math/mpx-mul4-*.S: Use more portable type syntax for ambiguous instructions.

[catacomb] / math / mpx-mul4-x86-sse2.S

diff --git a/math/mpx-mul4-x86-sse2.S b/math/mpx-mul4-x86-sse2.S
index 9e1d478..916adef 100644 (file)
--- a/math/mpx-mul4-x86-sse2.S
+++ b/math/mpx-mul4-x86-sse2.S
@@ -1054,15 +1054,17 @@ FUNC(mpxmont_redc4_x86_sse2)
        // Continue carry propagation until the end of the buffer.
 0:     add     [edi], eax
        mov     eax, 0                  // preserves flags
-       adcd    [edi + 4], 0
-       adcd    [edi + 8], 0
-       adcd    [edi + 12], 0
+       adc     dword ptr [edi + 4], 0
+       adc     dword ptr [edi + 8], 0
+       adc     dword ptr [edi + 12], 0
        adc     eax, 0
        add     edi, 16
        cmp     edi, esi
        jb      0b
 
-       // Deal with the tail end.
+       // Deal with the tail end.  Note that the actual destination length
+       // won't be an exact number of blocks of four, so it's safe to just
+       // drop through here.
 7:     add     [edi], eax
        mov     eax, 0
        add     edi, 4
@@ -1070,9 +1072,7 @@ FUNC(mpxmont_redc4_x86_sse2)
        cmp     edi, edx
        jb      7b
 
-       // All done for this iteration.  Start the next.  (This must have at
-       // least one follow-on iteration, or we'd not have started this outer
-       // loop.)
+       // All done for this iteration.  Start the next.
 8:     mov     edi, [SP + 0]           // -> dv[i - 1]
        mov     ebx, [BP + 28]          // -> X = nv[0]
        lea     edx, [SP + 44]          // -> space for Y
@@ -1286,6 +1286,8 @@ FUNC(test_mmul4)
        mov     edi, [BP + 28]
        movdqa  xmm0, [SP + 64]
        movdqa  xmm1, [SP + 80]
+       pshufd  xmm0, xmm0, SHUF(0, 2, 1, 3)
+       pshufd  xmm1, xmm1, SHUF(0, 2, 1, 3)
        movdqu  [edi], xmm0
        movdqu  [edi + 16], xmm1
        testcarryout [BP + 24]
@@ -1302,6 +1304,8 @@ FUNC(test_mmla4)
        mov     edi, [BP + 28]
        movdqa  xmm0, [SP + 64]
        movdqa  xmm1, [SP + 80]
+       pshufd  xmm0, xmm0, SHUF(0, 2, 1, 3)
+       pshufd  xmm1, xmm1, SHUF(0, 2, 1, 3)
        movdqu  [edi], xmm0
        movdqu  [edi + 16], xmm1
        testcarryout [BP + 24]
@@ -1318,6 +1322,8 @@ FUNC(test_mont4)
        mov     edi, [BP + 28]
        movdqa  xmm0, [SP + 64]
        movdqa  xmm1, [SP + 80]
+       pshufd  xmm0, xmm0, SHUF(0, 2, 1, 3)
+       pshufd  xmm1, xmm1, SHUF(0, 2, 1, 3)
        movdqu  [edi], xmm0
        movdqu  [edi + 16], xmm1
        testcarryout [BP + 24]