e2d7cd685d
This patch makes the dead_on_return parameter attribute optionally require a number
of bytes to be passed in to specify the number of bytes known to be dead
upon function return/unwind. This is aimed at enabling annotating the
this pointer in C++ destructors with dead_on_return in clang. We need
this to handle cases like the following:
```
struct X {
int n;
~X() {
this[n].n = 0;
}
};
void f() {
X xs[] = {42, -1};
}
```
Where we only certain that sizeof(X) bytes are dead upon return of ~X.
Otherwise DSE would be able to eliminate the store in ~X which would not
be correct.
This patch only does the wiring within IR. Future patches will make
clang emit correct sizing information and update DSE to only delete
stores to objects marked dead_on_return that are provably in bounds of
the number of bytes specified to be dead_on_return.
Reviewers: nikic, alinas, antoniofrighetto
Pull Request: https://github.com/llvm/llvm-project/pull/171712
1431 lines
113 KiB
C
1431 lines
113 KiB
C
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 4
|
|
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
|
|
// RUN: -o - | FileCheck %s --check-prefix=FULL
|
|
|
|
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
|
|
// RUN: -complex-range=basic -o - | FileCheck %s --check-prefix=BASIC
|
|
|
|
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
|
|
// RUN: -fno-cx-limited-range -o - | FileCheck %s --check-prefix=FULL
|
|
|
|
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
|
|
// RUN: -complex-range=improved -o - | FileCheck %s --check-prefix=IMPRVD
|
|
|
|
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
|
|
// RUN: -complex-range=promoted -o - | FileCheck %s --check-prefix=PRMTD
|
|
|
|
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
|
|
// RUN: -complex-range=full -o - | FileCheck %s --check-prefix=FULL
|
|
|
|
// RUN: %clang_cc1 -triple x86_64-windows-pc -complex-range=promoted \
|
|
// RUN: -emit-llvm -o - %s | FileCheck %s --check-prefix=X86WINPRMTD
|
|
|
|
// Fast math
|
|
// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu \
|
|
// RUN: -ffast-math -complex-range=basic -emit-llvm -o - %s \
|
|
// RUN: | FileCheck %s --check-prefix=BASIC_FAST
|
|
|
|
// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu \
|
|
// RUN: -ffast-math -complex-range=full -emit-llvm -o - %s \
|
|
// RUN: | FileCheck %s --check-prefix=FULL_FAST
|
|
|
|
// RUN: %clang_cc1 %s -O0 -emit-llvm -triple x86_64-unknown-unknown \
|
|
// RUN: -fno-cx-fortran-rules -o - | FileCheck %s --check-prefix=FULL
|
|
|
|
// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu \
|
|
// RUN: -ffast-math -complex-range=improved -emit-llvm -o - %s \
|
|
// RUN: | FileCheck %s --check-prefix=IMPRVD_FAST
|
|
|
|
// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu \
|
|
// RUN: -ffast-math -complex-range=promoted -emit-llvm -o - %s \
|
|
// RUN: | FileCheck %s --check-prefix=PRMTD_FAST
|
|
|
|
// FULL-LABEL: define dso_local <2 x half> @divf16(
|
|
// FULL-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// FULL-NEXT: entry:
|
|
// FULL-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[COERCE:%.*]] = alloca { float, float }, align 4
|
|
// FULL-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// FULL-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// FULL-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// FULL-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// FULL-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// FULL-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// FULL-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// FULL-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// FULL-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// FULL-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// FULL-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// FULL-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// FULL-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// FULL-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// FULL-NEXT: [[CALL:%.*]] = call <2 x float> @__divsc3(float noundef [[EXT]], float noundef [[EXT1]], float noundef [[EXT2]], float noundef [[EXT3]]) #[[ATTR1:[0-9]+]]
|
|
// FULL-NEXT: store <2 x float> [[CALL]], ptr [[COERCE]], align 4
|
|
// FULL-NEXT: [[COERCE_REALP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 0
|
|
// FULL-NEXT: [[COERCE_REAL:%.*]] = load float, ptr [[COERCE_REALP]], align 4
|
|
// FULL-NEXT: [[COERCE_IMAGP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 1
|
|
// FULL-NEXT: [[COERCE_IMAG:%.*]] = load float, ptr [[COERCE_IMAGP]], align 4
|
|
// FULL-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[COERCE_REAL]] to half
|
|
// FULL-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[COERCE_IMAG]] to half
|
|
// FULL-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// FULL-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// FULL-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// FULL-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// FULL-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// FULL-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// BASIC-LABEL: define dso_local <2 x half> @divf16(
|
|
// BASIC-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// BASIC-NEXT: entry:
|
|
// BASIC-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// BASIC-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// BASIC-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// BASIC-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// BASIC-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// BASIC-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// BASIC-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// BASIC-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// BASIC-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// BASIC-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// BASIC-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// BASIC-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// BASIC-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// BASIC-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// BASIC-NEXT: [[TMP0:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// BASIC-NEXT: [[TMP1:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// BASIC-NEXT: [[TMP2:%.*]] = fadd float [[TMP0]], [[TMP1]]
|
|
// BASIC-NEXT: [[TMP3:%.*]] = fmul float [[EXT2]], [[EXT2]]
|
|
// BASIC-NEXT: [[TMP4:%.*]] = fmul float [[EXT3]], [[EXT3]]
|
|
// BASIC-NEXT: [[TMP5:%.*]] = fadd float [[TMP3]], [[TMP4]]
|
|
// BASIC-NEXT: [[TMP6:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// BASIC-NEXT: [[TMP7:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// BASIC-NEXT: [[TMP8:%.*]] = fsub float [[TMP6]], [[TMP7]]
|
|
// BASIC-NEXT: [[TMP9:%.*]] = fdiv float [[TMP2]], [[TMP5]]
|
|
// BASIC-NEXT: [[TMP10:%.*]] = fdiv float [[TMP8]], [[TMP5]]
|
|
// BASIC-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP9]] to half
|
|
// BASIC-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[TMP10]] to half
|
|
// BASIC-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// BASIC-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// BASIC-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// BASIC-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// BASIC-NEXT: [[TMP11:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// BASIC-NEXT: ret <2 x half> [[TMP11]]
|
|
//
|
|
// IMPRVD-LABEL: define dso_local <2 x half> @divf16(
|
|
// IMPRVD-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// IMPRVD-NEXT: entry:
|
|
// IMPRVD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// IMPRVD-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// IMPRVD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// IMPRVD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// IMPRVD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// IMPRVD-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// IMPRVD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// IMPRVD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// IMPRVD-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// IMPRVD-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// IMPRVD-NEXT: [[TMP0:%.*]] = call float @llvm.fabs.f32(float [[EXT2]])
|
|
// IMPRVD-NEXT: [[TMP1:%.*]] = call float @llvm.fabs.f32(float [[EXT3]])
|
|
// IMPRVD-NEXT: [[ABS_CMP:%.*]] = fcmp ugt float [[TMP0]], [[TMP1]]
|
|
// IMPRVD-NEXT: br i1 [[ABS_CMP]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI:%.*]]
|
|
// IMPRVD: abs_rhsr_greater_or_equal_abs_rhsi:
|
|
// IMPRVD-NEXT: [[TMP2:%.*]] = fdiv float [[EXT3]], [[EXT2]]
|
|
// IMPRVD-NEXT: [[TMP3:%.*]] = fmul float [[TMP2]], [[EXT3]]
|
|
// IMPRVD-NEXT: [[TMP4:%.*]] = fadd float [[EXT2]], [[TMP3]]
|
|
// IMPRVD-NEXT: [[TMP5:%.*]] = fmul float [[EXT1]], [[TMP2]]
|
|
// IMPRVD-NEXT: [[TMP6:%.*]] = fadd float [[EXT]], [[TMP5]]
|
|
// IMPRVD-NEXT: [[TMP7:%.*]] = fdiv float [[TMP6]], [[TMP4]]
|
|
// IMPRVD-NEXT: [[TMP8:%.*]] = fmul float [[EXT]], [[TMP2]]
|
|
// IMPRVD-NEXT: [[TMP9:%.*]] = fsub float [[EXT1]], [[TMP8]]
|
|
// IMPRVD-NEXT: [[TMP10:%.*]] = fdiv float [[TMP9]], [[TMP4]]
|
|
// IMPRVD-NEXT: br label [[COMPLEX_DIV:%.*]]
|
|
// IMPRVD: abs_rhsr_less_than_abs_rhsi:
|
|
// IMPRVD-NEXT: [[TMP11:%.*]] = fdiv float [[EXT2]], [[EXT3]]
|
|
// IMPRVD-NEXT: [[TMP12:%.*]] = fmul float [[TMP11]], [[EXT2]]
|
|
// IMPRVD-NEXT: [[TMP13:%.*]] = fadd float [[EXT3]], [[TMP12]]
|
|
// IMPRVD-NEXT: [[TMP14:%.*]] = fmul float [[EXT]], [[TMP11]]
|
|
// IMPRVD-NEXT: [[TMP15:%.*]] = fadd float [[TMP14]], [[EXT1]]
|
|
// IMPRVD-NEXT: [[TMP16:%.*]] = fdiv float [[TMP15]], [[TMP13]]
|
|
// IMPRVD-NEXT: [[TMP17:%.*]] = fmul float [[EXT1]], [[TMP11]]
|
|
// IMPRVD-NEXT: [[TMP18:%.*]] = fsub float [[TMP17]], [[EXT]]
|
|
// IMPRVD-NEXT: [[TMP19:%.*]] = fdiv float [[TMP18]], [[TMP13]]
|
|
// IMPRVD-NEXT: br label [[COMPLEX_DIV]]
|
|
// IMPRVD: complex_div:
|
|
// IMPRVD-NEXT: [[TMP20:%.*]] = phi float [ [[TMP7]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP16]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD-NEXT: [[TMP21:%.*]] = phi float [ [[TMP10]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP19]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP20]] to half
|
|
// IMPRVD-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[TMP21]] to half
|
|
// IMPRVD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// IMPRVD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// IMPRVD-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[TMP22:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// IMPRVD-NEXT: ret <2 x half> [[TMP22]]
|
|
//
|
|
// PRMTD-LABEL: define dso_local <2 x half> @divf16(
|
|
// PRMTD-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// PRMTD-NEXT: entry:
|
|
// PRMTD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// PRMTD-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// PRMTD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// PRMTD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// PRMTD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// PRMTD-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// PRMTD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// PRMTD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// PRMTD-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// PRMTD-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// PRMTD-NEXT: [[TMP0:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// PRMTD-NEXT: [[TMP1:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// PRMTD-NEXT: [[TMP2:%.*]] = fadd float [[TMP0]], [[TMP1]]
|
|
// PRMTD-NEXT: [[TMP3:%.*]] = fmul float [[EXT2]], [[EXT2]]
|
|
// PRMTD-NEXT: [[TMP4:%.*]] = fmul float [[EXT3]], [[EXT3]]
|
|
// PRMTD-NEXT: [[TMP5:%.*]] = fadd float [[TMP3]], [[TMP4]]
|
|
// PRMTD-NEXT: [[TMP6:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// PRMTD-NEXT: [[TMP7:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// PRMTD-NEXT: [[TMP8:%.*]] = fsub float [[TMP6]], [[TMP7]]
|
|
// PRMTD-NEXT: [[TMP9:%.*]] = fdiv float [[TMP2]], [[TMP5]]
|
|
// PRMTD-NEXT: [[TMP10:%.*]] = fdiv float [[TMP8]], [[TMP5]]
|
|
// PRMTD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP9]] to half
|
|
// PRMTD-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[TMP10]] to half
|
|
// PRMTD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// PRMTD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// PRMTD-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[TMP11:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// PRMTD-NEXT: ret <2 x half> [[TMP11]]
|
|
//
|
|
// X86WINPRMTD-LABEL: define dso_local i32 @divf16(
|
|
// X86WINPRMTD-SAME: i32 noundef [[A_COERCE:%.*]], i32 noundef [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// X86WINPRMTD-NEXT: entry:
|
|
// X86WINPRMTD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: store i32 [[A_COERCE]], ptr [[A]], align 2
|
|
// X86WINPRMTD-NEXT: store i32 [[B_COERCE]], ptr [[B]], align 2
|
|
// X86WINPRMTD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// X86WINPRMTD-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// X86WINPRMTD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// X86WINPRMTD-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// X86WINPRMTD-NEXT: [[TMP0:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// X86WINPRMTD-NEXT: [[TMP1:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// X86WINPRMTD-NEXT: [[TMP2:%.*]] = fadd float [[TMP0]], [[TMP1]]
|
|
// X86WINPRMTD-NEXT: [[TMP3:%.*]] = fmul float [[EXT2]], [[EXT2]]
|
|
// X86WINPRMTD-NEXT: [[TMP4:%.*]] = fmul float [[EXT3]], [[EXT3]]
|
|
// X86WINPRMTD-NEXT: [[TMP5:%.*]] = fadd float [[TMP3]], [[TMP4]]
|
|
// X86WINPRMTD-NEXT: [[TMP6:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// X86WINPRMTD-NEXT: [[TMP7:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// X86WINPRMTD-NEXT: [[TMP8:%.*]] = fsub float [[TMP6]], [[TMP7]]
|
|
// X86WINPRMTD-NEXT: [[TMP9:%.*]] = fdiv float [[TMP2]], [[TMP5]]
|
|
// X86WINPRMTD-NEXT: [[TMP10:%.*]] = fdiv float [[TMP8]], [[TMP5]]
|
|
// X86WINPRMTD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP9]] to half
|
|
// X86WINPRMTD-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[TMP10]] to half
|
|
// X86WINPRMTD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[TMP11:%.*]] = load i32, ptr [[RETVAL]], align 2
|
|
// X86WINPRMTD-NEXT: ret i32 [[TMP11]]
|
|
//
|
|
// BASIC_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @divf16(
|
|
// BASIC_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// BASIC_FAST-NEXT: entry:
|
|
// BASIC_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// BASIC_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// BASIC_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// BASIC_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// BASIC_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// BASIC_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// BASIC_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// BASIC_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// BASIC_FAST-NEXT: [[TMP0:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT2]]
|
|
// BASIC_FAST-NEXT: [[TMP1:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT3]]
|
|
// BASIC_FAST-NEXT: [[TMP2:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP0]], [[TMP1]]
|
|
// BASIC_FAST-NEXT: [[TMP3:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT2]], [[EXT2]]
|
|
// BASIC_FAST-NEXT: [[TMP4:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT3]], [[EXT3]]
|
|
// BASIC_FAST-NEXT: [[TMP5:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP3]], [[TMP4]]
|
|
// BASIC_FAST-NEXT: [[TMP6:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT2]]
|
|
// BASIC_FAST-NEXT: [[TMP7:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT3]]
|
|
// BASIC_FAST-NEXT: [[TMP8:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[TMP6]], [[TMP7]]
|
|
// BASIC_FAST-NEXT: [[TMP9:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP2]], [[TMP5]]
|
|
// BASIC_FAST-NEXT: [[TMP10:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP8]], [[TMP5]]
|
|
// BASIC_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP9]] to half
|
|
// BASIC_FAST-NEXT: [[UNPROMOTION4:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP10]] to half
|
|
// BASIC_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// BASIC_FAST-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[TMP11:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// BASIC_FAST-NEXT: ret <2 x half> [[TMP11]]
|
|
//
|
|
// FULL_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @divf16(
|
|
// FULL_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// FULL_FAST-NEXT: entry:
|
|
// FULL_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[COERCE:%.*]] = alloca { float, float }, align 4
|
|
// FULL_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// FULL_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// FULL_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// FULL_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// FULL_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// FULL_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// FULL_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// FULL_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// FULL_FAST-NEXT: [[CALL:%.*]] = call reassoc nnan ninf nsz arcp afn nofpclass(nan inf) <2 x float> @__divsc3(float noundef nofpclass(nan inf) [[EXT]], float noundef nofpclass(nan inf) [[EXT1]], float noundef nofpclass(nan inf) [[EXT2]], float noundef nofpclass(nan inf) [[EXT3]]) #[[ATTR1:[0-9]+]]
|
|
// FULL_FAST-NEXT: store <2 x float> [[CALL]], ptr [[COERCE]], align 4
|
|
// FULL_FAST-NEXT: [[COERCE_REALP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[COERCE_REAL:%.*]] = load float, ptr [[COERCE_REALP]], align 4
|
|
// FULL_FAST-NEXT: [[COERCE_IMAGP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[COERCE_IMAG:%.*]] = load float, ptr [[COERCE_IMAGP]], align 4
|
|
// FULL_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[COERCE_REAL]] to half
|
|
// FULL_FAST-NEXT: [[UNPROMOTION4:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[COERCE_IMAG]] to half
|
|
// FULL_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// FULL_FAST-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// FULL_FAST-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// IMPRVD_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @divf16(
|
|
// IMPRVD_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// IMPRVD_FAST-NEXT: entry:
|
|
// IMPRVD_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// IMPRVD_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// IMPRVD_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// IMPRVD_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// IMPRVD_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// IMPRVD_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// IMPRVD_FAST-NEXT: [[TMP0:%.*]] = call reassoc nnan ninf nsz arcp afn float @llvm.fabs.f32(float [[EXT2]])
|
|
// IMPRVD_FAST-NEXT: [[TMP1:%.*]] = call reassoc nnan ninf nsz arcp afn float @llvm.fabs.f32(float [[EXT3]])
|
|
// IMPRVD_FAST-NEXT: [[ABS_CMP:%.*]] = fcmp reassoc nnan ninf nsz arcp afn ugt float [[TMP0]], [[TMP1]]
|
|
// IMPRVD_FAST-NEXT: br i1 [[ABS_CMP]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI:%.*]]
|
|
// IMPRVD_FAST: abs_rhsr_greater_or_equal_abs_rhsi:
|
|
// IMPRVD_FAST-NEXT: [[TMP2:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[EXT3]], [[EXT2]]
|
|
// IMPRVD_FAST-NEXT: [[TMP3:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[TMP2]], [[EXT3]]
|
|
// IMPRVD_FAST-NEXT: [[TMP4:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[EXT2]], [[TMP3]]
|
|
// IMPRVD_FAST-NEXT: [[TMP5:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[TMP2]]
|
|
// IMPRVD_FAST-NEXT: [[TMP6:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[EXT]], [[TMP5]]
|
|
// IMPRVD_FAST-NEXT: [[TMP7:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP6]], [[TMP4]]
|
|
// IMPRVD_FAST-NEXT: [[TMP8:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[TMP2]]
|
|
// IMPRVD_FAST-NEXT: [[TMP9:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[EXT1]], [[TMP8]]
|
|
// IMPRVD_FAST-NEXT: [[TMP10:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP9]], [[TMP4]]
|
|
// IMPRVD_FAST-NEXT: br label [[COMPLEX_DIV:%.*]]
|
|
// IMPRVD_FAST: abs_rhsr_less_than_abs_rhsi:
|
|
// IMPRVD_FAST-NEXT: [[TMP11:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[EXT2]], [[EXT3]]
|
|
// IMPRVD_FAST-NEXT: [[TMP12:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[TMP11]], [[EXT2]]
|
|
// IMPRVD_FAST-NEXT: [[TMP13:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[EXT3]], [[TMP12]]
|
|
// IMPRVD_FAST-NEXT: [[TMP14:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[TMP11]]
|
|
// IMPRVD_FAST-NEXT: [[TMP15:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP14]], [[EXT1]]
|
|
// IMPRVD_FAST-NEXT: [[TMP16:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP15]], [[TMP13]]
|
|
// IMPRVD_FAST-NEXT: [[TMP17:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[TMP11]]
|
|
// IMPRVD_FAST-NEXT: [[TMP18:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[TMP17]], [[EXT]]
|
|
// IMPRVD_FAST-NEXT: [[TMP19:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP18]], [[TMP13]]
|
|
// IMPRVD_FAST-NEXT: br label [[COMPLEX_DIV]]
|
|
// IMPRVD_FAST: complex_div:
|
|
// IMPRVD_FAST-NEXT: [[TMP20:%.*]] = phi reassoc nnan ninf nsz arcp afn float [ [[TMP7]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP16]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD_FAST-NEXT: [[TMP21:%.*]] = phi reassoc nnan ninf nsz arcp afn float [ [[TMP10]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP19]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP20]] to half
|
|
// IMPRVD_FAST-NEXT: [[UNPROMOTION4:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP21]] to half
|
|
// IMPRVD_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[TMP22:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// IMPRVD_FAST-NEXT: ret <2 x half> [[TMP22]]
|
|
//
|
|
// PRMTD_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @divf16(
|
|
// PRMTD_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// PRMTD_FAST-NEXT: entry:
|
|
// PRMTD_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// PRMTD_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// PRMTD_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// PRMTD_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// PRMTD_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// PRMTD_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// PRMTD_FAST-NEXT: [[TMP0:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT2]]
|
|
// PRMTD_FAST-NEXT: [[TMP1:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT3]]
|
|
// PRMTD_FAST-NEXT: [[TMP2:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP0]], [[TMP1]]
|
|
// PRMTD_FAST-NEXT: [[TMP3:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT2]], [[EXT2]]
|
|
// PRMTD_FAST-NEXT: [[TMP4:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT3]], [[EXT3]]
|
|
// PRMTD_FAST-NEXT: [[TMP5:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP3]], [[TMP4]]
|
|
// PRMTD_FAST-NEXT: [[TMP6:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT2]]
|
|
// PRMTD_FAST-NEXT: [[TMP7:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT3]]
|
|
// PRMTD_FAST-NEXT: [[TMP8:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[TMP6]], [[TMP7]]
|
|
// PRMTD_FAST-NEXT: [[TMP9:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP2]], [[TMP5]]
|
|
// PRMTD_FAST-NEXT: [[TMP10:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP8]], [[TMP5]]
|
|
// PRMTD_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP9]] to half
|
|
// PRMTD_FAST-NEXT: [[UNPROMOTION4:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP10]] to half
|
|
// PRMTD_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[TMP11:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// PRMTD_FAST-NEXT: ret <2 x half> [[TMP11]]
|
|
//
|
|
_Complex _Float16 divf16(_Complex _Float16 a, _Complex _Float16 b) {
|
|
return a / b;
|
|
}
|
|
|
|
// FULL-LABEL: define dso_local <2 x half> @mulf16(
|
|
// FULL-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// FULL-NEXT: entry:
|
|
// FULL-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[COERCE:%.*]] = alloca { float, float }, align 4
|
|
// FULL-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// FULL-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// FULL-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// FULL-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// FULL-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// FULL-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// FULL-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// FULL-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// FULL-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// FULL-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// FULL-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// FULL-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// FULL-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// FULL-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// FULL-NEXT: [[MUL_AC:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// FULL-NEXT: [[MUL_BD:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// FULL-NEXT: [[MUL_AD:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// FULL-NEXT: [[MUL_BC:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// FULL-NEXT: [[MUL_R:%.*]] = fsub float [[MUL_AC]], [[MUL_BD]]
|
|
// FULL-NEXT: [[MUL_I:%.*]] = fadd float [[MUL_AD]], [[MUL_BC]]
|
|
// FULL-NEXT: [[ISNAN_CMP:%.*]] = fcmp uno float [[MUL_R]], [[MUL_R]]
|
|
// FULL-NEXT: br i1 [[ISNAN_CMP]], label [[COMPLEX_MUL_IMAG_NAN:%.*]], label [[COMPLEX_MUL_CONT:%.*]], !prof [[PROF2:![0-9]+]]
|
|
// FULL: complex_mul_imag_nan:
|
|
// FULL-NEXT: [[ISNAN_CMP4:%.*]] = fcmp uno float [[MUL_I]], [[MUL_I]]
|
|
// FULL-NEXT: br i1 [[ISNAN_CMP4]], label [[COMPLEX_MUL_LIBCALL:%.*]], label [[COMPLEX_MUL_CONT]], !prof [[PROF2]]
|
|
// FULL: complex_mul_libcall:
|
|
// FULL-NEXT: [[CALL:%.*]] = call <2 x float> @__mulsc3(float noundef [[EXT]], float noundef [[EXT1]], float noundef [[EXT2]], float noundef [[EXT3]]) #[[ATTR1]]
|
|
// FULL-NEXT: store <2 x float> [[CALL]], ptr [[COERCE]], align 4
|
|
// FULL-NEXT: [[COERCE_REALP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 0
|
|
// FULL-NEXT: [[COERCE_REAL:%.*]] = load float, ptr [[COERCE_REALP]], align 4
|
|
// FULL-NEXT: [[COERCE_IMAGP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 1
|
|
// FULL-NEXT: [[COERCE_IMAG:%.*]] = load float, ptr [[COERCE_IMAGP]], align 4
|
|
// FULL-NEXT: br label [[COMPLEX_MUL_CONT]]
|
|
// FULL: complex_mul_cont:
|
|
// FULL-NEXT: [[REAL_MUL_PHI:%.*]] = phi float [ [[MUL_R]], [[ENTRY:%.*]] ], [ [[MUL_R]], [[COMPLEX_MUL_IMAG_NAN]] ], [ [[COERCE_REAL]], [[COMPLEX_MUL_LIBCALL]] ]
|
|
// FULL-NEXT: [[IMAG_MUL_PHI:%.*]] = phi float [ [[MUL_I]], [[ENTRY]] ], [ [[MUL_I]], [[COMPLEX_MUL_IMAG_NAN]] ], [ [[COERCE_IMAG]], [[COMPLEX_MUL_LIBCALL]] ]
|
|
// FULL-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[REAL_MUL_PHI]] to half
|
|
// FULL-NEXT: [[UNPROMOTION5:%.*]] = fptrunc float [[IMAG_MUL_PHI]] to half
|
|
// FULL-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// FULL-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// FULL-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// FULL-NEXT: store half [[UNPROMOTION5]], ptr [[RETVAL_IMAGP]], align 2
|
|
// FULL-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// FULL-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// BASIC-LABEL: define dso_local <2 x half> @mulf16(
|
|
// BASIC-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// BASIC-NEXT: entry:
|
|
// BASIC-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// BASIC-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// BASIC-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// BASIC-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// BASIC-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// BASIC-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// BASIC-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// BASIC-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// BASIC-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// BASIC-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// BASIC-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// BASIC-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// BASIC-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// BASIC-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// BASIC-NEXT: [[MUL_AC:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// BASIC-NEXT: [[MUL_BD:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// BASIC-NEXT: [[MUL_AD:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// BASIC-NEXT: [[MUL_BC:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// BASIC-NEXT: [[MUL_R:%.*]] = fsub float [[MUL_AC]], [[MUL_BD]]
|
|
// BASIC-NEXT: [[MUL_I:%.*]] = fadd float [[MUL_AD]], [[MUL_BC]]
|
|
// BASIC-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[MUL_R]] to half
|
|
// BASIC-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[MUL_I]] to half
|
|
// BASIC-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// BASIC-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// BASIC-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// BASIC-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// BASIC-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// BASIC-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// IMPRVD-LABEL: define dso_local <2 x half> @mulf16(
|
|
// IMPRVD-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// IMPRVD-NEXT: entry:
|
|
// IMPRVD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// IMPRVD-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// IMPRVD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// IMPRVD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// IMPRVD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// IMPRVD-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// IMPRVD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// IMPRVD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// IMPRVD-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// IMPRVD-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// IMPRVD-NEXT: [[MUL_AC:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// IMPRVD-NEXT: [[MUL_BD:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// IMPRVD-NEXT: [[MUL_AD:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// IMPRVD-NEXT: [[MUL_BC:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// IMPRVD-NEXT: [[MUL_R:%.*]] = fsub float [[MUL_AC]], [[MUL_BD]]
|
|
// IMPRVD-NEXT: [[MUL_I:%.*]] = fadd float [[MUL_AD]], [[MUL_BC]]
|
|
// IMPRVD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[MUL_R]] to half
|
|
// IMPRVD-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[MUL_I]] to half
|
|
// IMPRVD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// IMPRVD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// IMPRVD-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// IMPRVD-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// PRMTD-LABEL: define dso_local <2 x half> @mulf16(
|
|
// PRMTD-SAME: <2 x half> noundef [[A_COERCE:%.*]], <2 x half> noundef [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// PRMTD-NEXT: entry:
|
|
// PRMTD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// PRMTD-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// PRMTD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// PRMTD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// PRMTD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// PRMTD-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// PRMTD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// PRMTD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// PRMTD-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// PRMTD-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// PRMTD-NEXT: [[MUL_AC:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// PRMTD-NEXT: [[MUL_BD:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// PRMTD-NEXT: [[MUL_AD:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// PRMTD-NEXT: [[MUL_BC:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// PRMTD-NEXT: [[MUL_R:%.*]] = fsub float [[MUL_AC]], [[MUL_BD]]
|
|
// PRMTD-NEXT: [[MUL_I:%.*]] = fadd float [[MUL_AD]], [[MUL_BC]]
|
|
// PRMTD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[MUL_R]] to half
|
|
// PRMTD-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[MUL_I]] to half
|
|
// PRMTD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// PRMTD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// PRMTD-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// PRMTD-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// X86WINPRMTD-LABEL: define dso_local i32 @mulf16(
|
|
// X86WINPRMTD-SAME: i32 noundef [[A_COERCE:%.*]], i32 noundef [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// X86WINPRMTD-NEXT: entry:
|
|
// X86WINPRMTD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: store i32 [[A_COERCE]], ptr [[A]], align 2
|
|
// X86WINPRMTD-NEXT: store i32 [[B_COERCE]], ptr [[B]], align 2
|
|
// X86WINPRMTD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[EXT:%.*]] = fpext half [[A_REAL]] to float
|
|
// X86WINPRMTD-NEXT: [[EXT1:%.*]] = fpext half [[A_IMAG]] to float
|
|
// X86WINPRMTD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[EXT2:%.*]] = fpext half [[B_REAL]] to float
|
|
// X86WINPRMTD-NEXT: [[EXT3:%.*]] = fpext half [[B_IMAG]] to float
|
|
// X86WINPRMTD-NEXT: [[MUL_AC:%.*]] = fmul float [[EXT]], [[EXT2]]
|
|
// X86WINPRMTD-NEXT: [[MUL_BD:%.*]] = fmul float [[EXT1]], [[EXT3]]
|
|
// X86WINPRMTD-NEXT: [[MUL_AD:%.*]] = fmul float [[EXT]], [[EXT3]]
|
|
// X86WINPRMTD-NEXT: [[MUL_BC:%.*]] = fmul float [[EXT1]], [[EXT2]]
|
|
// X86WINPRMTD-NEXT: [[MUL_R:%.*]] = fsub float [[MUL_AC]], [[MUL_BD]]
|
|
// X86WINPRMTD-NEXT: [[MUL_I:%.*]] = fadd float [[MUL_AD]], [[MUL_BC]]
|
|
// X86WINPRMTD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[MUL_R]] to half
|
|
// X86WINPRMTD-NEXT: [[UNPROMOTION4:%.*]] = fptrunc float [[MUL_I]] to half
|
|
// X86WINPRMTD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[TMP0:%.*]] = load i32, ptr [[RETVAL]], align 2
|
|
// X86WINPRMTD-NEXT: ret i32 [[TMP0]]
|
|
//
|
|
// BASIC_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @mulf16(
|
|
// BASIC_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// BASIC_FAST-NEXT: entry:
|
|
// BASIC_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// BASIC_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// BASIC_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// BASIC_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// BASIC_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// BASIC_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// BASIC_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// BASIC_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// BASIC_FAST-NEXT: [[MUL_AC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT2]]
|
|
// BASIC_FAST-NEXT: [[MUL_BD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT3]]
|
|
// BASIC_FAST-NEXT: [[MUL_AD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT3]]
|
|
// BASIC_FAST-NEXT: [[MUL_BC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT2]]
|
|
// BASIC_FAST-NEXT: [[MUL_R:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[MUL_AC]], [[MUL_BD]]
|
|
// BASIC_FAST-NEXT: [[MUL_I:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[MUL_AD]], [[MUL_BC]]
|
|
// BASIC_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[MUL_R]] to half
|
|
// BASIC_FAST-NEXT: [[UNPROMOTION4:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[MUL_I]] to half
|
|
// BASIC_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// BASIC_FAST-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// BASIC_FAST-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// FULL_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @mulf16(
|
|
// FULL_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// FULL_FAST-NEXT: entry:
|
|
// FULL_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[COERCE:%.*]] = alloca { float, float }, align 4
|
|
// FULL_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// FULL_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// FULL_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// FULL_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// FULL_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// FULL_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// FULL_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// FULL_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// FULL_FAST-NEXT: [[MUL_AC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT2]]
|
|
// FULL_FAST-NEXT: [[MUL_BD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT3]]
|
|
// FULL_FAST-NEXT: [[MUL_AD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT3]]
|
|
// FULL_FAST-NEXT: [[MUL_BC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT2]]
|
|
// FULL_FAST-NEXT: [[MUL_R:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[MUL_AC]], [[MUL_BD]]
|
|
// FULL_FAST-NEXT: [[MUL_I:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[MUL_AD]], [[MUL_BC]]
|
|
// FULL_FAST-NEXT: [[ISNAN_CMP:%.*]] = fcmp reassoc nnan ninf nsz arcp afn uno float [[MUL_R]], [[MUL_R]]
|
|
// FULL_FAST-NEXT: br i1 [[ISNAN_CMP]], label [[COMPLEX_MUL_IMAG_NAN:%.*]], label [[COMPLEX_MUL_CONT:%.*]], !prof [[PROF2:![0-9]+]]
|
|
// FULL_FAST: complex_mul_imag_nan:
|
|
// FULL_FAST-NEXT: [[ISNAN_CMP4:%.*]] = fcmp reassoc nnan ninf nsz arcp afn uno float [[MUL_I]], [[MUL_I]]
|
|
// FULL_FAST-NEXT: br i1 [[ISNAN_CMP4]], label [[COMPLEX_MUL_LIBCALL:%.*]], label [[COMPLEX_MUL_CONT]], !prof [[PROF2]]
|
|
// FULL_FAST: complex_mul_libcall:
|
|
// FULL_FAST-NEXT: [[CALL:%.*]] = call reassoc nnan ninf nsz arcp afn nofpclass(nan inf) <2 x float> @__mulsc3(float noundef nofpclass(nan inf) [[EXT]], float noundef nofpclass(nan inf) [[EXT1]], float noundef nofpclass(nan inf) [[EXT2]], float noundef nofpclass(nan inf) [[EXT3]]) #[[ATTR1]]
|
|
// FULL_FAST-NEXT: store <2 x float> [[CALL]], ptr [[COERCE]], align 4
|
|
// FULL_FAST-NEXT: [[COERCE_REALP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[COERCE_REAL:%.*]] = load float, ptr [[COERCE_REALP]], align 4
|
|
// FULL_FAST-NEXT: [[COERCE_IMAGP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[COERCE_IMAG:%.*]] = load float, ptr [[COERCE_IMAGP]], align 4
|
|
// FULL_FAST-NEXT: br label [[COMPLEX_MUL_CONT]]
|
|
// FULL_FAST: complex_mul_cont:
|
|
// FULL_FAST-NEXT: [[REAL_MUL_PHI:%.*]] = phi reassoc nnan ninf nsz arcp afn float [ [[MUL_R]], [[ENTRY:%.*]] ], [ [[MUL_R]], [[COMPLEX_MUL_IMAG_NAN]] ], [ [[COERCE_REAL]], [[COMPLEX_MUL_LIBCALL]] ]
|
|
// FULL_FAST-NEXT: [[IMAG_MUL_PHI:%.*]] = phi reassoc nnan ninf nsz arcp afn float [ [[MUL_I]], [[ENTRY]] ], [ [[MUL_I]], [[COMPLEX_MUL_IMAG_NAN]] ], [ [[COERCE_IMAG]], [[COMPLEX_MUL_LIBCALL]] ]
|
|
// FULL_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[REAL_MUL_PHI]] to half
|
|
// FULL_FAST-NEXT: [[UNPROMOTION5:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[IMAG_MUL_PHI]] to half
|
|
// FULL_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// FULL_FAST-NEXT: store half [[UNPROMOTION5]], ptr [[RETVAL_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// FULL_FAST-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// IMPRVD_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @mulf16(
|
|
// IMPRVD_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// IMPRVD_FAST-NEXT: entry:
|
|
// IMPRVD_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// IMPRVD_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// IMPRVD_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// IMPRVD_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// IMPRVD_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// IMPRVD_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// IMPRVD_FAST-NEXT: [[MUL_AC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT2]]
|
|
// IMPRVD_FAST-NEXT: [[MUL_BD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT3]]
|
|
// IMPRVD_FAST-NEXT: [[MUL_AD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT3]]
|
|
// IMPRVD_FAST-NEXT: [[MUL_BC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT2]]
|
|
// IMPRVD_FAST-NEXT: [[MUL_R:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[MUL_AC]], [[MUL_BD]]
|
|
// IMPRVD_FAST-NEXT: [[MUL_I:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[MUL_AD]], [[MUL_BC]]
|
|
// IMPRVD_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[MUL_R]] to half
|
|
// IMPRVD_FAST-NEXT: [[UNPROMOTION4:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[MUL_I]] to half
|
|
// IMPRVD_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// IMPRVD_FAST-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
// PRMTD_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @mulf16(
|
|
// PRMTD_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], <2 x half> noundef nofpclass(nan inf) [[B_COERCE:%.*]]) #[[ATTR0]] {
|
|
// PRMTD_FAST-NEXT: entry:
|
|
// PRMTD_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: [[B:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// PRMTD_FAST-NEXT: store <2 x half> [[B_COERCE]], ptr [[B]], align 2
|
|
// PRMTD_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// PRMTD_FAST-NEXT: [[EXT1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// PRMTD_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[B_REAL:%.*]] = load half, ptr [[B_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[B]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: [[B_IMAG:%.*]] = load half, ptr [[B_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[EXT2:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_REAL]] to float
|
|
// PRMTD_FAST-NEXT: [[EXT3:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[B_IMAG]] to float
|
|
// PRMTD_FAST-NEXT: [[MUL_AC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT2]]
|
|
// PRMTD_FAST-NEXT: [[MUL_BD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT3]]
|
|
// PRMTD_FAST-NEXT: [[MUL_AD:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT3]]
|
|
// PRMTD_FAST-NEXT: [[MUL_BC:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT1]], [[EXT2]]
|
|
// PRMTD_FAST-NEXT: [[MUL_R:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[MUL_AC]], [[MUL_BD]]
|
|
// PRMTD_FAST-NEXT: [[MUL_I:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[MUL_AD]], [[MUL_BC]]
|
|
// PRMTD_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[MUL_R]] to half
|
|
// PRMTD_FAST-NEXT: [[UNPROMOTION4:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[MUL_I]] to half
|
|
// PRMTD_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: store half [[UNPROMOTION4]], ptr [[RETVAL_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[TMP0:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// PRMTD_FAST-NEXT: ret <2 x half> [[TMP0]]
|
|
//
|
|
_Complex _Float16 mulf16(_Complex _Float16 a, _Complex _Float16 b) {
|
|
return a * b;
|
|
}
|
|
|
|
// FULL-LABEL: define dso_local <2 x half> @f1(
|
|
// FULL-SAME: <2 x half> noundef [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// FULL-NEXT: entry:
|
|
// FULL-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// FULL-NEXT: [[COERCE:%.*]] = alloca { float, float }, align 4
|
|
// FULL-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// FULL-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// FULL-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// FULL-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// FULL-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// FULL-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// FULL-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// FULL-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// FULL-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// FULL-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// FULL-NEXT: [[CONV:%.*]] = fpext half [[C_REAL]] to x86_fp80
|
|
// FULL-NEXT: [[CONV1:%.*]] = fpext half [[C_IMAG]] to x86_fp80
|
|
// FULL-NEXT: [[CALL:%.*]] = call { x86_fp80, x86_fp80 } @__divxc3(x86_fp80 noundef [[B_REAL]], x86_fp80 noundef [[B_IMAG]], x86_fp80 noundef [[CONV]], x86_fp80 noundef [[CONV1]]) #[[ATTR1]]
|
|
// FULL-NEXT: [[TMP0:%.*]] = extractvalue { x86_fp80, x86_fp80 } [[CALL]], 0
|
|
// FULL-NEXT: [[TMP1:%.*]] = extractvalue { x86_fp80, x86_fp80 } [[CALL]], 1
|
|
// FULL-NEXT: [[CONV2:%.*]] = fptrunc x86_fp80 [[TMP0]] to half
|
|
// FULL-NEXT: [[CONV3:%.*]] = fptrunc x86_fp80 [[TMP1]] to half
|
|
// FULL-NEXT: [[EXT:%.*]] = fpext half [[CONV2]] to float
|
|
// FULL-NEXT: [[EXT4:%.*]] = fpext half [[CONV3]] to float
|
|
// FULL-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// FULL-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// FULL-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// FULL-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// FULL-NEXT: [[EXT5:%.*]] = fpext half [[A_REAL]] to float
|
|
// FULL-NEXT: [[EXT6:%.*]] = fpext half [[A_IMAG]] to float
|
|
// FULL-NEXT: [[CALL7:%.*]] = call <2 x float> @__divsc3(float noundef [[EXT]], float noundef [[EXT4]], float noundef [[EXT5]], float noundef [[EXT6]]) #[[ATTR1]]
|
|
// FULL-NEXT: store <2 x float> [[CALL7]], ptr [[COERCE]], align 4
|
|
// FULL-NEXT: [[COERCE_REALP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 0
|
|
// FULL-NEXT: [[COERCE_REAL:%.*]] = load float, ptr [[COERCE_REALP]], align 4
|
|
// FULL-NEXT: [[COERCE_IMAGP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 1
|
|
// FULL-NEXT: [[COERCE_IMAG:%.*]] = load float, ptr [[COERCE_IMAGP]], align 4
|
|
// FULL-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[COERCE_REAL]] to half
|
|
// FULL-NEXT: [[UNPROMOTION8:%.*]] = fptrunc float [[COERCE_IMAG]] to half
|
|
// FULL-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// FULL-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// FULL-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// FULL-NEXT: store half [[UNPROMOTION8]], ptr [[RETVAL_IMAGP]], align 2
|
|
// FULL-NEXT: [[TMP2:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// FULL-NEXT: ret <2 x half> [[TMP2]]
|
|
//
|
|
// BASIC-LABEL: define dso_local <2 x half> @f1(
|
|
// BASIC-SAME: <2 x half> noundef [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// BASIC-NEXT: entry:
|
|
// BASIC-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// BASIC-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// BASIC-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// BASIC-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// BASIC-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// BASIC-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// BASIC-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// BASIC-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// BASIC-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// BASIC-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// BASIC-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// BASIC-NEXT: [[CONV:%.*]] = fpext half [[C_REAL]] to x86_fp80
|
|
// BASIC-NEXT: [[CONV1:%.*]] = fpext half [[C_IMAG]] to x86_fp80
|
|
// BASIC-NEXT: [[TMP0:%.*]] = fmul x86_fp80 [[B_REAL]], [[CONV]]
|
|
// BASIC-NEXT: [[TMP1:%.*]] = fmul x86_fp80 [[B_IMAG]], [[CONV1]]
|
|
// BASIC-NEXT: [[TMP2:%.*]] = fadd x86_fp80 [[TMP0]], [[TMP1]]
|
|
// BASIC-NEXT: [[TMP3:%.*]] = fmul x86_fp80 [[CONV]], [[CONV]]
|
|
// BASIC-NEXT: [[TMP4:%.*]] = fmul x86_fp80 [[CONV1]], [[CONV1]]
|
|
// BASIC-NEXT: [[TMP5:%.*]] = fadd x86_fp80 [[TMP3]], [[TMP4]]
|
|
// BASIC-NEXT: [[TMP6:%.*]] = fmul x86_fp80 [[B_IMAG]], [[CONV]]
|
|
// BASIC-NEXT: [[TMP7:%.*]] = fmul x86_fp80 [[B_REAL]], [[CONV1]]
|
|
// BASIC-NEXT: [[TMP8:%.*]] = fsub x86_fp80 [[TMP6]], [[TMP7]]
|
|
// BASIC-NEXT: [[TMP9:%.*]] = fdiv x86_fp80 [[TMP2]], [[TMP5]]
|
|
// BASIC-NEXT: [[TMP10:%.*]] = fdiv x86_fp80 [[TMP8]], [[TMP5]]
|
|
// BASIC-NEXT: [[CONV2:%.*]] = fptrunc x86_fp80 [[TMP9]] to half
|
|
// BASIC-NEXT: [[CONV3:%.*]] = fptrunc x86_fp80 [[TMP10]] to half
|
|
// BASIC-NEXT: [[EXT:%.*]] = fpext half [[CONV2]] to float
|
|
// BASIC-NEXT: [[EXT4:%.*]] = fpext half [[CONV3]] to float
|
|
// BASIC-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// BASIC-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// BASIC-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// BASIC-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// BASIC-NEXT: [[EXT5:%.*]] = fpext half [[A_REAL]] to float
|
|
// BASIC-NEXT: [[EXT6:%.*]] = fpext half [[A_IMAG]] to float
|
|
// BASIC-NEXT: [[TMP11:%.*]] = fmul float [[EXT]], [[EXT5]]
|
|
// BASIC-NEXT: [[TMP12:%.*]] = fmul float [[EXT4]], [[EXT6]]
|
|
// BASIC-NEXT: [[TMP13:%.*]] = fadd float [[TMP11]], [[TMP12]]
|
|
// BASIC-NEXT: [[TMP14:%.*]] = fmul float [[EXT5]], [[EXT5]]
|
|
// BASIC-NEXT: [[TMP15:%.*]] = fmul float [[EXT6]], [[EXT6]]
|
|
// BASIC-NEXT: [[TMP16:%.*]] = fadd float [[TMP14]], [[TMP15]]
|
|
// BASIC-NEXT: [[TMP17:%.*]] = fmul float [[EXT4]], [[EXT5]]
|
|
// BASIC-NEXT: [[TMP18:%.*]] = fmul float [[EXT]], [[EXT6]]
|
|
// BASIC-NEXT: [[TMP19:%.*]] = fsub float [[TMP17]], [[TMP18]]
|
|
// BASIC-NEXT: [[TMP20:%.*]] = fdiv float [[TMP13]], [[TMP16]]
|
|
// BASIC-NEXT: [[TMP21:%.*]] = fdiv float [[TMP19]], [[TMP16]]
|
|
// BASIC-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP20]] to half
|
|
// BASIC-NEXT: [[UNPROMOTION7:%.*]] = fptrunc float [[TMP21]] to half
|
|
// BASIC-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// BASIC-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// BASIC-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// BASIC-NEXT: store half [[UNPROMOTION7]], ptr [[RETVAL_IMAGP]], align 2
|
|
// BASIC-NEXT: [[TMP22:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// BASIC-NEXT: ret <2 x half> [[TMP22]]
|
|
//
|
|
// IMPRVD-LABEL: define dso_local <2 x half> @f1(
|
|
// IMPRVD-SAME: <2 x half> noundef [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// IMPRVD-NEXT: entry:
|
|
// IMPRVD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// IMPRVD-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// IMPRVD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// IMPRVD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// IMPRVD-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// IMPRVD-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// IMPRVD-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// IMPRVD-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[CONV:%.*]] = fpext half [[C_REAL]] to x86_fp80
|
|
// IMPRVD-NEXT: [[CONV1:%.*]] = fpext half [[C_IMAG]] to x86_fp80
|
|
// IMPRVD-NEXT: [[TMP0:%.*]] = call x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV]])
|
|
// IMPRVD-NEXT: [[TMP1:%.*]] = call x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV1]])
|
|
// IMPRVD-NEXT: [[ABS_CMP:%.*]] = fcmp ugt x86_fp80 [[TMP0]], [[TMP1]]
|
|
// IMPRVD-NEXT: br i1 [[ABS_CMP]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI:%.*]]
|
|
// IMPRVD: abs_rhsr_greater_or_equal_abs_rhsi:
|
|
// IMPRVD-NEXT: [[TMP2:%.*]] = fdiv x86_fp80 [[CONV1]], [[CONV]]
|
|
// IMPRVD-NEXT: [[TMP3:%.*]] = fmul x86_fp80 [[TMP2]], [[CONV1]]
|
|
// IMPRVD-NEXT: [[TMP4:%.*]] = fadd x86_fp80 [[CONV]], [[TMP3]]
|
|
// IMPRVD-NEXT: [[TMP5:%.*]] = fmul x86_fp80 [[B_IMAG]], [[TMP2]]
|
|
// IMPRVD-NEXT: [[TMP6:%.*]] = fadd x86_fp80 [[B_REAL]], [[TMP5]]
|
|
// IMPRVD-NEXT: [[TMP7:%.*]] = fdiv x86_fp80 [[TMP6]], [[TMP4]]
|
|
// IMPRVD-NEXT: [[TMP8:%.*]] = fmul x86_fp80 [[B_REAL]], [[TMP2]]
|
|
// IMPRVD-NEXT: [[TMP9:%.*]] = fsub x86_fp80 [[B_IMAG]], [[TMP8]]
|
|
// IMPRVD-NEXT: [[TMP10:%.*]] = fdiv x86_fp80 [[TMP9]], [[TMP4]]
|
|
// IMPRVD-NEXT: br label [[COMPLEX_DIV:%.*]]
|
|
// IMPRVD: abs_rhsr_less_than_abs_rhsi:
|
|
// IMPRVD-NEXT: [[TMP11:%.*]] = fdiv x86_fp80 [[CONV]], [[CONV1]]
|
|
// IMPRVD-NEXT: [[TMP12:%.*]] = fmul x86_fp80 [[TMP11]], [[CONV]]
|
|
// IMPRVD-NEXT: [[TMP13:%.*]] = fadd x86_fp80 [[CONV1]], [[TMP12]]
|
|
// IMPRVD-NEXT: [[TMP14:%.*]] = fmul x86_fp80 [[B_REAL]], [[TMP11]]
|
|
// IMPRVD-NEXT: [[TMP15:%.*]] = fadd x86_fp80 [[TMP14]], [[B_IMAG]]
|
|
// IMPRVD-NEXT: [[TMP16:%.*]] = fdiv x86_fp80 [[TMP15]], [[TMP13]]
|
|
// IMPRVD-NEXT: [[TMP17:%.*]] = fmul x86_fp80 [[B_IMAG]], [[TMP11]]
|
|
// IMPRVD-NEXT: [[TMP18:%.*]] = fsub x86_fp80 [[TMP17]], [[B_REAL]]
|
|
// IMPRVD-NEXT: [[TMP19:%.*]] = fdiv x86_fp80 [[TMP18]], [[TMP13]]
|
|
// IMPRVD-NEXT: br label [[COMPLEX_DIV]]
|
|
// IMPRVD: complex_div:
|
|
// IMPRVD-NEXT: [[TMP20:%.*]] = phi x86_fp80 [ [[TMP7]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP16]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD-NEXT: [[TMP21:%.*]] = phi x86_fp80 [ [[TMP10]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP19]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD-NEXT: [[CONV2:%.*]] = fptrunc x86_fp80 [[TMP20]] to half
|
|
// IMPRVD-NEXT: [[CONV3:%.*]] = fptrunc x86_fp80 [[TMP21]] to half
|
|
// IMPRVD-NEXT: [[EXT:%.*]] = fpext half [[CONV2]] to float
|
|
// IMPRVD-NEXT: [[EXT4:%.*]] = fpext half [[CONV3]] to float
|
|
// IMPRVD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// IMPRVD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// IMPRVD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[EXT5:%.*]] = fpext half [[A_REAL]] to float
|
|
// IMPRVD-NEXT: [[EXT6:%.*]] = fpext half [[A_IMAG]] to float
|
|
// IMPRVD-NEXT: [[TMP22:%.*]] = call float @llvm.fabs.f32(float [[EXT5]])
|
|
// IMPRVD-NEXT: [[TMP23:%.*]] = call float @llvm.fabs.f32(float [[EXT6]])
|
|
// IMPRVD-NEXT: [[ABS_CMP7:%.*]] = fcmp ugt float [[TMP22]], [[TMP23]]
|
|
// IMPRVD-NEXT: br i1 [[ABS_CMP7]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI8:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI9:%.*]]
|
|
// IMPRVD: abs_rhsr_greater_or_equal_abs_rhsi8:
|
|
// IMPRVD-NEXT: [[TMP24:%.*]] = fdiv float [[EXT6]], [[EXT5]]
|
|
// IMPRVD-NEXT: [[TMP25:%.*]] = fmul float [[TMP24]], [[EXT6]]
|
|
// IMPRVD-NEXT: [[TMP26:%.*]] = fadd float [[EXT5]], [[TMP25]]
|
|
// IMPRVD-NEXT: [[TMP27:%.*]] = fmul float [[EXT4]], [[TMP24]]
|
|
// IMPRVD-NEXT: [[TMP28:%.*]] = fadd float [[EXT]], [[TMP27]]
|
|
// IMPRVD-NEXT: [[TMP29:%.*]] = fdiv float [[TMP28]], [[TMP26]]
|
|
// IMPRVD-NEXT: [[TMP30:%.*]] = fmul float [[EXT]], [[TMP24]]
|
|
// IMPRVD-NEXT: [[TMP31:%.*]] = fsub float [[EXT4]], [[TMP30]]
|
|
// IMPRVD-NEXT: [[TMP32:%.*]] = fdiv float [[TMP31]], [[TMP26]]
|
|
// IMPRVD-NEXT: br label [[COMPLEX_DIV10:%.*]]
|
|
// IMPRVD: abs_rhsr_less_than_abs_rhsi9:
|
|
// IMPRVD-NEXT: [[TMP33:%.*]] = fdiv float [[EXT5]], [[EXT6]]
|
|
// IMPRVD-NEXT: [[TMP34:%.*]] = fmul float [[TMP33]], [[EXT5]]
|
|
// IMPRVD-NEXT: [[TMP35:%.*]] = fadd float [[EXT6]], [[TMP34]]
|
|
// IMPRVD-NEXT: [[TMP36:%.*]] = fmul float [[EXT]], [[TMP33]]
|
|
// IMPRVD-NEXT: [[TMP37:%.*]] = fadd float [[TMP36]], [[EXT4]]
|
|
// IMPRVD-NEXT: [[TMP38:%.*]] = fdiv float [[TMP37]], [[TMP35]]
|
|
// IMPRVD-NEXT: [[TMP39:%.*]] = fmul float [[EXT4]], [[TMP33]]
|
|
// IMPRVD-NEXT: [[TMP40:%.*]] = fsub float [[TMP39]], [[EXT]]
|
|
// IMPRVD-NEXT: [[TMP41:%.*]] = fdiv float [[TMP40]], [[TMP35]]
|
|
// IMPRVD-NEXT: br label [[COMPLEX_DIV10]]
|
|
// IMPRVD: complex_div10:
|
|
// IMPRVD-NEXT: [[TMP42:%.*]] = phi float [ [[TMP29]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI8]] ], [ [[TMP38]], [[ABS_RHSR_LESS_THAN_ABS_RHSI9]] ]
|
|
// IMPRVD-NEXT: [[TMP43:%.*]] = phi float [ [[TMP32]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI8]] ], [ [[TMP41]], [[ABS_RHSR_LESS_THAN_ABS_RHSI9]] ]
|
|
// IMPRVD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP42]] to half
|
|
// IMPRVD-NEXT: [[UNPROMOTION11:%.*]] = fptrunc float [[TMP43]] to half
|
|
// IMPRVD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// IMPRVD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// IMPRVD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// IMPRVD-NEXT: store half [[UNPROMOTION11]], ptr [[RETVAL_IMAGP]], align 2
|
|
// IMPRVD-NEXT: [[TMP44:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// IMPRVD-NEXT: ret <2 x half> [[TMP44]]
|
|
//
|
|
// PRMTD-LABEL: define dso_local <2 x half> @f1(
|
|
// PRMTD-SAME: <2 x half> noundef [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// PRMTD-NEXT: entry:
|
|
// PRMTD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// PRMTD-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// PRMTD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// PRMTD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// PRMTD-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// PRMTD-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// PRMTD-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// PRMTD-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[CONV:%.*]] = fpext half [[C_REAL]] to x86_fp80
|
|
// PRMTD-NEXT: [[CONV1:%.*]] = fpext half [[C_IMAG]] to x86_fp80
|
|
// PRMTD-NEXT: [[TMP0:%.*]] = call x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV]])
|
|
// PRMTD-NEXT: [[TMP1:%.*]] = call x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV1]])
|
|
// PRMTD-NEXT: [[ABS_CMP:%.*]] = fcmp ugt x86_fp80 [[TMP0]], [[TMP1]]
|
|
// PRMTD-NEXT: br i1 [[ABS_CMP]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI:%.*]]
|
|
// PRMTD: abs_rhsr_greater_or_equal_abs_rhsi:
|
|
// PRMTD-NEXT: [[TMP2:%.*]] = fdiv x86_fp80 [[CONV1]], [[CONV]]
|
|
// PRMTD-NEXT: [[TMP3:%.*]] = fmul x86_fp80 [[TMP2]], [[CONV1]]
|
|
// PRMTD-NEXT: [[TMP4:%.*]] = fadd x86_fp80 [[CONV]], [[TMP3]]
|
|
// PRMTD-NEXT: [[TMP5:%.*]] = fmul x86_fp80 [[B_IMAG]], [[TMP2]]
|
|
// PRMTD-NEXT: [[TMP6:%.*]] = fadd x86_fp80 [[B_REAL]], [[TMP5]]
|
|
// PRMTD-NEXT: [[TMP7:%.*]] = fdiv x86_fp80 [[TMP6]], [[TMP4]]
|
|
// PRMTD-NEXT: [[TMP8:%.*]] = fmul x86_fp80 [[B_REAL]], [[TMP2]]
|
|
// PRMTD-NEXT: [[TMP9:%.*]] = fsub x86_fp80 [[B_IMAG]], [[TMP8]]
|
|
// PRMTD-NEXT: [[TMP10:%.*]] = fdiv x86_fp80 [[TMP9]], [[TMP4]]
|
|
// PRMTD-NEXT: br label [[COMPLEX_DIV:%.*]]
|
|
// PRMTD: abs_rhsr_less_than_abs_rhsi:
|
|
// PRMTD-NEXT: [[TMP11:%.*]] = fdiv x86_fp80 [[CONV]], [[CONV1]]
|
|
// PRMTD-NEXT: [[TMP12:%.*]] = fmul x86_fp80 [[TMP11]], [[CONV]]
|
|
// PRMTD-NEXT: [[TMP13:%.*]] = fadd x86_fp80 [[CONV1]], [[TMP12]]
|
|
// PRMTD-NEXT: [[TMP14:%.*]] = fmul x86_fp80 [[B_REAL]], [[TMP11]]
|
|
// PRMTD-NEXT: [[TMP15:%.*]] = fadd x86_fp80 [[TMP14]], [[B_IMAG]]
|
|
// PRMTD-NEXT: [[TMP16:%.*]] = fdiv x86_fp80 [[TMP15]], [[TMP13]]
|
|
// PRMTD-NEXT: [[TMP17:%.*]] = fmul x86_fp80 [[B_IMAG]], [[TMP11]]
|
|
// PRMTD-NEXT: [[TMP18:%.*]] = fsub x86_fp80 [[TMP17]], [[B_REAL]]
|
|
// PRMTD-NEXT: [[TMP19:%.*]] = fdiv x86_fp80 [[TMP18]], [[TMP13]]
|
|
// PRMTD-NEXT: br label [[COMPLEX_DIV]]
|
|
// PRMTD: complex_div:
|
|
// PRMTD-NEXT: [[TMP20:%.*]] = phi x86_fp80 [ [[TMP7]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP16]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// PRMTD-NEXT: [[TMP21:%.*]] = phi x86_fp80 [ [[TMP10]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP19]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// PRMTD-NEXT: [[CONV2:%.*]] = fptrunc x86_fp80 [[TMP20]] to half
|
|
// PRMTD-NEXT: [[CONV3:%.*]] = fptrunc x86_fp80 [[TMP21]] to half
|
|
// PRMTD-NEXT: [[EXT:%.*]] = fpext half [[CONV2]] to float
|
|
// PRMTD-NEXT: [[EXT4:%.*]] = fpext half [[CONV3]] to float
|
|
// PRMTD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// PRMTD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// PRMTD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[EXT5:%.*]] = fpext half [[A_REAL]] to float
|
|
// PRMTD-NEXT: [[EXT6:%.*]] = fpext half [[A_IMAG]] to float
|
|
// PRMTD-NEXT: [[TMP22:%.*]] = fmul float [[EXT]], [[EXT5]]
|
|
// PRMTD-NEXT: [[TMP23:%.*]] = fmul float [[EXT4]], [[EXT6]]
|
|
// PRMTD-NEXT: [[TMP24:%.*]] = fadd float [[TMP22]], [[TMP23]]
|
|
// PRMTD-NEXT: [[TMP25:%.*]] = fmul float [[EXT5]], [[EXT5]]
|
|
// PRMTD-NEXT: [[TMP26:%.*]] = fmul float [[EXT6]], [[EXT6]]
|
|
// PRMTD-NEXT: [[TMP27:%.*]] = fadd float [[TMP25]], [[TMP26]]
|
|
// PRMTD-NEXT: [[TMP28:%.*]] = fmul float [[EXT4]], [[EXT5]]
|
|
// PRMTD-NEXT: [[TMP29:%.*]] = fmul float [[EXT]], [[EXT6]]
|
|
// PRMTD-NEXT: [[TMP30:%.*]] = fsub float [[TMP28]], [[TMP29]]
|
|
// PRMTD-NEXT: [[TMP31:%.*]] = fdiv float [[TMP24]], [[TMP27]]
|
|
// PRMTD-NEXT: [[TMP32:%.*]] = fdiv float [[TMP30]], [[TMP27]]
|
|
// PRMTD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP31]] to half
|
|
// PRMTD-NEXT: [[UNPROMOTION7:%.*]] = fptrunc float [[TMP32]] to half
|
|
// PRMTD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// PRMTD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// PRMTD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// PRMTD-NEXT: store half [[UNPROMOTION7]], ptr [[RETVAL_IMAGP]], align 2
|
|
// PRMTD-NEXT: [[TMP33:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// PRMTD-NEXT: ret <2 x half> [[TMP33]]
|
|
//
|
|
// X86WINPRMTD-LABEL: define dso_local i32 @f1(
|
|
// X86WINPRMTD-SAME: i32 noundef [[A_COERCE:%.*]], ptr noundef dead_on_return [[B:%.*]], i32 noundef [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// X86WINPRMTD-NEXT: entry:
|
|
// X86WINPRMTD-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// X86WINPRMTD-NEXT: [[B_INDIRECT_ADDR:%.*]] = alloca ptr, align 8
|
|
// X86WINPRMTD-NEXT: store i32 [[A_COERCE]], ptr [[A]], align 2
|
|
// X86WINPRMTD-NEXT: store i32 [[C_COERCE]], ptr [[C]], align 2
|
|
// X86WINPRMTD-NEXT: store ptr [[B]], ptr [[B_INDIRECT_ADDR]], align 8
|
|
// X86WINPRMTD-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { double, double }, ptr [[B]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[B_REAL:%.*]] = load double, ptr [[B_REALP]], align 8
|
|
// X86WINPRMTD-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { double, double }, ptr [[B]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: [[B_IMAG:%.*]] = load double, ptr [[B_IMAGP]], align 8
|
|
// X86WINPRMTD-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[CONV:%.*]] = fpext half [[C_REAL]] to double
|
|
// X86WINPRMTD-NEXT: [[CONV1:%.*]] = fpext half [[C_IMAG]] to double
|
|
// X86WINPRMTD-NEXT: [[TMP0:%.*]] = call double @llvm.fabs.f64(double [[CONV]])
|
|
// X86WINPRMTD-NEXT: [[TMP1:%.*]] = call double @llvm.fabs.f64(double [[CONV1]])
|
|
// X86WINPRMTD-NEXT: [[ABS_CMP:%.*]] = fcmp ugt double [[TMP0]], [[TMP1]]
|
|
// X86WINPRMTD-NEXT: br i1 [[ABS_CMP]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI:%.*]]
|
|
// X86WINPRMTD: abs_rhsr_greater_or_equal_abs_rhsi:
|
|
// X86WINPRMTD-NEXT: [[TMP2:%.*]] = fdiv double [[CONV1]], [[CONV]]
|
|
// X86WINPRMTD-NEXT: [[TMP3:%.*]] = fmul double [[TMP2]], [[CONV1]]
|
|
// X86WINPRMTD-NEXT: [[TMP4:%.*]] = fadd double [[CONV]], [[TMP3]]
|
|
// X86WINPRMTD-NEXT: [[TMP5:%.*]] = fmul double [[B_IMAG]], [[TMP2]]
|
|
// X86WINPRMTD-NEXT: [[TMP6:%.*]] = fadd double [[B_REAL]], [[TMP5]]
|
|
// X86WINPRMTD-NEXT: [[TMP7:%.*]] = fdiv double [[TMP6]], [[TMP4]]
|
|
// X86WINPRMTD-NEXT: [[TMP8:%.*]] = fmul double [[B_REAL]], [[TMP2]]
|
|
// X86WINPRMTD-NEXT: [[TMP9:%.*]] = fsub double [[B_IMAG]], [[TMP8]]
|
|
// X86WINPRMTD-NEXT: [[TMP10:%.*]] = fdiv double [[TMP9]], [[TMP4]]
|
|
// X86WINPRMTD-NEXT: br label [[COMPLEX_DIV:%.*]]
|
|
// X86WINPRMTD: abs_rhsr_less_than_abs_rhsi:
|
|
// X86WINPRMTD-NEXT: [[TMP11:%.*]] = fdiv double [[CONV]], [[CONV1]]
|
|
// X86WINPRMTD-NEXT: [[TMP12:%.*]] = fmul double [[TMP11]], [[CONV]]
|
|
// X86WINPRMTD-NEXT: [[TMP13:%.*]] = fadd double [[CONV1]], [[TMP12]]
|
|
// X86WINPRMTD-NEXT: [[TMP14:%.*]] = fmul double [[B_REAL]], [[TMP11]]
|
|
// X86WINPRMTD-NEXT: [[TMP15:%.*]] = fadd double [[TMP14]], [[B_IMAG]]
|
|
// X86WINPRMTD-NEXT: [[TMP16:%.*]] = fdiv double [[TMP15]], [[TMP13]]
|
|
// X86WINPRMTD-NEXT: [[TMP17:%.*]] = fmul double [[B_IMAG]], [[TMP11]]
|
|
// X86WINPRMTD-NEXT: [[TMP18:%.*]] = fsub double [[TMP17]], [[B_REAL]]
|
|
// X86WINPRMTD-NEXT: [[TMP19:%.*]] = fdiv double [[TMP18]], [[TMP13]]
|
|
// X86WINPRMTD-NEXT: br label [[COMPLEX_DIV]]
|
|
// X86WINPRMTD: complex_div:
|
|
// X86WINPRMTD-NEXT: [[TMP20:%.*]] = phi double [ [[TMP7]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP16]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// X86WINPRMTD-NEXT: [[TMP21:%.*]] = phi double [ [[TMP10]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP19]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// X86WINPRMTD-NEXT: [[CONV2:%.*]] = fptrunc double [[TMP20]] to half
|
|
// X86WINPRMTD-NEXT: [[CONV3:%.*]] = fptrunc double [[TMP21]] to half
|
|
// X86WINPRMTD-NEXT: [[EXT:%.*]] = fpext half [[CONV2]] to float
|
|
// X86WINPRMTD-NEXT: [[EXT4:%.*]] = fpext half [[CONV3]] to float
|
|
// X86WINPRMTD-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[EXT5:%.*]] = fpext half [[A_REAL]] to float
|
|
// X86WINPRMTD-NEXT: [[EXT6:%.*]] = fpext half [[A_IMAG]] to float
|
|
// X86WINPRMTD-NEXT: [[TMP22:%.*]] = fmul float [[EXT]], [[EXT5]]
|
|
// X86WINPRMTD-NEXT: [[TMP23:%.*]] = fmul float [[EXT4]], [[EXT6]]
|
|
// X86WINPRMTD-NEXT: [[TMP24:%.*]] = fadd float [[TMP22]], [[TMP23]]
|
|
// X86WINPRMTD-NEXT: [[TMP25:%.*]] = fmul float [[EXT5]], [[EXT5]]
|
|
// X86WINPRMTD-NEXT: [[TMP26:%.*]] = fmul float [[EXT6]], [[EXT6]]
|
|
// X86WINPRMTD-NEXT: [[TMP27:%.*]] = fadd float [[TMP25]], [[TMP26]]
|
|
// X86WINPRMTD-NEXT: [[TMP28:%.*]] = fmul float [[EXT4]], [[EXT5]]
|
|
// X86WINPRMTD-NEXT: [[TMP29:%.*]] = fmul float [[EXT]], [[EXT6]]
|
|
// X86WINPRMTD-NEXT: [[TMP30:%.*]] = fsub float [[TMP28]], [[TMP29]]
|
|
// X86WINPRMTD-NEXT: [[TMP31:%.*]] = fdiv float [[TMP24]], [[TMP27]]
|
|
// X86WINPRMTD-NEXT: [[TMP32:%.*]] = fdiv float [[TMP30]], [[TMP27]]
|
|
// X86WINPRMTD-NEXT: [[UNPROMOTION:%.*]] = fptrunc float [[TMP31]] to half
|
|
// X86WINPRMTD-NEXT: [[UNPROMOTION7:%.*]] = fptrunc float [[TMP32]] to half
|
|
// X86WINPRMTD-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// X86WINPRMTD-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// X86WINPRMTD-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// X86WINPRMTD-NEXT: store half [[UNPROMOTION7]], ptr [[RETVAL_IMAGP]], align 2
|
|
// X86WINPRMTD-NEXT: [[TMP33:%.*]] = load i32, ptr [[RETVAL]], align 2
|
|
// X86WINPRMTD-NEXT: ret i32 [[TMP33]]
|
|
//
|
|
// BASIC_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @f1(
|
|
// BASIC_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef nofpclass(nan inf) [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// BASIC_FAST-NEXT: entry:
|
|
// BASIC_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// BASIC_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// BASIC_FAST-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// BASIC_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// BASIC_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// BASIC_FAST-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// BASIC_FAST-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[CONV:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_REAL]] to x86_fp80
|
|
// BASIC_FAST-NEXT: [[CONV1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_IMAG]] to x86_fp80
|
|
// BASIC_FAST-NEXT: [[TMP0:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[CONV]]
|
|
// BASIC_FAST-NEXT: [[TMP1:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[CONV1]]
|
|
// BASIC_FAST-NEXT: [[TMP2:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP0]], [[TMP1]]
|
|
// BASIC_FAST-NEXT: [[TMP3:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV]], [[CONV]]
|
|
// BASIC_FAST-NEXT: [[TMP4:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV1]], [[CONV1]]
|
|
// BASIC_FAST-NEXT: [[TMP5:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP3]], [[TMP4]]
|
|
// BASIC_FAST-NEXT: [[TMP6:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[CONV]]
|
|
// BASIC_FAST-NEXT: [[TMP7:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[CONV1]]
|
|
// BASIC_FAST-NEXT: [[TMP8:%.*]] = fsub reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP6]], [[TMP7]]
|
|
// BASIC_FAST-NEXT: [[TMP9:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP2]], [[TMP5]]
|
|
// BASIC_FAST-NEXT: [[TMP10:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP8]], [[TMP5]]
|
|
// BASIC_FAST-NEXT: [[CONV2:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP9]] to half
|
|
// BASIC_FAST-NEXT: [[CONV3:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP10]] to half
|
|
// BASIC_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV2]] to float
|
|
// BASIC_FAST-NEXT: [[EXT4:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV3]] to float
|
|
// BASIC_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// BASIC_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[EXT5:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// BASIC_FAST-NEXT: [[EXT6:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// BASIC_FAST-NEXT: [[TMP11:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT5]]
|
|
// BASIC_FAST-NEXT: [[TMP12:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT4]], [[EXT6]]
|
|
// BASIC_FAST-NEXT: [[TMP13:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP11]], [[TMP12]]
|
|
// BASIC_FAST-NEXT: [[TMP14:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT5]], [[EXT5]]
|
|
// BASIC_FAST-NEXT: [[TMP15:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT6]], [[EXT6]]
|
|
// BASIC_FAST-NEXT: [[TMP16:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP14]], [[TMP15]]
|
|
// BASIC_FAST-NEXT: [[TMP17:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT4]], [[EXT5]]
|
|
// BASIC_FAST-NEXT: [[TMP18:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT6]]
|
|
// BASIC_FAST-NEXT: [[TMP19:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[TMP17]], [[TMP18]]
|
|
// BASIC_FAST-NEXT: [[TMP20:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP13]], [[TMP16]]
|
|
// BASIC_FAST-NEXT: [[TMP21:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP19]], [[TMP16]]
|
|
// BASIC_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP20]] to half
|
|
// BASIC_FAST-NEXT: [[UNPROMOTION7:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP21]] to half
|
|
// BASIC_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// BASIC_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// BASIC_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// BASIC_FAST-NEXT: store half [[UNPROMOTION7]], ptr [[RETVAL_IMAGP]], align 2
|
|
// BASIC_FAST-NEXT: [[TMP22:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// BASIC_FAST-NEXT: ret <2 x half> [[TMP22]]
|
|
//
|
|
// FULL_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @f1(
|
|
// FULL_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef nofpclass(nan inf) [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// FULL_FAST-NEXT: entry:
|
|
// FULL_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// FULL_FAST-NEXT: [[COERCE:%.*]] = alloca { float, float }, align 4
|
|
// FULL_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// FULL_FAST-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// FULL_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// FULL_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// FULL_FAST-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// FULL_FAST-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[CONV:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_REAL]] to x86_fp80
|
|
// FULL_FAST-NEXT: [[CONV1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_IMAG]] to x86_fp80
|
|
// FULL_FAST-NEXT: [[CALL:%.*]] = call reassoc nnan ninf nsz arcp afn nofpclass(nan inf) { x86_fp80, x86_fp80 } @__divxc3(x86_fp80 noundef nofpclass(nan inf) [[B_REAL]], x86_fp80 noundef nofpclass(nan inf) [[B_IMAG]], x86_fp80 noundef nofpclass(nan inf) [[CONV]], x86_fp80 noundef nofpclass(nan inf) [[CONV1]]) #[[ATTR1]]
|
|
// FULL_FAST-NEXT: [[TMP0:%.*]] = extractvalue { x86_fp80, x86_fp80 } [[CALL]], 0
|
|
// FULL_FAST-NEXT: [[TMP1:%.*]] = extractvalue { x86_fp80, x86_fp80 } [[CALL]], 1
|
|
// FULL_FAST-NEXT: [[CONV2:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP0]] to half
|
|
// FULL_FAST-NEXT: [[CONV3:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP1]] to half
|
|
// FULL_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV2]] to float
|
|
// FULL_FAST-NEXT: [[EXT4:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV3]] to float
|
|
// FULL_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// FULL_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[EXT5:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// FULL_FAST-NEXT: [[EXT6:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// FULL_FAST-NEXT: [[CALL7:%.*]] = call reassoc nnan ninf nsz arcp afn nofpclass(nan inf) <2 x float> @__divsc3(float noundef nofpclass(nan inf) [[EXT]], float noundef nofpclass(nan inf) [[EXT4]], float noundef nofpclass(nan inf) [[EXT5]], float noundef nofpclass(nan inf) [[EXT6]]) #[[ATTR1]]
|
|
// FULL_FAST-NEXT: store <2 x float> [[CALL7]], ptr [[COERCE]], align 4
|
|
// FULL_FAST-NEXT: [[COERCE_REALP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[COERCE_REAL:%.*]] = load float, ptr [[COERCE_REALP]], align 4
|
|
// FULL_FAST-NEXT: [[COERCE_IMAGP:%.*]] = getelementptr inbounds nuw { float, float }, ptr [[COERCE]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: [[COERCE_IMAG:%.*]] = load float, ptr [[COERCE_IMAGP]], align 4
|
|
// FULL_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[COERCE_REAL]] to half
|
|
// FULL_FAST-NEXT: [[UNPROMOTION8:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[COERCE_IMAG]] to half
|
|
// FULL_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// FULL_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// FULL_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// FULL_FAST-NEXT: store half [[UNPROMOTION8]], ptr [[RETVAL_IMAGP]], align 2
|
|
// FULL_FAST-NEXT: [[TMP2:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// FULL_FAST-NEXT: ret <2 x half> [[TMP2]]
|
|
//
|
|
// IMPRVD_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @f1(
|
|
// IMPRVD_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef nofpclass(nan inf) [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// IMPRVD_FAST-NEXT: entry:
|
|
// IMPRVD_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// IMPRVD_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// IMPRVD_FAST-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// IMPRVD_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// IMPRVD_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// IMPRVD_FAST-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[CONV:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_REAL]] to x86_fp80
|
|
// IMPRVD_FAST-NEXT: [[CONV1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_IMAG]] to x86_fp80
|
|
// IMPRVD_FAST-NEXT: [[TMP0:%.*]] = call reassoc nnan ninf nsz arcp afn x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV]])
|
|
// IMPRVD_FAST-NEXT: [[TMP1:%.*]] = call reassoc nnan ninf nsz arcp afn x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV1]])
|
|
// IMPRVD_FAST-NEXT: [[ABS_CMP:%.*]] = fcmp reassoc nnan ninf nsz arcp afn ugt x86_fp80 [[TMP0]], [[TMP1]]
|
|
// IMPRVD_FAST-NEXT: br i1 [[ABS_CMP]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI:%.*]]
|
|
// IMPRVD_FAST: abs_rhsr_greater_or_equal_abs_rhsi:
|
|
// IMPRVD_FAST-NEXT: [[TMP2:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV1]], [[CONV]]
|
|
// IMPRVD_FAST-NEXT: [[TMP3:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP2]], [[CONV1]]
|
|
// IMPRVD_FAST-NEXT: [[TMP4:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV]], [[TMP3]]
|
|
// IMPRVD_FAST-NEXT: [[TMP5:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[TMP2]]
|
|
// IMPRVD_FAST-NEXT: [[TMP6:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[TMP5]]
|
|
// IMPRVD_FAST-NEXT: [[TMP7:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP6]], [[TMP4]]
|
|
// IMPRVD_FAST-NEXT: [[TMP8:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[TMP2]]
|
|
// IMPRVD_FAST-NEXT: [[TMP9:%.*]] = fsub reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[TMP8]]
|
|
// IMPRVD_FAST-NEXT: [[TMP10:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP9]], [[TMP4]]
|
|
// IMPRVD_FAST-NEXT: br label [[COMPLEX_DIV:%.*]]
|
|
// IMPRVD_FAST: abs_rhsr_less_than_abs_rhsi:
|
|
// IMPRVD_FAST-NEXT: [[TMP11:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV]], [[CONV1]]
|
|
// IMPRVD_FAST-NEXT: [[TMP12:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP11]], [[CONV]]
|
|
// IMPRVD_FAST-NEXT: [[TMP13:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV1]], [[TMP12]]
|
|
// IMPRVD_FAST-NEXT: [[TMP14:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[TMP11]]
|
|
// IMPRVD_FAST-NEXT: [[TMP15:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP14]], [[B_IMAG]]
|
|
// IMPRVD_FAST-NEXT: [[TMP16:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP15]], [[TMP13]]
|
|
// IMPRVD_FAST-NEXT: [[TMP17:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[TMP11]]
|
|
// IMPRVD_FAST-NEXT: [[TMP18:%.*]] = fsub reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP17]], [[B_REAL]]
|
|
// IMPRVD_FAST-NEXT: [[TMP19:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP18]], [[TMP13]]
|
|
// IMPRVD_FAST-NEXT: br label [[COMPLEX_DIV]]
|
|
// IMPRVD_FAST: complex_div:
|
|
// IMPRVD_FAST-NEXT: [[TMP20:%.*]] = phi reassoc nnan ninf nsz arcp afn x86_fp80 [ [[TMP7]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP16]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD_FAST-NEXT: [[TMP21:%.*]] = phi reassoc nnan ninf nsz arcp afn x86_fp80 [ [[TMP10]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP19]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// IMPRVD_FAST-NEXT: [[CONV2:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP20]] to half
|
|
// IMPRVD_FAST-NEXT: [[CONV3:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP21]] to half
|
|
// IMPRVD_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV2]] to float
|
|
// IMPRVD_FAST-NEXT: [[EXT4:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV3]] to float
|
|
// IMPRVD_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[EXT5:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// IMPRVD_FAST-NEXT: [[EXT6:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// IMPRVD_FAST-NEXT: [[TMP22:%.*]] = call reassoc nnan ninf nsz arcp afn float @llvm.fabs.f32(float [[EXT5]])
|
|
// IMPRVD_FAST-NEXT: [[TMP23:%.*]] = call reassoc nnan ninf nsz arcp afn float @llvm.fabs.f32(float [[EXT6]])
|
|
// IMPRVD_FAST-NEXT: [[ABS_CMP7:%.*]] = fcmp reassoc nnan ninf nsz arcp afn ugt float [[TMP22]], [[TMP23]]
|
|
// IMPRVD_FAST-NEXT: br i1 [[ABS_CMP7]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI8:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI9:%.*]]
|
|
// IMPRVD_FAST: abs_rhsr_greater_or_equal_abs_rhsi8:
|
|
// IMPRVD_FAST-NEXT: [[TMP24:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[EXT6]], [[EXT5]]
|
|
// IMPRVD_FAST-NEXT: [[TMP25:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[TMP24]], [[EXT6]]
|
|
// IMPRVD_FAST-NEXT: [[TMP26:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[EXT5]], [[TMP25]]
|
|
// IMPRVD_FAST-NEXT: [[TMP27:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT4]], [[TMP24]]
|
|
// IMPRVD_FAST-NEXT: [[TMP28:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[EXT]], [[TMP27]]
|
|
// IMPRVD_FAST-NEXT: [[TMP29:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP28]], [[TMP26]]
|
|
// IMPRVD_FAST-NEXT: [[TMP30:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[TMP24]]
|
|
// IMPRVD_FAST-NEXT: [[TMP31:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[EXT4]], [[TMP30]]
|
|
// IMPRVD_FAST-NEXT: [[TMP32:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP31]], [[TMP26]]
|
|
// IMPRVD_FAST-NEXT: br label [[COMPLEX_DIV10:%.*]]
|
|
// IMPRVD_FAST: abs_rhsr_less_than_abs_rhsi9:
|
|
// IMPRVD_FAST-NEXT: [[TMP33:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[EXT5]], [[EXT6]]
|
|
// IMPRVD_FAST-NEXT: [[TMP34:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[TMP33]], [[EXT5]]
|
|
// IMPRVD_FAST-NEXT: [[TMP35:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[EXT6]], [[TMP34]]
|
|
// IMPRVD_FAST-NEXT: [[TMP36:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[TMP33]]
|
|
// IMPRVD_FAST-NEXT: [[TMP37:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP36]], [[EXT4]]
|
|
// IMPRVD_FAST-NEXT: [[TMP38:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP37]], [[TMP35]]
|
|
// IMPRVD_FAST-NEXT: [[TMP39:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT4]], [[TMP33]]
|
|
// IMPRVD_FAST-NEXT: [[TMP40:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[TMP39]], [[EXT]]
|
|
// IMPRVD_FAST-NEXT: [[TMP41:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP40]], [[TMP35]]
|
|
// IMPRVD_FAST-NEXT: br label [[COMPLEX_DIV10]]
|
|
// IMPRVD_FAST: complex_div10:
|
|
// IMPRVD_FAST-NEXT: [[TMP42:%.*]] = phi reassoc nnan ninf nsz arcp afn float [ [[TMP29]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI8]] ], [ [[TMP38]], [[ABS_RHSR_LESS_THAN_ABS_RHSI9]] ]
|
|
// IMPRVD_FAST-NEXT: [[TMP43:%.*]] = phi reassoc nnan ninf nsz arcp afn float [ [[TMP32]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI8]] ], [ [[TMP41]], [[ABS_RHSR_LESS_THAN_ABS_RHSI9]] ]
|
|
// IMPRVD_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP42]] to half
|
|
// IMPRVD_FAST-NEXT: [[UNPROMOTION11:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP43]] to half
|
|
// IMPRVD_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// IMPRVD_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// IMPRVD_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// IMPRVD_FAST-NEXT: store half [[UNPROMOTION11]], ptr [[RETVAL_IMAGP]], align 2
|
|
// IMPRVD_FAST-NEXT: [[TMP44:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// IMPRVD_FAST-NEXT: ret <2 x half> [[TMP44]]
|
|
//
|
|
// PRMTD_FAST-LABEL: define dso_local nofpclass(nan inf) <2 x half> @f1(
|
|
// PRMTD_FAST-SAME: <2 x half> noundef nofpclass(nan inf) [[A_COERCE:%.*]], ptr noundef byval({ x86_fp80, x86_fp80 }) align 16 [[B:%.*]], <2 x half> noundef nofpclass(nan inf) [[C_COERCE:%.*]]) #[[ATTR0]] {
|
|
// PRMTD_FAST-NEXT: entry:
|
|
// PRMTD_FAST-NEXT: [[RETVAL:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: [[A:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: [[C:%.*]] = alloca { half, half }, align 2
|
|
// PRMTD_FAST-NEXT: store <2 x half> [[A_COERCE]], ptr [[A]], align 2
|
|
// PRMTD_FAST-NEXT: store <2 x half> [[C_COERCE]], ptr [[C]], align 2
|
|
// PRMTD_FAST-NEXT: [[B_REALP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[B_REAL:%.*]] = load x86_fp80, ptr [[B_REALP]], align 16
|
|
// PRMTD_FAST-NEXT: [[B_IMAGP:%.*]] = getelementptr inbounds nuw { x86_fp80, x86_fp80 }, ptr [[B]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: [[B_IMAG:%.*]] = load x86_fp80, ptr [[B_IMAGP]], align 16
|
|
// PRMTD_FAST-NEXT: [[C_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[C_REAL:%.*]] = load half, ptr [[C_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: [[C_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[C]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: [[C_IMAG:%.*]] = load half, ptr [[C_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[CONV:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_REAL]] to x86_fp80
|
|
// PRMTD_FAST-NEXT: [[CONV1:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[C_IMAG]] to x86_fp80
|
|
// PRMTD_FAST-NEXT: [[TMP0:%.*]] = call reassoc nnan ninf nsz arcp afn x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV]])
|
|
// PRMTD_FAST-NEXT: [[TMP1:%.*]] = call reassoc nnan ninf nsz arcp afn x86_fp80 @llvm.fabs.f80(x86_fp80 [[CONV1]])
|
|
// PRMTD_FAST-NEXT: [[ABS_CMP:%.*]] = fcmp reassoc nnan ninf nsz arcp afn ugt x86_fp80 [[TMP0]], [[TMP1]]
|
|
// PRMTD_FAST-NEXT: br i1 [[ABS_CMP]], label [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI:%.*]], label [[ABS_RHSR_LESS_THAN_ABS_RHSI:%.*]]
|
|
// PRMTD_FAST: abs_rhsr_greater_or_equal_abs_rhsi:
|
|
// PRMTD_FAST-NEXT: [[TMP2:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV1]], [[CONV]]
|
|
// PRMTD_FAST-NEXT: [[TMP3:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP2]], [[CONV1]]
|
|
// PRMTD_FAST-NEXT: [[TMP4:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV]], [[TMP3]]
|
|
// PRMTD_FAST-NEXT: [[TMP5:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[TMP2]]
|
|
// PRMTD_FAST-NEXT: [[TMP6:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[TMP5]]
|
|
// PRMTD_FAST-NEXT: [[TMP7:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP6]], [[TMP4]]
|
|
// PRMTD_FAST-NEXT: [[TMP8:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[TMP2]]
|
|
// PRMTD_FAST-NEXT: [[TMP9:%.*]] = fsub reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[TMP8]]
|
|
// PRMTD_FAST-NEXT: [[TMP10:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP9]], [[TMP4]]
|
|
// PRMTD_FAST-NEXT: br label [[COMPLEX_DIV:%.*]]
|
|
// PRMTD_FAST: abs_rhsr_less_than_abs_rhsi:
|
|
// PRMTD_FAST-NEXT: [[TMP11:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV]], [[CONV1]]
|
|
// PRMTD_FAST-NEXT: [[TMP12:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP11]], [[CONV]]
|
|
// PRMTD_FAST-NEXT: [[TMP13:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[CONV1]], [[TMP12]]
|
|
// PRMTD_FAST-NEXT: [[TMP14:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_REAL]], [[TMP11]]
|
|
// PRMTD_FAST-NEXT: [[TMP15:%.*]] = fadd reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP14]], [[B_IMAG]]
|
|
// PRMTD_FAST-NEXT: [[TMP16:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP15]], [[TMP13]]
|
|
// PRMTD_FAST-NEXT: [[TMP17:%.*]] = fmul reassoc nnan ninf nsz arcp afn x86_fp80 [[B_IMAG]], [[TMP11]]
|
|
// PRMTD_FAST-NEXT: [[TMP18:%.*]] = fsub reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP17]], [[B_REAL]]
|
|
// PRMTD_FAST-NEXT: [[TMP19:%.*]] = fdiv reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP18]], [[TMP13]]
|
|
// PRMTD_FAST-NEXT: br label [[COMPLEX_DIV]]
|
|
// PRMTD_FAST: complex_div:
|
|
// PRMTD_FAST-NEXT: [[TMP20:%.*]] = phi reassoc nnan ninf nsz arcp afn x86_fp80 [ [[TMP7]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP16]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// PRMTD_FAST-NEXT: [[TMP21:%.*]] = phi reassoc nnan ninf nsz arcp afn x86_fp80 [ [[TMP10]], [[ABS_RHSR_GREATER_OR_EQUAL_ABS_RHSI]] ], [ [[TMP19]], [[ABS_RHSR_LESS_THAN_ABS_RHSI]] ]
|
|
// PRMTD_FAST-NEXT: [[CONV2:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP20]] to half
|
|
// PRMTD_FAST-NEXT: [[CONV3:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn x86_fp80 [[TMP21]] to half
|
|
// PRMTD_FAST-NEXT: [[EXT:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV2]] to float
|
|
// PRMTD_FAST-NEXT: [[EXT4:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[CONV3]] to float
|
|
// PRMTD_FAST-NEXT: [[A_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[A_REAL:%.*]] = load half, ptr [[A_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: [[A_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[A]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: [[A_IMAG:%.*]] = load half, ptr [[A_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[EXT5:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_REAL]] to float
|
|
// PRMTD_FAST-NEXT: [[EXT6:%.*]] = fpext reassoc nnan ninf nsz arcp afn half [[A_IMAG]] to float
|
|
// PRMTD_FAST-NEXT: [[TMP22:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT5]]
|
|
// PRMTD_FAST-NEXT: [[TMP23:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT4]], [[EXT6]]
|
|
// PRMTD_FAST-NEXT: [[TMP24:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP22]], [[TMP23]]
|
|
// PRMTD_FAST-NEXT: [[TMP25:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT5]], [[EXT5]]
|
|
// PRMTD_FAST-NEXT: [[TMP26:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT6]], [[EXT6]]
|
|
// PRMTD_FAST-NEXT: [[TMP27:%.*]] = fadd reassoc nnan ninf nsz arcp afn float [[TMP25]], [[TMP26]]
|
|
// PRMTD_FAST-NEXT: [[TMP28:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT4]], [[EXT5]]
|
|
// PRMTD_FAST-NEXT: [[TMP29:%.*]] = fmul reassoc nnan ninf nsz arcp afn float [[EXT]], [[EXT6]]
|
|
// PRMTD_FAST-NEXT: [[TMP30:%.*]] = fsub reassoc nnan ninf nsz arcp afn float [[TMP28]], [[TMP29]]
|
|
// PRMTD_FAST-NEXT: [[TMP31:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP24]], [[TMP27]]
|
|
// PRMTD_FAST-NEXT: [[TMP32:%.*]] = fdiv reassoc nnan ninf nsz arcp afn float [[TMP30]], [[TMP27]]
|
|
// PRMTD_FAST-NEXT: [[UNPROMOTION:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP31]] to half
|
|
// PRMTD_FAST-NEXT: [[UNPROMOTION7:%.*]] = fptrunc reassoc nnan ninf nsz arcp afn float [[TMP32]] to half
|
|
// PRMTD_FAST-NEXT: [[RETVAL_REALP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 0
|
|
// PRMTD_FAST-NEXT: [[RETVAL_IMAGP:%.*]] = getelementptr inbounds nuw { half, half }, ptr [[RETVAL]], i32 0, i32 1
|
|
// PRMTD_FAST-NEXT: store half [[UNPROMOTION]], ptr [[RETVAL_REALP]], align 2
|
|
// PRMTD_FAST-NEXT: store half [[UNPROMOTION7]], ptr [[RETVAL_IMAGP]], align 2
|
|
// PRMTD_FAST-NEXT: [[TMP33:%.*]] = load <2 x half>, ptr [[RETVAL]], align 2
|
|
// PRMTD_FAST-NEXT: ret <2 x half> [[TMP33]]
|
|
//
|
|
_Complex _Float16 f1(_Complex _Float16 a, _Complex long double b, _Complex _Float16 c) {
|
|
return (_Complex _Float16)(b / c) / a;
|
|
}
|
|
//.
|
|
// FULL: [[PROF2]] = !{!"branch_weights", i32 1, i32 1048575}
|
|
//.
|
|
// FULL_FAST: [[PROF2]] = !{!"branch_weights", i32 1, i32 1048575}
|
|
//.
|