Changes between Version 31 and Version 32 of GcnInstrsVop1
 Timestamp:
 Jun 17, 2017, 2:00:30 PM (2 years ago)
Legend:
 Unmodified
 Added
 Removed
 Modified

GcnInstrsVop1
v31 v32 1164 1164 Syntax: V_CVT_F16_F32 VDST, SRC0<br /> 1165 1165 Description: 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.1166 MODE register (single FP rounding mode for GCN 1.0, double FP rounding modefor GCN 1.2), 1167 and store result to VDST. If absolute value is too high, then store /+infinity to VDST. 1168 1168 In GCN 1.2 flushing denormals controlled by MODE. In GCN 1.0/1.1, denormals are enabled.<br /> 1169 1169 Operation:<br /> … … 1379 1379 Opcode VOP3A: 385 (0x181) for GCN 1.2<br /> 1380 1380 Syntax: 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 /> 1381 Description: Approximate power of two from half FP value SRC0 and store it to VDST.<br /> 1383 1382 Operation:<br /> 1384 1383 <code>VDST = APPROX_POW2(ASHALF(SRC0))</code></p> … … 1600 1599 Syntax: V_LOG_F16 VDST, SRC0<br /> 1601 1600 Description: 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 /> 1601 result to VDST. If SRC0 is negative then store NaN to VDST.<br /> 1604 1602 Operation:<br /> 1605 1603 <code>HALF F = ASHALF(SRC0) … … 1822 1820 Syntax: V_RSQ_F16 VDST, SRC0<br /> 1823 1821 Description: 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 /> 1822 store it to VDST. If SRC0 is negative value, store NAN to VDST.<br /> 1826 1823 Operation:<br /> 1827 1824 <code>VDST = APPROX_RSQRT(ASHALF(SRC0))</code></p>