[llvm] r250794 - AMDGPU: Stop reserving v[254:255]
Matt Arsenault via llvm-commits
llvm-commits at lists.llvm.org
Mon Oct 19 20:59:58 PDT 2015
Author: arsenm
Date: Mon Oct 19 22:59:58 2015
New Revision: 250794
URL: http://llvm.org/viewvc/llvm-project?rev=250794&view=rev
Log:
AMDGPU: Stop reserving v[254:255]
This wasn't doing anything useful. They weren't explicitly used
anywhere, and the RegScavenger ignores reserved registers.
This for some reason caused a random scheduling change in the test.
Getting the check lines to pass is too frustrating, and there's probably
not too much value in checking the vector case's operands N times.
Modified:
llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
llvm/trunk/test/CodeGen/AMDGPU/udivrem.ll
Modified: llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp?rev=250794&r1=250793&r2=250794&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp Mon Oct 19 22:59:58 2015
@@ -41,10 +41,6 @@ BitVector SIRegisterInfo::getReservedReg
reserveRegisterTuples(Reserved, AMDGPU::EXEC);
reserveRegisterTuples(Reserved, AMDGPU::FLAT_SCR);
- // Reserve some VGPRs to use as temp registers in case we have to spill VGPRs
- reserveRegisterTuples(Reserved, AMDGPU::VGPR254);
- reserveRegisterTuples(Reserved, AMDGPU::VGPR255);
-
// Tonga and Iceland can only allocate a fixed number of SGPRs due
// to a hw bug.
if (MF.getSubtarget<AMDGPUSubtarget>().hasSGPRInitBug()) {
Modified: llvm/trunk/test/CodeGen/AMDGPU/udivrem.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/udivrem.ll?rev=250794&r1=250793&r2=250794&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/udivrem.ll (original)
+++ llvm/trunk/test/CodeGen/AMDGPU/udivrem.ll Mon Oct 19 22:59:58 2015
@@ -257,83 +257,83 @@ define void @test_udivrem_v2(<2 x i32> a
; EG-DAG: CNDE_INT
; EG-DAG: CNDE_INT
-; SI-DAG: v_rcp_iflag_f32_e32 [[FIRST_RCP:v[0-9]+]]
-; SI-DAG: v_mul_hi_u32 [[FIRST_RCP_HI:v[0-9]+]], [[FIRST_RCP]]
-; SI-DAG: v_mul_lo_i32 [[FIRST_RCP_LO:v[0-9]+]], [[FIRST_RCP]]
-; SI-DAG: v_sub_i32_e32 [[FIRST_NEG_RCP_LO:v[0-9]+]], vcc, 0, [[FIRST_RCP_LO]]
+; SI-DAG: v_rcp_iflag_f32_e32
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_mul_lo_i32
+; SI-DAG: v_sub_i32_e32
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_mul_hi_u32 [[FIRST_E:v[0-9]+]], {{v[0-9]+}}, [[FIRST_RCP]]
-; SI-DAG: v_add_i32_e32 [[FIRST_RCP_A_E:v[0-9]+]], vcc, [[FIRST_E]], [[FIRST_RCP]]
-; SI-DAG: v_subrev_i32_e32 [[FIRST_RCP_S_E:v[0-9]+]], vcc, [[FIRST_E]], [[FIRST_RCP]]
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_mul_hi_u32 [[FIRST_Quotient:v[0-9]+]]
-; SI-DAG: v_mul_lo_i32 [[FIRST_Num_S_Remainder:v[0-9]+]]
-; SI-DAG: v_subrev_i32_e32 [[FIRST_Remainder:v[l0-9]+]], vcc, [[FIRST_Num_S_Remainder]], v{{[0-9]+}}
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_mul_lo_i32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_and_b32_e32 [[FIRST_Tmp1:v[0-9]+]]
-; SI-DAG: v_add_i32_e32 [[FIRST_Quotient_A_One:v[0-9]+]], {{.*}}, [[FIRST_Quotient]]
-; SI-DAG: v_subrev_i32_e32 [[FIRST_Quotient_S_One:v[0-9]+]],
+; SI-DAG: v_and_b32_e32
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_add_i32_e32 [[FIRST_Remainder_A_Den:v[0-9]+]],
-; SI-DAG: v_subrev_i32_e32 [[FIRST_Remainder_S_Den:v[0-9]+]],
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_rcp_iflag_f32_e32 [[SECOND_RCP:v[0-9]+]]
-; SI-DAG: v_mul_hi_u32 [[SECOND_RCP_HI:v[0-9]+]], [[SECOND_RCP]]
-; SI-DAG: v_mul_lo_i32 [[SECOND_RCP_LO:v[0-9]+]], [[SECOND_RCP]]
-; SI-DAG: v_sub_i32_e32 [[SECOND_NEG_RCP_LO:v[0-9]+]], vcc, 0, [[SECOND_RCP_LO]]
+; SI-DAG: v_rcp_iflag_f32_e32
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_mul_lo_i32
+; SI-DAG: v_sub_i32_e32
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_mul_hi_u32 [[SECOND_E:v[0-9]+]], {{v[0-9]+}}, [[SECOND_RCP]]
-; SI-DAG: v_add_i32_e32 [[SECOND_RCP_A_E:v[0-9]+]], vcc, [[SECOND_E]], [[SECOND_RCP]]
-; SI-DAG: v_subrev_i32_e32 [[SECOND_RCP_S_E:v[0-9]+]], vcc, [[SECOND_E]], [[SECOND_RCP]]
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_mul_hi_u32 [[SECOND_Quotient:v[0-9]+]]
-; SI-DAG: v_mul_lo_i32 [[SECOND_Num_S_Remainder:v[0-9]+]]
-; SI-DAG: v_subrev_i32_e32 [[SECOND_Remainder:v[0-9]+]], vcc, [[SECOND_Num_S_Remainder]], v{{[0-9]+}}
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_mul_lo_i32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_and_b32_e32 [[SECOND_Tmp1:v[0-9]+]]
-; SI-DAG: v_add_i32_e32 [[SECOND_Quotient_A_One:v[0-9]+]], {{.*}}, [[SECOND_Quotient]]
-; SI-DAG: v_subrev_i32_e32 [[SECOND_Quotient_S_One:v[0-9]+]],
+; SI-DAG: v_and_b32_e32
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_add_i32_e32 [[SECOND_Remainder_A_Den:v[0-9]+]],
-; SI-DAG: v_subrev_i32_e32 [[SECOND_Remainder_S_Den:v[0-9]+]],
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_rcp_iflag_f32_e32 [[THIRD_RCP:v[0-9]+]]
-; SI-DAG: v_mul_hi_u32 [[THIRD_RCP_HI:v[0-9]+]], [[THIRD_RCP]]
-; SI-DAG: v_mul_lo_i32 [[THIRD_RCP_LO:v[0-9]+]], [[THIRD_RCP]]
-; SI-DAG: v_sub_i32_e32 [[THIRD_NEG_RCP_LO:v[0-9]+]], vcc, 0, [[THIRD_RCP_LO]]
+; SI-DAG: v_rcp_iflag_f32_e32
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_mul_lo_i32
+; SI-DAG: v_sub_i32_e32
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_mul_hi_u32 [[THIRD_E:v[0-9]+]], {{v[0-9]+}}, [[THIRD_RCP]]
-; SI-DAG: v_add_i32_e32 [[THIRD_RCP_A_E:v[0-9]+]], vcc, [[THIRD_E]], [[THIRD_RCP]]
-; SI-DAG: v_subrev_i32_e32 [[THIRD_RCP_S_E:v[0-9]+]], vcc, [[THIRD_E]], [[THIRD_RCP]]
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_mul_hi_u32 [[THIRD_Quotient:v[0-9]+]]
-; SI-DAG: v_mul_lo_i32 [[THIRD_Num_S_Remainder:v[0-9]+]]
-; SI-DAG: v_subrev_i32_e32 [[THIRD_Remainder:v[0-9]+]], vcc, [[THIRD_Num_S_Remainder]], {{v[0-9]+}}
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_mul_lo_i32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_and_b32_e32 [[THIRD_Tmp1:v[0-9]+]]
-; SI-DAG: v_add_i32_e32 [[THIRD_Quotient_A_One:v[0-9]+]], {{.*}}, [[THIRD_Quotient]]
-; SI-DAG: v_subrev_i32_e32 [[THIRD_Quotient_S_One:v[0-9]+]],
+; SI-DAG: v_and_b32_e32
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_add_i32_e32 [[THIRD_Remainder_A_Den:v[0-9]+]],
-; SI-DAG: v_subrev_i32_e32 [[THIRD_Remainder_S_Den:v[0-9]+]],
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_rcp_iflag_f32_e32 [[FOURTH_RCP:v[0-9]+]]
-; SI-DAG: v_mul_hi_u32 [[FOURTH_RCP_HI:v[0-9]+]], [[FOURTH_RCP]]
-; SI-DAG: v_mul_lo_i32 [[FOURTH_RCP_LO:v[0-9]+]], [[FOURTH_RCP]]
-; SI-DAG: v_sub_i32_e32 [[FOURTH_NEG_RCP_LO:v[0-9]+]], vcc, 0, [[FOURTH_RCP_LO]]
+; SI-DAG: v_rcp_iflag_f32_e32
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_mul_lo_i32
+; SI-DAG: v_sub_i32_e32
; SI-DAG: v_cndmask_b32_e64
-; SI-DAG: v_mul_hi_u32 [[FOURTH_E:v[0-9]+]], {{v[0-9]+}}, [[FOURTH_RCP]]
-; SI-DAG: v_add_i32_e32 [[FOURTH_RCP_A_E:v[0-9]+]], vcc, [[FOURTH_E]], [[FOURTH_RCP]]
-; SI-DAG: v_subrev_i32_e32 [[FOURTH_RCP_S_E:v[0-9]+]], vcc, [[FOURTH_E]], [[FOURTH_RCP]]
+; SI-DAG: v_mul_hi_u32
+; SI-DAG: v_add_i32_e32
+; SI-DAG: v_subrev_i32_e32
; SI-DAG: v_cndmask_b32_e64
; SI: s_endpgm
define void @test_udivrem_v4(<4 x i32> addrspace(1)* %out, <4 x i32> %x, <4 x i32> %y) {
More information about the llvm-commits
mailing list