@@ -160,7 +160,7 @@ specification we add four operations:
160160declare <[NUMo] x [TYo]> @dx.op.matvecmul.v[NUMo][TYo].v[NUMi][TYi](
161161 immarg i32 ; opcode
162162 <[NUMi] x [TYi]>, ; input vector
163- immarg i1, ; input signed op kind
163+ immarg i1, ; is input unsigned
164164 immarg i32, ; input interpretation
165165 %dx.types.Handle, ; matrix resource
166166 i32, ; matrix offset
@@ -170,12 +170,12 @@ declare <[NUMo] x [TYo]> @dx.op.matvecmul.v[NUMo][TYo].v[NUMi][TYi](
170170 immarg i32, ; matrix layout
171171 immarg i1, ; matrix transpose
172172 i32, ; matrix stride
173- immarg i1) ; output signed op kind
173+ immarg i1) ; is output unsigned
174174
175175declare <[NUMo] x [TYo]> @dx.op.matvecmuladd.v[NUMo][TYo].v[NUMi][TYi](
176176 immarg i32 ; opcode
177177 <[NUMi] x [TYi]>, ; input vector
178- immarg i1, ; input signed op kind
178+ immarg i1, ; is input unsigned
179179 immarg i32, ; input interpretation
180180 %dx.types.Handle, ; matrix resource
181181 i32, ; matrix offset
@@ -188,7 +188,7 @@ declare <[NUMo] x [TYo]> @dx.op.matvecmuladd.v[NUMo][TYo].v[NUMi][TYi](
188188 %dx.types.Handle, ; bias vector resource
189189 i32, ; bias vector offset
190190 immarg i32, ; bias vector interpretation
191- immarg i1) ; output signed op kind
191+ immarg i1) ; is output unsigned
192192```
193193
194194#### Overview
@@ -222,9 +222,9 @@ types are types such as **SignedInt8x4Packed** where each 32-bit element of the
222222vector corresponds to four 8-bit signed integers. See [ Type Interpretations]
223223for details.
224224
225- The ** input signed op kind ** is a value from the existing ` SignedOpKind ` enum.
226- ` 0 ` indicates that the input vector is a float or signed integer, ` 1 ` indicates
227- that the input vector is an unsigned integer.
225+ The ** is input unsigned ** is a boolean value; ` false ` indicates that the input
226+ vector is a float or signed integer, ` true ` indicates that the input vector is
227+ an unsigned integer.
228228
229229
230230##### Matrix
@@ -274,9 +274,9 @@ This operation returns a vector of size `NUMo` and contains elements of type
274274` TYo ` . The result vector does not have an interpretation parameter, its type is
275275the declared type.
276276
277- The ** output signed op kind ** is a value from the existing ` SignedOpKind ` enum.
278- ` 0 ` indicates that the output vector is a float or signed integer, ` 1 ` indicates
279- that the input vector is an unsigned integer.
277+ The ** is output unsigned ** is a boolean value; ` false ` indicates that the input
278+ vector is a float or signed integer, ` true ` indicates that the input vector is
279+ an unsigned integer.
280280
281281#### Validation
282282
@@ -533,7 +533,7 @@ Packed Case:
533533 OPCODE,
534534 %inputVector,
535535 16, ; input interpretation - ComponentType::PackedS8x32
536- 0 , ; input signed op kind = 0 = signed
536+ false , ; is input unsigned = false = signed
537537 %matrixResource,
538538 0, ; matrix offset
539539 19, ; matrix interpretation - ComponentType::I8
@@ -542,7 +542,7 @@ Packed Case:
542542 2, ; matrix layout - MulOptimal
543543 0, ; matrix transpose - false
544544 0, ; matrix stride
545- 0); ; output signed op kind = 0 = signed
545+ 0); ; is output unsigned = false = signed
546546```
547547
548548Non-Packed Case:
@@ -556,7 +556,7 @@ Non-Packed Case:
556556 OPCODE,
557557 %inputVector,
558558 19, ; input interpretation - ComponentType::I8
559- 0, ; input signed op kind = 0 = signed
559+ 0, ; is input unsigned = false = signed
560560 %matrixResource,
561561 0, ; matrix offset
562562 5, ; matrix interpretation - ComponentType::I8
@@ -565,7 +565,7 @@ Non-Packed Case:
565565 2, ; matrix layout - MulOptimal
566566 0, ; matrix transpose - false
567567 0, ; matrix stride
568- 0) ; output signed op kind = 0 = signed
568+ 0) ; is output unsigned = false = signed
569569```
570570
571571#### Precision Requirements
0 commit comments