Changes between Version 31 and Version 32 of GcnInstrsVop1


Ignore:
Timestamp:
Jun 17, 2017, 2:00:30 PM (2 years ago)
Author:
trac
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • GcnInstrsVop1

    v31 v32  
    11641164Syntax: V_CVT_F16_F32 VDST, SRC0<br />
    11651165Description: Convert single FP value to half floating point value with rounding from
    1166 MODE register (single FP rounding mode), and store result to VDST.
    1167 If absolute value is too high, then store -/+infinity to VDST.
     1166MODE register (single FP rounding mode for GCN 1.0, double FP rounding modefor GCN 1.2),
     1167and store result to VDST. If absolute value is too high, then store -/+infinity to VDST.
    11681168In GCN 1.2 flushing denormals controlled by MODE. In GCN 1.0/1.1, denormals are enabled.<br />
    11691169Operation:<br />
     
    13791379Opcode VOP3A: 385 (0x181) for GCN 1.2<br />
    13801380Syntax: V_EXP_F16 VDST, SRC0<br />
    1381 Description: Approximate power of two from half FP value SRC0 and store it to VDST.
    1382 Instruction always handles dernomals in output regardless floatmode in MODE register.<br />
     1381Description: Approximate power of two from half FP value SRC0 and store it to VDST.<br />
    13831382Operation:<br />
    13841383<code>VDST = APPROX_POW2(ASHALF(SRC0))</code></p>
     
    16001599Syntax: V_LOG_F16 VDST, SRC0<br />
    16011600Description: Approximate logarithm of base 2 from half floating point value SRC0, and store
    1602 result to VDST. If SRC0 is negative then store -NaN to VDST.
    1603 This instruction handle denormalized values regardless FLOAT MODE register setup.<br />
     1601result to VDST. If SRC0 is negative then store -NaN to VDST.<br />
    16041602Operation:<br />
    16051603<code>HALF F = ASHALF(SRC0)
     
    18221820Syntax: V_RSQ_F16 VDST, SRC0<br />
    18231821Description: Approximate reciprocal square root from half floating point value SRC0 and
    1824 store it to VDST. If SRC0 is negative value, store -NAN to VDST.
    1825 This instruction doesn't handle denormalized values regardless FLOAT MODE register setup.<br />
     1822store it to VDST. If SRC0 is negative value, store -NAN to VDST.<br />
    18261823Operation:<br />
    18271824<code>VDST = APPROX_RSQRT(ASHALF(SRC0))</code></p>