ppc64le: Trim dead callee-saved spills in poly_tomont

The prologue saved and restored v20-v30, but the body only clobbers
v23, v24, and v27-v30. v20, v21, v22, v25, and v26 were spilled and
reloaded but never used: five dead stxvx/lxvx pairs and 80 bytes of
frame. This saves only the registers the body touches and trims the
frame from 320 to 224 bytes. The Montgomery multiply is unchanged.
The mlkem/ copy was regenerated with scripts/simpasm.

Closes #1722.

Signed-off-by: Scott Boudreaux <scottbphone12@gmail.com>

Author: Crypteaucajun

dev/ppc64le/src/poly_tomont_ppc_asm.S

@@ -73,7 +73,7 @@
 .text
 
 /*
- * Barrett multiplication by R = 2^16 mod q, 
+ * Barrett multiplication by R = 2^16 mod q,
  * mapping each coefficient to the Montgomery domain.
  *   For each lane (a in vdata1..4):
  *     t      = vmhraddshs(a, V_FACTOR_TW, 0) = round(a*factor_tw / 2^15)
@@ -124,40 +124,26 @@
 .global MLK_ASM_NAMESPACE(poly_tomont_ppc_asm)
 .balign 16
 MLK_ASM_FN_SYMBOL(poly_tomont_ppc_asm)
-        stdu    1, -320(1)
+        stdu    1, -224(1)
         mflr    0
 
         /*
-         * TODO: this saves/restores v20-v30, but the body only touches
-         * v23 (vdata3), v24 (vtmp3) and v27-v30 (vresult1..4) in the
-         * callee-saved range. v20, v21, v22, v25, v26 are spilled and
-         * reloaded but never used -- 5 dead stxvx/lxvx pairs plus 80
-         * bytes of frame. Trim the save set (and frame size) to the
-         * registers actually clobbered. Separate change from the doc
-         * pass, since it alters emitted code.
+         * Save/restore only the callee-saved VRs the body clobbers:
+         * v23, v24, v27-v30. v20-v22, v25, v26 were spilled but never
+         * used. Frame trimmed from 320 to 224 bytes.
          */
         li      6, 128
         li      7, 144
         li      8, 160
         li      9, 176
         li      10, 192
         li      11, 208
-        li      12, 224
-        stxvx   32+20, 6, 1
-        stxvx   32+21, 7, 1
-        stxvx   32+22, 8, 1
-        stxvx   32+23, 9, 1
-        stxvx   32+24, 10, 1
-        stxvx   32+25, 11, 1
-        stxvx   32+26, 12, 1
-        li      6, 240
-        li      7, 256
-        li      8, 272
-        li      9, 288
-        stxvx   32+27, 6, 1
-        stxvx   32+28, 7, 1
-        stxvx   32+29, 8, 1
-        stxvx   32+30, 9, 1
+        stxvx   32+23, 6, 1
+        stxvx   32+24, 7, 1
+        stxvx   32+27, 8, 1
+        stxvx   32+28, 9, 1
+        stxvx   32+29, 10, 1
+        stxvx   32+30, 11, 1
 
         li      6, MLK_PPC_NQ_OFFSET
         li      7, MLK_PPC_TOMONT_OFFSET
@@ -199,24 +185,14 @@ MLK_ASM_FN_SYMBOL(poly_tomont_ppc_asm)
         li      9, 176
         li      10, 192
         li      11, 208
-        li      12, 224
-        lxvx    32+20, 6, 1
-        lxvx    32+21, 7, 1
-        lxvx    32+22, 8, 1
-        lxvx    32+23, 9, 1
-        lxvx    32+24, 10, 1
-        lxvx    32+25, 11, 1
-        lxvx    32+26, 12, 1
-        li      6, 240
-        li      7, 256
-        li      8, 272
-        li      9, 288
-        lxvx    32+27, 6, 1
-        lxvx    32+28, 7, 1
-        lxvx    32+29, 8, 1
-        lxvx    32+30, 9, 1
+        lxvx    32+23, 6, 1
+        lxvx    32+24, 7, 1
+        lxvx    32+27, 8, 1
+        lxvx    32+28, 9, 1
+        lxvx    32+29, 10, 1
+        lxvx    32+30, 11, 1
         mtlr    0
-        addi    1, 1, 320
+        addi    1, 1, 224
         blr
 
 /* To facilitate single-compilation-unit (SCU) builds, undefine all macros.

mlkem/src/native/ppc64le/src/poly_tomont_ppc_asm.S

@@ -48,30 +48,20 @@
 .global MLK_ASM_NAMESPACE(poly_tomont_ppc_asm)
 MLK_ASM_FN_SYMBOL(poly_tomont_ppc_asm)
 
-        stdu 1, -320(1)
+        stdu 1, -224(1)
         mflr 0
         li 6, 128
         li 7, 144
         li 8, 160
         li 9, 176
         li 10, 192
         li 11, 208
-        li 12, 224
-        stxvd2x 52, 6, 1
-        stxvd2x 53, 7, 1
-        stxvd2x 54, 8, 1
-        stxvd2x 55, 9, 1
-        stxvd2x 56, 10, 1
-        stxvd2x 57, 11, 1
-        stxvd2x 58, 12, 1
-        li 6, 240
-        li 7, 256
-        li 8, 272
-        li 9, 288
-        stxvd2x 59, 6, 1
-        stxvd2x 60, 7, 1
-        stxvd2x 61, 8, 1
-        stxvd2x 62, 9, 1
+        stxvd2x 55, 6, 1
+        stxvd2x 56, 7, 1
+        stxvd2x 59, 8, 1
+        stxvd2x 60, 9, 1
+        stxvd2x 61, 10, 1
+        stxvd2x 62, 11, 1
         li 6, 0
         li 7, 80
         li 8, 96
@@ -285,24 +275,14 @@ MLK_ASM_FN_SYMBOL(poly_tomont_ppc_asm)
         li 9, 176
         li 10, 192
         li 11, 208
-        li 12, 224
-        lxvd2x 52, 6, 1
-        lxvd2x 53, 7, 1
-        lxvd2x 54, 8, 1
-        lxvd2x 55, 9, 1
-        lxvd2x 56, 10, 1
-        lxvd2x 57, 11, 1
-        lxvd2x 58, 12, 1
-        li 6, 240
-        li 7, 256
-        li 8, 272
-        li 9, 288
-        lxvd2x 59, 6, 1
-        lxvd2x 60, 7, 1
-        lxvd2x 61, 8, 1
-        lxvd2x 62, 9, 1
+        lxvd2x 55, 6, 1
+        lxvd2x 56, 7, 1
+        lxvd2x 59, 8, 1
+        lxvd2x 60, 9, 1
+        lxvd2x 61, 10, 1
+        lxvd2x 62, 11, 1
         mtlr 0
-        addi 1, 1, 320
+        addi 1, 1, 224
         blr
 
 MLK_ASM_FN_SIZE(poly_tomont_ppc_asm)