Changes between Version 22 and Version 23 of GcnInstrsVop1


Ignore:
Timestamp:
05/29/16 10:00:30 (8 years ago)
Author:
trac
Comment:

--

Legend:

Unmodified
Added
Removed
Modified

  • GcnInstrsVop1

    v22 v23  
    12741274Description: Convert 32-bit floating point value from SRC0 to signed 32-bit integer, and
    12751275store result to VDST. Conversion adds 0.5 to value and rounds negative infinity (floor).
    1276 If value is higher/lower than maximal/minimal integer then store MAX_INT32/MIN_INT32 to VDST.
    1277 If input value is NaN/-NaN then store MAX_INT32/MIN_INT32 to VDST.<br />
     1276If value is higher/lower than maximal/minimal integer then store MAX_INT32/MIN_INT32 to
     1277VDST. If input value is NaN/-NaN then store MAX_INT32/MIN_INT32 to VDST.<br />
    12781278Operation:<br />
    12791279<code>FLOAT SF = ASFLOAT(SRC0)
    … …  
    14831483Opcode VOP3A: 423 (0x1a7) for GCN 1.0/1.1; 353 (0x161) for GCN 1.2<br />
    14841484Syntax: V_LOG_F32 VDST, SRC0<br />
    1485 Description: Approximate logarithm of base 2 from floating point value SRC0, and store result
    1486 to VDST. If SRC0 is negative then store -NaN to VDST.
     1485Description: Approximate logarithm of base 2 from floating point value SRC0, and store
     1486result to VDST. If SRC0 is negative then store -NaN to VDST.
    14871487This instruction doesn't handle denormalized values regardless FLOAT MODE register setup.<br />
    14881488Operation:<br />
    … …  
    14981498Opcode VOP3A: 453 (0x1c5) for GCN 1.1; 396 (0x18c) for GCN 1.2<br />
    14991499Syntax: V_LOG_LEGACY_F32 VDST, SRC0<br />
    1500 Description: Approximate logarithm of base 2 from floating point value SRC0, and store result
    1501 to VDST. If SRC0 is negative then store -NaN to VDST.
     1500Description: Approximate logarithm of base 2 from floating point value SRC0, and store
     1501result to VDST. If SRC0 is negative then store -NaN to VDST.
    15021502This instruction doesn't handle denormalized values regardless FLOAT MODE register setup.
    15031503This instruction returns slightly different results than V_LOG_F32.<br />
    … …  
    16071607Syntax: V_RCP_LEGACY_F32 VDST, SRC0<br />
    16081608Description: Approximate reciprocal from floating point value SRC0 and store it to VDST.
    1609 Guaranted error below 1ulp. If SRC0 or VDST is zero or infinity then store 0 with proper sign
    1610 to VDST.<br />
     1609Guaranted error below 1ulp. If SRC0 or VDST is zero or infinity then store 0 with proper
     1610sign to VDST.<br />
    16111611Operation:<br />
    16121612<code>FLOAT SF = ASFLOAT(SRC0)