maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/half.ll

942 lines
35 KiB
LLVM
Raw Normal View History

NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=-f16c -fixup-byte-word-insts=1 \
; RUN: | FileCheck %s -check-prefixes=CHECK,CHECK-LIBCALL,BWON,BWON-NOF16C
; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=-f16c -fixup-byte-word-insts=0 \
; RUN: | FileCheck %s -check-prefixes=CHECK,CHECK-LIBCALL,BWOFF
; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+f16c -fixup-byte-word-insts=1 \
; RUN: | FileCheck %s -check-prefixes=CHECK,BWON,BWON-F16C
; RUN: llc < %s -mtriple=i686-unknown-linux-gnu -mattr +sse2 -fixup-byte-word-insts=0 \
; RUN: | FileCheck %s -check-prefixes=CHECK-I686
define void @test_load_store(half* %in, half* %out) #0 {
; BWON-LABEL: test_load_store:
; BWON: # BB#0:
; BWON-NEXT: movzwl (%rdi), %eax
; BWON-NEXT: movw %ax, (%rsi)
; BWON-NEXT: retq
;
; BWOFF-LABEL: test_load_store:
; BWOFF: # BB#0:
; BWOFF-NEXT: movw (%rdi), %ax
; BWOFF-NEXT: movw %ax, (%rsi)
; BWOFF-NEXT: retq
;
; CHECK-I686-LABEL: test_load_store:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ecx
; CHECK-I686-NEXT: movw (%ecx), %cx
; CHECK-I686-NEXT: movw %cx, (%eax)
; CHECK-I686-NEXT: retl
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-28 05:17:42 +08:00
%val = load half, half* %in
CodeGen: soften f16 type by default instead of marking legal. Actual support for softening f16 operations is still limited, and can be added when it's needed. But Soften is much closer to being a useful thing to try than keeping it Legal when no registers can actually hold such values. Longer term, we probably want something between Soften and Promote semantics for most targets, it'll be more efficient to promote the 4 basic operations to f32 than libcall them. llvm-svn: 213372
2014-07-18 20:41:46 +08:00
store half %val, half* %out
ret void
}
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define i16 @test_bitcast_from_half(half* %addr) #0 {
; BWON-LABEL: test_bitcast_from_half:
; BWON: # BB#0:
; BWON-NEXT: movzwl (%rdi), %eax
; BWON-NEXT: retq
;
; BWOFF-LABEL: test_bitcast_from_half:
; BWOFF: # BB#0:
; BWOFF-NEXT: movw (%rdi), %ax
; BWOFF-NEXT: retq
;
; CHECK-I686-LABEL: test_bitcast_from_half:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-I686-NEXT: movw (%eax), %ax
; CHECK-I686-NEXT: retl
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-28 05:17:42 +08:00
%val = load half, half* %addr
CodeGen: soften f16 type by default instead of marking legal. Actual support for softening f16 operations is still limited, and can be added when it's needed. But Soften is much closer to being a useful thing to try than keeping it Legal when no registers can actually hold such values. Longer term, we probably want something between Soften and Promote semantics for most targets, it'll be more efficient to promote the 4 basic operations to f32 than libcall them. llvm-svn: 213372
2014-07-18 20:41:46 +08:00
%val_int = bitcast half %val to i16
ret i16 %val_int
}
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define void @test_bitcast_to_half(half* %addr, i16 %in) #0 {
CodeGen: soften f16 type by default instead of marking legal. Actual support for softening f16 operations is still limited, and can be added when it's needed. But Soften is much closer to being a useful thing to try than keeping it Legal when no registers can actually hold such values. Longer term, we probably want something between Soften and Promote semantics for most targets, it'll be more efficient to promote the 4 basic operations to f32 than libcall them. llvm-svn: 213372
2014-07-18 20:41:46 +08:00
; CHECK-LABEL: test_bitcast_to_half:
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK: # BB#0:
; CHECK-NEXT: movw %si, (%rdi)
; CHECK-NEXT: retq
;
; CHECK-I686-LABEL: test_bitcast_to_half:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: movw {{[0-9]+}}(%esp), %ax
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ecx
; CHECK-I686-NEXT: movw %ax, (%ecx)
; CHECK-I686-NEXT: retl
CodeGen: soften f16 type by default instead of marking legal. Actual support for softening f16 operations is still limited, and can be added when it's needed. But Soften is much closer to being a useful thing to try than keeping it Legal when no registers can actually hold such values. Longer term, we probably want something between Soften and Promote semantics for most targets, it'll be more efficient to promote the 4 basic operations to f32 than libcall them. llvm-svn: 213372
2014-07-18 20:41:46 +08:00
%val_fp = bitcast i16 %in to half
store half %val_fp, half* %addr
ret void
}
X86: support fpext/fptrunc operations to and from 16-bit floats. llvm-svn: 213374
2014-07-18 21:01:25 +08:00
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define float @test_extend32(half* %addr) #0 {
; CHECK-LIBCALL-LABEL: test_extend32:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi
; CHECK-LIBCALL-NEXT: jmp __gnu_h2f_ieee # TAILCALL
;
; BWON-F16C-LABEL: test_extend32:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: movswl (%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_extend32:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: subl $12, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-I686-NEXT: movzwl (%eax), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: addl $12, %esp
; CHECK-I686-NEXT: retl
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-28 05:17:42 +08:00
%val16 = load half, half* %addr
X86: support fpext/fptrunc operations to and from 16-bit floats. llvm-svn: 213374
2014-07-18 21:01:25 +08:00
%val32 = fpext half %val16 to float
ret float %val32
}
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define double @test_extend64(half* %addr) #0 {
; CHECK-LIBCALL-LABEL: test_extend64:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rax
; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: cvtss2sd %xmm0, %xmm0
; CHECK-LIBCALL-NEXT: popq %rax
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_extend64:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: movswl (%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: vcvtss2sd %xmm0, %xmm0, %xmm0
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_extend64:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: subl $12, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-I686-NEXT: movzwl (%eax), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: addl $12, %esp
; CHECK-I686-NEXT: retl
[opaque pointer type] Add textual IR support for explicit type parameter to load instruction Essentially the same as the GEP change in r230786. A similar migration script can be used to update test cases, though a few more test case improvements/changes were required this time around: (r229269-r229278) import fileinput import sys import re pat = re.compile(r"((?:=|:|^)\s*load (?:atomic )?(?:volatile )?(.*?))(| addrspace\(\d+\) *)\*($| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$)") for line in sys.stdin: sys.stdout.write(re.sub(pat, r"\1, \2\3*\4", line)) Reviewers: rafael, dexonsmith, grosser Differential Revision: http://reviews.llvm.org/D7649 llvm-svn: 230794
2015-02-28 05:17:42 +08:00
%val16 = load half, half* %addr
X86: support fpext/fptrunc operations to and from 16-bit floats. llvm-svn: 213374
2014-07-18 21:01:25 +08:00
%val32 = fpext half %val16 to double
ret double %val32
}
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define void @test_trunc32(float %in, half* %addr) #0 {
; CHECK-LIBCALL-LABEL: test_trunc32:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rbx
; CHECK-LIBCALL-NEXT: movq %rdi, %rbx
; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee
; CHECK-LIBCALL-NEXT: movw %ax, (%rbx)
; CHECK-LIBCALL-NEXT: popq %rbx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_trunc32:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0
; BWON-F16C-NEXT: vmovd %xmm0, %eax
; BWON-F16C-NEXT: movw %ax, (%rdi)
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_trunc32:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $8, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi
; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: movss %xmm0, (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movw %ax, (%esi)
; CHECK-I686-NEXT: addl $8, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: retl
X86: support fpext/fptrunc operations to and from 16-bit floats. llvm-svn: 213374
2014-07-18 21:01:25 +08:00
%val16 = fptrunc float %in to half
store half %val16, half* %addr
ret void
}
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define void @test_trunc64(double %in, half* %addr) #0 {
X86: support fpext/fptrunc operations to and from 16-bit floats. llvm-svn: 213374
2014-07-18 21:01:25 +08:00
; CHECK-LABEL: test_trunc64:
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK: # BB#0:
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: movq %rdi, %rbx
; CHECK-NEXT: callq __truncdfhf2
; CHECK-NEXT: movw %ax, (%rbx)
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: retq
;
; CHECK-I686-LABEL: test_trunc64:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $8, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi
; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-I686-NEXT: movsd %xmm0, (%esp)
; CHECK-I686-NEXT: calll __truncdfhf2
; CHECK-I686-NEXT: movw %ax, (%esi)
; CHECK-I686-NEXT: addl $8, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: retl
X86: support fpext/fptrunc operations to and from 16-bit floats. llvm-svn: 213374
2014-07-18 21:01:25 +08:00
%val16 = fptrunc double %in to half
store half %val16, half* %addr
ret void
}
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
define i64 @test_fptosi_i64(half* %p) #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_fptosi_i64:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rax
; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: cvttss2si %xmm0, %rax
; CHECK-LIBCALL-NEXT: popq %rcx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_fptosi_i64:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: movswl (%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: vcvttss2si %xmm0, %rax
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_fptosi_i64:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: subl $12, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-I686-NEXT: movzwl (%eax), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstps (%esp)
; CHECK-I686-NEXT: calll __fixsfdi
; CHECK-I686-NEXT: addl $12, %esp
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%a = load half, half* %p, align 2
%r = fptosi half %a to i64
ret i64 %r
}
define void @test_sitofp_i64(i64 %a, half* %p) #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_sitofp_i64:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rbx
; CHECK-LIBCALL-NEXT: movq %rsi, %rbx
; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, %xmm0
; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee
; CHECK-LIBCALL-NEXT: movw %ax, (%rbx)
; CHECK-LIBCALL-NEXT: popq %rbx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_sitofp_i64:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0
; BWON-F16C-NEXT: vmovd %xmm0, %eax
; BWON-F16C-NEXT: movw %ax, (%rsi)
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_sitofp_i64:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $24, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi
; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-I686-NEXT: movlps %xmm0, {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fildll {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: movss %xmm0, (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movw %ax, (%esi)
; CHECK-I686-NEXT: addl $24, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%r = sitofp i64 %a to half
store half %r, half* %p
ret void
}
define i64 @test_fptoui_i64(half* %p) #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_fptoui_i64:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rax
; CHECK-LIBCALL-NEXT: movzwl (%rdi), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-LIBCALL-NEXT: movaps %xmm0, %xmm2
; CHECK-LIBCALL-NEXT: subss %xmm1, %xmm2
; CHECK-LIBCALL-NEXT: cvttss2si %xmm2, %rax
; CHECK-LIBCALL-NEXT: movabsq $-9223372036854775808, %rcx # imm = 0x8000000000000000
; CHECK-LIBCALL-NEXT: xorq %rax, %rcx
; CHECK-LIBCALL-NEXT: cvttss2si %xmm0, %rax
; CHECK-LIBCALL-NEXT: ucomiss %xmm1, %xmm0
; CHECK-LIBCALL-NEXT: cmovaeq %rcx, %rax
; CHECK-LIBCALL-NEXT: popq %rcx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_fptoui_i64:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: movswl (%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; BWON-F16C-NEXT: vsubss %xmm1, %xmm0, %xmm2
; BWON-F16C-NEXT: vcvttss2si %xmm2, %rax
; BWON-F16C-NEXT: movabsq $-9223372036854775808, %rcx # imm = 0x8000000000000000
; BWON-F16C-NEXT: xorq %rax, %rcx
; BWON-F16C-NEXT: vcvttss2si %xmm0, %rax
; BWON-F16C-NEXT: vucomiss %xmm1, %xmm0
; BWON-F16C-NEXT: cmovaeq %rcx, %rax
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_fptoui_i64:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: subl $12, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-I686-NEXT: movzwl (%eax), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstps (%esp)
; CHECK-I686-NEXT: calll __fixunssfdi
; CHECK-I686-NEXT: addl $12, %esp
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%a = load half, half* %p, align 2
%r = fptoui half %a to i64
ret i64 %r
}
define void @test_uitofp_i64(i64 %a, half* %p) #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_uitofp_i64:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rbx
; CHECK-LIBCALL-NEXT: movq %rsi, %rbx
; CHECK-LIBCALL-NEXT: testq %rdi, %rdi
; CHECK-LIBCALL-NEXT: js .LBB10_1
; CHECK-LIBCALL-NEXT: # BB#2:
; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, %xmm0
; CHECK-LIBCALL-NEXT: jmp .LBB10_3
; CHECK-LIBCALL-NEXT: .LBB10_1:
; CHECK-LIBCALL-NEXT: movq %rdi, %rax
; CHECK-LIBCALL-NEXT: shrq %rax
; CHECK-LIBCALL-NEXT: andl $1, %edi
; CHECK-LIBCALL-NEXT: orq %rax, %rdi
; CHECK-LIBCALL-NEXT: cvtsi2ssq %rdi, %xmm0
; CHECK-LIBCALL-NEXT: addss %xmm0, %xmm0
; CHECK-LIBCALL-NEXT: .LBB10_3:
; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee
; CHECK-LIBCALL-NEXT: movw %ax, (%rbx)
; CHECK-LIBCALL-NEXT: popq %rbx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_uitofp_i64:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: testq %rdi, %rdi
; BWON-F16C-NEXT: js .LBB10_1
; BWON-F16C-NEXT: # BB#2:
; BWON-F16C-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0
; BWON-F16C-NEXT: jmp .LBB10_3
; BWON-F16C-NEXT: .LBB10_1:
; BWON-F16C-NEXT: movq %rdi, %rax
; BWON-F16C-NEXT: shrq %rax
; BWON-F16C-NEXT: andl $1, %edi
; BWON-F16C-NEXT: orq %rax, %rdi
; BWON-F16C-NEXT: vcvtsi2ssq %rdi, %xmm0, %xmm0
; BWON-F16C-NEXT: vaddss %xmm0, %xmm0, %xmm0
; BWON-F16C-NEXT: .LBB10_3:
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0
; BWON-F16C-NEXT: vmovd %xmm0, %eax
; BWON-F16C-NEXT: movw %ax, (%rsi)
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_uitofp_i64:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $24, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi
; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-I686-NEXT: movlps %xmm0, {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: xorl %eax, %eax
; CHECK-I686-NEXT: cmpl $0, {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: setns %al
; CHECK-I686-NEXT: fildll {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fadds {{\.LCPI.*}}(,%eax,4)
; CHECK-I686-NEXT: fstps (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movw %ax, (%esi)
; CHECK-I686-NEXT: addl $24, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%r = uitofp i64 %a to half
store half %r, half* %p
ret void
}
define <4 x float> @test_extend32_vec4(<4 x half>* %p) #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_extend32_vec4:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rbx
; CHECK-LIBCALL-NEXT: subq $48, %rsp
; CHECK-LIBCALL-NEXT: movq %rdi, %rbx
Revert "[MachineCopyPropagation] Extend pass to do COPY source forwarding" round 2 This reverts commit r311135. sanitizer-x86_64-linux-android buildbot is timing out with just this patch applied. llvm-svn: 311142
2017-08-18 09:43:11 +08:00
; CHECK-LIBCALL-NEXT: movzwl (%rbx), %edi
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill
; CHECK-LIBCALL-NEXT: movzwl 2(%rbx), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill
; CHECK-LIBCALL-NEXT: movzwl 4(%rbx), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-LIBCALL-NEXT: movzwl 6(%rbx), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movaps (%rsp), %xmm1 # 16-byte Reload
; CHECK-LIBCALL-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; CHECK-LIBCALL-NEXT: movaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
; CHECK-LIBCALL-NEXT: unpcklps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Folded Reload
; CHECK-LIBCALL-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-LIBCALL-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-LIBCALL-NEXT: addq $48, %rsp
; CHECK-LIBCALL-NEXT: popq %rbx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_extend32_vec4:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: movswl 6(%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: movswl 4(%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm1
; BWON-F16C-NEXT: vcvtph2ps %xmm1, %xmm1
; BWON-F16C-NEXT: movswl (%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm2
; BWON-F16C-NEXT: vcvtph2ps %xmm2, %xmm2
; BWON-F16C-NEXT: movswl 2(%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm3
; BWON-F16C-NEXT: vcvtph2ps %xmm3, %xmm3
; BWON-F16C-NEXT: vinsertps {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[2,3]
; BWON-F16C-NEXT: vinsertps {{.*#+}} xmm1 = xmm2[0,1],xmm1[0],xmm2[3]
; BWON-F16C-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_extend32_vec4:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $56, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi
; CHECK-I686-NEXT: movzwl 2(%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill
; CHECK-I686-NEXT: movzwl 4(%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill
; CHECK-I686-NEXT: movzwl 6(%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: movzwl (%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload
; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload
; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; CHECK-I686-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; CHECK-I686-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-I686-NEXT: addl $56, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%a = load <4 x half>, <4 x half>* %p, align 8
%b = fpext <4 x half> %a to <4 x float>
ret <4 x float> %b
}
define <4 x double> @test_extend64_vec4(<4 x half>* %p) #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_extend64_vec4:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rbx
; CHECK-LIBCALL-NEXT: subq $16, %rsp
; CHECK-LIBCALL-NEXT: movq %rdi, %rbx
Revert "[MachineCopyPropagation] Extend pass to do COPY source forwarding" round 2 This reverts commit r311135. sanitizer-x86_64-linux-android buildbot is timing out with just this patch applied. llvm-svn: 311142
2017-08-18 09:43:11 +08:00
; CHECK-LIBCALL-NEXT: movzwl 4(%rbx), %edi
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill
; CHECK-LIBCALL-NEXT: movzwl 6(%rbx), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill
; CHECK-LIBCALL-NEXT: movzwl (%rbx), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill
; CHECK-LIBCALL-NEXT: movzwl 2(%rbx), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: cvtss2sd %xmm0, %xmm1
; CHECK-LIBCALL-NEXT: movss {{[0-9]+}}(%rsp), %xmm0 # 4-byte Reload
; CHECK-LIBCALL-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-LIBCALL-NEXT: cvtss2sd %xmm0, %xmm0
; CHECK-LIBCALL-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; CHECK-LIBCALL-NEXT: movss {{[0-9]+}}(%rsp), %xmm1 # 4-byte Reload
; CHECK-LIBCALL-NEXT: # xmm1 = mem[0],zero,zero,zero
; CHECK-LIBCALL-NEXT: cvtss2sd %xmm1, %xmm2
; CHECK-LIBCALL-NEXT: movss {{[0-9]+}}(%rsp), %xmm1 # 4-byte Reload
; CHECK-LIBCALL-NEXT: # xmm1 = mem[0],zero,zero,zero
; CHECK-LIBCALL-NEXT: cvtss2sd %xmm1, %xmm1
; CHECK-LIBCALL-NEXT: unpcklpd {{.*#+}} xmm1 = xmm1[0],xmm2[0]
; CHECK-LIBCALL-NEXT: addq $16, %rsp
; CHECK-LIBCALL-NEXT: popq %rbx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_extend64_vec4:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: movswl (%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: movswl 2(%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm1
; BWON-F16C-NEXT: vcvtph2ps %xmm1, %xmm1
; BWON-F16C-NEXT: movswl 4(%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm2
; BWON-F16C-NEXT: vcvtph2ps %xmm2, %xmm2
; BWON-F16C-NEXT: movswl 6(%rdi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm3
; BWON-F16C-NEXT: vcvtph2ps %xmm3, %xmm3
; BWON-F16C-NEXT: vcvtss2sd %xmm3, %xmm3, %xmm3
; BWON-F16C-NEXT: vcvtss2sd %xmm2, %xmm2, %xmm2
; BWON-F16C-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
; BWON-F16C-NEXT: vcvtss2sd %xmm1, %xmm1, %xmm1
; BWON-F16C-NEXT: vcvtss2sd %xmm0, %xmm0, %xmm0
; BWON-F16C-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; BWON-F16C-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_extend64_vec4:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $88, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %esi
; CHECK-I686-NEXT: movzwl 6(%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill
; CHECK-I686-NEXT: movzwl 4(%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill
; CHECK-I686-NEXT: movzwl 2(%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstpt {{[0-9]+}}(%esp) # 10-byte Folded Spill
; CHECK-I686-NEXT: movzwl (%esi), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload
; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload
; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: fldt {{[0-9]+}}(%esp) # 10-byte Folded Reload
; CHECK-I686-NEXT: fstpl {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-I686-NEXT: movhpd {{.*#+}} xmm0 = xmm0[0],mem[0]
; CHECK-I686-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
; CHECK-I686-NEXT: movhpd {{.*#+}} xmm1 = xmm1[0],mem[0]
; CHECK-I686-NEXT: addl $88, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%a = load <4 x half>, <4 x half>* %p, align 8
%b = fpext <4 x half> %a to <4 x double>
ret <4 x double> %b
}
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define void @test_trunc32_vec4(<4 x float> %a, <4 x half>* %p) #0 {
; BWON-NOF16C-LABEL: test_trunc32_vec4:
; BWON-NOF16C: # BB#0:
; BWON-NOF16C-NEXT: pushq %rbp
; BWON-NOF16C-NEXT: pushq %r15
; BWON-NOF16C-NEXT: pushq %r14
; BWON-NOF16C-NEXT: pushq %rbx
; BWON-NOF16C-NEXT: subq $24, %rsp
; BWON-NOF16C-NEXT: movq %rdi, %rbx
; BWON-NOF16C-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; BWON-NOF16C-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,2,3]
; BWON-NOF16C-NEXT: callq __gnu_f2h_ieee
; BWON-NOF16C-NEXT: movl %eax, %r14d
; BWON-NOF16C-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWON-NOF16C-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; BWON-NOF16C-NEXT: callq __gnu_f2h_ieee
; BWON-NOF16C-NEXT: movl %eax, %r15d
; BWON-NOF16C-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWON-NOF16C-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
; BWON-NOF16C-NEXT: callq __gnu_f2h_ieee
; BWON-NOF16C-NEXT: movl %eax, %ebp
; BWON-NOF16C-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWON-NOF16C-NEXT: callq __gnu_f2h_ieee
; BWON-NOF16C-NEXT: movw %ax, (%rbx)
; BWON-NOF16C-NEXT: movw %bp, 6(%rbx)
; BWON-NOF16C-NEXT: movw %r15w, 4(%rbx)
; BWON-NOF16C-NEXT: movw %r14w, 2(%rbx)
; BWON-NOF16C-NEXT: addq $24, %rsp
; BWON-NOF16C-NEXT: popq %rbx
; BWON-NOF16C-NEXT: popq %r14
; BWON-NOF16C-NEXT: popq %r15
; BWON-NOF16C-NEXT: popq %rbp
; BWON-NOF16C-NEXT: retq
;
; BWOFF-LABEL: test_trunc32_vec4:
; BWOFF: # BB#0:
; BWOFF-NEXT: pushq %rbp
; BWOFF-NEXT: pushq %r15
; BWOFF-NEXT: pushq %r14
; BWOFF-NEXT: pushq %rbx
; BWOFF-NEXT: subq $24, %rsp
; BWOFF-NEXT: movq %rdi, %rbx
; BWOFF-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; BWOFF-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,2,3]
; BWOFF-NEXT: callq __gnu_f2h_ieee
; BWOFF-NEXT: movw %ax, %r14w
; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWOFF-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; BWOFF-NEXT: callq __gnu_f2h_ieee
; BWOFF-NEXT: movw %ax, %r15w
; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWOFF-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
; BWOFF-NEXT: callq __gnu_f2h_ieee
; BWOFF-NEXT: movw %ax, %bp
; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWOFF-NEXT: callq __gnu_f2h_ieee
; BWOFF-NEXT: movw %ax, (%rbx)
; BWOFF-NEXT: movw %bp, 6(%rbx)
; BWOFF-NEXT: movw %r15w, 4(%rbx)
; BWOFF-NEXT: movw %r14w, 2(%rbx)
; BWOFF-NEXT: addq $24, %rsp
; BWOFF-NEXT: popq %rbx
; BWOFF-NEXT: popq %r14
; BWOFF-NEXT: popq %r15
; BWOFF-NEXT: popq %rbp
; BWOFF-NEXT: retq
;
; BWON-F16C-LABEL: test_trunc32_vec4:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1
; BWON-F16C-NEXT: vmovd %xmm1, %eax
; BWON-F16C-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1
; BWON-F16C-NEXT: vmovd %xmm1, %ecx
; BWON-F16C-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[3,1,2,3]
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1
; BWON-F16C-NEXT: vmovd %xmm1, %edx
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0
; BWON-F16C-NEXT: vmovd %xmm0, %esi
; BWON-F16C-NEXT: movw %si, (%rdi)
; BWON-F16C-NEXT: movw %dx, 6(%rdi)
; BWON-F16C-NEXT: movw %cx, 4(%rdi)
; BWON-F16C-NEXT: movw %ax, 2(%rdi)
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_trunc32_vec4:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %ebp
; CHECK-I686-NEXT: pushl %ebx
; CHECK-I686-NEXT: pushl %edi
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $44, %esp
; CHECK-I686-NEXT: movaps %xmm0, {{[0-9]+}}(%esp) # 16-byte Spill
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ebp
; CHECK-I686-NEXT: movaps %xmm0, %xmm1
Revert "[MachineCopyPropagation] Extend pass to do COPY source forwarding" round 2 This reverts commit r311135. sanitizer-x86_64-linux-android buildbot is timing out with just this patch applied. llvm-svn: 311142
2017-08-18 09:43:11 +08:00
; CHECK-I686-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1,2,3]
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-I686-NEXT: movss %xmm1, (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movw %ax, %si
; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload
; CHECK-I686-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-I686-NEXT: movss %xmm0, (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movw %ax, %di
; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload
; CHECK-I686-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3]
; CHECK-I686-NEXT: movss %xmm0, (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movw %ax, %bx
; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload
; CHECK-I686-NEXT: movss %xmm0, (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movw %ax, (%ebp)
; CHECK-I686-NEXT: movw %bx, 6(%ebp)
; CHECK-I686-NEXT: movw %di, 4(%ebp)
; CHECK-I686-NEXT: movw %si, 2(%ebp)
; CHECK-I686-NEXT: addl $44, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: popl %edi
; CHECK-I686-NEXT: popl %ebx
; CHECK-I686-NEXT: popl %ebp
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%v = fptrunc <4 x float> %a to <4 x half>
store <4 x half> %v, <4 x half>* %p
ret void
}
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
define void @test_trunc64_vec4(<4 x double> %a, <4 x half>* %p) #0 {
; BWON-NOF16C-LABEL: test_trunc64_vec4:
; BWON-NOF16C: # BB#0:
; BWON-NOF16C-NEXT: pushq %rbp
; BWON-NOF16C-NEXT: pushq %r15
; BWON-NOF16C-NEXT: pushq %r14
; BWON-NOF16C-NEXT: pushq %rbx
; BWON-NOF16C-NEXT: subq $40, %rsp
; BWON-NOF16C-NEXT: movq %rdi, %rbx
; BWON-NOF16C-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill
; BWON-NOF16C-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill
; BWON-NOF16C-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; BWON-NOF16C-NEXT: callq __truncdfhf2
; BWON-NOF16C-NEXT: movl %eax, %r14d
; BWON-NOF16C-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWON-NOF16C-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; BWON-NOF16C-NEXT: callq __truncdfhf2
; BWON-NOF16C-NEXT: movl %eax, %r15d
; BWON-NOF16C-NEXT: movaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
; BWON-NOF16C-NEXT: callq __truncdfhf2
; BWON-NOF16C-NEXT: movl %eax, %ebp
; BWON-NOF16C-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWON-NOF16C-NEXT: callq __truncdfhf2
; BWON-NOF16C-NEXT: movw %ax, 4(%rbx)
; BWON-NOF16C-NEXT: movw %bp, (%rbx)
; BWON-NOF16C-NEXT: movw %r15w, 6(%rbx)
; BWON-NOF16C-NEXT: movw %r14w, 2(%rbx)
; BWON-NOF16C-NEXT: addq $40, %rsp
; BWON-NOF16C-NEXT: popq %rbx
; BWON-NOF16C-NEXT: popq %r14
; BWON-NOF16C-NEXT: popq %r15
; BWON-NOF16C-NEXT: popq %rbp
; BWON-NOF16C-NEXT: retq
;
; BWOFF-LABEL: test_trunc64_vec4:
; BWOFF: # BB#0:
; BWOFF-NEXT: pushq %rbp
; BWOFF-NEXT: pushq %r15
; BWOFF-NEXT: pushq %r14
; BWOFF-NEXT: pushq %rbx
; BWOFF-NEXT: subq $40, %rsp
; BWOFF-NEXT: movq %rdi, %rbx
; BWOFF-NEXT: movaps %xmm1, (%rsp) # 16-byte Spill
; BWOFF-NEXT: movaps %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill
; BWOFF-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; BWOFF-NEXT: callq __truncdfhf2
; BWOFF-NEXT: movw %ax, %r14w
; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWOFF-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; BWOFF-NEXT: callq __truncdfhf2
; BWOFF-NEXT: movw %ax, %r15w
; BWOFF-NEXT: movaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
; BWOFF-NEXT: callq __truncdfhf2
; BWOFF-NEXT: movw %ax, %bp
; BWOFF-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; BWOFF-NEXT: callq __truncdfhf2
; BWOFF-NEXT: movw %ax, 4(%rbx)
; BWOFF-NEXT: movw %bp, (%rbx)
; BWOFF-NEXT: movw %r15w, 6(%rbx)
; BWOFF-NEXT: movw %r14w, 2(%rbx)
; BWOFF-NEXT: addq $40, %rsp
; BWOFF-NEXT: popq %rbx
; BWOFF-NEXT: popq %r14
; BWOFF-NEXT: popq %r15
; BWOFF-NEXT: popq %rbp
; BWOFF-NEXT: retq
;
; BWON-F16C-LABEL: test_trunc64_vec4:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: pushq %rbp
; BWON-F16C-NEXT: pushq %r15
; BWON-F16C-NEXT: pushq %r14
; BWON-F16C-NEXT: pushq %rbx
; BWON-F16C-NEXT: subq $88, %rsp
; BWON-F16C-NEXT: movq %rdi, %rbx
; BWON-F16C-NEXT: vmovupd %ymm0, {{[0-9]+}}(%rsp) # 32-byte Spill
; BWON-F16C-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
; BWON-F16C-NEXT: vzeroupper
; BWON-F16C-NEXT: callq __truncdfhf2
; BWON-F16C-NEXT: movl %eax, %r14d
; BWON-F16C-NEXT: vmovups {{[0-9]+}}(%rsp), %ymm0 # 32-byte Reload
; BWON-F16C-NEXT: vextractf128 $1, %ymm0, %xmm0
; BWON-F16C-NEXT: vmovapd %xmm0, {{[0-9]+}}(%rsp) # 16-byte Spill
; BWON-F16C-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
; BWON-F16C-NEXT: vzeroupper
; BWON-F16C-NEXT: callq __truncdfhf2
; BWON-F16C-NEXT: movl %eax, %r15d
; BWON-F16C-NEXT: vmovups {{[0-9]+}}(%rsp), %ymm0 # 32-byte Reload
; BWON-F16C-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<kill>
; BWON-F16C-NEXT: vzeroupper
; BWON-F16C-NEXT: callq __truncdfhf2
; BWON-F16C-NEXT: movl %eax, %ebp
; BWON-F16C-NEXT: vmovaps {{[0-9]+}}(%rsp), %xmm0 # 16-byte Reload
; BWON-F16C-NEXT: callq __truncdfhf2
; BWON-F16C-NEXT: movw %ax, 4(%rbx)
; BWON-F16C-NEXT: movw %bp, (%rbx)
; BWON-F16C-NEXT: movw %r15w, 6(%rbx)
; BWON-F16C-NEXT: movw %r14w, 2(%rbx)
; BWON-F16C-NEXT: addq $88, %rsp
; BWON-F16C-NEXT: popq %rbx
; BWON-F16C-NEXT: popq %r14
; BWON-F16C-NEXT: popq %r15
; BWON-F16C-NEXT: popq %rbp
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_trunc64_vec4:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: pushl %ebp
; CHECK-I686-NEXT: pushl %ebx
; CHECK-I686-NEXT: pushl %edi
; CHECK-I686-NEXT: pushl %esi
; CHECK-I686-NEXT: subl $60, %esp
; CHECK-I686-NEXT: movaps %xmm1, {{[0-9]+}}(%esp) # 16-byte Spill
; CHECK-I686-NEXT: movaps %xmm0, {{[0-9]+}}(%esp) # 16-byte Spill
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %ebp
; CHECK-I686-NEXT: movlps %xmm0, (%esp)
; CHECK-I686-NEXT: calll __truncdfhf2
; CHECK-I686-NEXT: movw %ax, %si
; CHECK-I686-NEXT: movapd {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload
; CHECK-I686-NEXT: movhpd %xmm0, (%esp)
; CHECK-I686-NEXT: calll __truncdfhf2
; CHECK-I686-NEXT: movw %ax, %di
; CHECK-I686-NEXT: movaps {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload
; CHECK-I686-NEXT: movlps %xmm0, (%esp)
; CHECK-I686-NEXT: calll __truncdfhf2
; CHECK-I686-NEXT: movw %ax, %bx
; CHECK-I686-NEXT: movapd {{[0-9]+}}(%esp), %xmm0 # 16-byte Reload
; CHECK-I686-NEXT: movhpd %xmm0, (%esp)
; CHECK-I686-NEXT: calll __truncdfhf2
; CHECK-I686-NEXT: movw %ax, 6(%ebp)
; CHECK-I686-NEXT: movw %bx, 4(%ebp)
; CHECK-I686-NEXT: movw %di, 2(%ebp)
; CHECK-I686-NEXT: movw %si, (%ebp)
; CHECK-I686-NEXT: addl $60, %esp
; CHECK-I686-NEXT: popl %esi
; CHECK-I686-NEXT: popl %edi
; CHECK-I686-NEXT: popl %ebx
; CHECK-I686-NEXT: popl %ebp
; CHECK-I686-NEXT: retl
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
%v = fptrunc <4 x double> %a to <4 x half>
store <4 x half> %v, <4 x half>* %p
ret void
}
Don't combine fp_round (fp_round x) if f80 to f16 is generated Summary: This patch skips DAG combine of fp_round (fp_round x) if it results in an fp_round from f80 to f16. fp_round from f80 to f16 always generates an expensive (and as yet, unimplemented) libcall to __truncxfhf2. This prevents selection of native f16 conversion instructions from f32 or f64. Moreover, the first (value-preserving) fp_round from f80 to either f32 or f64 may become a NOP in platforms like x86. Reviewers: ab Subscribers: srhines, llvm-commits Differential Revision: http://reviews.llvm.org/D17221 llvm-svn: 260769
2016-02-13 08:08:05 +08:00
declare float @test_floatret();
Reverting commit 306414 on behalf of @gadi.haber llvm-svn: 306532
2017-06-28 19:23:31 +08:00
; On i686, if SSE2 is available, the return value from test_floatret is loaded
; to f80 and then rounded to f32. The DAG combiner should not combine this
; fp_round and the subsequent fptrunc from float to half.
Don't combine fp_round (fp_round x) if f80 to f16 is generated Summary: This patch skips DAG combine of fp_round (fp_round x) if it results in an fp_round from f80 to f16. fp_round from f80 to f16 always generates an expensive (and as yet, unimplemented) libcall to __truncxfhf2. This prevents selection of native f16 conversion instructions from f32 or f64. Moreover, the first (value-preserving) fp_round from f80 to either f32 or f64 may become a NOP in platforms like x86. Reviewers: ab Subscribers: srhines, llvm-commits Differential Revision: http://reviews.llvm.org/D17221 llvm-svn: 260769
2016-02-13 08:08:05 +08:00
define half @test_f80trunc_nodagcombine() #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_f80trunc_nodagcombine:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rax
; CHECK-LIBCALL-NEXT: callq test_floatret
; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee
; CHECK-LIBCALL-NEXT: movzwl %ax, %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: popq %rax
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_f80trunc_nodagcombine:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: pushq %rax
; BWON-F16C-NEXT: callq test_floatret
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm0, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: popq %rax
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_f80trunc_nodagcombine:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: subl $12, %esp
; CHECK-I686-NEXT: calll test_floatret
; CHECK-I686-NEXT: fstps (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movzwl %ax, %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: addl $12, %esp
; CHECK-I686-NEXT: retl
Don't combine fp_round (fp_round x) if f80 to f16 is generated Summary: This patch skips DAG combine of fp_round (fp_round x) if it results in an fp_round from f80 to f16. fp_round from f80 to f16 always generates an expensive (and as yet, unimplemented) libcall to __truncxfhf2. This prevents selection of native f16 conversion instructions from f32 or f64. Moreover, the first (value-preserving) fp_round from f80 to either f32 or f64 may become a NOP in platforms like x86. Reviewers: ab Subscribers: srhines, llvm-commits Differential Revision: http://reviews.llvm.org/D17221 llvm-svn: 260769
2016-02-13 08:08:05 +08:00
%1 = call float @test_floatret()
%2 = fptrunc float %1 to half
ret half %2
}
[CodeGen] Round [SU]INT_TO_FP result when promoting from f16. If we don't, values that aren't precisely representable in f16 could be used as-is in a promoted f32 operation, which would produce incorrect results. AArch64 had the correct behavior; add a focused test. Fixes http://llvm.org/PR26871 llvm-svn: 268700
2016-05-06 08:58:00 +08:00
define float @test_sitofp_fadd_i32(i32 %a, half* %b) #0 {
NFC. Made some updates to the half.ll test under CodeGen to make it friendly to the update_llc_test_checks .py tool as follows: 1.Removing the llc flag -asm-verbose=false 2.Grouping the multiple check-prefix directives 3.Apply update_llc_test_checks.py tool on the test This change is needed to easily update scheduling changes in an upcoming patch. Reviewers: zvi, RKSimon, craig.topper Differential Revision: https://reviews.llvm.org/D34934 llvm-svn: 307108
2017-07-05 05:51:05 +08:00
; CHECK-LIBCALL-LABEL: test_sitofp_fadd_i32:
; CHECK-LIBCALL: # BB#0:
; CHECK-LIBCALL-NEXT: pushq %rbx
; CHECK-LIBCALL-NEXT: subq $16, %rsp
; CHECK-LIBCALL-NEXT: movl %edi, %ebx
; CHECK-LIBCALL-NEXT: movzwl (%rsi), %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: movss %xmm0, {{[0-9]+}}(%rsp) # 4-byte Spill
; CHECK-LIBCALL-NEXT: cvtsi2ssl %ebx, %xmm0
; CHECK-LIBCALL-NEXT: callq __gnu_f2h_ieee
; CHECK-LIBCALL-NEXT: movzwl %ax, %edi
; CHECK-LIBCALL-NEXT: callq __gnu_h2f_ieee
; CHECK-LIBCALL-NEXT: addss {{[0-9]+}}(%rsp), %xmm0 # 4-byte Folded Reload
; CHECK-LIBCALL-NEXT: addq $16, %rsp
; CHECK-LIBCALL-NEXT: popq %rbx
; CHECK-LIBCALL-NEXT: retq
;
; BWON-F16C-LABEL: test_sitofp_fadd_i32:
; BWON-F16C: # BB#0:
; BWON-F16C-NEXT: movswl (%rsi), %eax
; BWON-F16C-NEXT: vmovd %eax, %xmm0
; BWON-F16C-NEXT: vcvtph2ps %xmm0, %xmm0
; BWON-F16C-NEXT: vcvtsi2ssl %edi, %xmm1, %xmm1
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm1, %xmm1
; BWON-F16C-NEXT: vcvtph2ps %xmm1, %xmm1
; BWON-F16C-NEXT: vaddss %xmm1, %xmm0, %xmm0
; BWON-F16C-NEXT: retq
;
; CHECK-I686-LABEL: test_sitofp_fadd_i32:
; CHECK-I686: # BB#0:
; CHECK-I686-NEXT: subl $28, %esp
; CHECK-I686-NEXT: movl {{[0-9]+}}(%esp), %eax
; CHECK-I686-NEXT: movzwl (%eax), %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: movss %xmm0, {{[0-9]+}}(%esp) # 4-byte Spill
; CHECK-I686-NEXT: xorps %xmm0, %xmm0
; CHECK-I686-NEXT: cvtsi2ssl {{[0-9]+}}(%esp), %xmm0
; CHECK-I686-NEXT: movss %xmm0, (%esp)
; CHECK-I686-NEXT: calll __gnu_f2h_ieee
; CHECK-I686-NEXT: movzwl %ax, %eax
; CHECK-I686-NEXT: movl %eax, (%esp)
; CHECK-I686-NEXT: calll __gnu_h2f_ieee
; CHECK-I686-NEXT: fstps {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: movss {{[0-9]+}}(%esp), %xmm0 # 4-byte Reload
; CHECK-I686-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-I686-NEXT: addss {{[0-9]+}}(%esp), %xmm0
; CHECK-I686-NEXT: movss %xmm0, {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: flds {{[0-9]+}}(%esp)
; CHECK-I686-NEXT: addl $28, %esp
; CHECK-I686-NEXT: retl
[CodeGen] Round [SU]INT_TO_FP result when promoting from f16. If we don't, values that aren't precisely representable in f16 could be used as-is in a promoted f32 operation, which would produce incorrect results. AArch64 had the correct behavior; add a focused test. Fixes http://llvm.org/PR26871 llvm-svn: 268700
2016-05-06 08:58:00 +08:00
%tmp0 = load half, half* %b
%tmp1 = sitofp i32 %a to half
%tmp2 = fadd half %tmp0, %tmp1
%tmp3 = fpext half %tmp2 to float
ret float %tmp3
}
[X86] Updates to X86 backend for f16 promotion Summary: r235215 adds support for f16 to be considered as a load/store type and promote f16 operations to f32. This patch has miscellaneous fixes for the X86 backend so all f16 operations are handled: 1. Set loadextaction for f16 vectors to expand. 2. Handle FP_EXTEND in a switch statement when handling v2f32 3. Do not fold (FP_TO_SINT (load f16)) into FP_TO_INT*_IN_MEM or (store (SINT_TO_FP )) to a FILD. Tests included. Reviewers: ab, srhines, delena Subscribers: llvm-commits Differential Revision: http://reviews.llvm.org/D9092 llvm-svn: 237004
2015-05-12 01:14:39 +08:00
attributes #0 = { nounwind }