R600: Implement f64 ftrunc, ffloor and fceil.

CI has instructions for these, so this fixes them for older hardware.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@211183 91177308-0d34-0410-b5e6-96231b3b80d8

3 files changed, 87 insertions, 22 deletions

diff --git a/test/CodeGen/R600/fceil.ll b/test/CodeGen/R600/fceil.ll
index b8b945f46f..b42aefa173 100644
--- a/test/CodeGen/R600/fceil.ll
+++ b/test/CodeGen/R600/fceil.ll
@@ -1,4 +1,5 @@
-; RUN: llc -march=r600 -mcpu=bonaire < %s | FileCheck -check-prefix=CI %s
+; RUN: llc -march=r600 -mcpu=bonaire < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
 
 declare double @llvm.ceil.f64(double) nounwind readnone
 declare <2 x double> @llvm.ceil.v2f64(<2 x double>) nounwind readnone
@@ -7,15 +8,33 @@ declare <4 x double> @llvm.ceil.v4f64(<4 x double>) nounwind readnone
 declare <8 x double> @llvm.ceil.v8f64(<8 x double>) nounwind readnone
 declare <16 x double> @llvm.ceil.v16f64(<16 x double>) nounwind readnone
 
-; CI-LABEL: @fceil_f64:
-; CI: V_CEIL_F64_e32
+; FUNC-LABEL: @fceil_f64:
+; CI: V_CEIL_F64_e32
+; SI: S_BFE_I32 [[SEXP:s[0-9]+]], {{s[0-9]+}}, 0xb0014
+; SI: S_ADD_I32 s{{[0-9]+}}, [[SEXP]], 0xfffffc01
+; SI: S_LSHR_B64
+; SI: S_NOT_B64
+; SI: S_AND_B64
+; SI: S_AND_B32 s{{[0-9]+}}, s{{[0-9]+}}, 0x80000000
+; SI: CMP_LT_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: CMP_GT_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: CMP_GT_F64
+; SI: CNDMASK_B32
+; SI: CMP_NE_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: V_ADD_F64
 define void @fceil_f64(double addrspace(1)* %out, double %x) {
   %y = call double @llvm.ceil.f64(double %x) nounwind readnone
   store double %y, double addrspace(1)* %out
   ret void
 }
 
-; CI-LABEL: @fceil_v2f64:
+; FUNC-LABEL: @fceil_v2f64:
 ; CI: V_CEIL_F64_e32
 ; CI: V_CEIL_F64_e32
 define void @fceil_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
@@ -24,7 +43,7 @@ define void @fceil_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
   ret void
 }
 
-; FIXME-CI-LABEL: @fceil_v3f64:
+; FIXME-FUNC-LABEL: @fceil_v3f64:
 ; FIXME-CI: V_CEIL_F64_e32
 ; FIXME-CI: V_CEIL_F64_e32
 ; FIXME-CI: V_CEIL_F64_e32
@@ -34,7 +53,7 @@ define void @fceil_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
 ;   ret void
 ; }
 
-; CI-LABEL: @fceil_v4f64:
+; FUNC-LABEL: @fceil_v4f64:
 ; CI: V_CEIL_F64_e32
 ; CI: V_CEIL_F64_e32
 ; CI: V_CEIL_F64_e32
@@ -45,7 +64,7 @@ define void @fceil_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %x) {
   ret void
 }
 
-; CI-LABEL: @fceil_v8f64:
+; FUNC-LABEL: @fceil_v8f64:
 ; CI: V_CEIL_F64_e32
 ; CI: V_CEIL_F64_e32
 ; CI: V_CEIL_F64_e32
@@ -60,7 +79,7 @@ define void @fceil_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %x) {
   ret void
 }
 
-; CI-LABEL: @fceil_v16f64:
+; FUNC-LABEL: @fceil_v16f64:
 ; CI: V_CEIL_F64_e32
 ; CI: V_CEIL_F64_e32
 ; CI: V_CEIL_F64_e32
diff --git a/test/CodeGen/R600/ffloor.ll b/test/CodeGen/R600/ffloor.ll
index 51d2b89615..31c6116988 100644
--- a/test/CodeGen/R600/ffloor.ll
+++ b/test/CodeGen/R600/ffloor.ll
@@ -1,4 +1,5 @@
-; RUN: llc -march=r600 -mcpu=bonaire < %s | FileCheck -check-prefix=CI %s
+; RUN: llc -march=r600 -mcpu=bonaire < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
 
 declare double @llvm.floor.f64(double) nounwind readnone
 declare <2 x double> @llvm.floor.v2f64(<2 x double>) nounwind readnone
@@ -7,15 +8,34 @@ declare <4 x double> @llvm.floor.v4f64(<4 x double>) nounwind readnone
 declare <8 x double> @llvm.floor.v8f64(<8 x double>) nounwind readnone
 declare <16 x double> @llvm.floor.v16f64(<16 x double>) nounwind readnone
 
-; CI-LABEL: @ffloor_f64:
+; FUNC-LABEL: @ffloor_f64:
 ; CI: V_FLOOR_F64_e32
+
+; SI: S_BFE_I32 [[SEXP:s[0-9]+]], {{s[0-9]+}}, 0xb0014
+; SI: S_ADD_I32 s{{[0-9]+}}, [[SEXP]], 0xfffffc01
+; SI: S_LSHR_B64
+; SI: S_NOT_B64
+; SI: S_AND_B64
+; SI: S_AND_B32 s{{[0-9]+}}, s{{[0-9]+}}, 0x80000000
+; SI: CMP_LT_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: CMP_GT_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: CMP_LT_F64
+; SI: CNDMASK_B32
+; SI: CMP_NE_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: V_ADD_F64
 define void @ffloor_f64(double addrspace(1)* %out, double %x) {
   %y = call double @llvm.floor.f64(double %x) nounwind readnone
   store double %y, double addrspace(1)* %out
   ret void
 }
 
-; CI-LABEL: @ffloor_v2f64:
+; FUNC-LABEL: @ffloor_v2f64:
 ; CI: V_FLOOR_F64_e32
 ; CI: V_FLOOR_F64_e32
 define void @ffloor_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
@@ -24,7 +44,7 @@ define void @ffloor_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
   ret void
 }
 
-; FIXME-CI-LABEL: @ffloor_v3f64:
+; FIXME-FUNC-LABEL: @ffloor_v3f64:
 ; FIXME-CI: V_FLOOR_F64_e32
 ; FIXME-CI: V_FLOOR_F64_e32
 ; FIXME-CI: V_FLOOR_F64_e32
@@ -34,7 +54,7 @@ define void @ffloor_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
 ;   ret void
 ; }
 
-; CI-LABEL: @ffloor_v4f64:
+; FUNC-LABEL: @ffloor_v4f64:
 ; CI: V_FLOOR_F64_e32
 ; CI: V_FLOOR_F64_e32
 ; CI: V_FLOOR_F64_e32
@@ -45,7 +65,7 @@ define void @ffloor_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %x) {
   ret void
 }
 
-; CI-LABEL: @ffloor_v8f64:
+; FUNC-LABEL: @ffloor_v8f64:
 ; CI: V_FLOOR_F64_e32
 ; CI: V_FLOOR_F64_e32
 ; CI: V_FLOOR_F64_e32
@@ -60,7 +80,7 @@ define void @ffloor_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %x) {
   ret void
 }
 
-; CI-LABEL: @ffloor_v16f64:
+; FUNC-LABEL: @ffloor_v16f64:
 ; CI: V_FLOOR_F64_e32
 ; CI: V_FLOOR_F64_e32
 ; CI: V_FLOOR_F64_e32
diff --git a/test/CodeGen/R600/ftrunc.ll b/test/CodeGen/R600/ftrunc.ll
index 6b235ffbd9..3cd1deb921 100644
--- a/test/CodeGen/R600/ftrunc.ll
+++ b/test/CodeGen/R600/ftrunc.ll
@@ -1,4 +1,5 @@
-; RUN: llc -march=r600 -mcpu=bonaire < %s | FileCheck -check-prefix=CI %s
+; RUN: llc -march=r600 -mcpu=bonaire < %s | FileCheck -check-prefix=CI -check-prefix=FUNC %s
+; RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
 
 declare double @llvm.trunc.f64(double) nounwind readnone
 declare <2 x double> @llvm.trunc.v2f64(<2 x double>) nounwind readnone
@@ -7,15 +8,40 @@ declare <4 x double> @llvm.trunc.v4f64(<4 x double>) nounwind readnone
 declare <8 x double> @llvm.trunc.v8f64(<8 x double>) nounwind readnone
 declare <16 x double> @llvm.trunc.v16f64(<16 x double>) nounwind readnone
 
-; CI-LABEL: @ftrunc_f64:
+; FUNC-LABEL: @v_ftrunc_f64:
 ; CI: V_TRUNC_F64_e32
+; SI: V_BFE_I32 {{v[0-9]+}}, {{v[0-9]+}}, 20, 11
+; SI: S_ENDPGM
+define void @v_ftrunc_f64(double addrspace(1)* %out, double addrspace(1)* %in) {
+  %x = load double addrspace(1)* %in, align 8
+  %y = call double @llvm.trunc.f64(double %x) nounwind readnone
+  store double %y, double addrspace(1)* %out, align 8
+  ret void
+}
+
+; FUNC-LABEL: @ftrunc_f64:
+; CI: V_TRUNC_F64_e32
+
+; SI: S_BFE_I32 [[SEXP:s[0-9]+]], {{s[0-9]+}}, 0xb0014
+; SI: S_ADD_I32 s{{[0-9]+}}, [[SEXP]], 0xfffffc01
+; SI: S_LSHR_B64
+; SI: S_NOT_B64
+; SI: S_AND_B64
+; SI: S_AND_B32 s{{[0-9]+}}, s{{[0-9]+}}, 0x80000000
+; SI: CMP_LT_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: CMP_GT_I32
+; SI: CNDMASK_B32
+; SI: CNDMASK_B32
+; SI: S_ENDPGM
 define void @ftrunc_f64(double addrspace(1)* %out, double %x) {
   %y = call double @llvm.trunc.f64(double %x) nounwind readnone
   store double %y, double addrspace(1)* %out
   ret void
 }
 
-; CI-LABEL: @ftrunc_v2f64:
+; FUNC-LABEL: @ftrunc_v2f64:
 ; CI: V_TRUNC_F64_e32
 ; CI: V_TRUNC_F64_e32
 define void @ftrunc_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
@@ -24,7 +50,7 @@ define void @ftrunc_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
   ret void
 }
 
-; FIXME-CI-LABEL: @ftrunc_v3f64:
+; FIXME-FUNC-LABEL: @ftrunc_v3f64:
 ; FIXME-CI: V_TRUNC_F64_e32
 ; FIXME-CI: V_TRUNC_F64_e32
 ; FIXME-CI: V_TRUNC_F64_e32
@@ -34,7 +60,7 @@ define void @ftrunc_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %x) {
 ;   ret void
 ; }
 
-; CI-LABEL: @ftrunc_v4f64:
+; FUNC-LABEL: @ftrunc_v4f64:
 ; CI: V_TRUNC_F64_e32
 ; CI: V_TRUNC_F64_e32
 ; CI: V_TRUNC_F64_e32
@@ -45,7 +71,7 @@ define void @ftrunc_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %x) {
   ret void
 }
 
-; CI-LABEL: @ftrunc_v8f64:
+; FUNC-LABEL: @ftrunc_v8f64:
 ; CI: V_TRUNC_F64_e32
 ; CI: V_TRUNC_F64_e32
 ; CI: V_TRUNC_F64_e32
@@ -60,7 +86,7 @@ define void @ftrunc_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %x) {
   ret void
 }
 
-; CI-LABEL: @ftrunc_v16f64:
+; FUNC-LABEL: @ftrunc_v16f64:
 ; CI: V_TRUNC_F64_e32
 ; CI: V_TRUNC_F64_e32
 ; CI: V_TRUNC_F64_e32