diff options
| author | Diana Picus <diana.picus@linaro.org> | 2018-01-12 11:30:45 +0000 |
|---|---|---|
| committer | Diana Picus <diana.picus@linaro.org> | 2018-01-12 11:30:45 +0000 |
| commit | e74243d4736efcecf794dd5693cd49864b2c5b03 (patch) | |
| tree | cedc4ae6c7396c3ec69de03042f19460e609ee91 /llvm/test | |
| parent | a006baa620985e7974eda829c6bf6e3d63823b97 (diff) | |
| download | bcm5719-llvm-e74243d4736efcecf794dd5693cd49864b2c5b03.tar.gz bcm5719-llvm-e74243d4736efcecf794dd5693cd49864b2c5b03.zip | |
[ARM GlobalISel] Legalize G_FMA
For hard float with VFP4, it is legal. Otherwise, we use libcalls.
This needs a bit of support in the LegalizerHelper for soft float
because we didn't handle G_FMA libcalls yet. The support is trivial, as
the only difference between G_FMA and other libcalls that we already
handle is that it has 3 input operands rather than just 2.
llvm-svn: 322366
Diffstat (limited to 'llvm/test')
| -rw-r--r-- | llvm/test/CodeGen/ARM/GlobalISel/arm-legalize-vfp4.mir | 121 |
1 files changed, 121 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/ARM/GlobalISel/arm-legalize-vfp4.mir b/llvm/test/CodeGen/ARM/GlobalISel/arm-legalize-vfp4.mir new file mode 100644 index 00000000000..5fe0d86b2b4 --- /dev/null +++ b/llvm/test/CodeGen/ARM/GlobalISel/arm-legalize-vfp4.mir @@ -0,0 +1,121 @@ +# RUN: llc -mtriple arm-linux-gnueabihf -mattr=+vfp4 -float-abi=hard -global-isel -run-pass=legalizer %s -o - | FileCheck %s -check-prefix CHECK -check-prefix HARD +# RUN: llc -mtriple arm-linux-gnueabihf -mattr=+vfp2 -float-abi=hard -global-isel -run-pass=legalizer %s -o - | FileCheck %s -check-prefix CHECK -check-prefix SOFT -check-prefix HARD-ABI +# RUN: llc -mtriple arm-linux-gnueabi -mattr=+vfp4,+soft-float -float-abi=soft -global-isel -run-pass=legalizer %s -o - | FileCheck %s -check-prefix CHECK -check-prefix SOFT -check-prefix SOFT-ABI +# RUN: llc -mtriple arm-linux-gnu -mattr=+vfp4,+soft-float -float-abi=soft -global-isel -run-pass=legalizer %s -o - | FileCheck %s -check-prefix CHECK -check-prefix SOFT -check-prefix SOFT-ABI +--- | + define void @test_fma_float() { ret void } + define void @test_fma_double() { ret void } +... +--- +name: test_fma_float +# CHECK-LABEL: name: test_fma_float +legalized: false +# CHECK: legalized: true +regBankSelected: false +selected: false +tracksRegLiveness: true +registers: + - { id: 0, class: _ } + - { id: 1, class: _ } + - { id: 2, class: _ } + - { id: 3, class: _ } +body: | + bb.0: + liveins: %r0, %r1, %r2 + + ; CHECK-DAG: [[X:%[0-9]+]]:_(s32) = COPY %r0 + ; CHECK-DAG: [[Y:%[0-9]+]]:_(s32) = COPY %r1 + ; CHECK-DAG: [[Z:%[0-9]+]]:_(s32) = COPY %r2 + %0(s32) = COPY %r0 + %1(s32) = COPY %r1 + %2(s32) = COPY %r2 + ; HARD: [[R:%[0-9]+]]:_(s32) = G_FMA [[X]], [[Y]], [[Z]] + ; SOFT-NOT: G_FMA + ; SOFT: ADJCALLSTACKDOWN + ; SOFT-ABI-DAG: %r0 = COPY [[X]] + ; SOFT-ABI-DAG: %r1 = COPY [[Y]] + ; SOFT-ABI-DAG: %r2 = COPY [[Z]] + ; SOFT-ABI: BL &fmaf, {{.*}}, implicit %r0, implicit %r1, implicit %r2, implicit-def %r0 + ; SOFT-ABI: [[R:%[0-9]+]]:_(s32) = COPY %r0 + ; HARD-ABI-DAG: %s0 = COPY [[X]] + ; HARD-ABI-DAG: %s1 = COPY [[Y]] + ; HARD-ABI-DAG: %s2 = COPY [[Z]] + ; HARD-ABI: BL &fmaf, {{.*}}, implicit %s0, implicit %s1, implicit %s2, implicit-def %s0 + ; HARD-ABI: [[R:%[0-9]+]]:_(s32) = COPY %s0 + ; SOFT: ADJCALLSTACKUP + ; SOFT-NOT: G_FMA + %3(s32) = G_FMA %0, %1, %2 + ; CHECK: %r0 = COPY [[R]] + %r0 = COPY %3(s32) + BX_RET 14, %noreg, implicit %r0 +... +--- +name: test_fma_double +# CHECK-LABEL: name: test_fma_double +legalized: false +# CHECK: legalized: true +regBankSelected: false +selected: false +tracksRegLiveness: true +registers: + - { id: 0, class: _ } + - { id: 1, class: _ } + - { id: 2, class: _ } + - { id: 3, class: _ } + - { id: 4, class: _ } + - { id: 5, class: _ } + - { id: 6, class: _ } + - { id: 7, class: _ } + - { id: 8, class: _ } +body: | + bb.0: + liveins: %r0, %r1, %r2, %r3 + + ; CHECK-DAG: [[X0:%[0-9]+]]:_(s32) = COPY %r0 + ; CHECK-DAG: [[X1:%[0-9]+]]:_(s32) = COPY %r1 + ; CHECK-DAG: [[Y0:%[0-9]+]]:_(s32) = COPY %r2 + ; CHECK-DAG: [[Y1:%[0-9]+]]:_(s32) = COPY %r3 + %0(s32) = COPY %r0 + %1(s32) = COPY %r1 + %2(s32) = COPY %r2 + %3(s32) = COPY %r3 + ; HARD-DAG: [[X:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[X0]] + ; HARD-DAG: [[Y:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[Y0]] + ; HARD-ABI-DAG: [[X:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[X0]] + ; HARD-ABI-DAG: [[Y:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[Y0]] + %4(s64) = G_MERGE_VALUES %0(s32), %1(s32) + %5(s64) = G_MERGE_VALUES %2(s32), %3(s32) + ; HARD: [[R:%[0-9]+]]:_(s64) = G_FMA [[X]], [[X]], [[Y]] + ; SOFT-NOT: G_FMA + ; SOFT: ADJCALLSTACKDOWN + ; SOFT-ABI-DAG: %r{{[0-1]}} = COPY [[X0]] + ; SOFT-ABI-DAG: %r{{[0-1]}} = COPY [[X1]] + ; SOFT-ABI-DAG: %r{{[2-3]}} = COPY [[X0]] + ; SOFT-ABI-DAG: %r{{[2-3]}} = COPY [[X1]] + ; SOFT-ABI: [[SP1:%[0-9]+]]:_(p0) = COPY %sp + ; SOFT-ABI: [[OFF1:%[0-9]+]]:_(s32) = G_CONSTANT i32 0 + ; SOFT-ABI: [[FI1:%[0-9]+]]:_(p0) = G_GEP [[SP1]], [[OFF1]](s32) + ; SOFT-ABI: G_STORE [[Y0]](s32), [[FI1]](p0){{.*}}store 8 into stack + ; SOFT-ABI: [[OFF2:%[0-9]+]]:_(s32) = G_CONSTANT i32 4 + ; SOFT-ABI: [[FI2:%[0-9]+]]:_(p0) = G_GEP [[FI1]], [[OFF2]](s32) + ; SOFT-ABI: G_STORE [[Y1]](s32), [[FI2]](p0){{.*}}store 8 into stack + ; SOFT-ABI: BL &fma, {{.*}}, implicit %r0, implicit %r1, implicit %r2, implicit %r3, implicit-def %r0, implicit-def %r1 + ; SOFT-ABI-DAG: [[R0:%[0-9]+]]:_(s32) = COPY %r0 + ; SOFT-ABI-DAG: [[R1:%[0-9]+]]:_(s32) = COPY %r1 + ; HARD-ABI-DAG: %d0 = COPY [[X]] + ; HARD-ABI-DAG: %d1 = COPY [[X]] + ; HARD-ABI-DAG: %d2 = COPY [[Y]] + ; HARD-ABI: BL &fma, {{.*}}, implicit %d0, implicit %d1, implicit %d2, implicit-def %d0 + ; HARD-ABI: [[R:%[0-9]+]]:_(s64) = COPY %d0 + ; SOFT: ADJCALLSTACKUP + ; SOFT-NOT: G_FMA + %6(s64) = G_FMA %4, %4, %5 + ; HARD: [[R0:%[0-9]+]]:_(s32), [[R1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[R]](s64) + ; HARD-ABI: [[R0:%[0-9]+]]:_(s32), [[R1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[R]](s64) + %7(s32),%8(s32) = G_UNMERGE_VALUES %6(s64) + ; CHECK-DAG: %r0 = COPY [[R0]] + ; CHECK-DAG: %r1 = COPY [[R1]] + %r0 = COPY %7(s32) + %r1 = COPY %8(s32) + BX_RET 14, %noreg, implicit %r0, implicit %r1 +... |

