[PATCH] [x86][AVX512] add vscalef support
Elena Demikhovsky
elena.demikhovsky at intel.com
Sun Jun 28 02:36:17 PDT 2015
REPOSITORY
rL LLVM
================
Comment at: lib/Target/X86/X86ISelLowering.cpp:15093
@@ +15092,3 @@
+ // We specify 2 possible modes for intrinsics, with/without rounding modes.
+ // First, we check if the intrinsic may have non-default rounding mode,
+ // (IntrData->Opc1 != 0), if not, we set rounding mode to "current".
----------------
Change comments, please
================
Comment at: lib/Target/X86/X86ISelLowering.cpp:15101
@@ +15100,3 @@
+ return getVectorMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT,
+ Src1, Src2, Rnd),
+ Mask, PassThru, Subtarget, DAG);
----------------
braces alignment
================
Comment at: lib/Target/X86/X86InstrAVX512.td:3449
@@ -3448,3 +3448,3 @@
multiclass avx512_fp_sae_packed<bits<8> opc, string OpcodeStr, SDNode OpNodeRnd,
X86VectorVTInfo _, bit IsCommutable> {
defm rb: AVX512_maskable<opc, MRMSrcReg, _, (outs _.RC:$dst),
----------------
Set IsCommutable to default 0 and do not pass zeroes.
================
Comment at: lib/Target/X86/X86InstrAVX512.td:3517
@@ +3516,3 @@
+multiclass avx512_fp_scalef_p<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ X86VectorVTInfo _, bit IsCommutable> {
+ defm rr: AVX512_maskable<opc, MRMSrcReg, _, (outs _.RC:$dst),
----------------
Why do you need IsCommutable ?
http://reviews.llvm.org/D10730
EMAIL PREFERENCES
http://reviews.llvm.org/settings/panel/emailpreferences/
More information about the llvm-commits
mailing list