CMPSS—Compare Scalar Single-Precision Floating-Point Value

Instruction Operand Encoding

Description

Compares the low single-precision floating-point values in the second source operand and the first source operand and returns the results of the comparison to the destination operand. The comparison predicate operand (imme-diate operand) specifies the type of comparison performed.

128-bit Legacy SSE version: The first source and destination operand (first operand) is an XMM register. The second source operand (second operand) can be an XMM register or 32-bit memory location. Bits (MAX_VL-1:32) of the corresponding YMM destination register remain unchanged. The comparison result is a doubleword mask of all 1s (comparison true) or all 0s (comparison false).

VEX.128 encoded version: The first source operand (second operand) is an XMM register. The second source operand (third operand) can be an XMM register or a 32-bit memory location. The result is stored in the low 32 bits of the destination operand; bits 128:32 of the destination operand are copied from the first source operand. Bits (MAX_VL-1:128) of the destination ZMM register are zeroed. The comparison result is a doubleword mask of all 1s (comparison true) or all 0s (comparison false).

EVEX encoded version: The first source operand (second operand) is an XMM register. The second source operand can be a XMM register or a 32-bit memory location. The destination operand (first operand) is an opmask register. The comparison result is a single mask bit of 1 (comparison true) or 0 (comparison false), written to the destination starting from the LSB according to the writemask k2. Bits (MAX_KL-1:128) of the destination register are cleared.

The comparison predicate operand is an 8-bit immediate:

The unordered relationship is true when at least one of the two source operands being compared is a NaN; the ordered relationship is true when neither source operand is a NaN.

A subsequent computational instruction that uses the mask result in the destination operand as an input operand will not generate an exception, because a mask of all 0s corresponds to a floating-point value of +0.0 and a mask of all 1s corresponds to a QNaN.

Note that processors with “CPUID.1H:ECX.AVX =0” do not implement the “greater-than”, “greater-than-or-equal”, “not-greater than”, and “not-greater-than-or-equal relations” predicates. These comparisons can be made either

by using the inverse relationship (that is, use the “not-less-than-or-equal” to make a “greater-than” comparison) or by using software emulation. When using software emulation, the program must swap the operands (copying registers when necessary to protect the data that will now be in the destination), and then perform the compare using a different predicate. The predicate to be used for these emulations is listed in the first 8 rows of Table 3-7 (Intel 64 and IA-32 Architectures Software Developer’s Manual Volume 2A) under the heading Emulation.

Compilers and assemblers may implement the following two-operand pseudo-ops in addition to the three-operand CMPSS instruction, for processors with “CPUID.1H:ECX.AVX =0”. See Table 3-8. Compiler should treat reserved Imm8 values as illegal syntax.

Table 3-8. Pseudo-Op and CMPSS Implementation

:

The greater-than relations that the processor does not implement require more than one instruction to emulate in software and therefore should not be implemented as pseudo-ops. (For these, the programmer should reverse the operands of the corresponding less than relations and use move instructions to ensure that the mask is moved to the correct destination register and that the source operand is left intact.)

Processors with “CPUID.1H:ECX.AVX =1” implement the full complement of 32 predicates shown in Table 3-7, soft-ware emulation is no longer needed. Compilers and assemblers may implement the following three-operand pseudo-ops in addition to the four-operand VCMPSS instruction. See Table 3-9, where the notations of reg1 reg2, and reg3 represent either XMM registers or YMM registers. Compiler should treat reserved Imm8 values as illegal syntax. Alternately, intrinsics can map the pseudo-ops to pre-defined constants to support a simpler intrinsic inter-face. Compilers and assemblers may implement three-operand pseudo-ops for EVEX encoded VCMPSS instructions in a similar fashion by extending the syntax listed in Table 3-9.

Table 3-9. Pseudo-Op and VCMPSS Implementation

:

Table 3-9. Pseudo-Op and VCMPSS Implementation

Software should ensure VCMPSS is encoded with VEX.L=0. Encoding VCMPSS with VEX.L=1 may encounter unpre-dictable behavior across different processor generations.

Operation

CASE (COMPARISON PREDICATE) OF
0: OP3 (cid:197)EQ_OQ; OP5 (cid:197)EQ_OQ;
1: OP3 (cid:197)LT_OS; OP5 (cid:197)LT_OS;
2: OP3 (cid:197)LE_OS; OP5 (cid:197)LE_OS;
3: OP3 (cid:197)UNORD_Q; OP5 (cid:197)UNORD_Q;
4: OP3 (cid:197)NEQ_UQ; OP5 (cid:197)NEQ_UQ;
5: OP3 (cid:197)NLT_US; OP5 (cid:197)NLT_US;
6: OP3 (cid:197)NLE_US; OP5 (cid:197)NLE_US;
7: OP3 (cid:197)ORD_Q; OP5 (cid:197)ORD_Q;
8: OP5 (cid:197)EQ_UQ;
9: OP5 (cid:197)NGE_US;
10: OP5 (cid:197)NGT_US;
11: OP5 (cid:197)FALSE_OQ;
12: OP5 (cid:197)NEQ_OQ;
13: OP5 (cid:197)GE_OS;
14: OP5 (cid:197)GT_OS;
15: OP5 (cid:197)TRUE_UQ;
16: OP5 (cid:197)EQ_OS;
17: OP5 (cid:197)LT_OQ;
18: OP5 (cid:197)LE_OQ;
19: OP5 (cid:197)UNORD_S;
20: OP5 (cid:197)NEQ_US;
21: OP5 (cid:197)NLT_UQ;
22: OP5 (cid:197)NLE_UQ;
23: OP5 (cid:197)ORD_S;
24: OP5 (cid:197)EQ_US;
25: OP5 (cid:197)NGE_UQ;
26: OP5 (cid:197)NGT_UQ;
27: OP5 (cid:197)FALSE_OS;
28: OP5 (cid:197)NEQ_OS;
29: OP5 (cid:197)GE_OQ;
30: OP5 (cid:197)GT_OQ;
31: OP5 (cid:197)TRUE_US;
DEFAULT: Reserved
ESAC;

CMP0 (cid:197) SRC1[31:0] OP5 SRC2[31:0];
IF k2[0] or *no writemask*
    THEN
    IF CMP0 = TRUE
        THEN DEST[0] (cid:197) 1;
        ELSE DEST[0] (cid:197) 0; FI;
    ELSE
    DEST[0] (cid:197) 0
    ; zeroing-masking only
FI;
DEST[MAX_KL-1:1] (cid:197) 0

CMP0 (cid:197)DEST[31:0] OP3 SRC[31:0];
IF CMP0 = TRUE
    THEN DEST[31:0] (cid:197)FFFFFFFFH;
    ELSE DEST[31:0] (cid:197)00000000H; FI;
DEST[MAX_VL-1:32] (Unmodified)

CMP0 (cid:197)SRC1[31:0] OP5 SRC2[31:0];
IF CMP0 = TRUE
    THEN DEST[31:0] (cid:197)FFFFFFFFH;
    ELSE DEST[31:0] (cid:197)00000000H; FI;
DEST[127:32] (cid:197)SRC1[127:32]
DEST[MAX_VL-1:128] (cid:197)0

Intel C/C++ Compiler Intrinsic Equivalent

VCMPSS __mmask8 _mm_cmp_ss_mask( __m128 a, __m128 b, int imm);
VCMPSS __mmask8 _mm_cmp_round_ss_mask( __m128 a, __m128 b, int imm, int sae);
VCMPSS __mmask8 _mm_mask_cmp_ss_mask( __mmask8 k1, __m128 a, __m128 b, int imm);
VCMPSS __mmask8 _mm_mask_cmp_round_ss_mask( __mmask8 k1, __m128 a, __m128 b, int imm, int sae);
(V)CMPSS __m128 _mm_cmp_ss(__m128 a, __m128 b, const int imm)

SIMD Floating-Point Exceptions

Invalid if SNaN operand, Invalid if QNaN and predicate as listed in Table 3-1, Denormal.

Other Exceptions