin_ia32_scalefpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_scalefps512_mask(A, B, C, D, E) __builtin_ia32_scalefps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_scalefsd_round(A, B, C) __builtin_ia32_scalefsd_round(A, B, 1)
-#define __builtin_ia32_scalefss_round(A, B, C) __builtin_ia32_scalefss_round(A, B, 1)
-#define __builtin_ia32_scatterdiv8df(A, B, C, D, F) __builtin_ia32_scatterdiv8df(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv8di(A, B, C, D, F) __builtin_ia32_scatterdiv8di(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv16sf(A, B, C, D, F) __builtin_ia32_scatterdiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_scatterdiv16si(A, B, C, D, F) __builtin_ia32_scatterdiv16si(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv16sf(A, B, C, D, F) __builtin_ia32_scattersiv16sf(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv16si(A, B, C, D, F) __builtin_ia32_scattersiv16si(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv8df(A, B, C, D, F) __builtin_ia32_scattersiv8df(A, B, C, D, 1)
-#define __builtin_ia32_scattersiv8di(A, B, C, D, F) __builtin_ia32_scattersiv8di(A, B, C, D, 1)
+#define __builtin_ia32_scalefpd512_mask(A, B, C, D, E) __builtin_ia32_scalefpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_scalefps512_mask(A, B, C, D, E) __builtin_ia32_scalefps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_scalefsd_round(A, B, C) __builtin_ia32_scalefsd_round(A, B, 8)
+#define __builtin_ia32_scalefss_round(A, B, C) __builtin_ia32_scalefss_round(A, B, 8)
+#define __builtin_ia32_scatterdiv8df(A, B, C, D, F) __builtin_ia32_scatterdiv8df(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv8di(A, B, C, D, F) __builtin_ia32_scatterdiv8di(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv16sf(A, B, C, D, F) __builtin_ia32_scatterdiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_scatterdiv16si(A, B, C, D, F) __builtin_ia32_scatterdiv16si(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv16sf(A, B, C, D, F) __builtin_ia32_scattersiv16sf(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv16si(A, B, C, D, F) __builtin_ia32_scattersiv16si(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv8df(A, B, C, D, F) __builtin_ia32_scattersiv8df(A, B, C, D, 8)
+#define __builtin_ia32_scattersiv8di(A, B, C, D, F) __builtin_ia32_scattersiv8di(A, B, C, D, 8)
#define __builtin_ia32_shuf_f32x4_mask(A, B, F, D, E) __builtin_ia32_shuf_f32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_f64x2_mask(A, B, F, D, E) __builtin_ia32_shuf_f64x2_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_i32x4_mask(A, B, F, D, E) __builtin_ia32_shuf_i32x4_mask(A, B, 1, D, E)
#define __builtin_ia32_shuf_i64x2_mask(A, B, F, D, E) __builtin_ia32_shuf_i64x2_mask(A, B, 1, D, E)
#define __builtin_ia32_shufpd512_mask(A, B, F, D, E) __builtin_ia32_shufpd512_mask(A, B, 1, D, E)
#define __builtin_ia32_shufps512_mask(A, B, F, D, E) __builtin_ia32_shufps512_mask(A, B, 1, D, E)
-#define __builtin_ia32_sqrtpd512_mask(A, B, C, D) __builtin_ia32_sqrtpd512_mask(A, B, C, 1)
-#define __builtin_ia32_sqrtps512_mask(A, B, C, D) __builtin_ia32_sqrtps512_mask(A, B, C, 1)
-#define __builtin_ia32_sqrtss_round(A, B, C) __builtin_ia32_sqrtss_round(A, B, 1)
-#define __builtin_ia32_sqrtsd_round(A, B, C) __builtin_ia32_sqrtsd_round(A, B, 1)
-#define __builtin_ia32_subpd512_mask(A, B, C, D, E) __builtin_ia32_subpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_subps512_mask(A, B, C, D, E) __builtin_ia32_subps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_subsd_round(A, B, C) __builtin_ia32_subsd_round(A, B, 1)
-#define __builtin_ia32_subss_round(A, B, C) __builtin_ia32_subss_round(A, B, 1)
+#define __builtin_ia32_sqrtpd512_mask(A, B, C, D) __builtin_ia32_sqrtpd512_mask(A, B, C, 8)
+#define __builtin_ia32_sqrtps512_mask(A, B, C, D) __builtin_ia32_sqrtps512_mask(A, B, C, 8)
+#define __builtin_ia32_sqrtss_round(A, B, C) __builtin_ia32_sqrtss_round(A, B, 8)
+#define __builtin_ia32_sqrtsd_round(A, B, C) __builtin_ia32_sqrtsd_round(A, B, 8)
+#define __builtin_ia32_subpd512_mask(A, B, C, D, E) __builtin_ia32_subpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_subps512_mask(A, B, C, D, E) __builtin_ia32_subps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_subsd_round(A, B, C) __builtin_ia32_subsd_round(A, B, 8)
+#define __builtin_ia32_subss_round(A, B, C) __builtin_ia32_subss_round(A, B, 8)
#define __builtin_ia32_ucmpd512_mask(A, B, E, D) __builtin_ia32_ucmpd512_mask(A, B, 1, D)
#define __builtin_ia32_ucmpq512_mask(A, B, E, D) __builtin_ia32_ucmpq512_mask(A, B, 1, D)
-#define __builtin_ia32_vcomisd(A, B, C, D) __builtin_ia32_vcomisd(A, B, 1, 5)
-#define __builtin_ia32_vcomiss(A, B, C, D) __builtin_ia32_vcomiss(A, B, 1, 5)
-#define __builtin_ia32_vcvtph2ps512_mask(A, B, C, D) __builtin_ia32_vcvtph2ps512_mask(A, B, C, 5)
+#define __builtin_ia32_vcomisd(A, B, C, D) __builtin_ia32_vcomisd(A, B, 1, 8)
+#define __builtin_ia32_vcomiss(A, B, C, D) __builtin_ia32_vcomiss(A, B, 1, 8)
+#define __builtin_ia32_vcvtph2ps512_mask(A, B, C, D) __builtin_ia32_vcvtph2ps512_mask(A, B, C, 8)
#define __builtin_ia32_vcvtps2ph512_mask(A, E, C, D) __builtin_ia32_vcvtps2ph512_mask(A, 1, C, D)
-#define __builtin_ia32_vcvtsd2si32(A, B) __builtin_ia32_vcvtsd2si32(A, 1)
-#define __builtin_ia32_vcvtsd2si64(A, B) __builtin_ia32_vcvtsd2si64(A, 1)
-#define __builtin_ia32_vcvtsd2usi32(A, B) __builtin_ia32_vcvtsd2usi32(A, 1)
-#define __builtin_ia32_vcvtsd2usi64(A, B) __builtin_ia32_vcvtsd2usi64(A, 1)
-#define __builtin_ia32_vcvtss2si32(A, B) __builtin_ia32_vcvtss2si32(A, 1)
-#define __builtin_ia32_vcvtss2si64(A, B) __builtin_ia32_vcvtss2si64(A, 1)
-#define __builtin_ia32_vcvtss2usi32(A, B) __builtin_ia32_vcvtss2usi32(A, 1)
-#define __builtin_ia32_vcvtss2usi64(A, B) __builtin_ia32_vcvtss2usi64(A, 1)
-#define __builtin_ia32_vcvttsd2si32(A, B) __builtin_ia32_vcvttsd2si32(A, 5)
-#define __builtin_ia32_vcvttsd2si64(A, B) __builtin_ia32_vcvttsd2si64(A, 5)
-#define __builtin_ia32_vcvttsd2usi32(A, B) __builtin_ia32_vcvttsd2usi32(A, 5)
-#define __builtin_ia32_vcvttsd2usi64(A, B) __builtin_ia32_vcvttsd2usi64(A, 5)
-#define __builtin_ia32_vcvttss2si32(A, B) __builtin_ia32_vcvttss2si32(A, 5)
-#define __builtin_ia32_vcvttss2si64(A, B) __builtin_ia32_vcvttss2si64(A, 5)
-#define __builtin_ia32_vcvttss2usi32(A, B) __builtin_ia32_vcvttss2usi32(A, 5)
-#define __builtin_ia32_vcvttss2usi64(A, B) __builtin_ia32_vcvttss2usi64(A, 5)
-#define __builtin_ia32_vfmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddps512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsd3_round(A, B, C, D) __builtin_ia32_vfmaddsd3_round(A, B, C, 1)
-#define __builtin_ia32_vfmaddss3_round(A, B, C, D) __builtin_ia32_vfmaddss3_round(A, B, C, 1)
-#define __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubps512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubps512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask(A, B, C, D, 1)
-#define __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, 1)
+#define __builtin_ia32_vcvtsd2si32(A, B) __builtin_ia32_vcvtsd2si32(A, 8)
+#define __builtin_ia32_vcvtsd2si64(A, B) __builtin_ia32_vcvtsd2si64(A, 8)
+#define __builtin_ia32_vcvtsd2usi32(A, B) __builtin_ia32_vcvtsd2usi32(A, 8)
+#define __builtin_ia32_vcvtsd2usi64(A, B) __builtin_ia32_vcvtsd2usi64(A, 8)
+#define __builtin_ia32_vcvtss2si32(A, B) __builtin_ia32_vcvtss2si32(A, 8)
+#define __builtin_ia32_vcvtss2si64(A, B) __builtin_ia32_vcvtss2si64(A, 8)
+#define __builtin_ia32_vcvtss2usi32(A, B) __builtin_ia32_vcvtss2usi32(A, 8)
+#define __builtin_ia32_vcvtss2usi64(A, B) __builtin_ia32_vcvtss2usi64(A, 8)
+#define __builtin_ia32_vcvttsd2si32(A, B) __builtin_ia32_vcvttsd2si32(A, 8)
+#define __builtin_ia32_vcvttsd2si64(A, B) __builtin_ia32_vcvttsd2si64(A, 8)
+#define __builtin_ia32_vcvttsd2usi32(A, B) __builtin_ia32_vcvttsd2usi32(A, 8)
+#define __builtin_ia32_vcvttsd2usi64(A, B) __builtin_ia32_vcvttsd2usi64(A, 8)
+#define __builtin_ia32_vcvttss2si32(A, B) __builtin_ia32_vcvttss2si32(A, 8)
+#define __builtin_ia32_vcvttss2si64(A, B) __builtin_ia32_vcvttss2si64(A, 8)
+#define __builtin_ia32_vcvttss2usi32(A, B) __builtin_ia32_vcvttss2usi32(A, 8)
+#define __builtin_ia32_vcvttss2usi64(A, B) __builtin_ia32_vcvttss2usi64(A, 8)
+#define __builtin_ia32_vfmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddpd512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddps512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsd3_round(A, B, C, D) __builtin_ia32_vfmaddsd3_round(A, B, C, 8)
+#define __builtin_ia32_vfmaddss3_round(A, B, C, D) __builtin_ia32_vfmaddss3_round(A, B, C, 8)
+#define __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubpd512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, E) __builtin_ia32_vfmaddsubps512_maskz(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubaddps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfmsubps512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmaddps512_mask(A, B, C, D, E) __builtin_ia32_vfnmaddps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubpd512_mask3(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubps512_mask(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask(A, B, C, D, 8)
+#define __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, E) __builtin_ia32_vfnmsubps512_mask3(A, B, C, D, 8)
#define __builtin_ia32_vpermilpd512_mask(A, E, C, D) __builtin_ia32_vpermilpd512_mask(A, 1, C, D)
#define __builtin_ia32_vpermilps512_mask(A, E, C, D) __builtin_ia32_vpermilps512_mask(A, 1, C, D)
@@ -371,16 +371,16 @@
#define __builtin_ia32_scatterpfqpd(A, B, C, D, E) __builtin_ia32_scatterpfqpd(A, B, C, 1, 1)
/* avx512erintrin.h */
-#define __builtin_ia32_exp2pd_mask(A, B, C, D) __builtin_ia32_exp2pd_mask (A, B, C, 5)
-#define __builtin_ia32_exp2ps_mask(A, B, C, D) __builtin_ia32_exp2ps_mask (A, B, C, 5)
-#define __builtin_ia32_rcp28pd_mask(A, B, C, D) __builtin_ia32_rcp28pd_mask (A, B, C, 5)
-#define __builtin_ia32_rcp28ps_mask(A, B, C, D) __builtin_ia32_rcp28ps_mask (A, B, C, 5)
-#define __builtin_ia32_rsqrt28pd_mask(A, B, C, D) __builtin_ia32_rsqrt28pd_mask (A, B, C, 5)
-#define __builtin_ia32_rsqrt28ps_mask(A, B, C, D) __builtin_ia32_rsqrt28ps_mask (A, B, C, 5)
-#define __builtin_ia32_rcp28sd_round(A, B, C) __builtin_ia32_rcp28sd_round(A, B, 5)
-#define __builtin_ia32_rcp28ss_round(A, B, C) __builtin_ia32_rcp28ss_round(A, B, 5)
-#define __builtin_ia32_rsqrt28sd_round(A, B, C) __builtin_ia32_rsqrt28sd_round(A, B, 5)
-#define __builtin_ia32_rsqrt28ss_round(A, B, C) __builtin_ia32_rsqrt28ss_round(A, B, 5)
+#define __builtin_ia32_exp2pd_mask(A, B, C, D) __builtin_ia32_exp2pd_mask (A, B, C, 8)
+#define __builtin_ia32_exp2ps_mask(A, B, C, D) __builtin_ia32_exp2ps_mask (A, B, C, 8)
+#define __builtin_ia32_rcp28pd_mask(A, B, C, D) __builtin_ia32_rcp28pd_mask (A, B, C, 8)
+#define __builtin_ia32_rcp28ps_mask(A, B, C, D) __builtin_ia32_rcp28ps_mask (A, B, C, 8)
+#define __builtin_ia32_rsqrt28pd_mask(A, B, C, D) __builtin_ia32_rsqrt28pd_mask (A, B, C, 8)
+#define __builtin_ia32_rsqrt28ps_mask(A, B, C, D) __builtin_ia32_rsqrt28ps_mask (A, B, C, 8)
+#define __builtin_ia32_rcp28sd_round(A, B, C) __builtin_ia32_rcp28sd_round(A, B, 8)
+#define __builtin_ia32_rcp28ss_round(A, B, C) __builtin_ia32_rcp28ss_round(A, B, 8)
+#define __builtin_ia32_rsqrt28sd_round(A, B, C) __builtin_ia32_rsqrt28sd_round(A, B, 8)
+#define __builtin_ia32_rsqrt28ss_round(A, B, C) __builtin_ia32_rsqrt28ss_round(A, B, 8)
/* shaintrin.h */
#define __builtin_ia32_sha1rnds4(A, B, C) __builtin_ia32_sha1rnds4(A, B, 1)