RPCSX/llvm
1
0
Fork 0
mirror of https://github.com/RPCSX/llvm.git synced 2025-02-27 00:15:58 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[X86] Remove some intrinsic instructions from hasPartialRegUpdate

Browse Source 
Summary:
These intrinsic instructions are all selected from intrinsics that have well defined behavior for where the upper bits come from. It's not the same place as the lower bits.

As you can see we were suppressing load folding for these instructions in some cases. In none of the cases was the separate load helping avoid a partial dependency on the destination register. So we should just go ahead and allow the load to be folded.

Only foldMemoryOperand was suppressing folding for these. They all have patterns for folding sse_load_f32/f64 that aren't gated with OptForSize, but sse_load_f32/f64 doesn't allow 128-bit vector loads. It only allows scalar_to_vector and vzmovl of scalar loads to match. There's no reason we can't allow a 128-bit vector load to be narrowed so I would like to fix sse_load_f32/f64 to allow that. And if I do that it changes some of these same test cases to fold the load too.

Reviewers: spatel, zvi, RKSimon

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D27611

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@289419 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Craig Topper 2016-12-12 05:07:17 +00:00
parent b226b5d5c0
commit 4f6cd5addf
6 changed files with 25 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
lib/Target/X86/X86InstrInfo.cpp
View File

@ -7242,12 +7242,8 @@ static bool hasPartialRegUpdate(unsigned Opcode) {
case X86::CVTSI2SD64rm:
case X86::CVTSD2SSrr:
case X86::CVTSD2SSrm:
case X86::Int_CVTSD2SSrr:
case X86::Int_CVTSD2SSrm:
case X86::CVTSS2SDrr:
case X86::CVTSS2SDrm:
case X86::Int_CVTSS2SDrr:
case X86::Int_CVTSS2SDrm:
case X86::MOVHPDrm:
case X86::MOVHPSrm:
case X86::MOVLPDrm:
@ -7258,12 +7254,8 @@ static bool hasPartialRegUpdate(unsigned Opcode) {
case X86::RCPSSm_Int:
case X86::ROUNDSDr:
case X86::ROUNDSDm:
case X86::ROUNDSDr_Int:
case X86::ROUNDSDm_Int:
case X86::ROUNDSSr:
case X86::ROUNDSSm:
case X86::ROUNDSSr_Int:
case X86::ROUNDSSm_Int:
case X86::RSQRTSSr:
case X86::RSQRTSSm:
case X86::RSQRTSSr_Int:

6
test/CodeGen/X86/sse2-intrinsics-fast-isel.ll
View File

@ -1227,14 +1227,12 @@ define <4 x float> @test_mm_cvtsd_ss_load(<4 x float> %a0, <2 x double>* %p1) {
; X32-LABEL: test_mm_cvtsd_ss_load:
; X32: # BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-NEXT: movaps (%eax), %xmm1
; X32-NEXT: cvtsd2ss %xmm1, %xmm0
; X32-NEXT: cvtsd2ss (%eax), %xmm0
; X32-NEXT: retl
;
; X64-LABEL: test_mm_cvtsd_ss_load:
; X64: # BB#0:
; X64-NEXT: movaps (%rdi), %xmm1
; X64-NEXT: cvtsd2ss %xmm1, %xmm0
; X64-NEXT: cvtsd2ss (%rdi), %xmm0
; X64-NEXT: retq
%a1 = load <2 x double>, <2 x double>* %p1
%res = call <4 x float> @llvm.x86.sse2.cvtsd2ss(<4 x float> %a0, <2 x double> %a1)

6
test/CodeGen/X86/sse2-intrinsics-x86.ll
View File

@ -370,8 +370,7 @@ define <4 x float> @test_x86_sse2_cvtsd2ss_load(<4 x float> %a0, <2 x double>* %
; SSE-LABEL: test_x86_sse2_cvtsd2ss_load:
; SSE: ## BB#0:
; SSE-NEXT: movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
; SSE-NEXT: movaps (%eax), %xmm1 ## encoding: [0x0f,0x28,0x08]
; SSE-NEXT: cvtsd2ss %xmm1, %xmm0 ## encoding: [0xf2,0x0f,0x5a,0xc1]
; SSE-NEXT: cvtsd2ss (%eax), %xmm0 ## encoding: [0xf2,0x0f,0x5a,0x00]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; VCHECK-LABEL: test_x86_sse2_cvtsd2ss_load:
@ -444,8 +443,7 @@ define <2 x double> @test_x86_sse2_cvtss2sd_load(<2 x double> %a0, <4 x float>*
; SSE-LABEL: test_x86_sse2_cvtss2sd_load:
; SSE: ## BB#0:
; SSE-NEXT: movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
; SSE-NEXT: movaps (%eax), %xmm1 ## encoding: [0x0f,0x28,0x08]
; SSE-NEXT: cvtss2sd %xmm1, %xmm0 ## encoding: [0xf3,0x0f,0x5a,0xc1]
; SSE-NEXT: cvtss2sd (%eax), %xmm0 ## encoding: [0xf3,0x0f,0x5a,0x00]
; SSE-NEXT: retl ## encoding: [0xc3]
;
; VCHECK-LABEL: test_x86_sse2_cvtss2sd_load:

18
test/CodeGen/X86/sse41-intrinsics-x86.ll
View File

@ -467,6 +467,24 @@ define <2 x double> @test_x86_sse41_round_sd(<2 x double> %a0, <2 x double> %a1)
declare <2 x double> @llvm.x86.sse41.round.sd(<2 x double>, <2 x double>, i32) nounwind readnone
define <2 x double> @test_x86_sse41_round_sd_load(<2 x double> %a0, <2 x double>* %a1) {
; SSE41-LABEL: test_x86_sse41_round_sd_load:
; SSE41: ## BB#0:
; SSE41-NEXT: movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
; SSE41-NEXT: roundsd $7, (%eax), %xmm0 ## encoding: [0x66,0x0f,0x3a,0x0b,0x00,0x07]
; SSE41-NEXT: retl ## encoding: [0xc3]
;
; VCHECK-LABEL: test_x86_sse41_round_sd_load:
; VCHECK: ## BB#0:
; VCHECK-NEXT: movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
; VCHECK-NEXT: vroundsd $7, (%eax), %xmm0, %xmm0 ## encoding: [0xc4,0xe3,0x79,0x0b,0x00,0x07]
; VCHECK-NEXT: retl ## encoding: [0xc3]
%a1b = load <2 x double>, <2 x double>* %a1
%res = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %a1b, i32 7) ; <<2 x double>> [#uses=1]
ret <2 x double> %res
}
define <4 x float> @test_x86_sse41_round_ss(<4 x float> %a0, <4 x float> %a1) {
; SSE41-LABEL: test_x86_sse41_round_ss:
; SSE41: ## BB#0:

3
test/CodeGen/X86/sse_partial_update.ll
View File

@ -98,9 +98,8 @@ declare <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double>) nounwind readnone
define <2 x double> @load_fold_cvtss2sd_int(<4 x float> *%a) {
; CHECK-LABEL: load_fold_cvtss2sd_int:
; CHECK: ## BB#0:
; CHECK-NEXT: movaps (%rdi), %xmm1
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: cvtss2sd %xmm1, %xmm0
; CHECK-NEXT: cvtss2sd (%rdi), %xmm0
; CHECK-NEXT: retq
%ld = load <4 x float>, <4 x float> *%a
%x = call <2 x double> @llvm.x86.sse2.cvtss2sd(<2 x double> <double 0x0, double 0x0>, <4 x float> %ld)

6
test/CodeGen/X86/vec_ss_load_fold.ll
View File

@ -200,8 +200,7 @@ define <4 x float> @test4(<4 x float> %A, float *%b, i32 %C) nounwind {
; X32-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X32-NEXT: movaps %xmm0, (%esp) ## 16-byte Spill
; X32-NEXT: calll _f
; X32-NEXT: movaps (%esp), %xmm1 ## 16-byte Reload
; X32-NEXT: roundss $4, %xmm1, %xmm0
; X32-NEXT: roundss $4, (%esp), %xmm0 ## 16-byte Folded Reload
; X32-NEXT: addl $28, %esp
; X32-NEXT: retl
;
@ -211,8 +210,7 @@ define <4 x float> @test4(<4 x float> %A, float *%b, i32 %C) nounwind {
; X64-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
; X64-NEXT: movaps %xmm0, (%rsp) ## 16-byte Spill
; X64-NEXT: callq _f
; X64-NEXT: movaps (%rsp), %xmm1 ## 16-byte Reload
; X64-NEXT: roundss $4, %xmm1, %xmm0
; X64-NEXT: roundss $4, (%rsp), %xmm0 ## 16-byte Folded Reload
; X64-NEXT: addq $24, %rsp
; X64-NEXT: retq
;