[X86] Fix SLM scheduler model for PMULLD (PR37059)

Adjust the PMULLD entry to match the Intel AoM numbers - PMULLD is a uop nightmare on SLM and we should model it as such.

We had reports of internal regressions the last time this was attempted (rG13a0f83a05ff), but no public repros, and tests I did last year when I had access to a SLM box failed to see anything. My hunch is that the more aggressive PMULLD -> PMADDWD folds we now perform might have helped. We can revisit this again if we ever receive an actual repro.

Fixes #36407

llvm/lib/Target/X86/X86ScheduleSLM.td

@@ -377,10 +377,8 @@ defm : SLMWriteResPair<WriteVecIMul,  [SLM_FPC_RSV0],   4>;
 defm : SLMWriteResPair<WriteVecIMulX, [SLM_FPC_RSV0],   5, [2]>;
 defm : SLMWriteResPair<WriteVecIMulY, [SLM_FPC_RSV0],   5, [2]>;
 defm : X86WriteResPairUnsupported<WriteVecIMulZ>;
-// FIXME: The below is closer to correct, but caused some perf regressions.
-//defm : SLMWriteResPair<WritePMULLD,  [SLM_FPC_RSV0],   11, [11], 7>;
-defm : SLMWriteResPair<WritePMULLD,  [SLM_FPC_RSV0],   4>;
-defm : SLMWriteResPair<WritePMULLDY, [SLM_FPC_RSV0],   4>;
+defm : SLMWriteResPair<WritePMULLD,  [SLM_FPC_RSV0],   11, [11], 7>;
+defm : X86WriteResPairUnsupported<WritePMULLDY>;
 defm : X86WriteResPairUnsupported<WritePMULLDZ>;
 defm : SLMWriteResPair<WriteShuffle,  [SLM_FPC_RSV0],  1>;
 defm : SLMWriteResPair<WriteShuffleY, [SLM_FPC_RSV0],  1>;

llvm/test/CodeGen/X86/slow-pmulld.ll

@@ -901,10 +901,10 @@ define <4 x i32> @test_mul_v4i32_v4i16_minsize(<4 x i16> %A) minsize {
 define <8 x i32> @test_mul_v8i32_v8i16_minsize(<8 x i16> %A) minsize {
 ; SLM-LABEL: test_mul_v8i32_v8i16_minsize:
 ; SLM:       # %bb.0:
-; SLM-NEXT:    pxor %xmm1, %xmm1
-; SLM-NEXT:    pmovzxwd {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; SLM-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
 ; SLM-NEXT:    movdqa {{.*#+}} xmm1 = [18778,18778,18778,18778]
+; SLM-NEXT:    pmovzxwd {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
+; SLM-NEXT:    pxor %xmm3, %xmm3
+; SLM-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm3[4],xmm0[5],xmm3[5],xmm0[6],xmm3[6],xmm0[7],xmm3[7]
 ; SLM-NEXT:    pmulld %xmm1, %xmm2
 ; SLM-NEXT:    pmulld %xmm0, %xmm1
 ; SLM-NEXT:    movdqa %xmm2, %xmm0
@@ -946,13 +946,13 @@ define <8 x i32> @test_mul_v8i32_v8i16_minsize(<8 x i16> %A) minsize {
 define <16 x i32> @test_mul_v16i32_v16i16_minsize(<16 x i16> %A) minsize {
 ; SLM-LABEL: test_mul_v16i32_v16i16_minsize:
 ; SLM:       # %bb.0:
+; SLM-NEXT:    movdqa {{.*#+}} xmm3 = [18778,18778,18778,18778]
 ; SLM-NEXT:    movdqa %xmm0, %xmm4
-; SLM-NEXT:    pxor %xmm3, %xmm3
+; SLM-NEXT:    pxor %xmm5, %xmm5
 ; SLM-NEXT:    pmovzxwd {{.*#+}} xmm2 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero
 ; SLM-NEXT:    pmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
-; SLM-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7]
-; SLM-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; SLM-NEXT:    movdqa {{.*#+}} xmm3 = [18778,18778,18778,18778]
+; SLM-NEXT:    punpckhwd {{.*#+}} xmm4 = xmm4[4],xmm5[4],xmm4[5],xmm5[5],xmm4[6],xmm5[6],xmm4[7],xmm5[7]
+; SLM-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm5[4],xmm1[5],xmm5[5],xmm1[6],xmm5[6],xmm1[7],xmm5[7]
 ; SLM-NEXT:    pmulld %xmm3, %xmm4
 ; SLM-NEXT:    pmulld %xmm3, %xmm0
 ; SLM-NEXT:    pmulld %xmm3, %xmm2

llvm/test/tools/llvm-mca/X86/SLM/resources-sse41.s

@@ -239,8 +239,8 @@ roundss     $1, (%rax), %xmm2
 # CHECK-NEXT:  1      4     1.00    *                   pmovzxwq	(%rax), %xmm2
 # CHECK-NEXT:  1      5     2.00                        pmuldq	%xmm0, %xmm2
 # CHECK-NEXT:  1      8     2.00    *                   pmuldq	(%rax), %xmm2
-# CHECK-NEXT:  1      4     1.00                        pmulld	%xmm0, %xmm2
-# CHECK-NEXT:  1      7     1.00    *                   pmulld	(%rax), %xmm2
+# CHECK-NEXT:  7      11    11.00                       pmulld	%xmm0, %xmm2
+# CHECK-NEXT:  7      14    11.00   *                   pmulld	(%rax), %xmm2
 # CHECK-NEXT:  1      1     0.50                        ptest	%xmm0, %xmm1
 # CHECK-NEXT:  1      4     1.00    *                   ptest	(%rax), %xmm1
 # CHECK-NEXT:  1      3     1.00                        roundpd	$1, %xmm0, %xmm2
@@ -264,7 +264,7 @@ roundss     $1, (%rax), %xmm2
 
 # CHECK:      Resource pressure per iteration:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]
-# CHECK-NEXT:  -      -      -     108.00 65.00   -      -     54.00
+# CHECK-NEXT:  -      -      -     128.00 65.00   -      -     54.00
 
 # CHECK:      Resource pressure by instruction:
 # CHECK-NEXT: [0]    [1]    [2]    [3]    [4]    [5]    [6]    [7]    Instructions:
@@ -352,8 +352,8 @@ roundss     $1, (%rax), %xmm2
 # CHECK-NEXT:  -      -      -     1.00    -      -      -     1.00   pmovzxwq	(%rax), %xmm2
 # CHECK-NEXT:  -      -      -     2.00    -      -      -      -     pmuldq	%xmm0, %xmm2
 # CHECK-NEXT:  -      -      -     2.00    -      -      -     1.00   pmuldq	(%rax), %xmm2
-# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     pmulld	%xmm0, %xmm2
-# CHECK-NEXT:  -      -      -     1.00    -      -      -     1.00   pmulld	(%rax), %xmm2
+# CHECK-NEXT:  -      -      -     11.00   -      -      -      -     pmulld	%xmm0, %xmm2
+# CHECK-NEXT:  -      -      -     11.00   -      -      -     1.00   pmulld	(%rax), %xmm2
 # CHECK-NEXT:  -      -      -     0.50   0.50    -      -      -     ptest	%xmm0, %xmm1
 # CHECK-NEXT:  -      -      -     0.50   0.50    -      -     1.00   ptest	(%rax), %xmm1
 # CHECK-NEXT:  -      -      -      -     1.00    -      -      -     roundpd	$1, %xmm0, %xmm2