================
@@ -722,6 +722,37 @@ let hasSideEffects = false in {
defm CVT_f16x2 : CVT_FROM_FLOAT_V2_SM80<"f16x2", Int32Regs>;
defm CVT_bf16x2 : CVT_FROM_FLOAT_V2_SM80<"bf16x2", Int32Regs>;
+
+ // FP8 conversions.
+ multiclass CVT_TO_F8X2<string F8Name> {
+ def _f32 :
+ NVPTXInst<(outs Int16Regs:$dst),
+ (ins Float32Regs:$src1, Float32Regs:$src2, CvtMode:$mode),
+ !strconcat("cvt${mode:base}.satfinite${mode:relu}.",
+ F8Name, "x2.f32 \t$dst, $src1, $src2;"), []>,
+ Requires<[hasPTX<81>, hasSM<89>]>;
+ def _f16x2 :
+ NVPTXInst<(outs Int16Regs:$dst),
+ (ins Int32Regs:$src, CvtMode:$mode),
+ !strconcat("cvt${mode:base}.satfinite${mode:relu}.",
+ F8Name, "x2.f16x2 \t$dst, $src;"), []>,
+ Requires<[hasPTX<81>, hasSM<89>]>;
+ }
+
+ defm CVT_e4m3x2 : CVT_TO_F8X2<"e4m3">;
+ defm CVT_e5m2x2 : CVT_TO_F8X2<"e5m2">;
+
+ multiclass CVT_FROM_F8X2<string F8Name> {
+ def x2 :
+ NVPTXInst<(outs Int32Regs:$dst),
----------------
sergey-kozub wrote:
This one is confusing, "x2" is only there only because (for some reason) I
cannot specify an empty suffix.
All the classes in this file use the `CVT_<to>_<from>` pattern. I need to
define two classes: CVT_f16x2_e4m3x2 and CVT_f16x2_e5m2.
You're suggesting to rename to `_f16x2` but I guess "multiclass" only supports
appending the suffixes, whereas I need "f16x2" to appear in the middle of the
class name (in order to conform to the common naming pattern).
Also, "outs" is Int32Regs, and I cannot specify <2xf16> here. This is in line
with the other class definitions in this file.
https://github.com/llvm/llvm-project/pull/102969
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
