[llvm] 0ba6d6d - [AArch64] Correct am_indexed used in bitcast loadext patterns. (#164588)

via llvm-commits llvm-commits at lists.llvm.org
Thu Oct 23 07:03:30 PDT 2025 

Author: David Green
Date: 2025-10-23T15:03:26+01:00
New Revision: 0ba6d6d41659fd375a43f1a646ffab7c075637f8

URL: https://github.com/llvm/llvm-project/commit/0ba6d6d41659fd375a43f1a646ffab7c075637f8
DIFF: https://github.com/llvm/llvm-project/commit/0ba6d6d41659fd375a43f1a646ffab7c075637f8.diff

LOG: [AArch64] Correct am_indexed used in bitcast loadext patterns. (#164588)

The i8 versions were using the wrong index modes, we didn't have enough
test cases.

Added: 
    

Modified: 
    llvm/lib/Target/AArch64/AArch64InstrInfo.td
    llvm/test/CodeGen/AArch64/load-zext-bitcast.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.td b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
index f788c7510f80c..92f260f408674 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.td
@@ -4005,24 +4005,20 @@ def : Pat<(i64 (zextloadi32 (am_indexed32 GPR64sp:$Rn, uimm12s4:$offset))),
       (SUBREG_TO_REG (i64 0), (LDRWui GPR64sp:$Rn, uimm12s4:$offset), sub_32)>;
 
 // load zero-extended i32, bitcast to f64
-def : Pat <(f64 (bitconvert (i64 (zextloadi32 (am_indexed32 GPR64sp:$Rn, uimm12s4:$offset))))),
-       (SUBREG_TO_REG (i64 0), (LDRSui GPR64sp:$Rn, uimm12s4:$offset), ssub)>;
-
+def : Pat<(f64 (bitconvert (i64 (zextloadi32 (am_indexed32 GPR64sp:$Rn, uimm12s4:$offset))))),
+          (SUBREG_TO_REG (i64 0), (LDRSui GPR64sp:$Rn, uimm12s4:$offset), ssub)>;
 // load zero-extended i16, bitcast to f64
-def : Pat <(f64 (bitconvert (i64 (zextloadi16 (am_indexed32 GPR64sp:$Rn, uimm12s2:$offset))))),
-           (SUBREG_TO_REG (i64 0), (LDRHui GPR64sp:$Rn, uimm12s2:$offset), hsub)>;
-
+def : Pat<(f64 (bitconvert (i64 (zextloadi16 (am_indexed16 GPR64sp:$Rn, uimm12s2:$offset))))),
+          (SUBREG_TO_REG (i64 0), (LDRHui GPR64sp:$Rn, uimm12s2:$offset), hsub)>;
 // load zero-extended i8, bitcast to f64
-def : Pat <(f64 (bitconvert (i64 (zextloadi8 (am_indexed32 GPR64sp:$Rn, uimm12s1:$offset))))),
-           (SUBREG_TO_REG (i64 0), (LDRBui GPR64sp:$Rn, uimm12s1:$offset), bsub)>;
-
+def : Pat<(f64 (bitconvert (i64 (zextloadi8 (am_indexed8 GPR64sp:$Rn, uimm12s1:$offset))))),
+          (SUBREG_TO_REG (i64 0), (LDRBui GPR64sp:$Rn, uimm12s1:$offset), bsub)>;
 // load zero-extended i16, bitcast to f32
-def : Pat <(f32 (bitconvert (i32 (zextloadi16 (am_indexed16 GPR64sp:$Rn, uimm12s2:$offset))))),
-           (SUBREG_TO_REG (i32 0), (LDRHui GPR64sp:$Rn, uimm12s2:$offset), hsub)>;
-
+def : Pat<(f32 (bitconvert (i32 (zextloadi16 (am_indexed16 GPR64sp:$Rn, uimm12s2:$offset))))),
+          (SUBREG_TO_REG (i32 0), (LDRHui GPR64sp:$Rn, uimm12s2:$offset), hsub)>;
 // load zero-extended i8, bitcast to f32
-def : Pat <(f32 (bitconvert (i32 (zextloadi8 (am_indexed16 GPR64sp:$Rn, uimm12s1:$offset))))),
-           (SUBREG_TO_REG (i32 0), (LDRBui GPR64sp:$Rn, uimm12s1:$offset), bsub)>;
+def : Pat<(f32 (bitconvert (i32 (zextloadi8 (am_indexed8 GPR64sp:$Rn, uimm12s1:$offset))))),
+          (SUBREG_TO_REG (i32 0), (LDRBui GPR64sp:$Rn, uimm12s1:$offset), bsub)>;
 
 // Pre-fetch.
 def PRFMui : PrefetchUI<0b11, 0, 0b10, "prfm",

diff  --git a/llvm/test/CodeGen/AArch64/load-zext-bitcast.ll b/llvm/test/CodeGen/AArch64/load-zext-bitcast.ll
index 9193025264cc0..6177ae5f05020 100644
--- a/llvm/test/CodeGen/AArch64/load-zext-bitcast.ll
+++ b/llvm/test/CodeGen/AArch64/load-zext-bitcast.ll
@@ -112,8 +112,7 @@ entry:
 define double @load_u64_from_u8_off1(ptr %n){
 ; CHECK-LABEL: load_u64_from_u8_off1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrb w8, [x0, #1]
-; CHECK-NEXT:    fmov d0, x8
+; CHECK-NEXT:    ldr b0, [x0, #1]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 1
@@ -140,8 +139,7 @@ entry:
 define float @load_u32_from_u8_off1(ptr %n){
 ; CHECK-LABEL: load_u32_from_u8_off1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrb w8, [x0, #1]
-; CHECK-NEXT:    fmov s0, w8
+; CHECK-NEXT:    ldr b0, [x0, #1]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 1
@@ -154,8 +152,7 @@ entry:
 define half @load_u16_from_u8_off1(ptr %n){
 ; CHECK-LABEL: load_u16_from_u8_off1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrb w8, [x0, #1]
-; CHECK-NEXT:    fmov s0, w8
+; CHECK-NEXT:    ldr b0, [x0, #1]
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $s0
 ; CHECK-NEXT:    ret
 entry:
@@ -185,8 +182,7 @@ entry:
 define double @load_u64_from_u16_off2(ptr %n){
 ; CHECK-LABEL: load_u64_from_u16_off2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrh w8, [x0, #2]
-; CHECK-NEXT:    fmov d0, x8
+; CHECK-NEXT:    ldr h0, [x0, #2]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 2
@@ -199,8 +195,7 @@ entry:
 define double @load_u64_from_u8_off2(ptr %n){
 ; CHECK-LABEL: load_u64_from_u8_off2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrb w8, [x0, #2]
-; CHECK-NEXT:    fmov d0, x8
+; CHECK-NEXT:    ldr b0, [x0, #2]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 2
@@ -226,7 +221,7 @@ entry:
 define float @load_u32_from_u8_off2(ptr %n){
 ; CHECK-LABEL: load_u32_from_u8_off2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #1]
+; CHECK-NEXT:    ldr b0, [x0, #2]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 2
@@ -239,7 +234,7 @@ entry:
 define half @load_u16_from_u8_off2(ptr %n){
 ; CHECK-LABEL: load_u16_from_u8_off2:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #1]
+; CHECK-NEXT:    ldr b0, [x0, #2]
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $s0
 ; CHECK-NEXT:    ret
 entry:
@@ -283,8 +278,7 @@ entry:
 define double @load_u64_from_u8_off255(ptr %n){
 ; CHECK-LABEL: load_u64_from_u8_off255:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrb w8, [x0, #255]
-; CHECK-NEXT:    fmov d0, x8
+; CHECK-NEXT:    ldr b0, [x0, #255]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 255
@@ -311,8 +305,7 @@ entry:
 define float @load_u32_from_u8_off255(ptr %n){
 ; CHECK-LABEL: load_u32_from_u8_off255:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrb w8, [x0, #255]
-; CHECK-NEXT:    fmov s0, w8
+; CHECK-NEXT:    ldr b0, [x0, #255]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 255
@@ -325,8 +318,7 @@ entry:
 define half @load_u16_from_u8_off255(ptr %n){
 ; CHECK-LABEL: load_u16_from_u8_off255:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldrb w8, [x0, #255]
-; CHECK-NEXT:    fmov s0, w8
+; CHECK-NEXT:    ldr b0, [x0, #255]
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $s0
 ; CHECK-NEXT:    ret
 entry:
@@ -354,7 +346,7 @@ entry:
 define double @load_u64_from_u16_off256(ptr %n){
 ; CHECK-LABEL: load_u64_from_u16_off256:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr h0, [x0, #128]
+; CHECK-NEXT:    ldr h0, [x0, #256]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 256
@@ -367,7 +359,7 @@ entry:
 define double @load_u64_from_u8_off256(ptr %n){
 ; CHECK-LABEL: load_u64_from_u8_off256:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #64]
+; CHECK-NEXT:    ldr b0, [x0, #256]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 256
@@ -393,7 +385,7 @@ entry:
 define float @load_u32_from_u8_off256(ptr %n){
 ; CHECK-LABEL: load_u32_from_u8_off256:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #128]
+; CHECK-NEXT:    ldr b0, [x0, #256]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 256
@@ -406,7 +398,7 @@ entry:
 define half @load_u16_from_u8_off256(ptr %n){
 ; CHECK-LABEL: load_u16_from_u8_off256:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #128]
+; CHECK-NEXT:    ldr b0, [x0, #256]
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $s0
 ; CHECK-NEXT:    ret
 entry:
@@ -435,8 +427,7 @@ entry:
 define double @load_u64_from_u16_offn(ptr %n){
 ; CHECK-LABEL: load_u64_from_u16_offn:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    mov w8, #8190 // =0x1ffe
-; CHECK-NEXT:    ldr h0, [x0, x8]
+; CHECK-NEXT:    ldr h0, [x0, #8190]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 8190
@@ -517,7 +508,8 @@ entry:
 define double @load_u64_from_u16_offnp1(ptr %n){
 ; CHECK-LABEL: load_u64_from_u16_offnp1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr h0, [x0, #4096]
+; CHECK-NEXT:    add x8, x0, #2, lsl #12 // =8192
+; CHECK-NEXT:    ldr h0, [x8]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 8192
@@ -530,7 +522,8 @@ entry:
 define double @load_u64_from_u8_offnp1(ptr %n){
 ; CHECK-LABEL: load_u64_from_u8_offnp1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #1024]
+; CHECK-NEXT:    add x8, x0, #1, lsl #12 // =4096
+; CHECK-NEXT:    ldr b0, [x8]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 4096
@@ -557,7 +550,8 @@ entry:
 define float @load_u32_from_u8_offnp1(ptr %n){
 ; CHECK-LABEL: load_u32_from_u8_offnp1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #2048]
+; CHECK-NEXT:    add x8, x0, #1, lsl #12 // =4096
+; CHECK-NEXT:    ldr b0, [x8]
 ; CHECK-NEXT:    ret
 entry:
   %p = getelementptr i8, ptr %n, i64 4096
@@ -570,7 +564,8 @@ entry:
 define half @load_u16_from_u8_offnp1(ptr %n){
 ; CHECK-LABEL: load_u16_from_u8_offnp1:
 ; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    ldr b0, [x0, #2048]
+; CHECK-NEXT:    add x8, x0, #1, lsl #12 // =4096
+; CHECK-NEXT:    ldr b0, [x8]
 ; CHECK-NEXT:    // kill: def $h0 killed $h0 killed $s0
 ; CHECK-NEXT:    ret
 entry:

More information about the llvm-commits mailing list