[X86] Add support expanding multiplies by constant where the constant is -3/-5/-9 multplied by a power of 2.

These can be replaced with an LEA, a shift, and a negate. This seems to match what gcc and icc would do.

llvm-svn: 338174

3 files changed, 294 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/mul-constant-i16.ll b/llvm/test/CodeGen/X86/mul-constant-i16.ll
index 737bcc7c864..cf367ecbb98 100644
--- a/llvm/test/CodeGen/X86/mul-constant-i16.ll
+++ b/llvm/test/CodeGen/X86/mul-constant-i16.ll
@@ -766,6 +766,50 @@ define i16 @test_mul_by_520(i16 %x) {
   ret i16 %mul
 }
 
+define i16 @test_mul_by_neg10(i16 %x) {
+; X86-LABEL: test_mul_by_neg10:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl %eax, %eax
+; X86-NEXT:    leal (%eax,%eax,4), %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: test_mul_by_neg10:
+; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    addl %edi, %edi
+; X64-NEXT:    leal (%rdi,%rdi,4), %eax
+; X64-NEXT:    negl %eax
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+  %mul = mul nsw i16 %x, -10
+  ret i16 %mul
+}
+
+define i16 @test_mul_by_neg36(i16 %x) {
+; X86-LABEL: test_mul_by_neg36:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shll $2, %eax
+; X86-NEXT:    leal (%eax,%eax,8), %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    # kill: def $ax killed $ax killed $eax
+; X86-NEXT:    retl
+;
+; X64-LABEL: test_mul_by_neg36:
+; X64:       # %bb.0:
+; X64-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-NEXT:    shll $2, %edi
+; X64-NEXT:    leal (%rdi,%rdi,8), %eax
+; X64-NEXT:    negl %eax
+; X64-NEXT:    # kill: def $ax killed $ax killed $eax
+; X64-NEXT:    retq
+  %mul = mul nsw i16 %x, -36
+  ret i16 %mul
+}
+
 ; (x*9+42)*(x*5+2)
 define i16 @test_mul_spec(i16 %x) nounwind {
 ; X86-LABEL: test_mul_spec:
diff --git a/llvm/test/CodeGen/X86/mul-constant-i32.ll b/llvm/test/CodeGen/X86/mul-constant-i32.ll
index 356d5a00abf..04f867bb4e1 100644
--- a/llvm/test/CodeGen/X86/mul-constant-i32.ll
+++ b/llvm/test/CodeGen/X86/mul-constant-i32.ll
@@ -1997,6 +1997,118 @@ define i32 @test_mul_by_520(i32 %x) {
   ret i32 %mul
 }
 
+define i32 @test_mul_by_neg10(i32 %x) {
+; X86-LABEL: test_mul_by_neg10:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl %eax, %eax
+; X86-NEXT:    leal (%eax,%eax,4), %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    retl
+;
+; X64-HSW-LABEL: test_mul_by_neg10:
+; X64-HSW:       # %bb.0:
+; X64-HSW-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-HSW-NEXT:    addl %edi, %edi # sched: [1:0.25]
+; X64-HSW-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:0.50]
+; X64-HSW-NEXT:    negl %eax # sched: [1:0.25]
+; X64-HSW-NEXT:    retq # sched: [7:1.00]
+;
+; X64-JAG-LABEL: test_mul_by_neg10:
+; X64-JAG:       # %bb.0:
+; X64-JAG-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-JAG-NEXT:    addl %edi, %edi # sched: [1:0.50]
+; X64-JAG-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [2:1.00]
+; X64-JAG-NEXT:    negl %eax # sched: [1:0.50]
+; X64-JAG-NEXT:    retq # sched: [4:1.00]
+;
+; X86-NOOPT-LABEL: test_mul_by_neg10:
+; X86-NOOPT:       # %bb.0:
+; X86-NOOPT-NEXT:    imull $-10, {{[0-9]+}}(%esp), %eax
+; X86-NOOPT-NEXT:    retl
+;
+; HSW-NOOPT-LABEL: test_mul_by_neg10:
+; HSW-NOOPT:       # %bb.0:
+; HSW-NOOPT-NEXT:    imull $-10, %edi, %eax # sched: [3:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [7:1.00]
+;
+; JAG-NOOPT-LABEL: test_mul_by_neg10:
+; JAG-NOOPT:       # %bb.0:
+; JAG-NOOPT-NEXT:    imull $-10, %edi, %eax # sched: [3:1.00]
+; JAG-NOOPT-NEXT:    retq # sched: [4:1.00]
+;
+; X64-SLM-LABEL: test_mul_by_neg10:
+; X64-SLM:       # %bb.0:
+; X64-SLM-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-SLM-NEXT:    addl %edi, %edi # sched: [1:0.50]
+; X64-SLM-NEXT:    leal (%rdi,%rdi,4), %eax # sched: [1:1.00]
+; X64-SLM-NEXT:    negl %eax # sched: [1:0.50]
+; X64-SLM-NEXT:    retq # sched: [4:1.00]
+;
+; SLM-NOOPT-LABEL: test_mul_by_neg10:
+; SLM-NOOPT:       # %bb.0:
+; SLM-NOOPT-NEXT:    imull $-10, %edi, %eax # sched: [3:1.00]
+; SLM-NOOPT-NEXT:    retq # sched: [4:1.00]
+  %mul = mul nsw i32 %x, -10
+  ret i32 %mul
+}
+
+define i32 @test_mul_by_neg36(i32 %x) {
+; X86-LABEL: test_mul_by_neg36:
+; X86:       # %bb.0:
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    shll $2, %eax
+; X86-NEXT:    leal (%eax,%eax,8), %eax
+; X86-NEXT:    negl %eax
+; X86-NEXT:    retl
+;
+; X64-HSW-LABEL: test_mul_by_neg36:
+; X64-HSW:       # %bb.0:
+; X64-HSW-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-HSW-NEXT:    shll $2, %edi # sched: [1:0.50]
+; X64-HSW-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
+; X64-HSW-NEXT:    negl %eax # sched: [1:0.25]
+; X64-HSW-NEXT:    retq # sched: [7:1.00]
+;
+; X64-JAG-LABEL: test_mul_by_neg36:
+; X64-JAG:       # %bb.0:
+; X64-JAG-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-JAG-NEXT:    shll $2, %edi # sched: [1:0.50]
+; X64-JAG-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [2:1.00]
+; X64-JAG-NEXT:    negl %eax # sched: [1:0.50]
+; X64-JAG-NEXT:    retq # sched: [4:1.00]
+;
+; X86-NOOPT-LABEL: test_mul_by_neg36:
+; X86-NOOPT:       # %bb.0:
+; X86-NOOPT-NEXT:    imull $-36, {{[0-9]+}}(%esp), %eax
+; X86-NOOPT-NEXT:    retl
+;
+; HSW-NOOPT-LABEL: test_mul_by_neg36:
+; HSW-NOOPT:       # %bb.0:
+; HSW-NOOPT-NEXT:    imull $-36, %edi, %eax # sched: [3:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [7:1.00]
+;
+; JAG-NOOPT-LABEL: test_mul_by_neg36:
+; JAG-NOOPT:       # %bb.0:
+; JAG-NOOPT-NEXT:    imull $-36, %edi, %eax # sched: [3:1.00]
+; JAG-NOOPT-NEXT:    retq # sched: [4:1.00]
+;
+; X64-SLM-LABEL: test_mul_by_neg36:
+; X64-SLM:       # %bb.0:
+; X64-SLM-NEXT:    # kill: def $edi killed $edi def $rdi
+; X64-SLM-NEXT:    shll $2, %edi # sched: [1:1.00]
+; X64-SLM-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:1.00]
+; X64-SLM-NEXT:    negl %eax # sched: [1:0.50]
+; X64-SLM-NEXT:    retq # sched: [4:1.00]
+;
+; SLM-NOOPT-LABEL: test_mul_by_neg36:
+; SLM-NOOPT:       # %bb.0:
+; SLM-NOOPT-NEXT:    imull $-36, %edi, %eax # sched: [3:1.00]
+; SLM-NOOPT-NEXT:    retq # sched: [4:1.00]
+  %mul = mul nsw i32 %x, -36
+  ret i32 %mul
+}
+
 ; (x*9+42)*(x*5+2)
 define i32 @test_mul_spec(i32 %x) nounwind {
 ; X86-LABEL: test_mul_spec:
diff --git a/llvm/test/CodeGen/X86/mul-constant-i64.ll b/llvm/test/CodeGen/X86/mul-constant-i64.ll
index 332ad7f0129..761ca67ab31 100644
--- a/llvm/test/CodeGen/X86/mul-constant-i64.ll
+++ b/llvm/test/CodeGen/X86/mul-constant-i64.ll
@@ -2107,6 +2107,144 @@ define i64 @test_mul_by_520(i64 %x) {
   ret i64 %mul
 }
 
+define i64 @test_mul_by_neg10(i64 %x) {
+; X86-LABEL: test_mul_by_neg10:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl $-10, %edx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    mull %edx
+; X86-NEXT:    subl %ecx, %edx
+; X86-NEXT:    leal (%esi,%esi,4), %ecx
+; X86-NEXT:    addl %ecx, %ecx
+; X86-NEXT:    subl %ecx, %edx
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-HSW-LABEL: test_mul_by_neg10:
+; X64-HSW:       # %bb.0:
+; X64-HSW-NEXT:    addq %rdi, %rdi # sched: [1:0.25]
+; X64-HSW-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:0.50]
+; X64-HSW-NEXT:    negq %rax # sched: [1:0.25]
+; X64-HSW-NEXT:    retq # sched: [7:1.00]
+;
+; X64-JAG-LABEL: test_mul_by_neg10:
+; X64-JAG:       # %bb.0:
+; X64-JAG-NEXT:    addq %rdi, %rdi # sched: [1:0.50]
+; X64-JAG-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [2:1.00]
+; X64-JAG-NEXT:    negq %rax # sched: [1:0.50]
+; X64-JAG-NEXT:    retq # sched: [4:1.00]
+;
+; X86-NOOPT-LABEL: test_mul_by_neg10:
+; X86-NOOPT:       # %bb.0:
+; X86-NOOPT-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NOOPT-NEXT:    movl $-10, %edx
+; X86-NOOPT-NEXT:    movl %ecx, %eax
+; X86-NOOPT-NEXT:    mull %edx
+; X86-NOOPT-NEXT:    subl %ecx, %edx
+; X86-NOOPT-NEXT:    imull $-10, {{[0-9]+}}(%esp), %ecx
+; X86-NOOPT-NEXT:    addl %ecx, %edx
+; X86-NOOPT-NEXT:    retl
+;
+; HSW-NOOPT-LABEL: test_mul_by_neg10:
+; HSW-NOOPT:       # %bb.0:
+; HSW-NOOPT-NEXT:    imulq $-10, %rdi, %rax # sched: [3:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [7:1.00]
+;
+; JAG-NOOPT-LABEL: test_mul_by_neg10:
+; JAG-NOOPT:       # %bb.0:
+; JAG-NOOPT-NEXT:    imulq $-10, %rdi, %rax # sched: [6:4.00]
+; JAG-NOOPT-NEXT:    retq # sched: [4:1.00]
+;
+; X64-SLM-LABEL: test_mul_by_neg10:
+; X64-SLM:       # %bb.0:
+; X64-SLM-NEXT:    addq %rdi, %rdi # sched: [1:0.50]
+; X64-SLM-NEXT:    leaq (%rdi,%rdi,4), %rax # sched: [1:1.00]
+; X64-SLM-NEXT:    negq %rax # sched: [1:0.50]
+; X64-SLM-NEXT:    retq # sched: [4:1.00]
+;
+; SLM-NOOPT-LABEL: test_mul_by_neg10:
+; SLM-NOOPT:       # %bb.0:
+; SLM-NOOPT-NEXT:    imulq $-10, %rdi, %rax # sched: [3:1.00]
+; SLM-NOOPT-NEXT:    retq # sched: [4:1.00]
+  %mul = mul nsw i64 %x, -10
+  ret i64 %mul
+}
+
+define i64 @test_mul_by_neg36(i64 %x) {
+; X86-LABEL: test_mul_by_neg36:
+; X86:       # %bb.0:
+; X86-NEXT:    pushl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 8
+; X86-NEXT:    .cfi_offset %esi, -8
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
+; X86-NEXT:    movl $-36, %edx
+; X86-NEXT:    movl %ecx, %eax
+; X86-NEXT:    mull %edx
+; X86-NEXT:    subl %ecx, %edx
+; X86-NEXT:    leal (%esi,%esi,8), %ecx
+; X86-NEXT:    shll $2, %ecx
+; X86-NEXT:    subl %ecx, %edx
+; X86-NEXT:    popl %esi
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+;
+; X64-HSW-LABEL: test_mul_by_neg36:
+; X64-HSW:       # %bb.0:
+; X64-HSW-NEXT:    shlq $2, %rdi # sched: [1:0.50]
+; X64-HSW-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
+; X64-HSW-NEXT:    negq %rax # sched: [1:0.25]
+; X64-HSW-NEXT:    retq # sched: [7:1.00]
+;
+; X64-JAG-LABEL: test_mul_by_neg36:
+; X64-JAG:       # %bb.0:
+; X64-JAG-NEXT:    shlq $2, %rdi # sched: [1:0.50]
+; X64-JAG-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [2:1.00]
+; X64-JAG-NEXT:    negq %rax # sched: [1:0.50]
+; X64-JAG-NEXT:    retq # sched: [4:1.00]
+;
+; X86-NOOPT-LABEL: test_mul_by_neg36:
+; X86-NOOPT:       # %bb.0:
+; X86-NOOPT-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-NOOPT-NEXT:    movl $-36, %edx
+; X86-NOOPT-NEXT:    movl %ecx, %eax
+; X86-NOOPT-NEXT:    mull %edx
+; X86-NOOPT-NEXT:    subl %ecx, %edx
+; X86-NOOPT-NEXT:    imull $-36, {{[0-9]+}}(%esp), %ecx
+; X86-NOOPT-NEXT:    addl %ecx, %edx
+; X86-NOOPT-NEXT:    retl
+;
+; HSW-NOOPT-LABEL: test_mul_by_neg36:
+; HSW-NOOPT:       # %bb.0:
+; HSW-NOOPT-NEXT:    imulq $-36, %rdi, %rax # sched: [3:1.00]
+; HSW-NOOPT-NEXT:    retq # sched: [7:1.00]
+;
+; JAG-NOOPT-LABEL: test_mul_by_neg36:
+; JAG-NOOPT:       # %bb.0:
+; JAG-NOOPT-NEXT:    imulq $-36, %rdi, %rax # sched: [6:4.00]
+; JAG-NOOPT-NEXT:    retq # sched: [4:1.00]
+;
+; X64-SLM-LABEL: test_mul_by_neg36:
+; X64-SLM:       # %bb.0:
+; X64-SLM-NEXT:    shlq $2, %rdi # sched: [1:1.00]
+; X64-SLM-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:1.00]
+; X64-SLM-NEXT:    negq %rax # sched: [1:0.50]
+; X64-SLM-NEXT:    retq # sched: [4:1.00]
+;
+; SLM-NOOPT-LABEL: test_mul_by_neg36:
+; SLM-NOOPT:       # %bb.0:
+; SLM-NOOPT-NEXT:    imulq $-36, %rdi, %rax # sched: [3:1.00]
+; SLM-NOOPT-NEXT:    retq # sched: [4:1.00]
+  %mul = mul nsw i64 %x, -36
+  ret i64 %mul
+}
+
 ; (x*9+42)*(x*5+2)
 define i64 @test_mul_spec(i64 %x) nounwind {
 ; X86-LABEL: test_mul_spec: