diff --git a/runtime/arch/arm/quick_entrypoints_arm.S b/runtime/arch/arm/quick_entrypoints_arm.S
index c4ec726..11357b5 100644
--- a/runtime/arch/arm/quick_entrypoints_arm.S
+++ b/runtime/arch/arm/quick_entrypoints_arm.S
@@ -191,7 +191,7 @@
     .cfi_rel_offset r11, 44
     .cfi_rel_offset ip, 48
     .cfi_rel_offset lr, 52
-    vpush {d0-d15}                      @ 32 words of float args.
+    vpush {d0-d15}                      @ 32 words, 2 for each of the 16 saved doubles.
     .cfi_adjust_cfa_offset 128
     sub sp, #8                          @ 2 words of space, alignment padding and Method*
     .cfi_adjust_cfa_offset 8
diff --git a/runtime/arch/arm64/quick_entrypoints_arm64.S b/runtime/arch/arm64/quick_entrypoints_arm64.S
index 4289cab..3e6fbaf 100644
--- a/runtime/arch/arm64/quick_entrypoints_arm64.S
+++ b/runtime/arch/arm64/quick_entrypoints_arm64.S
@@ -331,6 +331,7 @@
 #endif
 
     // Save FP registers.
+    // For better performance, store d0 and d31 separately, so that all STPs are 16-byte aligned.
     str d0,       [sp, #8]
     stp d1, d2,   [sp, #16]
     stp d3, d4,   [sp, #32]
@@ -431,6 +432,7 @@
 
 .macro RESTORE_SAVE_EVERYTHING_FRAME
     // Restore FP registers.
+    // For better performance, load d0 and d31 separately, so that all LDPs are 16-byte aligned.
     ldr d0,       [sp, #8]
     ldp d1, d2,   [sp, #16]
     ldp d3, d4,   [sp, #32]