Skip to content

Commit baecc9e

Browse files
RKSimonRKSimon
committed
[CostModel][X86] getShuffleCost - add fallback (to half vector) for bfloat vector shuffle costs
Add initial half/bfloat broadcast shuffles test coverage (more to follow) Fixes llvm#68117 - which was stuck in a loop between getting scalarized insert/extract costs for the shuffle and then trying to convert a bfloat insert into a shuffle again......
1 parent abb9eb2 commit baecc9e

File tree

5 files changed

+364
-0
lines changed

5 files changed

+364
-0
lines changed

‎llvm/lib/Target/X86/X86TargetTransformInfo.cpp

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1481,6 +1481,10 @@ InstructionCost X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
14811481
if (Kind == TTI::SK_Broadcast)
14821482
LT.first = 1;
14831483

1484+
// Treat <X x bfloat> shuffles as <X x half>.
1485+
if (LT.second.isVector() && LT.second.getScalarType() == MVT::bf16)
1486+
LT.second = LT.second.changeVectorElementType(MVT::f16);
1487+
14841488
// Subvector extractions are free if they start at the beginning of a
14851489
// vector and cheap if the subvectors are aligned.
14861490
if (Kind == TTI::SK_ExtractSubvector && LT.second.isVector()) {

‎llvm/test/Analysis/CostModel/X86/shuffle-broadcast-codesize.ll

Lines changed: 90 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -150,6 +150,96 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
150150
ret void
151151
}
152152

153+
define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
154+
; SSE2-LABEL: 'test_vXf16'
155+
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
156+
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
157+
; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
158+
; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
159+
; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
160+
; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
161+
;
162+
; SSSE3-LABEL: 'test_vXf16'
163+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
164+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
165+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
166+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
167+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
168+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
169+
;
170+
; SSE42-LABEL: 'test_vXf16'
171+
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
172+
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
173+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
174+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
175+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
176+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
177+
;
178+
; AVX2-LABEL: 'test_vXf16'
179+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
180+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
181+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
182+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
183+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
184+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
185+
;
186+
; AVX512-LABEL: 'test_vXf16'
187+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
188+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
189+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
190+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
191+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
192+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
193+
;
194+
%V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
195+
%V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
196+
%V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
197+
%V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
198+
%V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
199+
ret void
200+
}
201+
202+
define void @test_vXbf16(<2 x bfloat> %src32, <4 x bfloat> %src64, <8 x bfloat> %src128, <16 x bfloat> %src256, <32 x bfloat> %src512) {
203+
; SSE-LABEL: 'test_vXbf16'
204+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
205+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
206+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
207+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
208+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
209+
; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
210+
;
211+
; AVX1-LABEL: 'test_vXbf16'
212+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
213+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
214+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
215+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
216+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
217+
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
218+
;
219+
; AVX2-LABEL: 'test_vXbf16'
220+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
221+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
222+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
223+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
224+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
225+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
226+
;
227+
; AVX512-LABEL: 'test_vXbf16'
228+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
229+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
230+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
231+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
232+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
233+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
234+
;
235+
%V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
236+
%V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
237+
%V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
238+
%V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
239+
%V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
240+
ret void
241+
}
242+
153243
define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512) {
154244
; SSE2-LABEL: 'test_vXi16'
155245
; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer

‎llvm/test/Analysis/CostModel/X86/shuffle-broadcast-latency.ll

Lines changed: 90 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -150,6 +150,96 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
150150
ret void
151151
}
152152

153+
define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
154+
; SSE2-LABEL: 'test_vXf16'
155+
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
156+
; SSE2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
157+
; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
158+
; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
159+
; SSE2-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
160+
; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
161+
;
162+
; SSSE3-LABEL: 'test_vXf16'
163+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
164+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
165+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
166+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
167+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
168+
; SSSE3-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
169+
;
170+
; SSE42-LABEL: 'test_vXf16'
171+
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
172+
; SSE42-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
173+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
174+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
175+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
176+
; SSE42-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
177+
;
178+
; AVX2-LABEL: 'test_vXf16'
179+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
180+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
181+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
182+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
183+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
184+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
185+
;
186+
; AVX512-LABEL: 'test_vXf16'
187+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
188+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
189+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
190+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
191+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
192+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
193+
;
194+
%V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
195+
%V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
196+
%V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
197+
%V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
198+
%V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
199+
ret void
200+
}
201+
202+
define void @test_vXbf16(<2 x bfloat> %src32, <4 x bfloat> %src64, <8 x bfloat> %src128, <16 x bfloat> %src256, <32 x bfloat> %src512) {
203+
; SSE-LABEL: 'test_vXbf16'
204+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
205+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
206+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
207+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
208+
; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
209+
; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
210+
;
211+
; AVX1-LABEL: 'test_vXbf16'
212+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
213+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
214+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
215+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
216+
; AVX1-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
217+
; AVX1-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
218+
;
219+
; AVX2-LABEL: 'test_vXbf16'
220+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
221+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
222+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
223+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
224+
; AVX2-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
225+
; AVX2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
226+
;
227+
; AVX512-LABEL: 'test_vXbf16'
228+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
229+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
230+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
231+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
232+
; AVX512-NEXT: Cost Model: Found an estimated cost of 0 for instruction: %V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
233+
; AVX512-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
234+
;
235+
%V32 = shufflevector <2 x bfloat> %src32, <2 x bfloat> undef, <2 x i32> zeroinitializer
236+
%V64 = shufflevector <4 x bfloat> %src64, <4 x bfloat> undef, <4 x i32> zeroinitializer
237+
%V128 = shufflevector <8 x bfloat> %src128, <8 x bfloat> undef, <8 x i32> zeroinitializer
238+
%V256 = shufflevector <16 x bfloat> %src256, <16 x bfloat> undef, <16 x i32> zeroinitializer
239+
%V512 = shufflevector <32 x bfloat> %src512, <32 x bfloat> undef, <32 x i32> zeroinitializer
240+
ret void
241+
}
242+
153243
define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <16 x i16> %src256, <32 x i16> %src512) {
154244
; SSE2-LABEL: 'test_vXi16'
155245
; SSE2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer

0 commit comments

Comments
 (0)