1 ; RUN: llc < %s -mtriple=thumbv7-none-eabi -mcpu=cortex-m3 | FileCheck %s -check-prefix=CHECK -check-prefix=NONE -check-prefix=NOREGS
2 ; RUN: llc < %s -mtriple=thumbv7-none-eabihf -mcpu=cortex-m4 | FileCheck %s -check-prefix=CHECK -check-prefix=HARD -check-prefix=SP -check-prefix=VFP4-ALL
3 ; RUN: llc < %s -mtriple=thumbv7-none-eabihf -mcpu=cortex-m7 | FileCheck %s -check-prefix=CHECK -check-prefix=HARD -check-prefix=DP -check-prefix=FP-ARMv8
4 ; RUN: llc < %s -mtriple=thumbv7-none-eabihf -mcpu=cortex-a8 | FileCheck %s -check-prefix=CHECK -check-prefix=HARD -check-prefix=DP -check-prefix=VFP4-ALL -check-prefix=VFP4-DP
5 ; RUN: llc < %s -mtriple=thumbv8.1m.main-none-eabihf -mattr=+mve | FileCheck %s -check-prefix=CHECK -check-prefix=NONE -check-prefix=ONLYREGS
7 define float @add_f(float %a, float %b) {
10 ; NONE: {{b|bl}} __aeabi_fadd
11 ; HARD: vadd.f32 s0, s0, s1
12 %0 = fadd float %a, %b
16 define double @add_d(double %a, double %b) {
19 ; NONE: {{b|bl}} __aeabi_dadd
20 ; SP: {{b|bl}} __aeabi_dadd
21 ; DP: vadd.f64 d0, d0, d1
22 %0 = fadd double %a, %b
26 define float @sub_f(float %a, float %b) {
29 ; NONE: {{b|bl}} __aeabi_fsub
31 %0 = fsub float %a, %b
35 define double @sub_d(double %a, double %b) {
38 ; NONE: {{b|bl}} __aeabi_dsub
39 ; SP: {{b|bl}} __aeabi_dsub
40 ; DP: vsub.f64 d0, d0, d1
41 %0 = fsub double %a, %b
45 define float @mul_f(float %a, float %b) {
48 ; NONE: {{b|bl}} __aeabi_fmul
50 %0 = fmul float %a, %b
54 define double @mul_d(double %a, double %b) {
57 ; NONE: {{b|bl}} __aeabi_dmul
58 ; SP: {{b|bl}} __aeabi_dmul
59 ; DP: vmul.f64 d0, d0, d1
60 %0 = fmul double %a, %b
64 define float @div_f(float %a, float %b) {
67 ; NONE: {{b|bl}} __aeabi_fdiv
69 %0 = fdiv float %a, %b
73 define double @div_d(double %a, double %b) {
76 ; NONE: {{b|bl}} __aeabi_ddiv
77 ; SP: {{b|bl}} __aeabi_ddiv
78 ; DP: vdiv.f64 d0, d0, d1
79 %0 = fdiv double %a, %b
83 define float @rem_f(float %a, float %b) {
88 %0 = frem float %a, %b
92 define double @rem_d(double %a, double %b) {
97 %0 = frem double %a, %b
101 define float @load_f(float* %a) {
103 ; CHECK-LABEL: load_f:
105 ; HARD: vldr s0, [r0]
106 %0 = load float, float* %a, align 4
110 define double @load_d(double* %a) {
112 ; CHECK-LABEL: load_d:
113 ; NOREGS: ldm r0, {r0, r1}
114 ; ONLYREGS: vldr d0, [r0]
115 ; HARD: vldr d0, [r0]
116 %0 = load double, double* %a, align 8
120 define void @store_f(float* %a, float %b) {
122 ; CHECK-LABEL: store_f:
124 ; HARD: vstr s0, [r0]
125 store float %b, float* %a, align 4
129 define void @store_d(double* %a, double %b) {
131 ; CHECK-LABEL: store_d:
132 ; NOREGS: strd r2, r3, [r0]
133 ; ONLYREGS: vstr d0, [r0]
134 ; HARD: vstr d0, [r0]
135 store double %b, double* %a, align 8
139 define double @f_to_d(float %a) {
140 ; CHECK-LABEL: f_to_d:
141 ; NONE: bl __aeabi_f2d
143 ; DP: vcvt.f64.f32 d0, s0
144 %1 = fpext float %a to double
148 define float @d_to_f(double %a) {
149 ; CHECK-LABEL: d_to_f:
150 ; NONE: bl __aeabi_d2f
152 ; DP: vcvt.f32.f64 s0, d0
153 %1 = fptrunc double %a to float
157 define i32 @f_to_si(float %a) {
158 ; CHECK-LABEL: f_to_si:
159 ; NONE: bl __aeabi_f2iz
160 ; HARD: vcvt.s32.f32 s0, s0
162 %1 = fptosi float %a to i32
166 define i32 @d_to_si(double %a) {
167 ; CHECK-LABEL: d_to_si:
168 ; NONE: bl __aeabi_d2iz
169 ; SP: vmov r0, r1, d0
170 ; SP: bl __aeabi_d2iz
171 ; DP: vcvt.s32.f64 s0, d0
173 %1 = fptosi double %a to i32
177 define i32 @f_to_ui(float %a) {
178 ; CHECK-LABEL: f_to_ui:
179 ; NONE: bl __aeabi_f2uiz
180 ; HARD: vcvt.u32.f32 s0, s0
182 %1 = fptoui float %a to i32
186 define i32 @d_to_ui(double %a) {
187 ; CHECK-LABEL: d_to_ui:
188 ; NONE: bl __aeabi_d2uiz
189 ; SP: vmov r0, r1, d0
190 ; SP: bl __aeabi_d2uiz
191 ; DP: vcvt.u32.f64 s0, d0
193 %1 = fptoui double %a to i32
197 define float @si_to_f(i32 %a) {
198 ; CHECK-LABEL: si_to_f:
199 ; NONE: bl __aeabi_i2f
200 ; HARD: vcvt.f32.s32 s0, s0
201 %1 = sitofp i32 %a to float
205 define double @si_to_d(i32 %a) {
206 ; CHECK-LABEL: si_to_d:
207 ; NONE: bl __aeabi_i2d
209 ; DP: vcvt.f64.s32 d0, s0
210 %1 = sitofp i32 %a to double
214 define float @ui_to_f(i32 %a) {
215 ; CHECK-LABEL: ui_to_f:
216 ; NONE: bl __aeabi_ui2f
217 ; HARD: vcvt.f32.u32 s0, s0
218 %1 = uitofp i32 %a to float
222 define double @ui_to_d(i32 %a) {
223 ; CHECK-LABEL: ui_to_d:
224 ; NONE: bl __aeabi_ui2d
225 ; SP: bl __aeabi_ui2d
226 ; DP: vcvt.f64.u32 d0, s0
227 %1 = uitofp i32 %a to double
231 define float @bitcast_i_to_f(i32 %a) {
232 ; CHECK-LABEL: bitcast_i_to_f:
235 %1 = bitcast i32 %a to float
239 define double @bitcast_i_to_d(i64 %a) {
240 ; CHECK-LABEL: bitcast_i_to_d:
242 ; HARD: vmov d0, r0, r1
243 %1 = bitcast i64 %a to double
247 define i32 @bitcast_f_to_i(float %a) {
248 ; CHECK-LABEL: bitcast_f_to_i:
251 %1 = bitcast float %a to i32
255 define i64 @bitcast_d_to_i(double %a) {
256 ; CHECK-LABEL: bitcast_d_to_i:
258 ; HARD: vmov r0, r1, d0
259 %1 = bitcast double %a to i64
263 define float @select_f(float %a, float %b, i1 %c) {
264 ; CHECK-LABEL: select_f:
265 ; NOREGS: lsls r2, r2, #31
266 ; NOREGS: moveq r0, r1
267 ; ONLYREGS: lsls r2, r2, #31
268 ; ONLYREGS: vmovne.f32 s2, s0
269 ; HARD: lsls r0, r0, #31
270 ; VFP4-ALL: vmovne.f32 s1, s0
271 ; VFP4-ALL: vmov.f32 s0, s1
272 ; FP-ARMv8: vseleq.f32 s0, s1, s0
273 %1 = select i1 %c, float %a, float %b
277 define double @select_d(double %a, double %b, i1 %c) {
278 ; CHECK-LABEL: select_d:
279 ; NONE: ldr{{(.w)?}} [[REG:r[0-9]+]], [sp]
280 ; NONE ands [[REG]], [[REG]], #1
281 ; NONE-DAG: moveq r0, r2
282 ; NONE-DAG: moveq r1, r3
283 ; SP: ands r0, r0, #1
284 ; SP-DAG: vmov [[ALO:r[0-9]+]], [[AHI:r[0-9]+]], d0
285 ; SP-DAG: vmov [[BLO:r[0-9]+]], [[BHI:r[0-9]+]], d1
287 ; SP-DAG: movne [[BLO]], [[ALO]]
288 ; SP-DAG: movne [[BHI]], [[AHI]]
289 ; SP: vmov d0, [[BLO]], [[BHI]]
290 ; DP: lsls r0, r0, #31
291 ; VFP4-DP: vmovne.f64 d1, d0
292 ; VFP4-DP: vmov.f64 d0, d1
293 ; FP-ARMV8: vseleq.f64 d0, d1, d0
294 %1 = select i1 %c, double %a, double %b